Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
awesome-generative-ai-guide
A one stop repository for generative AI research updates, interview resources, notebooks and much more!
https://github.com/aishwaryanr/awesome-generative-ai-guide
Last synced: 1 day ago
JSON representation
-
:star: Best Gen AI Papers List (June 2024)
- SELF-TUNING: Instructing LLMs to Effectively Acquire New Knowledge through Self-Teaching - to-date information due to their one-time training and the constantly evolving nature of the world. To keep LLMs current, existing approaches typically involve continued pre-training on new documents. However, they frequently face difficulties in extracting stored knowledge. Motivated by the remarkable success of the Feynman Technique in efficient human learning, we introduce SELFTUNING, a learning framework aimed at improving an LLM’s ability to effectively acquire new knowledge from raw documents through self-teaching. Specifically, we develop a SELFTEACHING strategy that augments the documents with a set of knowledge-intensive tasks created in a self-supervised manner, focusing on three crucial aspects: memorization, comprehension, and self-reflection. In addition, we introduce three Wiki-Newpages-2023-QA datasets to facilitate an in-depth analysis of an LLM’s knowledge acquisition ability concerning memorization, extraction, and reasoning. Extensive experimental results on LLAMA2 family models reveal that SELF-TUNING consistently exhibits superior performance across all knowledge acquisition tasks and excels in preserving previous knowledge.1 | LLM Training, Knowledge Integration |
- DeepSeek-Coder-V2 - Coder-V2, an open-source Mixture-of-Experts (MoE) code language model that achieves performance comparable to GPT4-Turbo in code-specific tasks. Specifically, DeepSeek-Coder-V2 is further pre-trained from an intermediate checkpoint of DeepSeek-V2 with additional 6 trillion tokens. Through this continued pre-training, DeepSeek-Coder-V2 substantially enhances the coding and mathematical reasoning capabilities of DeepSeek-V2, while maintaining comparable performance in general language tasks. Compared to DeepSeek-Coder-33B, DeepSeek-Coder-V2 demonstrates significant advancements in various aspects of code-related tasks, as well as reasoning and general capabilities. Additionally, DeepSeek-Coder-V2 expands its support for programming languages from 86 to 338, while extending the context length from 16K to 128K. | Domain Specific LLMs |
- Never Miss A Beat: An Efficient Recipe for Context Window Extension of Large Language Models with Consistent "Middle" Enhancement - trained large language models (LLMs), but they often require fine-tuning at the target length (≫4K) and struggle to effectively utilize information from the middle part of the context. To address these issues, we propose Continuity-Relativity indExing with gAussian Middle (CREAM), which interpolates positional encodings by manipulating position indices. Apart from being simple, CREAM is training-efficient: it only requires fine-tuning at the pre-trained context window (eg, Llama 2-4K) and can extend LLMs to a much longer target context length (eg, 256K). To ensure that the model focuses more on the information in the middle, we introduce a truncated Gaussian to encourage sampling from the middle part of the context during fine-tuning, thus alleviating the ``Lost-in-the-Middle'' problem faced by long-context LLMs. Experimental results show that CREAM successfully extends LLMs to the target length for both Base and Chat versions of 𝙻𝚕𝚊𝚖𝚊𝟸-𝟽𝙱 with ``Never Miss A Beat''. Our code will be publicly available soon. | Context Length |
- Step-DPO: Step-wise Preference Optimization for Long-chain Reasoning of LLMs - chain mathematical reasoning, as models employing DPO struggle to identify detailed errors in incorrect answers. This limitation stems from a lack of fine-grained process supervision. We propose a simple, effective, and data-efficient method called Step-DPO, which treats individual reasoning steps as units for preference optimization rather than evaluating answers holistically. Additionally, we have developed a data construction pipeline for Step-DPO, enabling the creation of a high-quality dataset containing 10K step-wise preference pairs. We also observe that in DPO, self-generated data is more effective than data generated by humans or GPT-4, due to the latter's out-of-distribution nature. Our findings demonstrate that as few as 10K preference data pairs and fewer than 500 Step-DPO training steps can yield a nearly 3% gain in accuracy on MATH for models with over 70B parameters. Notably, Step-DPO, when applied to Qwen2-72B-Instruct, achieves scores of 70.8% and 94.0% on the test sets of MATH and GSM8K, respectively, surpassing a series of closed-source models, including GPT-4-1106, Claude-3-Opus, and Gemini-1.5-Pro. | Mathematical Reasoning, Optimization |
- Scaling Synthetic Data Creation with 1,000,000,000 Personas - driven data synthesis methodology that leverages various perspectives within a large language model (LLM) to create diverse synthetic data. To fully exploit this methodology at scale, we introduce Persona Hub -- a collection of 1 billion diverse personas automatically curated from web data. These 1 billion personas (~13% of the world's total population), acting as distributed carriers of world knowledge, can tap into almost every perspective encapsulated within the LLM, thereby facilitating the creation of diverse synthetic data at scale for various scenarios. By showcasing Persona Hub's use cases in synthesizing high-quality mathematical and logical reasoning problems, instructions (i.e., user prompts), knowledge-rich texts, game NPCs and tools (functions) at scale, we demonstrate persona-driven data synthesis is versatile, scalable, flexible, and easy to use, potentially driving a paradigm shift in synthetic data creation and applications in practice, which may have a profound impact on LLM research and development | Synthetic Data Generation |
- WildTeaming at Scale: From In-the-Wild Jailbreaks to (Adversarially) Safer Language Models - teaming framework that mines in-the-wild user-chatbot interactions to discover 5.7K unique clusters of novel jailbreak tactics, and then composes multiple tactics for systematic exploration of novel jailbreaks. Compared to prior work that performed red-teaming via recruited human workers, gradient-based optimization, or iterative revision with LLMs, our work investigates jailbreaks from chatbot users who were not specifically instructed to break the system. WildTeaming reveals previously unidentified vulnerabilities of frontier LLMs, resulting in up to 4.6x more diverse and successful adversarial attacks compared to state-of-the-art jailbreak methods. While many datasets exist for jailbreak evaluation, very few open-source datasets exist for jailbreak training, as safety training data has been closed even when model weights are open. With WildTeaming we create WildJailbreak, a large-scale open-source synthetic safety dataset with 262K vanilla (direct request) and adversarial (complex jailbreak) prompt-response pairs. To mitigate exaggerated safety behaviors, WildJailbreak provides two contrastive types of queries: 1) harmful queries (vanilla & adversarial) and 2) benign queries that resemble harmful queries in form but contain no harm. As WildJailbreak considerably upgrades the quality and scale of existing safety resources, it uniquely enables us to examine the scaling effects of data and the interplay of data properties and model capabilities during safety training. Through extensive experiments, we identify the training properties that enable an ideal balance of safety behaviors: appropriate safeguarding without over-refusal, effective handling of vanilla and adversarial queries, and minimal, if any, decrease in general capabilities. All components of WildJailbeak contribute to achieving balanced safety behaviors of models. | Red Teaming, LLM Attacks |
- LiveBench: A Challenging, Contamination-Free LLM Benchmark - documented obstacle for fair LLM evaluation and can quickly render benchmarks obsolete. To mitigate this, many recent benchmarks crowdsource new prompts and evaluations from human or LLM judges; however, these can introduce significant biases, and break down when scoring hard questions. In this work, we introduce a new benchmark for LLMs designed to be immune to both test set contamination and the pitfalls of LLM judging and human crowdsourcing. We release LiveBench, the first benchmark that (1) contains frequently-updated questions from recent information sources, (2) scores answers automatically according to objective ground-truth values, and (3) contains a wide variety of challenging tasks, spanning math, coding, reasoning, language, instruction following, and data analysis. To achieve this, LiveBench contains questions that are based on recently-released math competitions, arXiv papers, news articles, and datasets, and it contains harder, contamination-free versions of tasks from previous benchmarks such as Big-Bench Hard, AMPS, and IFEval. We evaluate many prominent closed-source models, as well as dozens of open-source models ranging from 0.5B to 110B in size. LiveBench is difficult, with top models achieving below 65% accuracy. We release all questions, code, and model answers. Questions will be added and updated on a monthly basis, and we will release new tasks and harder versions of tasks over time so that LiveBench can distinguish between the capabilities of LLMs as they improve in the future. We welcome community engagement and collaboration for expanding the benchmark tasks and models. | Benchmark, Dataset |
- Understand What LLM Needs: Dual Preference Alignment for Retrieval-Augmented Generation - augmented generation (RAG) has demonstrated effectiveness in mitigating the hallucination problem of large language models (LLMs). However, the difficulty of aligning the retriever with the diverse LLMs' knowledge preferences inevitably poses an inevitable challenge in developing a reliable RAG system. To address this issue, we propose DPA-RAG, a universal framework designed to align diverse knowledge preferences within RAG systems. Specifically, we initially introduce a preference knowledge construction pipline and incorporate five novel query augmentation strategies to alleviate preference data scarcity. Based on preference data, DPA-RAG accomplishes both external and internal preference alignment: 1) It jointly integrate pair-wise, point-wise, and contrastive preference alignment abilities into the reranker, achieving external preference alignment among RAG components. 2) It further introduces a pre-aligned stage before vanilla Supervised Fine-tuning (SFT), enabling LLMs to implicitly capture knowledge aligned with their reasoning preferences, achieving LLMs' internal alignment. Experimental results across four knowledge-intensive QA datasets demonstrate that DPA-RAG outperforms all baselines and seamlessly integrates both black-box and open-sourced LLM readers. Further qualitative analysis and discussions also provide empirical guidance for achieving reliable RAG systems. | RAG, Alignment |
- LongRAG: Enhancing Retrieval-Augmented Generation with Long-context LLMs - word Wikipedia paragraphs. Such a design forces the retriever to search over a large corpus to find the `needle' unit. In contrast, the readers only need to extract answers from the short retrieved units. Such an imbalanced `heavy' retriever and `light' reader design can lead to sub-optimal performance. In order to alleviate the imbalance, we propose a new framework LongRAG, consisting of a `long retriever' and a `long reader'. LongRAG processes the entire Wikipedia into 4K-token units, which is 30x longer than before. By increasing the unit size, we significantly reduce the total units from 22M to 700K. This significantly lowers the burden of retriever, which leads to a remarkable retrieval score: answer recall@1=71% on NQ (previously 52%) and answer recall@2=72% (previously 47%) on HotpotQA (full-wiki). Then we feed the top-k retrieved units (approx 30K tokens) to an existing long-context LLM to perform zero-shot answer extraction. Without requiring any training, LongRAG achieves an EM of 62.7% on NQ, which is the best known result. LongRAG also achieves 64.3% on HotpotQA (full-wiki), which is on par of the SoTA model. Our study offers insights into the future roadmap for combining RAG with long-context LLMs. | RAG |
- Claude 3.5 Sonnet - tier model, Claude 3 Sonnet. | Foundational LLM |
- Can LLMs Learn by Teaching? A Preliminary Study - produced data or stronger models. In this paper, we provide a preliminary exploration of this ambitious agenda. We show that LbT ideas can be incorporated into existing LLM training/prompting pipelines and provide noticeable improvements. Specifically, we design three methods, each mimicking one of the three levels of LbT in humans: observing students' feedback, learning from the feedback, and learning iteratively, with the goals of improving answer accuracy without training and improving models' inherent capability with fine-tuning. The findings are encouraging. For example, similar to LbT in human, we see that: (1) LbT can induce weak-to-strong generalization: strong models can improve themselves by teaching other weak models; (2) Diversity in students might help: teaching multiple students could be better than teaching one student or the teacher itself. We hope that this early promise can inspire future research on LbT and more broadly adopting the advanced techniques in education to improve LLMs. The code is available at https://github.com/imagination-research/lbt. | LLM learning |
- Can Long-Context Language Models Subsume Retrieval, RAG, SQL, and More? - context language models (LCLMs) have the potential to revolutionize our approach to tasks traditionally reliant on external tools like retrieval systems or databases. Leveraging LCLMs' ability to natively ingest and process entire corpora of information offers numerous advantages. It enhances user-friendliness by eliminating the need for specialized knowledge of tools, provides robust end-to-end modeling that minimizes cascading errors in complex pipelines, and allows for the application of sophisticated prompting techniques across the entire system. To assess this paradigm shift, we introduce LOFT, a benchmark of real-world tasks requiring context up to millions of tokens designed to evaluate LCLMs' performance on in-context retrieval and reasoning. Our findings reveal LCLMs' surprising ability to rival state-of-the-art retrieval and RAG systems, despite never having been explicitly trained for these tasks. However, LCLMs still face challenges in areas like compositional reasoning that are required in SQL-like tasks. Notably, prompting strategies significantly influence performance, emphasizing the need for continued research as context lengths grow. Overall, LOFT provides a rigorous testing ground for LCLMs, showcasing their potential to supplant existing paradigms and tackle novel tasks as model capabilities scale. | Long Context, Analysis |
- Judging the Judges: Evaluating Alignment and Vulnerabilities in LLMs-as-Judges - as-a-judge paradigm is rapidly gaining traction as an approach to evaluating large language models (LLMs). However, there are still many open questions about the strengths and weaknesses of this paradigm, and what potential biases it may hold. In this paper, we present a comprehensive study of the performance of various LLMs acting as judges. We leverage TriviaQA as a benchmark for assessing objective knowledge reasoning of LLMs and evaluate them alongside human annotations which we found to have a high inter-annotator agreement. Our study includes 9 judge models and 9 exam taker models -- both base and instruction-tuned. We assess the judge model's alignment across different model sizes, families, and judge prompts. Among other results, our research rediscovers the importance of using Cohen's kappa as a metric of alignment as opposed to simple percent agreement, showing that judges with high percent agreement can still assign vastly different scores. We find that both Llama-3 70B and GPT-4 Turbo have an excellent alignment with humans, but in terms of ranking exam taker models, they are outperformed by both JudgeLM-7B and the lexical judge Contains, which have up to 34 points lower human alignment. Through error analysis and various other studies, including the effects of instruction length and leniency bias, we hope to provide valuable lessons for using LLMs as judges in the future. | Evaluation |
- From RAGs to rich parameters: Probing how language models utilize external knowledge over parametric information for factual queries - bots. However, the exact nature of how this approach works isn't clearly understood. In this paper, we mechanistically examine the RAG pipeline to highlight that language models take shortcut and have a strong bias towards utilizing only the context information to answer the question, while relying minimally on their parametric memory. We probe this mechanistic behavior in language models with: (i) Causal Mediation Analysis to show that the parametric memory is minimally utilized when answering a question and (ii) Attention Contributions and Knockouts to show that the last token residual stream do not get enriched from the subject token in the question, but gets enriched from other informative tokens in the context. We find this pronounced shortcut behaviour true across both LLaMa and Phi family of models. | RAG, Knowledge Integration |
- PlanRAG: A Plan-then-Retrieval Augmented Generation for Generative Large Language Models as Decision Makers - then-retrieval augmented generation (PlanRAG). Our PlanRAG-based LM generates the plan for decision making as the first step, and the retriever generates the queries for data analysis as the second step. The proposed method outperforms the state-of-the-art iterative RAG method by 15.8% in the Locating scenario and by 7.4% in the Building scenario, respectively. We release our code and benchmark at this https URL. | RAG, Knowledge Integration |
- Self-MoE: Towards Compositional Large Language Models with Self-Specialized Experts - MoE, an approach that transforms a monolithic LLM into a compositional, modular system of self-specialized experts, named MiXSE (MiXture of Self-specialized Experts). Our approach leverages self-specialization, which constructs expert modules using self-generated synthetic data, each equipped with a shared base LLM and incorporating self-optimized routing. This allows for dynamic and capability-specific handling of various target tasks, enhancing overall capabilities, without extensive human-labeled data and added parameters. Our empirical results reveal that specializing LLMs may exhibit potential trade-offs in performances on non-specialized tasks. On the other hand, our Self-MoE demonstrates substantial improvements over the base LLM across diverse benchmarks such as knowledge, reasoning, math, and coding. It also consistently outperforms other methods, including instance merging and weight merging, while offering better flexibility and interpretability by design with semantic experts and routing. Our findings highlight the critical role of modularity and the potential of self-improvement in achieving efficient, scalable, and adaptable systems. | Mixture of Experts, LLM Architecture |
- mDPO: Conditional Preference Optimization for Multimodal Large Language Models - prioritization of language-only preferences by also optimizing image preference. Moreover, we introduce a reward anchor that forces the reward to be positive for chosen responses, thereby avoiding the decrease in their likelihood -- an intrinsic problem of relative preference optimization. Experiments on two multimodal LLMs of different sizes and three widely used benchmarks demonstrate that mDPO effectively addresses the unconditional preference problem in multimodal preference optimization and significantly improves model performance, particularly in reducing hallucination. | Optimization |
- Nemotron-4 340B Technical Report - 4 340B model family, including Nemotron-4-340B-Base, Nemotron-4- 340B-Instruct, and Nemotron-4-340B-Reward. Our models are open access under the NVIDIA Open Model License Agreement, a permissive model license that allows distribution, modification, and use of the models and its outputs. These models perform competitively to open access models on a wide range of evaluation benchmarks, and were sized to fit on a single DGX H100 with 8 GPUs when deployed in FP8 precision. We believe that the community can benefit from these models in various research studies and commercial applications, especially for generating synthetic data to train smaller language models. Notably, over 98% of data used in our model alignment process is synthetically generated, showcasing the effectiveness of these models in generating synthetic data. To further support open research and facilitate model development, w | Foundational LLM |
- Open-Sora 1.2 - Sora, an initiative dedicated to efficiently producing high-quality video. We hope to make the model, tools and all details accessible to all. By embracing open-source principles, Open-Sora not only democratizes access to advanced video generation techniques, but also offers a streamlined and user-friendly platform that simplifies the complexities of video generation. With Open-Sora, our goal is to foster innovation, creativity, and inclusivity within the field of content creation. | Multimodal foundational model |
- Be like a Goldfish, Don't Memorize! Mitigating Memorization in Generative LLMs - token training objective that we call the goldfish loss. During training, a randomly sampled subset of tokens are excluded from the loss computation. These dropped tokens are not memorized by the model, which prevents verbatim reproduction of a complete chain of tokens from the training set. We run extensive experiments training billion-scale Llama-2 models, both pre-trained and trained from scratch, and demonstrate significant reductions in extractable memorization with little to no impact on downstream benchmarks. | New Loss |
- Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality - space models (SSMs) such as Mamba have recently been shown to match or outperform Transformers at small to medium scale. We show that these families of models are actually quite closely related, and develop a rich framework of theoretical connections between SSMs and variants of attention, connected through various decompositions of a well-studied class of structured semiseparable matrices. Our state space duality (SSD) framework allows us to design a new architecture (Mamba-2) whose core layer is an a refinement of Mamba's selective SSM that is 2-8X faster, while continuing to be competitive with Transformers on language modeling. | LLM Architecture |
- An Image is Worth More Than 16x16 Patches: Exploring Transformers on Individual Pixels - - locality in modern computer vision architectures. Concretely, we find that vanilla Transformers can operate by directly treating each individual pixel as a token and achieve highly performant results. This is substantially different from the popular design in Vision Transformer, which maintains the inductive bias from ConvNets towards local neighborhoods (e.g. by treating each 16x16 patch as a token). We mainly showcase the effectiveness of pixels-as-tokens across three well-studied tasks in computer vision: supervised learning for object classification, self-supervised learning via masked autoencoding, and image generation with diffusion models. Although directly operating on individual pixels is less computationally practical, we believe the community must be aware of this surprising piece of knowledge when devising the next generation of neural architectures for computer vision. | Convolutional Networks, Transformers |
- Visual Sketchpad: Sketching as a Visual Chain of Thought for Multimodal Language Models - capacity working memory. However, such actions are missing in current multimodal language models (LMs). Current chain-of-thought and tool-use paradigms only use text as intermediate reasoning steps. In this work, we introduce Sketchpad, a framework that gives multimodal LMs a visual sketchpad and tools to draw on the sketchpad. The LM conducts planning and reasoning according to the visual artifacts it has drawn. Different from prior work, which uses text-to-image models to enable LMs to draw, Sketchpad enables LMs to draw with lines, boxes, marks, etc., which is closer to human sketching and better facilitates reasoning. Sketchpad can also use specialist vision models during the sketching process (e.g., draw bounding boxes with object detection models, draw masks with segmentation models), to further enhance visual perception and reasoning. We experiment with a wide range of math tasks (including geometry, functions, graphs, and chess) and complex visual reasoning tasks. Sketchpad substantially improves performance on all tasks over strong base models with no sketching, yielding an average gain of 12.7% on math tasks, and 8.6% on vision tasks. GPT-4o with Sketchpad sets a new state of the art on all tasks, including V*Bench (80.3%), BLINK spatial reasoning (83.9%), and visual correspondence (80.8%). | Multimodal models, Prompt Engineering |
- Multimodal Table Understanding - quality textual table representations in some real-world scenarios, and table images are much more accessible. Therefore, how to directly understand tables using intuitive visual information is a crucial and urgent challenge for developing more practical applications. In this paper, we propose a new problem, multimodal table understanding, where the model needs to generate correct responses to various table-related requests based on the given table image. To facilitate both the model training and evaluation, we construct a large-scale dataset named MMTab, which covers a wide spectrum of table images, instructions and tasks. On this basis, we develop Table-LLaVA, a generalist tabular multimodal large language model (MLLM), which significantly outperforms recent open-source MLLM baselines on 23 benchmarks under held-in and held-out settings. | Domain Specific LLMs |
- TextGrad: Automatic "Differentiation" via Text - key. Inspired by this, we introduce TextGrad, a powerful framework performing automatic ``differentiation'' via text. TextGrad backpropagates textual feedback provided by LLMs to improve individual components of a compound AI system. In our framework, LLMs provide rich, general, natural language suggestions to optimize variables in computation graphs, ranging from code snippets to molecular structures. TextGrad follows PyTorch's syntax and abstraction and is flexible and easy-to-use. It works out-of-the-box for a variety of tasks, where the users only provide the objective function without tuning components or prompts of the framework. We showcase TextGrad's effectiveness and generality across a diverse range of applications, from question answering and molecule optimization to radiotherapy treatment planning. Without modifying the framework, TextGrad improves the zero-shot accuracy of GPT-4o in Google-Proof Question Answering from 51% to 55%, yields 20% relative performance gain in optimizing LeetCode-Hard coding problem solutions, improves prompts for reasoning, designs new druglike small molecules with desirable in silico binding, and designs radiation oncology treatment plans with high specificity. TextGrad lays a foundation to accelerate the development of the next-generation of AI systems. | Optimization Algorithm |
- Needle In A Multimodal Haystack - world applications, remains underexplored. In this work, we present Needle In A Multimodal Haystack (MM-NIAH), the first benchmark specifically designed to systematically evaluate the capability of existing MLLMs to comprehend long multimodal documents. Our benchmark includes three types of evaluation tasks: multimodal retrieval, counting, and reasoning. In each task, the model is required to answer the questions according to different key information scattered throughout the given multimodal document. Evaluating the leading MLLMs on MM-NIAH, we observe that existing models still have significant room for improvement on these tasks, especially on vision-centric evaluation. We hope this work can provide a platform for further research on long multimodal document comprehension and contribute to the advancement of MLLMs. | Multimodal models |
- Estimating the Hallucination Rate of Generative AI - context learning (ICL) with Generative AI. In ICL, a conditional generative model (CGM) is prompted with a dataset and asked to make a prediction based on that dataset. The Bayesian interpretation of ICL assumes that the CGM is calculating a posterior predictive distribution over an unknown Bayesian model of a latent parameter and data. With this perspective, we define a hallucination as a generated prediction that has low-probability under the true latent parameter. We develop a new method that takes an ICL problem -- that is, a CGM, a dataset, and a prediction question -- and estimates the probability that a CGM will generate a hallucination. Our method only requires generating queries and responses from the model and evaluating its response log probability. We empirically evaluate our method on synthetic regression and natural language ICL tasks using large language models. | Hallucination |
- Simple and Effective Masked Diffusion Language Models - quality images, prior work reports a significant performance gap between diffusion and autoregressive (AR) methods in language modeling. In this work, we show that simple masked discrete diffusion is more performant than previously thought. We apply an effective training recipe that improves the performance of masked diffusion models and derive a simplified, Rao-Blackwellized objective that results in additional improvements. Our objective has a simple form -- it is a mixture of classical masked language modeling losses -- and can be used to train encoder-only language models that admit efficient samplers, including ones that can generate arbitrary lengths of text semi-autoregressively like a traditional language model. On language modeling benchmarks, a range of masked diffusion models trained with modern engineering practices achieves a new state-of-the-art among diffusion models, and approaches AR perplexity. We release our code at: https://github.com/kuleshov-group/mdlm | Diffusion Models |
- Merging Improves Self-Critique Against Jailbreak Attacks - critique capability of the LLM and further fine-tunes it over sanitized synthetic data. This is done with the addition of an external critic model that can be merged with the original, thus bolstering self-critique capabilities and improving the robustness of the LLMs response to adversarial prompts. Our results demonstrate that the combination of merging and self-critique can reduce the attack success rate of adversaries significantly, thus offering a promising defense mechanism against jailbreak attacks. Code, data and models released at https://github.com/vicgalle/merging-self-critique-jailbreaks . | Adversarial Attacks, Jailbreaking |
- NATURAL PLAN: Benchmarking LLMs on Natural Language Planning - use environment for evaluating LLMs on Planning. We observe that NATURAL PLAN is a challenging benchmark for state of the art models. For example, in Trip Planning, GPT-4 and Gemini 1.5 Pro could only achieve 31.1% and 34.8% solve rate respectively. We find that model performance drops drastically as the complexity of the problem increases: all models perform below 5% when there are 10 cities, highlighting a significant gap in planning in natural language for SoTA LLMs. We also conduct extensive ablation studies on NATURAL PLAN to further shed light on the (in)effectiveness of approaches such as self-correction, few-shot generalization, and in-context planning with long-contexts on improving LLM planning. | Agents, Planning |
- SelfGoal: Your Language Agents Already Know How to Achieve High-level Goals - making tools in domains such as gaming and programming. However, these agents often face challenges in achieving high-level goals without detailed instructions and in adapting to environments where feedback is delayed. In this paper, we present SelfGoal, a novel automatic approach designed to enhance agents' capabilities to achieve high-level goals with limited human prior and environmental feedback. The core concept of SelfGoal involves adaptively breaking down a high-level goal into a tree structure of more practical subgoals during the interaction with environments while identifying the most useful subgoals and progressively updating this structure. Experimental results demonstrate that SelfGoal significantly enhances the performance of language agents across various tasks, including competitive, cooperative, and deferred feedback environments. | Agents, Task Decomposition |
- CRAG -- Comprehensive RAG Benchmark - Augmented Generation (RAG) has recently emerged as a promising solution to alleviate Large Language Model (LLM)'s deficiency in lack of knowledge. Existing RAG datasets, however, do not adequately represent the diverse and dynamic nature of real-world Question Answering (QA) tasks. To bridge this gap, we introduce the Comprehensive RAG Benchmark (CRAG), a factual question answering benchmark of 4,409 question-answer pairs and mock APIs to simulate web and Knowledge Graph (KG) search. CRAG is designed to encapsulate a diverse array of questions across five domains and eight question categories, reflecting varied entity popularity from popular to long-tail, and temporal dynamisms ranging from years to seconds. Our evaluation on this benchmark highlights the gap to fully trustworthy QA. Whereas most advanced LLMs achieve <=34% accuracy on CRAG, adding RAG in a straightforward manner improves the accuracy only to 44%. State-of-the-art industry RAG solutions only answer 63% questions without any hallucination. CRAG also reveals much lower accuracy in answering questions regarding facts with higher dynamism, lower popularity, or higher complexity, suggesting future research directions. The CRAG benchmark laid the groundwork for a KDD Cup 2024 challenge, attracting thousands of participants and submissions within the first 50 days of the competition. We commit to maintaining CRAG to serve research communities in advancing RAG solutions and general QA solutions. | RAG, Benchmark |
- Mixture-of-Agents Enhances Large Language Model Capabilities - of-Agents (MoA) methodology. In our approach, we construct a layered MoA architecture wherein each layer comprises multiple LLM agents. Each agent takes all the outputs from agents in the previous layer as auxiliary information in generating its response. MoA models achieves state-of-art performance on AlpacaEval 2.0, MT-Bench and FLASK, surpassing GPT-4 Omni. For example, our MoA using only open-source LLMs is the leader of AlpacaEval 2.0 by a substantial gap, achieving a score of 65.1% compared to 57.5% by GPT-4 Omni. | Agents, Multi-Agents |
- AgentGym: Evolving Large Language Model-based Agents across Diverse Environments - term goal in the AI community. Large language models (LLMs) are considered a promising foundation to build such agents due to their generalized capabilities. Current approaches either have LLM-based agents imitate expert-provided trajectories step-by-step, requiring human supervision, which is hard to scale and limits environmental exploration; or they let agents explore and learn in isolated environments, resulting in specialist agents with limited generalization. In this paper, we take the first step towards building generally-capable LLM-based agents with self-evolution ability. We identify a trinity of ingredients: 1) diverse environments for agent exploration and learning, 2) a trajectory set to equip agents with basic capabilities and prior knowledge, and 3) an effective and scalable evolution method. We propose AgentGym, a new framework featuring a variety of environments and tasks for broad, real-time, uni-format, and concurrent agent exploration. AgentGym also includes a database with expanded instructions, a benchmark suite, and high-quality trajectories across environments. Next, we propose a novel method, AgentEvol, to investigate the potential of agent self-evolution beyond previously seen data across tasks and environments. Experimental results show that the evolved agents can achieve results comparable to SOTA models. We release the AgentGym suite, including the platform, dataset, benchmark, checkpoints, and algorithm implementations. | Agents |
- Are We Done with MMLU? - Redux, which is a subset of 3,000 manually re-annotated questions across 30 MMLU subjects. Using MMLU-Redux, we demonstrate significant discrepancies with the model performance metrics that were originally reported. Our results strongly advocate for revising MMLU's error-ridden questions to enhance its future utility and reliability as a benchmark. Therefore, we open up MMLU-Redux for additional annotation https://huggingface.co/datasets/edinburgh-dawg/mmlu-redux. | Evaluation, Task Benchmarks |
- GenAI Arena: An Open Evaluation Platform for Generative Models - Arena to evaluate different image and video generative models, where users can actively participate in evaluating these models. By leveraging collective user feedback and votes, GenAI-Arena aims to provide a more democratic and accurate measure of model performance. It covers three arenas for text-to-image generation, text-to-video generation, and image editing respectively. Currently, we cover a total of 27 open-source generative models. GenAI-Arena has been operating for four months, amassing over 6000 votes from the community. We describe our platform, analyze the data, and explain the statistical methods for ranking the models. To further promote the research in building model-based evaluation metrics, we release a cleaned version of our preference data for the three tasks, namely GenAI-Bench. We prompt the existing multi-modal models like Gemini, GPT-4o to mimic human voting. We compute the correlation between model voting with human voting to understand their judging abilities. Our results show existing multimodal models are still lagging in assessing the generated visual content, even the best model GPT-4o only achieves a Pearson correlation of 0.22 in the quality subscore, and behaves like random guessing in others. | Evaluation |
- Scaling and evaluating sparse autoencoders - sparse autoencoders [Makhzani and Frey, 2013] to directly control sparsity, simplifying tuning and improving the reconstruction-sparsity frontier. Additionally, we find modifications that result in few dead latents, even at the largest scales we tried. Using these techniques, we find clean scaling laws with respect to autoencoder size and sparsity. We also introduce several new metrics for evaluating feature quality based on the recovery of hypothesized features, the explainability of activation patterns, and the sparsity of downstream effects. These metrics all generally improve with autoencoder size. To demonstrate the scalability of our approach, we train a 16 million latent autoencoder on GPT-4 activations for 40 billion tokens. We release training code and autoencoders for open-source models, as well as a visualizer. | LLM Architecture |
- Buffer of Thoughts: Thought-Augmented Reasoning with Large Language Models - augmented reasoning approach for enhancing accuracy, efficiency and robustness of large language models (LLMs). Specifically, we propose meta-buffer to store a series of informative high-level thoughts, namely thought-template, distilled from the problem-solving processes across various tasks. Then for each problem, we retrieve a relevant thought-template and adaptively instantiate it with specific reasoning structures to conduct efficient reasoning. To guarantee the scalability and stability, we further propose buffer-manager to dynamically update the meta-buffer, thus enhancing the capacity of meta-buffer as more tasks are solved. We conduct extensive experiments on 10 challenging reasoning-intensive tasks, and achieve significant performance improvements over previous SOTA methods: 11% on Game of 24, 20% on Geometric Shapes and 51% on Checkmate-in-One. Further analysis demonstrate the superior generalization ability and model robustness of our BoT, while requiring only 12% of the cost of multi-query prompting methods (e.g., tree/graph of thoughts) on average. Notably, we find that our Llama3-8B+BoT has the potential to surpass Llama3-70B model. | RAG, Knowledge Integration |
- Improve Mathematical Reasoning in Language Models by Automated Process Supervision - step reasoning tasks, such as solving mathematical problems or generating code, remain a significant hurdle for even the most advanced large language models (LLMs). Verifying LLM outputs with an Outcome Reward Model (ORM) is a standard inference-time technique aimed at enhancing the reasoning performance of LLMs. However, this still proves insufficient for reasoning tasks with a lengthy or multi-hop reasoning chain, where the intermediate outcomes are neither properly rewarded nor penalized. Process supervision addresses this limitation by assigning intermediate rewards during the reasoning process. To date, the methods used to collect process supervision data have relied on either human annotation or per-step Monte Carlo estimation, both prohibitively expensive to scale, thus hindering the broad application of this technique. In response to this challenge, we propose a novel divide-and-conquer style Monte Carlo Tree Search (MCTS) algorithm named OmegaPRM for the efficient collection of high-quality process supervision data. This algorithm swiftly identifies the first error in the Chain of Thought (CoT) with binary search and balances the positive and negative examples, thereby ensuring both efficiency and quality. As a result, we are able to collect over 1.5 million process supervision annotations to train a Process Reward Model (PRM). Utilizing this fully automated process supervision alongside the weighted self-consistency algorithm, we have enhanced the instruction tuned Gemini Pro model's math reasoning performance, achieving a 69.4\% success rate on the MATH benchmark, a 36\% relative improvement from the 51\% base model performance. Additionally, the entire process operates without any human intervention, making our method both financially and computationally cost-effective compared to existing methods. | Math Reasoning |
- SaySelf: Teaching LLMs to Express Confidence with Self-Reflective Rationales - consistency prompting, or constructing specific datasets for supervised finetuning. The prompting-based approaches have inferior performance, and the training-based approaches are limited to binary or inaccurate group-level confidence estimates. In this work, we present the advanced SaySelf, a training framework that teaches LLMs to express more accurate fine-grained confidence estimates. In addition, beyond the confidence scores, SaySelf initiates the process of directing LLMs to produce self-reflective rationales that clearly identify gaps in their parametric knowledge and explain their uncertainty. This is achieved by using an LLM to automatically summarize the uncertainties in specific knowledge via natural language. The summarization is based on the analysis of the inconsistency in multiple sampled reasoning chains, and the resulting data is utilized for supervised fine-tuning. Moreover, we utilize reinforcement learning with a meticulously crafted reward function to calibrate the confidence estimates, motivating LLMs to deliver accurate, high-confidence predictions and to penalize overconfidence in erroneous outputs. Experimental results in both in-distribution and out-of-distribution datasets demonstrate the effectiveness of SaySelf in reducing the confidence calibration error and maintaining the task performance. We show that the generated self-reflective rationales are reasonable and can further contribute to the calibration. | LLM Training, LLM Challenges |
- Guiding a Diffusion Model with a Bad Version of Itself - generating diffusion models are image quality, the amount of variation in the results, and how well the results align with a given condition, e.g., a class label or a text prompt. The popular classifier-free guidance approach uses an unconditional model to guide a conditional model, leading to simultaneously better prompt alignment and higher-quality images at the cost of reduced variation. These effects seem inherently entangled, and thus hard to control. We make the surprising observation that it is possible to obtain disentangled control over image quality without compromising the amount of variation by guiding generation using a smaller, less-trained version of the model itself rather than an unconditional model. This leads to significant improvements in ImageNet generation, setting record FIDs of 1.01 for 64x64 and 1.25 for 512x512, using publicly available networks. Furthermore, the method is also applicable to unconditional diffusion models, drastically improving their quality. | Diffusion Models |
- To Believe or Not to Believe Your LLM - theoretic metric that allows to reliably detect when only epistemic uncertainty is large, in which case the output of the model is unreliable. This condition can be computed based solely on the output of the model obtained simply by some special iterative prompting based on the previous responses. Such quantification, for instance, allows to detect hallucinations (cases when epistemic uncertainty is high) in both single- and multi-answer responses. This is in contrast to many standard uncertainty quantification strategies (such as thresholding the log-likelihood of a response) where hallucinations in the multi-answer case cannot be detected. We conduct a series of experiments which demonstrate the advantage of our formulation. Further, our investigations shed some light on how the probabilities assigned to a given output by an LLM can be amplified by iterative prompting, which might be of independent interest. | Hallucinations, Uncertainty Estimation |
- Self-Improving Robust Preference Optimization - dependent (i.e., not robust to out-of-distribution (OOD) tasks). Here we address this challenge by proposing Self-Improving Robust Preference Optimization SRPO, a practical and mathematically principled offline RLHF framework that is completely robust to the changes in the task. The key idea of SRPO is to cast the problem of learning from human preferences as a self-improvement process, which can be mathematically expressed in terms of a min-max objective that aims at joint optimization of self-improvement policy and the generative policy in an adversarial fashion. The solution for this optimization problem is independent of the training task and thus it is robust to its changes. We then show that this objective can be re-expressed in the form of a non-adversarial offline loss which can be optimized using standard supervised optimization techniques at scale without any need for reward model and online inference. We show the effectiveness of SRPO in terms of AI Win-Rate (WR) against human (GOLD) completions. In particular, when SRPO is evaluated on the OOD XSUM dataset, it outperforms the celebrated DPO by a clear margin of 15% after 5 self-revisions, achieving WR of 90%. | Optimization, Alignment |
- Towards Scalable Automated Alignment of LLMs: A Survey - annotation are increasingly unable to meet the scalability demands. Therefore, there is an urgent need to explore new sources of automated alignment signals and technical approaches. In this paper, we systematically review the recently emerging methods of automated alignment, attempting to explore how to achieve effective, scalable, automated alignment once the capabilities of LLMs exceed those of humans. Specifically, we categorize existing automated alignment methods into 4 major categories based on the sources of alignment signals and discuss the current status and potential development of each category. Additionally, we explore the underlying mechanisms that enable automated alignment and discuss the essential factors that make automated alignment technologies feasible and effective from the fundamental role of alignment. | Alignment |
- Show, Don't Tell: Aligning Language Models with Demonstrated Feedback - hoc tasks. We argue that it is instead possible to align an LLM to a specific setting by leveraging a very small number (<10) of demonstrations as feedback. Our method, Demonstration ITerated Task Optimization (DITTO), directly aligns language model outputs to a user's demonstrated behaviors. Derived using ideas from online imitation learning, DITTO cheaply generates online comparison data by treating users' demonstrations as preferred over output from the LLM and its intermediate checkpoints. We evaluate DITTO's ability to learn fine-grained style and task alignment across domains such as news articles, emails, and blog posts. Additionally, we conduct a user study soliciting a range of demonstrations from participants (N=16). Across our benchmarks and user study, we find that win-rates for DITTO outperform few-shot prompting, supervised fine-tuning, and other self-play methods by an average of 19% points. By using demonstrations as feedback directly, DITTO offers a novel method for effective customization of LLMs. | Alignment |
- Artificial Generational Intelligence: Cultural Accumulation in Reinforcement Learning - ended and diverse progress in capabilities spanning human history. It builds an expanding body of knowledge and skills by combining individual exploration with inter-generational information transmission. Despite its widespread success among humans, the capacity for artificial learning agents to accumulate culture remains under-explored. In particular, approaches to reinforcement learning typically strive for improvements over only a single lifetime. Generational algorithms that do exist fail to capture the open-ended, emergent nature of cultural accumulation, which allows individuals to trade-off innovation and imitation. Building on the previously demonstrated ability for reinforcement learning agents to perform social learning, we find that training setups which balance this with independent learning give rise to cultural accumulation. These accumulating agents outperform those trained for a single lifetime with the same cumulative experience. We explore this accumulation by constructing two models under two distinct notions of a generation: episodic generations, in which accumulation occurs via in-context learning and train-time generations, in which accumulation occurs via in-weights learning. In-context and in-weights cultural accumulation can be interpreted as analogous to knowledge and skill accumulation, respectively. To the best of our knowledge, this work is the first to present general models that achieve emergent cultural accumulation in reinforcement learning, opening up new avenues towards more open-ended learning systems, as well as presenting new opportunities for modelling human culture. | Cultural Adaptation |
-
:mortar_board: Courses
- CS324 - Large Language Models
- Generative AI for Beginners
- Quantization Fundamentals with Hugging Face
- Preprocessing Unstructured Data for LLM Applications
- Retrieval Augmented Generation for Production with LangChain & LlamaIndex
- [1hr Talk
- Training & Fine-Tuning LLMs for Production
- LangChain for LLM Application Development
- Automated Testing for LLMOps
- Reinforcement Learning from Human Feedback
- Finetuning Large Language Models
- Evaluating and Debugging Generative AI Models Using Weights and Biases
- Large Language Model Agents MOOC, Fall 2024
- LLM University
- 11-777: Multimodal Machine Learning
- Functions, Tools and Agents with LangChain
- LLMOps
- Building Applications with Vector Databases
- LangChain: Chat with Your Data
- Building Systems with the ChatGPT API
- ChatGPT Prompt Engineering for Developers
- Building and Evaluating Advanced RAG Applications
- LLM Foundation Models from the Ground Up | Primer
- Generative AI Explained
- Transformer Models and BERT Model
- Generative AI Learning Plan for Decision Makers
- Introduction to Responsible AI
- Fundamentals of Generative AI
- ChatGPT for Beginners: The Ultimate Use Cases for Everyone
- Building Generative AI Applications Using Amazon Bedrock
- Efficiently Serving LLMs
- Building Systems with the ChatGPT API
- Building RAG Agents with LLMs
- Serverless LLM apps with Amazon Bedrock
- Building Applications with Vector Databases
- Automated Testing for LLMOps
- LLMOps
- Build LLM Apps with LangChain.js
- Advanced Retrieval for AI with Chroma
- Prompt Engineering with Llama 2
- Prompt Engineering Specialization
- Augment your LLM Using Retrieval Augmented Generation
- Knowledge Graphs for RAG
- Open Source Models with Hugging Face
- Vector Databases: from Embeddings to Applications
- Understanding and Applying Text Embeddings
- Quality and Safety for LLM Applications
- How Diffusion Models Work
- How to Use Midjourney, AI Art and ChatGPT to Create an Amazing Website
- Build AI Apps with ChatGPT, DALL-E and GPT-4
- LLM University
- LLM Learning Lab
- Functions, Tools and Agents with LangChain
- LangChain for LLM Application Development
- LLMOps
- Automated Testing for LLMOps
- LangChain & Vector Databases in Production
- Reinforcement Learning from Human Feedback
- Building Applications with Vector Databases
- How Diffusion Models Work
- Finetuning Large Language Models
- LangChain: Chat with Your Data
- Building Systems with the ChatGPT API
- Advanced RAG Orchestration series
- Building and Evaluating Advanced RAG Applications
- Evaluating and Debugging Generative AI Models Using Weights and Biases
- Large Language Models: Application through Production
- Generative AI Foundations
- Introduction to Generative AI Community Course
- NLP course
- Introduction to Generative AI
- Introduction to Large Language Models
- Introduction to Generative AI
- Generative AI Concepts
- 1 Hour Introduction to LLM (Large Language Models)
- Link
- Feb 2024
- Large Language Models
- Understanding Large Language Models
- Generative AI with Large Language Models
- LLMOps: Building Real-World Applications With Large Language Models
- Full Stack LLM Bootcamp
- Generative AI for beginners
- Operationalizing LLMs on Azure
- Generative AI Full Course – Gemini Pro, OpenAI, Llama, Langchain, Pinecone, Vector Databases & More
- JavaScript RAG Web Apps with LlamaIndex
- Red Teaming LLM Applications
- 11-777: Multimodal Machine Learning
- Avoiding AI Harm
- Developing AI Policy
- Prompt Engineering for Vision Models
- AI Agents in LangGraph
- AI Agentic Design Patterns with AutoGen
- Multi AI Agent Systems with crewAI
- Building Agentic RAG with LlamaIndex
- LLM Observability: Agents, Tools, and Chains
- ChatGPT for Everyone
- Large Language Models (LLMs) (In English)
- Quantization in Depth
- Generative AI Fundamentals
- Large Language Models (LLMs) (In English)
- Agents Tools & Function Calling with Amazon Bedrock (How-to)
- ChatGPT & Zapier: Agentic AI for Everyone
- Multi-Agent Systems with AutoGen
-
:notebook: Code Notebooks
- AWS Bedrock Workshop Tutorials
- Langchain Tutorials
- LLM Hub
- LLM Fine-tuning tutorials
- Free LLM Fine-Tuning Notebooks
- LLM-PlayLab
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- LLM Applications for production - project
- LLM tutorials
- PEFT
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
- Free LLM Fine-Tuning Notebooks
-
:star: Best Gen AI Papers List (Oct 2024)
- What Happened in LLMs Layers when Trained for Fast vs. Slow Thinking: A Gradient Perspective - training of LLMs? We investigate the training patterns of different layers in large language models (LLMs), through the lens of gradient, when training with different responses and initial models. We are specifically interested in how fast vs. slow thinking affects the layer-wise gradients, given the recent popularity of training LLMs on reasoning paths such as chain-of-thoughts (CoT) and process rewards. In our study, fast thinking without CoT leads to larger gradients and larger differences of gradients across layers than slow thinking (Detailed CoT), indicating the learning stability brought by the latter. Moreover, pre-trained LLMs are less affected by the instability of fast thinking than instruction-tuned LLMs. Additionally, we study whether the gradient patterns can reflect the correctness of responses when training different LLMs using slow vs. fast thinking paths. The results show that the gradients of slow thinking can distinguish correct and irrelevant reasoning paths. As a comparison, we conduct similar gradient analyses on non-reasoning knowledge learning tasks, on which, however, trivially increasing the response length does not lead to similar behaviors of slow thinking. Our study strengthens fundamental understandings of LLM training and sheds novel insights on its efficiency and stability, which pave the way towards building a generalizable System-2 agent. Our code, data, and gradient statistics can be found in: https://github.com/MingLiiii/Layer_Gradient. |
- ReferEverything: Towards Segmenting Everything We Can Speak of in Videos - language representations learned by video diffusion models on Internet-scale datasets. A key insight of our approach is preserving as much of the generative model's original representation as possible, while fine-tuning it on narrow-domain Referral Object Segmentation datasets. As a result, our framework can accurately segment and track rare and unseen objects, despite being trained on object masks from a limited set of categories. Additionally, it can generalize to non-object dynamic concepts, such as waves crashing in the ocean, as demonstrated in our newly introduced benchmark for Referral Video Process Segmentation (Ref-VPS). Our experiments show that REM performs on par with state-of-the-art approaches on in-domain datasets, like Ref-DAVIS, while outperforming them by up to twelve points in terms of region similarity on out-of-domain data, leveraging the power of Internet-scale pre-training. |
- TokenFormer: Rethinking Transformer Scaling with Tokenized Model Parameters - parameter attention layer, where input tokens act as queries and model parameters as keys and values. This reformulation allows for progressive and efficient scaling without necessitating retraining from scratch. Our model scales from 124M to 1.4B parameters by incrementally adding new key-value parameter pairs, achieving performance comparable to Transformers trained from scratch while greatly reducing training costs. Code and models are available at \url{https://github.com/Haiyang-W/TokenFormer}. |
- CORAL: Benchmarking Multi-turn Conversational Retrieval-Augmentation Generation - Augmented Generation (RAG) has become a powerful paradigm for enhancing large language models (LLMs) through external knowledge retrieval. Despite its widespread attention, existing academic research predominantly focuses on single-turn RAG, leaving a significant gap in addressing the complexities of multi-turn conversations found in real-world applications. To bridge this gap, we introduce CORAL, a large-scale benchmark designed to assess RAG systems in realistic multi-turn conversational settings. CORAL includes diverse information-seeking conversations automatically derived from Wikipedia and tackles key challenges such as open-domain coverage, knowledge intensity, free-form responses, and topic shifts. It supports three core tasks of conversational RAG: passage retrieval, response generation, and citation labeling. We propose a unified framework to standardize various conversational RAG methods and conduct a comprehensive evaluation of these methods on CORAL, demonstrating substantial opportunities for improving existing approaches. |
- A Pointer Network-based Approach for Joint Extraction and Detection of Multi-Label Multi-Class Intents - oriented dialogue systems, intent detection is crucial for interpreting user queries and providing appropriate responses. Existing research primarily addresses simple queries with a single intent, lacking effective systems for handling complex queries with multiple intents and extracting different intent spans. Additionally, there is a notable absence of multilingual, multi-intent datasets. This study addresses three critical tasks: extracting multiple intent spans from queries, detecting multiple intents, and developing a multi-lingual multi-label intent dataset. We introduce a novel multi-label multi-class intent detection dataset (MLMCID-dataset) curated from existing benchmark datasets. We also propose a pointer network-based architecture (MLMCID) to extract intent spans and detect multiple intents with coarse and fine-grained labels in the form of sextuplets. Comprehensive analysis demonstrates the superiority of our pointer network-based system over baseline approaches in terms of accuracy and F1-score across various datasets. |
- Unpacking SDXL Turbo: Interpreting Text-to-Image Models with Sparse Autoencoders - language models (LLMs). For LLMs, they have been shown to decompose intermediate representations that often are not interpretable directly into sparse sums of interpretable features, facilitating better control and subsequent analysis. However, similar analyses and approaches have been lacking for text-to-image models. We investigated the possibility of using SAEs to learn interpretable features for a few-step text-to-image diffusion models, such as SDXL Turbo. To this end, we train SAEs on the updates performed by transformer blocks within SDXL Turbo's denoising U-net. We find that their learned features are interpretable, causally influence the generation process, and reveal specialization among the blocks. In particular, we find one block that deals mainly with image composition, one that is mainly responsible for adding local details, and one for color, illumination, and style. Therefore, our work is an important first step towards better understanding the internals of generative text-to-image models like SDXL Turbo and showcases the potential of features learned by SAEs for the visual domain. Code is available at https://github.com/surkovv/sdxl-unbox |
- Document Parsing Unveiled: Techniques, Challenges, and Prospects for Structured Information Extraction - structured documents-such as contracts, academic papers, and invoices-into structured, machine-readable data. Document parsing extract reliable structured data from unstructured inputs, providing huge convenience for numerous applications. Especially with recent achievements in Large Language Models, document parsing plays an indispensable role in both knowledge base construction and training data generation. This survey presents a comprehensive review of the current state of document parsing, covering key methodologies, from modular pipeline systems to end-to-end models driven by large vision-language models. Core components such as layout detection, content extraction (including text, tables, and mathematical expressions), and multi-modal data integration are examined in detail. Additionally, this paper discusses the challenges faced by modular document parsing systems and vision-language models in handling complex layouts, integrating multiple modules, and recognizing high-density text. It emphasizes the importance of developing larger and more diverse datasets and outlines future research directions. |
- AutoKaggle: A Multi-Agent Framework for Autonomous Data Science Competitions - solving approaches. We propose AutoKaggle, a powerful and user-centric framework that assists data scientists in completing daily data pipelines through a collaborative multi-agent system. AutoKaggle implements an iterative development process that combines code execution, debugging, and comprehensive unit testing to ensure code correctness and logic consistency. The framework offers highly customizable workflows, allowing users to intervene at each phase, thus integrating automated intelligence with human expertise. Our universal data science toolkit, comprising validated functions for data cleaning, feature engineering, and modeling, forms the foundation of this solution, enhancing productivity by streamlining common tasks. We selected 8 Kaggle competitions to simulate data processing workflows in real-world application scenarios. Evaluation results demonstrate that AutoKaggle achieves a validation submission rate of 0.85 and a comprehensive score of 0.82 in typical data science pipelines, fully proving its effectiveness and practicality in handling complex data science tasks. |
- A Survey of Small Language Models - device, mobile, edge devices, among many others. In this article, we present a comprehensive survey on SLMs, focusing on their architectures, training techniques, and model compression techniques. We propose a novel taxonomy for categorizing the methods used to optimize SLMs, including model compression, pruning, and quantization techniques. We summarize the benchmark datasets that are useful for benchmarking SLMs along with the evaluation metrics commonly used. Additionally, we highlight key open challenges that remain to be addressed. Our survey aims to serve as a valuable resource for researchers and practitioners interested in developing and deploying small yet efficient language models. |
- GPT-4o System Card - 4o is an autoregressive omni model that accepts as input any combination of text, audio, image, and video, and generates any combination of text, audio, and image outputs. It's trained end-to-end across text, vision, and audio, meaning all inputs and outputs are processed by the same neural network. GPT-4o can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in conversation. It matches GPT-4 Turbo performance on text in English and code, with significant improvement on text in non-English languages, while also being much faster and 50\% cheaper in the API. GPT-4o is especially better at vision and audio understanding compared to existing models. In line with our commitment to building AI safely and consistent with our voluntary commitments to the White House, we are sharing the GPT-4o System Card, which includes our Preparedness Framework evaluations. In this System Card, we provide a detailed look at GPT-4o's capabilities, limitations, and safety evaluations across multiple categories, focusing on speech-to-speech while also evaluating text and image capabilities, and measures we've implemented to ensure the model is safe and aligned. We also include third-party assessments on dangerous capabilities, as well as discussion of potential societal impacts of GPT-4o's text and vision capabilities. |
- MMAU: A Massive Multi-Task Audio Understanding and Reasoning Benchmark - -which includes speech, non-speech sounds, and music--is crucial for AI agents to interact effectively with the world. We present MMAU, a novel benchmark designed to evaluate multimodal audio understanding models on tasks requiring expert-level knowledge and complex reasoning. MMAU comprises 10k carefully curated audio clips paired with human-annotated natural language questions and answers spanning speech, environmental sounds, and music. It includes information extraction and reasoning questions, requiring models to demonstrate 27 distinct skills across unique and challenging tasks. Unlike existing benchmarks, MMAU emphasizes advanced perception and reasoning with domain-specific knowledge, challenging models to tackle tasks akin to those faced by experts. We assess 18 open-source and proprietary (Large) Audio-Language Models, demonstrating the significant challenges posed by MMAU. Notably, even the most advanced Gemini Pro v1.5 achieves only 52.97% accuracy, and the state-of-the-art open-source Qwen2-Audio achieves only 52.50%, highlighting considerable room for improvement. We believe MMAU will drive the audio and multimodal research community to develop more advanced audio understanding models capable of solving complex audio tasks. |
- Unleashing Reasoning Capability of LLMs via Scalable Question Synthesis from Scratch - quality data is one of the most important factors in improving the reasoning capability of LLMs. Existing works have demonstrated the effectiveness of creating more instruction data from seed questions or knowledge bases. Recent research indicates that continually scaling up data synthesis from strong models (e.g., GPT-4) can further elicit reasoning performance. Though promising, the open-sourced community still lacks high-quality data at scale and scalable data synthesis methods with affordable costs. To address this, we introduce ScaleQuest, a scalable and novel data synthesis method that utilizes "small-size" (e.g., 7B) open-source models to generate questions from scratch without the need for seed data with complex augmentation constraints. With the efficient ScaleQuest, we automatically constructed a mathematical reasoning dataset consisting of 1 million problem-solution pairs, which are more effective than existing open-sourced datasets. It can universally increase the performance of mainstream open-source models (i.e., Mistral, Llama3, DeepSeekMath, and Qwen2-Math) by achieving 29.2% to 46.4% gains on MATH. Notably, simply fine-tuning the Qwen2-Math-7B-Base model with our dataset can even surpass Qwen2-Math-7B-Instruct, a strong and well-aligned model on closed-source data, and proprietary models such as GPT-4-Turbo and Claude-3.5 Sonnet. |
- AgentStore: Scalable Integration of Heterogeneous Agents As Specialized Generalist Computer Assistant - computer interaction. However, existing agent methods exhibit deficiencies in their generalization and specialization capabilities, especially in handling open-ended computer tasks in real-world environments. Inspired by the rich functionality of the App store, we present AgentStore, a scalable platform designed to dynamically integrate heterogeneous agents for automating computer tasks. AgentStore empowers users to integrate third-party agents, allowing the system to continuously enrich its capabilities and adapt to rapidly evolving operating systems. Additionally, we propose a novel core \textbf{MetaAgent} with the \textbf{AgentToken} strategy to efficiently manage diverse agents and utilize their specialized and generalist abilities for both domain-specific and system-wide tasks. Extensive experiments on three challenging benchmarks demonstrate that AgentStore surpasses the limitations of previous systems with narrow capabilities, particularly achieving a significant improvement from 11.21\% to 23.85\% on the OSWorld benchmark, more than doubling the previous results. Comprehensive quantitative and qualitative results further demonstrate AgentStore's ability to enhance agent systems in both generalization and specialization, underscoring its potential for developing the specialized generalist computer assistant. All our codes will be made publicly available in https://chengyou-jia.github.io/AgentStore-Home. |
- CLEAR: Character Unlearning in Textual and Visual Modalities - source benchmark. To address this, we introduce CLEAR, a new benchmark designed to evaluate MMU methods. CLEAR contains 200 fictitious individuals and 3,700 images linked with corresponding question-answer pairs, enabling a thorough evaluation across modalities. We assess 10 MU methods, adapting them for MMU, and highlight new challenges specific to multimodal forgetting. We also demonstrate that simple $\ell_1$ regularization on LoRA weights significantly mitigates catastrophic forgetting, preserving model performance on retained data. The dataset is available at https://huggingface.co/datasets/therem/CLEAR |
- Breaking the Memory Barrier: Near Infinite Batch Size Scaling for Contrastive Loss - based computation strategy that partitions the contrastive loss calculation into arbitrary small blocks, avoiding full materialization of the similarity matrix. Furthermore, we introduce a multi-level tiling strategy to leverage the hierarchical structure of distributed systems, employing ring-based communication at the GPU level to optimize synchronization and fused kernels at the CUDA core level to reduce I/O overhead. Experimental results show that the proposed method scales batch sizes to unprecedented levels. For instance, it enables contrastive training of a CLIP-ViT-L/14 model with a batch size of 4M or 12M using 8 or 32 A800 80GB without sacrificing any accuracy. Compared to SOTA memory-efficient solutions, it achieves a two-order-of-magnitude reduction in memory while maintaining comparable speed. The code will be made publicly available. |
- Aligning Large Language Models via Self-Steering Optimization - Steering Optimization ($SSO$), an algorithm that autonomously generates high-quality preference signals based on predefined principles during iterative training, eliminating the need for manual annotation. $SSO$ maintains the accuracy of signals by ensuring a consistent gap between chosen and rejected responses while keeping them both on-policy to suit the current policy model's learning capacity. $SSO$ can benefit the online and offline training of the policy model, as well as enhance the training of reward models. We validate the effectiveness of $SSO$ with two foundation models, Qwen2 and Llama3.1, indicating that it provides accurate, on-policy preference signals throughout iterative training. Without any manual annotation or external models, $SSO$ leads to significant performance improvements across six subjective or objective benchmarks. Besides, the preference data generated by $SSO$ significantly enhanced the performance of the reward model on Rewardbench. Our work presents a scalable approach to preference optimization, paving the way for more efficient and effective automated alignment. |
- CompassJudger-1: All-in-one Judge Model Helps Model Evaluation and Evolution - world usage scenarios and human preferences. However, human-based evaluations are costly and lack reproducibility, making precise automated evaluators (judgers) vital in this process. In this report, we introduce \textbf{CompassJudger-1}, the first open-source \textbf{all-in-one} judge LLM. CompassJudger-1 is a general-purpose LLM that demonstrates remarkable versatility. It is capable of: 1. Performing unitary scoring and two-model comparisons as a reward model; 2. Conducting evaluations according to specified formats; 3. Generating critiques; 4. Executing diverse tasks like a general LLM. To assess the evaluation capabilities of different judge models under a unified setting, we have also established \textbf{JudgerBench}, a new benchmark that encompasses various subjective evaluation tasks and covers a wide range of topics. CompassJudger-1 offers a comprehensive solution for various evaluation tasks while maintaining the flexibility to adapt to diverse requirements. Both CompassJudger and JudgerBench are released and available to the research community athttps://github.com/open-compass/CompassJudger. We believe that by open-sourcing these tools, we can foster collaboration and accelerate progress in LLM evaluation methodologies. |
- Can Knowledge Editing Really Correct Hallucinations? - factual information in generated content, despite their superior capacities across tasks. Meanwhile, knowledge editing has been developed as a new popular paradigm to correct the erroneous factual knowledge encoded in LLMs with the advantage of avoiding retraining from scratch. However, one common issue of existing evaluation datasets for knowledge editing is that they do not ensure LLMs actually generate hallucinated answers to the evaluation questions before editing. When LLMs are evaluated on such datasets after being edited by different techniques, it is hard to directly adopt the performance to assess the effectiveness of different knowledge editing methods in correcting hallucinations. Thus, the fundamental question remains insufficiently validated: Can knowledge editing really correct hallucinations in LLMs? We proposed HalluEditBench to holistically benchmark knowledge editing methods in correcting real-world hallucinations. First, we rigorously construct a massive hallucination dataset with 9 domains, 26 topics and more than 6,000 hallucinations. Then, we assess the performance of knowledge editing methods in a holistic way on five dimensions including Efficacy, Generalization, Portability, Locality, and Robustness. Through HalluEditBench, we have provided new insights into the potentials and limitations of different knowledge editing methods in correcting hallucinations, which could inspire future improvements and facilitate the progress in the field of knowledge editing. |
- Are AI Detectors Good Enough? A Survey on Quality of Datasets With Machine-Generated Texts - generated text detectors. A huge number of detectors and collections with AI fragments have emerged, and several detection methods even showed recognition quality up to 99.9% according to the target metrics in such collections. However, the quality of such detectors tends to drop dramatically in the wild, posing a question: Are detectors actually highly trustworthy or do their high benchmark scores come from the poor quality of evaluation datasets? In this paper, we emphasise the need for robust and qualitative methods for evaluating generated data to be secure against bias and low generalising ability of future model. We present a systematic review of datasets from competitions dedicated to AI-generated content detection and propose methods for evaluating the quality of datasets containing AI-generated fragments. In addition, we discuss the possibility of using high-quality generated data to achieve two goals: improving the training of detection models and improving the training datasets themselves. Our contribution aims to facilitate a better understanding of the dynamics between human and machine text, which will ultimately support the integrity of information in an increasingly automated world. |
- Movie Gen: A Cast of Media Foundation Models - quality, 1080p HD videos with different aspect ratios and synchronized audio. We also show additional capabilities such as precise instruction-based video editing and generation of personalized videos based on a user's image. Our models set a new state-of-the-art on multiple tasks: text-to-video synthesis, video personalization, video editing, video-to-audio generation, and text-to-audio generation. Our largest video generation model is a 30B parameter transformer trained with a maximum context length of 73K video tokens, corresponding to a generated video of 16 seconds at 16 frames-per-second. We show multiple technical innovations and simplifications on the architecture, latent spaces, training objectives and recipes, data curation, evaluation protocols, parallelization techniques, and inference optimizations that allow us to reap the benefits of scaling pre-training data, model size, and training compute for training large scale media generation models. We hope this paper helps the research community to accelerate progress and innovation in media generation models. All videos from this paper are available at https://go.fb.me/MovieGenResearchVideos. |
- Web Agents with World Models: Learning and Leveraging Environment Dynamics in Web Navigation - based web agents in long-horizon tasks is far from optimal, often yielding errors such as repeatedly buying a non-refundable flight ticket. By contrast, humans can avoid such an irreversible mistake, as we have an awareness of the potential outcomes (e.g., losing money) of our actions, also known as the "world model". Motivated by this, our study first starts with preliminary analyses, confirming the absence of world models in current LLMs (e.g., GPT-4o, Claude-3.5-Sonnet, etc.). Then, we present a World-model-augmented (WMA) web agent, which simulates the outcomes of its actions for better decision-making. To overcome the challenges in training LLMs as world models predicting next observations, such as repeated elements across observations and long HTML inputs, we propose a transition-focused observation abstraction, where the prediction objectives are free-form natural language descriptions exclusively highlighting important state differences between time steps. Experiments on WebArena and Mind2Web show that our world models improve agents' policy selection without training and demonstrate our agents' cost- and time-efficiency compared to recent tree-search-based agents. |
- The Curse of Multi-Modalities: Evaluating Hallucinations of Large Multimodal Models across Language, Visual, and Audio - world scenarios. This paper presents the first systematic investigation of hallucinations in LMMs involving the three most common modalities: language, visual, and audio. Our study reveals two key contributors to hallucinations: overreliance on unimodal priors and spurious inter-modality correlations. To address these challenges, we introduce the benchmark The Curse of Multi-Modalities (CMM), which comprehensively evaluates hallucinations in LMMs, providing a detailed analysis of their underlying issues. Our findings highlight key vulnerabilities, including imbalances in modality integration and biases from training data, underscoring the need for balanced cross-modal learning and enhanced hallucination mitigation strategies. Based on our observations and findings, we suggest potential research directions that could enhance the reliability of LMMs. |
- JudgeBench: A Benchmark for Evaluating LLM-based Judges - based judges have emerged as a scalable alternative to human evaluation and are increasingly used to assess, compare, and improve models. However, the reliability of LLM-based judges themselves is rarely scrutinized. As LLMs become more advanced, their responses grow more sophisticated, requiring stronger judges to evaluate them. Existing benchmarks primarily focus on a judge's alignment with human preferences, but often fail to account for more challenging tasks where crowdsourced human preference is a poor indicator of factual and logical correctness. To address this, we propose a novel evaluation framework to objectively evaluate LLM-based judges. Based on this framework, we propose JudgeBench, a benchmark for evaluating LLM-based judges on challenging response pairs spanning knowledge, reasoning, math, and coding. JudgeBench leverages a novel pipeline for converting existing difficult datasets into challenging response pairs with preference labels reflecting objective correctness. Our comprehensive evaluation on a collection of prompted judges, fine-tuned judges, multi-agent judges, and reward models shows that JudgeBench poses a significantly greater challenge than previous benchmarks, with many strong models (e.g., GPT-4o) performing just slightly better than random guessing. Overall, JudgeBench offers a reliable platform for assessing increasingly advanced LLM-based judges. Data and code are available at https://github.com/ScalerLab/JudgeBench . |
- Revealing the Barriers of Language Agents in Planning - level planning abilities. Even the state-of-the-art reasoning model, OpenAI o1, achieves only 15.6% on one of the complex real-world planning benchmarks. This highlights a critical question: What hinders language agents from achieving human-level planning? Although existing studies have highlighted weak performance in agent planning, the deeper underlying issues and the mechanisms and limitations of the strategies proposed to address them remain insufficiently understood. In this work, we apply the feature attribution study and identify two key factors that hinder agent planning: the limited role of constraints and the diminishing influence of questions. We also find that although current strategies help mitigate these challenges, they do not fully resolve them, indicating that agents still have a long way to go before reaching human-level intelligence. |
- HumanEval-V: Evaluating Visual Understanding and Reasoning Abilities of Large Multimodal Models Through Coding Tasks - level instructions, complex reasoning, and the implementation of functional programs -- core capabilities for advancing Artificial General Intelligence. Despite the progress in Large Multimodal Models (LMMs), which extend LLMs with visual perception and understanding capabilities, there remains a notable lack of coding benchmarks that rigorously assess these models, particularly in tasks that emphasize visual reasoning. To address this gap, we introduce HumanEval-V, a novel and lightweight benchmark specifically designed to evaluate LMMs' visual understanding and reasoning capabilities through code generation. HumanEval-V includes 108 carefully crafted, entry-level Python coding tasks derived from platforms like CodeForces and Stack Overflow. Each task is adapted by modifying the context and algorithmic patterns of the original problems, with visual elements redrawn to ensure distinction from the source, preventing potential data leakage. LMMs are required to complete the code solution based on the provided visual context and a predefined Python function signature outlining the task requirements. Every task is equipped with meticulously handcrafted test cases to ensure a thorough and reliable evaluation of model-generated solutions. We evaluate 19 state-of-the-art LMMs using HumanEval-V, uncovering significant challenges. Proprietary models like GPT-4o achieve only 13% pass@1 and 36.4% pass@10, while open-weight models with 70B parameters score below 4% pass@1. Ablation studies further reveal the limitations of current LMMs in vision reasoning and coding capabilities. These results underscore key areas for future research to enhance LMMs' capabilities. We have open-sourced our code and benchmark at https://github.com/HumanEval-V/HumanEval-V-Benchmark. |
- Your Mixture-of-Experts LLM Is Secretly an Embedding Model For Free - only architecture often limits their potential as embedding models if no further representation finetuning is applied. Does this contradict their claim of generalists? To answer the question, we take a closer look at Mixture-of-Experts (MoE) LLMs. Our study shows that the expert routers in MoE LLMs can serve as an off-the-shelf embedding model with promising performance on a diverse class of embedding-focused tasks, without requiring any finetuning. Moreover, our extensive analysis shows that the MoE routing weights (RW) is complementary to the hidden state (HS) of LLMs, a widely-used embedding. Compared to HS, we find that RW is more robust to the choice of prompts and focuses on high-level semantics. Motivated by the analysis, we propose MoEE combining RW and HS, which achieves better performance than using either separately. Our exploration of their combination and prompting strategy shed several novel insights, e.g., a weighted sum of RW and HS similarities outperforms the similarity on their concatenation. Our experiments are conducted on 6 embedding tasks with 20 datasets from the Massive Text Embedding Benchmark (MTEB). The results demonstrate the significant improvement brought by MoEE to LLM-based embedding without further finetuning. |
- LOKI: A Comprehensive Synthetic Data Detection Benchmark using Large Multimodal Models - generated content, the future internet may be inundated with synthetic data, making the discrimination of authentic and credible multimodal data increasingly challenging. Synthetic data detection has thus garnered widespread attention, and the performance of large multimodal models (LMMs) in this task has attracted significant interest. LMMs can provide natural language explanations for their authenticity judgments, enhancing the explainability of synthetic content detection. Simultaneously, the task of distinguishing between real and synthetic data effectively tests the perception, knowledge, and reasoning capabilities of LMMs. In response, we introduce LOKI, a novel benchmark designed to evaluate the ability of LMMs to detect synthetic data across multiple modalities. LOKI encompasses video, image, 3D, text, and audio modalities, comprising 18K carefully curated questions across 26 subcategories with clear difficulty levels. The benchmark includes coarse-grained judgment and multiple-choice questions, as well as fine-grained anomaly selection and explanation tasks, allowing for a comprehensive analysis of LMMs. We evaluated 22 open-source LMMs and 6 closed-source models on LOKI, highlighting their potential as synthetic data detectors and also revealing some limitations in the development of LMM capabilities. More information about LOKI can be found at https://opendatalab.github.io/LOKI/ |
- Toward General Instruction-Following Alignment for Retrieval-Augmented Generation - Augmented Generation (RAG) systems. Despite recent advancements in Large Language Models (LLMs), research on assessing and improving instruction-following (IF) alignment within the RAG domain remains limited. To address this issue, we propose VIF-RAG, the first automated, scalable, and verifiable synthetic pipeline for instruction-following alignment in RAG systems. We start by manually crafting a minimal set of atomic instructions (<100) and developing combination rules to synthesize and verify complex instructions for a seed set. We then use supervised models for instruction rewriting while simultaneously generating code to automate the verification of instruction quality via a Python executor. Finally, we integrate these instructions with extensive RAG and general data samples, scaling up to a high-quality VIF-RAG-QA dataset (>100k) through automated processes. To further bridge the gap in instruction-following auto-evaluation for RAG systems, we introduce FollowRAG Benchmark, which includes approximately 3K test samples, covering 22 categories of general instruction constraints and four knowledge-intensive QA datasets. Due to its robust pipeline design, FollowRAG can seamlessly integrate with different RAG benchmarks. Using FollowRAG and eight widely-used IF and foundational abilities benchmarks for LLMs, we demonstrate that VIF-RAG markedly enhances LLM performance across a broad range of general instruction constraints while effectively leveraging its capabilities in RAG scenarios. Further analysis offers practical insights for achieving IF alignment in RAG systems. Our code and datasets are released at https://FollowRAG.github.io. |
- StructRAG: Boosting Knowledge Intensive Reasoning of LLMs via Inference-time Hybrid Information Structurization - augmented generation (RAG) is a key means to effectively enhance large language models (LLMs) in many knowledge-based tasks. However, existing RAG methods struggle with knowledge-intensive reasoning tasks, because useful information required to these tasks are badly scattered. This characteristic makes it difficult for existing RAG methods to accurately identify key information and perform global reasoning with such noisy augmentation. In this paper, motivated by the cognitive theories that humans convert raw information into various structured knowledge when tackling knowledge-intensive reasoning, we proposes a new framework, StructRAG, which can identify the optimal structure type for the task at hand, reconstruct original documents into this structured format, and infer answers based on the resulting structure. Extensive experiments across various knowledge-intensive tasks show that StructRAG achieves state-of-the-art performance, particularly excelling in challenging scenarios, demonstrating its potential as an effective solution for enhancing LLMs in complex real-world applications. |
- Ocean-omni: To Understand the World with Omni-modality - 4o highlight its critical role in practical applications, yet it lacks a high-performing open-source counterpart. In this paper, we introduce Ocean-omni, the first open-source 7B Multimodal Large Language Model (MLLM) adept at concurrently processing and analyzing modalities of image, video, audio, and text, while delivering an advanced multimodal interactive experience and strong performance. We propose an effective multimodal training schema starting with 7B model and proceeding through two stages of multimodal alignment and multitask fine-tuning across audio, image, video, and text modal. This approach equips the language model with the ability to handle visual and audio data effectively. Demonstrating strong performance across various omni-modal and multimodal benchmarks, we aim for this contribution to serve as a competitive baseline for the open-source community in advancing multimodal understanding and real-time interaction. |
- Agent S: An Open Agentic Framework that Uses Computers Like a Human - computer interaction by automating complex, multi-step tasks. Agent S aims to address three key challenges in automating computer tasks: acquiring domain-specific knowledge, planning over long task horizons, and handling dynamic, non-uniform interfaces. To this end, Agent S introduces experience-augmented hierarchical planning, which learns from external knowledge search and internal experience retrieval at multiple levels, facilitating efficient task planning and subtask execution. In addition, it employs an Agent-Computer Interface (ACI) to better elicit the reasoning and control capabilities of GUI agents based on Multimodal Large Language Models (MLLMs). Evaluation on the OSWorld benchmark shows that Agent S outperforms the baseline by 9.37% on success rate (an 83.6% relative improvement) and achieves a new state-of-the-art. Comprehensive analysis highlights the effectiveness of individual components and provides insights for future improvements. Furthermore, Agent S demonstrates broad generalizability to different operating systems on a newly-released WindowsAgentArena benchmark. Code available at https://github.com/simular-ai/Agent-S. |
- Multi-Agent Collaborative Data Selection for Efficient LLM Pretraining - agent collaborative data selection mechanism. In this framework, each data selection method serves as an independent agent, and an agent console is designed to dynamically integrate the information from all agents throughout the LLM training process. We conduct extensive empirical studies to evaluate our multi-agent framework. The experimental results demonstrate that our approach significantly improves data efficiency, accelerates convergence in LLM training, and achieves an average performance gain up to 10.5% across multiple language model benchmarks compared to the state-of-the-art methods. |
- Benchmarking Agentic Workflow Generation - faceted scenarios and intricate graph workflow structures. Additionally, we present WorFEval, a systemic evaluation protocol utilizing subsequence and subgraph matching algorithms to accurately quantify the LLM agent's workflow generation capabilities. Through comprehensive evaluations across different types of LLMs, we discover distinct gaps between the sequence planning capabilities and graph planning capabilities of LLM agents, with even GPT-4 exhibiting a gap of around 15%. We also train two open-source models and evaluate their generalization abilities on held-out tasks. Furthermore, we observe that the generated workflows can enhance downstream tasks, enabling them to achieve superior performance with less time during inference. Code and dataset are available at https://github.com/zjunlp/WorFBench. |
- WALL-E: World Alignment by Rule Learning Improves World Model-based LLM Agents - based agents? While the gaps between the prior knowledge of LLMs and the specified environment's dynamics do exist, our study reveals that the gaps can be bridged by aligning an LLM with its deployed environment and such "world alignment" can be efficiently achieved by rule learning on LLMs. Given the rich prior knowledge of LLMs, only a few additional rules suffice to align LLM predictions with the specified environment dynamics. To this end, we propose a neurosymbolic approach to learn these rules gradient-free through LLMs, by inducing, updating, and pruning rules based on comparisons of agent-explored trajectories and world model predictions. The resulting world model is composed of the LLM and the learned rules. Our embodied LLM agent "WALL-E" is built upon model-predictive control (MPC). By optimizing look-ahead actions based on the precise world model, MPC significantly improves exploration and learning efficiency. Compared to existing LLM agents, WALL-E's reasoning only requires a few principal rules rather than verbose buffered trajectories being included in the LLM input. On open-world challenges in Minecraft and ALFWorld, WALL-E achieves higher success rates than existing methods, with lower costs on replanning time and the number of tokens used for reasoning. In Minecraft, WALL-E exceeds baselines by 15-30% in success rate while costing 8-20 fewer replanning rounds and only 60-80% of tokens. In ALFWorld, its success rate surges to a new record high of 95% only after 6 iterations. |
- Personalized Visual Instruction Tuning - modal) large language models. To evaluate the personalized potential of MLLMs, we present a benchmark called P-Bench, which encompasses various question types with different levels of difficulty. The experiments demonstrate a substantial personalized performance enhancement after fine-tuning with our curated dataset. |
- From Generalist to Specialist: Adapting Vision Language Models via Task-Specific Visual Instruction Tuning - specific applications due to domain gaps between pre-training and fine-tuning. We introduce VITask, a novel framework that enhances task-specific adaptability of VLMs by integrating task-specific models (TSMs). VITask employs three key strategies: exemplar prompting (EP), response distribution alignment (RDA), and contrastive response tuning (CRT) to improve the task-specific performance of VLMs by adjusting their response distributions. EP allows TSM features to guide VLMs, while RDA enables VLMs to adapt without TSMs during inference by learning from exemplar-prompted models. CRT further optimizes the ranking of correct image-response pairs, thereby reducing the risk of generating undesired responses. Experiments on 12 medical diagnosis datasets across 9 imaging modalities show that VITask outperforms both vanilla instruction-tuned VLMs and TSMs, showcasing its ability to integrate complementary features from both models effectively. Additionally, VITask offers practical advantages such as flexible TSM integration and robustness to incomplete instructions, making it a versatile and efficient solution for task-specific VLM tuning. Our code are available at https://github.com/baiyang4/VITask. |
- Aria: An Open Multimodal Native Mixture-of-Experts Model - world information and deliver comprehensive understanding. While proprietary multimodal native models exist, their lack of openness imposes obstacles for adoptions, let alone adaptations. To fill this gap, we introduce Aria, an open multimodal native model with best-in-class performance across a wide range of multimodal, language, and coding tasks. Aria is a mixture-of-expert model with 3.9B and 3.5B activated parameters per visual token and text token, respectively. It outperforms Pixtral-12B and Llama3.2-11B, and is competitive against the best proprietary models on various multimodal tasks. We pre-train Aria from scratch following a 4-stage pipeline, which progressively equips the model with strong capabilities in language understanding, multimodal understanding, long context window, and instruction following. We open-source the model weights along with a codebase that facilitates easy adoptions and adaptations of Aria in real-world applications. |
- Differential Transformer - context modeling, key information retrieval, hallucination mitigation, in-context learning, and reduction of activation outliers. By being less distracted by irrelevant context, Diff Transformer can mitigate hallucination in question answering and text summarization. For in-context learning, Diff Transformer not only enhances accuracy but is also more robust to order permutation, which was considered as a chronic robustness issue. The results position Diff Transformer as a highly effective and promising architecture to advance large language models. |
- Intriguing Properties of Large Language and Vision Models - related tasks (e.g., MMVP) remains surprisingly low. This discrepancy raises the question of how LLVMs truly perceive images and exploit the advantages of the vision encoder. To address this, we systematically investigate this question regarding several aspects: permutation invariance, robustness, math reasoning, alignment preserving and importance, by evaluating the most common LLVM's families (i.e., LLaVA) across 10 evaluation benchmarks. Our extensive experiments reveal several intriguing properties of current LLVMs: (1) they internally process the image in a global manner, even when the order of visual patch sequences is randomly permuted; (2) they are sometimes able to solve math problems without fully perceiving detailed numerical information; (3) the cross-modal alignment is overfitted to complex reasoning tasks, thereby, causing them to lose some of the original perceptual capabilities of their vision encoder; (4) the representation space in the lower layers (<25%) plays a crucial role in determining performance and enhancing visual understanding. Lastly, based on the above observations, we suggest potential future directions for building better LLVMs and constructing more challenging evaluation benchmarks. |
- LongGenBench: Long-context Generation Benchmark - context benchmarks primarily focus on retrieval-based tests, requiring Large Language Models (LLMs) to locate specific information within extensive input contexts, such as the needle-in-a-haystack (NIAH) benchmark. Long-context generation refers to the ability of a language model to generate coherent and contextually accurate text that spans across lengthy passages or documents. While recent studies show strong performance on NIAH and other retrieval-based long-context benchmarks, there is a significant lack of benchmarks for evaluating long-context generation capabilities. To bridge this gap and offer a comprehensive assessment, we introduce a synthetic benchmark, LongGenBench, which allows for flexible configurations of customized generation context lengths. LongGenBench advances beyond traditional benchmarks by redesigning the format of questions and necessitating that LLMs respond with a single, cohesive long-context answer. Upon extensive evaluation using LongGenBench, we observe that: (1) both API accessed and open source models exhibit performance degradation in long-context generation scenarios, ranging from 1.2% to 47.1%; (2) different series of LLMs exhibit varying trends of performance degradation, with the Gemini-1.5-Flash model showing the least degradation among API accessed models, and the Qwen2 series exhibiting the least degradation in LongGenBench among open source models. |
- MLLM as Retriever: Interactively Learning Multimodal Retrieval for Embodied Agents - relevant trajectory data. However, current retrieval methods primarily focus on surface-level similarities of textual or visual cues in trajectories, neglecting their effectiveness for the specific task at hand. To address this issue, we propose a novel method, MLLM as ReTriever (MART), which enhances the performance of embodied agents by utilizing interaction data to fine-tune an MLLM retriever based on preference learning, such that the retriever fully considers the effectiveness of trajectories and prioritize them for unseen tasks. We also introduce Trajectory Abstraction, a mechanism that leverages MLLMs' summarization capabilities to represent trajectories with fewer tokens while preserving key information, enabling agents to better comprehend milestones in the trajectory. Experimental results across various environments demonstrate our method significantly improves task success rates in unseen scenes compared to baseline methods. This work presents a new paradigm for multimodal retrieval in embodied agents, by fine-tuning a general-purpose MLLM as the retriever to assess trajectory effectiveness. All benchmark task sets and simulator code modifications for action and observation spaces will be released. |
- Revisit Large-Scale Image-Caption Data in Pre-training Multimodal Foundation Models - text alignment, it is not clear whether they can fully replace AltTexts: the role of synthetic captions and their interaction with original web-crawled AltTexts in pre-training is still not well understood. Moreover, different multimodal foundation models may have unique preferences for specific caption formats, but efforts to identify the optimal captions for each model remain limited. In this work, we propose a novel, controllable, and scalable captioning pipeline designed to generate diverse caption formats tailored to various multimodal models. By examining Short Synthetic Captions (SSC) towards Dense Synthetic Captions (DSC+) as case studies, we systematically explore their effects and interactions with AltTexts across models such as CLIP, multimodal LLMs, and diffusion models. Our findings reveal that a hybrid approach that keeps both synthetic captions and AltTexts can outperform the use of synthetic captions alone, improving both alignment and performance, with each model demonstrating preferences for particular caption formats. This comprehensive analysis provides valuable insights into optimizing captioning strategies, thereby advancing the pre-training of multimodal foundation models. |
- LLMs Know More Than They Show: On the Intrinsic Representation of LLM Hallucinations - - contrary to prior claims -- truthfulness encoding is not universal but rather multifaceted. Next, we show that internal representations can also be used for predicting the types of errors the model is likely to make, facilitating the development of tailored mitigation strategies. Lastly, we reveal a discrepancy between LLMs' internal encoding and external behavior: they may encode the correct answer, yet consistently generate an incorrect one. Taken together, these insights deepen our understanding of LLM errors from the model's internal perspective, which can guide future research on enhancing error analysis and mitigation. |
- PHI-S: Distribution Balancing for Label-Free Multi-Teacher Distillation - teacher knowledge distillation without labels, termed "agglomerative models." We build upon this body of work by studying the effect of the teachers' activation statistics, particularly the impact of the loss function on the resulting student model quality. We explore a standard toolkit of statistical normalization techniques to better align the different distributions and assess their effects. Further, we examine the impact on downstream teacher-matching metrics, which motivates the use of Hadamard matrices. With these matrices, we demonstrate useful properties, showing how they can be used for isotropic standardization, where each dimension of a multivariate distribution is standardized using the same scale. We call this technique "PHI Standardization" (PHI-S) and empirically demonstrate that it produces the best student model across the suite of methods studied. |
- RATIONALYST: Pre-training Process-Supervision for Improving Reasoning - training data: underlying rationales are frequently left implicit (unstated). To address this challenge, we introduce RATIONALYST, a model for process-supervision of reasoning based on pre-training on a vast collection of rationale annotations extracted from unlabeled data. We extract 79k rationales from web-scale unlabelled dataset (the Pile) and a combination of reasoning datasets with minimal human intervention. This web-scale pre-training for reasoning allows RATIONALYST to consistently generalize across diverse reasoning tasks, including mathematical, commonsense, scientific, and logical reasoning. Fine-tuned from LLaMa-3-8B, RATIONALYST improves the accuracy of reasoning by an average of 3.9% on 7 representative reasoning benchmarks. It also demonstrates superior performance compared to significantly larger verifiers like GPT-4 and similarly sized models fine-tuned on matching training sets. |
- Addition is All You Need for Energy-efficient Language Models - complexity multiplication L-Mul algorithm that approximates floating point number multiplication with integer addition operations. The new algorithm costs significantly less computation resource than 8-bit floating point multiplication but achieves higher precision. Compared to 8-bit floating point multiplications, the proposed method achieves higher precision but consumes significantly less bit-level computation. Since multiplying floating point numbers requires substantially higher energy compared to integer addition operations, applying the L-Mul operation in tensor processing hardware can potentially reduce 95% energy cost by element-wise floating point tensor multiplications and 80% energy cost of dot products. We calculated the theoretical error expectation of L-Mul, and evaluated the algorithm on a wide range of textual, visual, and symbolic tasks, including natural language understanding, structural reasoning, mathematics, and commonsense question answering. Our numerical analysis experiments agree with the theoretical error estimation, which indicates that L-Mul with 4-bit mantissa achieves comparable precision as float8_e4m3 multiplications, and L-Mul with 3-bit mantissa outperforms float8_e5m2. Evaluation results on popular benchmarks show that directly applying L-Mul to the attention mechanism is almost lossless. We further show that replacing all floating point multiplications with 3-bit mantissa L-Mul in a transformer model achieves equivalent precision as using float8_e4m3 as accumulation precision in both fine-tuning and inference. |
-
:star: Best Gen AI Papers List (July 2024)
- The Llama 3 Herd of Models - 4 on a plethora of tasks. We publicly release Llama 3, including pre-trained and post-trained versions of the 405B parameter language model and our Llama Guard 3 model for input and output safety. The paper also presents the results of experiments in which we integrate image, video, and speech capabilities into Llama 3 via a compositional approach. We observe this approach performs competitively with the state-of-the-art on image, video, and speech recognition tasks. The resulting models are not yet being broadly released as they are still under development. | Foundation Model |
- ShieldGemma: Generative AI Content Moderation Based on Gemma - based safety content moderation models built upon Gemma2. These models provide robust, state-of-the-art predictions of safety risks across key harm types (sexually explicit, dangerous content, harassment, hate speech) in both user input and LLM-generated output. By evaluating on both public and internal benchmarks, we demonstrate superior performance compared to existing models, such as Llama Guard (+10.8\% AU-PRC on public benchmarks) and WildCard (+4.3\%). Additionally, we present a novel LLM-based data curation pipeline, adaptable to a variety of safety-related tasks and beyond. We have shown strong generalization performance for model trained mainly on synthetic data. By releasing ShieldGemma, we provide a valuable resource to the research community, advancing LLM safety and enabling the creation of more effective content moderation solutions for developers. | Content Moderation |
- MoMa: Efficient Early-Fusion Pre-training with Mixture of Modality-Aware Experts - aware mixture-of-experts (MoE) architecture designed for pre-training mixed-modal, early-fusion language models. MoMa processes images and text in arbitrary sequences by dividing expert modules into modality-specific groups. These groups exclusively process designated tokens while employing learned routing within each group to maintain semantically informed adaptivity. Our empirical results reveal substantial pre-training efficiency gains through this modality-specific parameter allocation. Under a 1-trillion-token training budget, the MoMa 1.4B model, featuring 4 text experts and 4 image experts, achieves impressive FLOPs savings: 3.7x overall, with 2.6x for text and 5.2x for image processing compared to a compute-equivalent dense baseline, measured by pre-training loss. This outperforms the standard expert-choice MoE with 8 mixed-modal experts, which achieves 3x overall FLOPs savings (3x for text, 2.8x for image). Combining MoMa with mixture-of-depths (MoD) further improves pre-training FLOPs savings to 4.2x overall (text: 3.4x, image: 5.3x), although this combination hurts performance in causal inference due to increased sensitivity to router accuracy. These results demonstrate MoMa's potential to significantly advance the efficiency of mixed-modal, early-fusion language model pre-training, paving the way for more resource-efficient and capable multimodal AI systems. | LLM Architecture |
- Very Large-Scale Multi-Agent Simulation in AgentScope - agent systems in very large-scale simulations. However, there remain several challenges when conducting multi-agent simulations with existing platforms, such as limited scalability and low efficiency, unsatisfied agent diversity, and effort-intensive management processes. To address these challenges, we develop several new features and components for AgentScope, a user-friendly multi-agent platform, enhancing its convenience and flexibility for supporting very large-scale multi-agent simulations. Specifically, we propose an actor-based distributed mechanism as the underlying technological infrastructure towards great scalability and high efficiency, and provide flexible environment support for simulating various real-world scenarios, which enables parallel execution of multiple agents, centralized workflow orchestration, and both inter-agent and agent-environment interactions among agents. Moreover, we integrate an easy-to-use configurable tool and an automatic background generation pipeline in AgentScope, simplifying the process of creating agents with diverse yet detailed background settings. Last but not least, we provide a web-based interface for conveniently monitoring and managing a large number of agents that might deploy across multiple devices. We conduct a comprehensive simulation to demonstrate the effectiveness of the proposed enhancements in AgentScope, and provide detailed observations and discussions to highlight the great potential of applying multi-agent systems in large-scale simulations. The source code is released on GitHub at https://github.com/modelscope/agentscope to inspire further research and development in large-scale multi-agent simulations. | Agents |
- OpenDevin: An Open Platform for AI Software Developers as Generalist Agents - Bench) and web browsing (e.g., WebArena), among others. Released under the permissive MIT license, OpenDevin is a community project spanning academia and industry with more than 1.3K contributions from over 160 contributors and will improve going forward. | Task Specific LLMs |
- The Vision of Autonomic Computing: Can LLMs Make It a Reality? - manage akin to biological organisms, adapting seamlessly to changing environments. Despite decades of research, achieving ACV remains challenging due to the dynamic and complex nature of modern computing systems. Recent advancements in Large Language Models (LLMs) offer promising solutions to these challenges by leveraging their extensive knowledge, language understanding, and task automation capabilities. This paper explores the feasibility of realizing ACV through an LLM-based multi-agent framework for microservice management. We introduce a five-level taxonomy for autonomous service maintenance and present an online evaluation benchmark based on the Sock Shop microservice demo project to assess our framework's performance. Our findings demonstrate significant progress towards achieving Level 3 autonomy, highlighting the effectiveness of LLMs in detecting and resolving issues within microservice architectures. This study contributes to advancing autonomic computing by pioneering the integration of LLMs into microservice management frameworks, paving the way for more adaptive and self-managing computing systems. The code will be made available at https://aka.ms/ACV-LLM. | Multi-Agent Systems, Position Paper |
- Compact Language Models via Pruning and Knowledge Distillation - intensive. In this paper, we investigate if pruning an existing LLM and then re-training it with a fraction (<3%) of the original training data can be a suitable alternative to repeated, full retraining. To this end, we develop a set of practical and effective compression best practices for LLMs that combine depth, width, attention and MLP pruning with knowledge distillation-based retraining; we arrive at these best practices through a detailed empirical exploration of pruning strategies for each axis, methods to combine axes, distillation strategies, and search techniques for arriving at optimal compressed architectures. We use this guide to compress the Nemotron-4 family of LLMs by a factor of 2-4x, and compare their performance to similarly-sized models on a variety of language modeling tasks. Deriving 8B and 4B models from an already pretrained 15B model using our approach requires up to 40x fewer training tokens per model compared to training from scratch; this results in compute cost savings of 1.8x for training the full model family (15B, 8B, and 4B). Minitron models exhibit up to a 16% improvement in MMLU scores compared to training from scratch, perform comparably to other community models such as Mistral 7B, Gemma 7B and Llama-3 8B, and outperform state-of-the-art compression techniques from the literature. We have open-sourced Minitron model weights on Huggingface, with corresponding supplementary material including example code available on GitHub. | Knowledge Distillation |
- Internal Consistency and Self-Feedback in Large Language Models: A Survey - '' such as Self-Consistency, Self-Improve, and Self-Refine have been initiated. They share a commonality: involving LLMs evaluating and updating itself to mitigate the issues. Nonetheless, these efforts lack a unified perspective on summarization, as existing surveys predominantly focus on categorization without examining the motivations behind these works. In this paper, we summarize a theoretical framework, termed Internal Consistency, which offers unified explanations for phenomena such as the lack of reasoning and the presence of hallucinations. Internal Consistency assesses the coherence among LLMs' latent layer, decoding layer, and response layer based on sampling methodologies. Expanding upon the Internal Consistency framework, we introduce a streamlined yet effective theoretical framework capable of mining Internal Consistency, named Self-Feedback. The Self-Feedback framework consists of two modules: Self-Evaluation and Self-Update. This framework has been employed in numerous studies. We systematically classify these studies by tasks and lines of work; summarize relevant evaluation methods and benchmarks; and delve into the concern, `Does Self-Feedback Really Work?'' We propose several critical viewpoints, including the `Hourglass Evolution of Internal Consistency'', `Consistency Is (Almost) Correctness'' hypothesis, and ``The Paradox of Latent and Explicit Reasoning''. Furthermore, we outline promising directions for future research. We have open-sourced the experimental code, reference list, and statistical data, available at \url{https://github.com/IAAR-Shanghai/ICSFSurvey}. | Self-Feedback in LLMs |
- LazyLLM: Dynamic Token Pruning for Efficient Long Context LLM Inference - based large language models consists of two sequential stages: 1) a prefilling stage to compute the KV cache of prompts and generate the first token, and 2) a decoding stage to generate subsequent tokens. For long prompts, the KV cache must be computed for all tokens during the prefilling stage, which can significantly increase the time needed to generate the first token. Consequently, the prefilling stage may become a bottleneck in the generation process. An open question remains whether all prompt tokens are essential for generating the first token. To answer this, we introduce a novel method, LazyLLM, that selectively computes the KV for tokens important for the next token prediction in both the prefilling and decoding stages. Contrary to static pruning approaches that prune the prompt at once, LazyLLM allows language models to dynamically select different subsets of tokens from the context in different generation steps, even though they might be pruned in previous steps. Extensive experiments on standard datasets across various tasks demonstrate that LazyLLM is a generic method that can be seamlessly integrated with existing language models to significantly accelerate the generation without fine-tuning. For instance, in the multi-document question-answering task, LazyLLM accelerates the prefilling stage of the LLama 2 7B model by 2.34x while maintaining accuracy. | LLM Pruning |
- Scaling Laws with Vocabulary: Larger Models Deserve Larger Vocabularies - optimal vocabulary size: IsoFLOPs analysis, derivative estimation, and parametric fit of the loss function. Our approaches converge on the same result that the optimal vocabulary size depends on the available compute budget and that larger models deserve larger vocabularies. However, most LLMs use too small vocabulary sizes. For example, we predict that the optimal vocabulary size of Llama2-70B should have been at least 216K, 7 times larger than its vocabulary of 32K. We validate our predictions empirically by training models with 3B parameters across different FLOPs budgets. Adopting our predicted optimal vocabulary size consistently improves downstream performance over commonly used vocabulary sizes. By increasing the vocabulary size from the conventional 32K to 43K, we improve performance on ARC-Challenge from 29.1 to 32.0 with the same 2.3e21 FLOPs. Our work emphasizes the necessity of jointly considering model parameters and vocabulary size for efficient scaling. | LLM Vocabulary |
- AgentPoison: Red-teaming LLM Agents via Poisoning Memory or Knowledge Bases - augmented generation (RAG) mechanism, retrieving past knowledge and instances with similar embeddings from knowledge bases to inform task planning and execution. However, the reliance on unverified knowledge bases raises significant concerns about their safety and trustworthiness. To uncover such vulnerabilities, we propose a novel red teaming approach AgentPoison, the first backdoor attack targeting generic and RAG-based LLM agents by poisoning their long-term memory or RAG knowledge base. In particular, we form the trigger generation process as a constrained optimization to optimize backdoor triggers by mapping the triggered instances to a unique embedding space, so as to ensure that whenever a user instruction contains the optimized backdoor trigger, the malicious demonstrations are retrieved from the poisoned memory or knowledge base with high probability. In the meantime, benign instructions without the trigger will still maintain normal performance. Unlike conventional backdoor attacks, AgentPoison requires no additional model training or fine-tuning, and the optimized backdoor trigger exhibits superior transferability, in-context coherence, and stealthiness. Extensive experiments demonstrate AgentPoison's effectiveness in attacking three types of real-world LLM agents: RAG-based autonomous driving agent, knowledge-intensive QA agent, and healthcare EHRAgent. On each agent, AgentPoison achieves an average attack success rate higher than 80% with minimal impact on benign performance (less than 1%) with a poison rate less than 0.1%. | Red-Teaming |
- Towards Understanding Unsafe Video Generation - quality output. It is important to understand their potential to produce unsafe content, such as violent or terrifying videos. In this work, we provide a comprehensive understanding of unsafe video generation. First, to confirm the possibility that these models could indeed generate unsafe videos, we choose unsafe content generation prompts collected from 4chan and Lexica, and three open-source SOTA VGMs to generate unsafe videos. After filtering out duplicates and poorly generated content, we created an initial set of 2112 unsafe videos from an original pool of 5607 videos. Through clustering and thematic coding analysis of these generated videos, we identify 5 unsafe video categories: Distorted/Weird, Terrifying, Pornographic, Violent/Bloody, and Political. With IRB approval, we then recruit online participants to help label the generated videos. Based on the annotations submitted by 403 participants, we identified 937 unsafe videos from the initial video set. With the labeled information and the corresponding prompts, we created the first dataset of unsafe videos generated by VGMs. We then study possible defense mechanisms to prevent the generation of unsafe videos. Existing defense methods in image generation focus on filtering either input prompt or output results. We propose a new approach called Latent Variable Defense (LVD), which works within the model's internal sampling process. LVD can achieve 0.90 defense accuracy while reducing time and computing resources by 10x when sampling a large number of unsafe prompts. | Moderating Video Generation |
- Spectra: A Comprehensive Study of Ternary, Quantized, and FP16 Language Models - training quantization is the leading method for addressing memory-related bottlenecks in LLM inference, but unfortunately, it suffers from significant performance degradation below 4-bit precision. An alternative approach involves training compressed models directly at a low bitwidth (e.g., binary or ternary models). However, the performance, training dynamics, and scaling trends of such models are not yet well understood. To address this issue, we train and openly release the Spectra LLM suite consisting of 54 language models ranging from 99M to 3.9B parameters, trained on 300B tokens. Spectra includes FloatLMs, post-training quantized QuantLMs (3, 4, 6, and 8 bits), and ternary LLMs (TriLMs) - our improved architecture for ternary language modeling, which significantly outperforms previously proposed ternary models of a given size (in bits), matching half-precision models at scale. For example, TriLM 3.9B is (bit-wise) smaller than the half-precision FloatLM 830M, but matches half-precision FloatLM 3.9B in commonsense reasoning and knowledge benchmarks. However, TriLM 3.9B is also as toxic and stereotyping as FloatLM 3.9B, a model six times larger in size. Additionally, TriLM 3.9B lags behind FloatLM in perplexity on validation splits and web-based corpora but performs better on less noisy datasets like Lambada and PennTreeBank. To enhance understanding of low-bitwidth models, we are releasing 500+ intermediate checkpoints of the Spectra suite at \href{https://github.com/NolanoOrg/SpectraSuite}{https://github.com/NolanoOrg/SpectraSuite}. | Small Language Models |
- NeedleBench: Can LLMs Do Retrieval and Reasoning in 1 Million Context Window? - context capabilities of large language models (LLMs), identifying content relevant to a user's query from original long documents is a crucial prerequisite for any LLM to answer questions based on long text. We present NeedleBench, a framework consisting of a series of progressively more challenging tasks for assessing bilingual long-context capabilities, spanning multiple length intervals (4k, 8k, 32k, 128k, 200k, 1000k, and beyond) and different depth ranges, allowing the strategic insertion of critical data points in different text depth zones to rigorously test the retrieval and reasoning capabilities of models in diverse contexts. We use the NeedleBench framework to assess how well the leading open-source models can identify key information relevant to the question and apply that information to reasoning in bilingual long texts. Furthermore, we propose the Ancestral Trace Challenge (ATC) to mimic the complexity of logical reasoning challenges that are likely to be present in real-world long-context tasks, providing a simple method for evaluating LLMs in dealing with complex long-context situations. Our results suggest that current LLMs have significant room for improvement in practical long-context applications, as they struggle with the complexity of logical reasoning challenges that are likely to be present in real-world long-context tasks. All codes and resources are available at OpenCompass: https://github.com/open-compass/opencompass. | Long Context |
- MMM: Multilingual Mutual Reinforcement Effect Mix Datasets & Test with Open-domain Information Extraction Large Language Models - datasets in English, Japanese, and Chinese. In this paper, we also propose a method for dataset translation assisted by Large Language Models (LLMs), which significantly reduces the manual annotation time required for dataset construction by leveraging LLMs to translate the original Japanese datasets. Additionally, we have enriched the dataset by incorporating open-domain Named Entity Recognition (NER) and sentence classification tasks. Utilizing this expanded dataset, we developed a unified input-output framework to train an Open-domain Information Extraction Large Language Model (OIELLM). The OIELLM model demonstrates the capability to effectively process novel MMM datasets, exhibiting significant improvements in performance. | Multilingual LLMs |
- Qwen2-Audio Technical Report - Audio, a large-scale audio-language model called Qwen2-Audio, which is capable of accepting various audio signal inputs and performing audio analysis or direct textual responses with regard to speech instructions. In contrast to complex hierarchical tags, we have simplified the pre-training process by utilizing natural language prompts for different data and tasks, and have further expanded the data volume. We have boosted the instruction-following capability of Qwen2-Audio and implemented two distinct audio interaction modes for voice chat and audio analysis. In the voice chat mode, users can freely engage in voice interactions with Qwen2-Audio without text input. In the audio analysis mode, users could provide audio and text instructions for analysis during the interaction. Note that we do not use any system prompts to switch between voice chat and audio analysis modes. Qwen2-Audio is capable of intelligently comprehending the content within audio and following voice commands to respond appropriately. For instance, in an audio segment that simultaneously contains sounds, multi-speaker conversations, and a voice command, Qwen2-Audio can directly understand the command and provide an interpretation and response to the audio. Additionally, DPO has optimized the model's performance in terms of factuality and adherence to desired behavior. According to the evaluation results from AIR-Bench, Qwen2-Audio outperformed previous SOTAs, such as Gemini-1.5-pro, in tests focused on audio-centric instruction-following capabilities. Qwen2-Audio is open-sourced with the aim of fostering the advancement of the multi-modal language community. | Foundation Model |
- Qwen2 Technical Report - tuned language models, encompassing a parameter range from 0.5 to 72 billion, featuring dense models and a Mixture-of-Experts model. Qwen2 surpasses most prior open-weight models, including its predecessor Qwen1.5, and exhibits competitive performance relative to proprietary models across diverse benchmarks on language understanding, generation, multilingual proficiency, coding, mathematics, and reasoning. The flagship model, Qwen2-72B, showcases remarkable performance: 84.2 on MMLU, 37.9 on GPQA, 64.6 on HumanEval, 89.5 on GSM8K, and 82.4 on BBH as a base language model. The instruction-tuned variant, Qwen2-72B-Instruct, attains 9.1 on MT-Bench, 48.1 on Arena-Hard, and 35.7 on LiveCodeBench. Moreover, Qwen2 demonstrates robust multilingual capabilities, proficient in approximately 30 languages, spanning English, Chinese, Spanish, French, German, Arabic, Russian, Korean, Japanese, Thai, Vietnamese, and more, underscoring its versatility and global reach. To foster community innovation and accessibility, we have made the Qwen2 model weights openly available on Hugging Face and ModelScope, and the supplementary materials including example code on GitHub. These platforms also include resources for quantization, fine-tuning, and deployment, facilitating a wide range of applications and research endeavors. | Foundation Model |
- Learning to Refuse: Towards Mitigating Privacy Risks in LLMs - world pErsonal daTa UnleaRNing dataset, comprising 2,492 individuals from Wikipedia with associated QA pairs, to evaluate machine unlearning (MU) methods for protecting personal data in a realistic scenario. Additionally, we introduce the Name-Aware Unlearning Framework (NAUF) for Privacy Protection, which enables the model to learn which individuals' information should be protected without affecting its ability to answer questions related to other unrelated individuals. Our extensive experiments demonstrate that NAUF achieves a state-of-the-art average unlearning score, surpassing the best baseline method by 5.65 points, effectively protecting target individuals' personal data while maintaining the model's general capabilities. | Privacy in LLMs |
- Human-like Episodic Memory for Infinite Context LLMs - LLM, a novel approach that integrates key aspects of human episodic memory and event cognition into LLMs, enabling them to effectively handle practically infinite context lengths while maintaining computational efficiency. EM-LLM organises sequences of tokens into coherent episodic events using a combination of Bayesian surprise and graph-theoretic boundary refinement in an on-line fashion. When needed, these events are retrieved through a two-stage memory process, combining similarity-based and temporally contiguous retrieval for efficient and human-like access to relevant information. Experiments on the LongBench dataset demonstrate EM-LLM's superior performance, outperforming the state-of-the-art InfLLM model with an overall relative improvement of 4.3% across various tasks, including a 33% improvement on the PassageRetrieval task. Furthermore, our analysis reveals strong correlations between EM-LLM's event segmentation and human-perceived events, suggesting a bridge between this artificial system and its biological counterpart. This work not only advances LLM capabilities in processing extended contexts but also provides a computational framework for exploring human memory mechanisms, opening new avenues for interdisciplinary research in AI and cognitive science. | LLM Memory |
- Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Training - known models such as GPT-4 in defending against attacks. Importantly, our approach successfully defends recent advanced attack methods (e.g., CodeAttack) that have jailbroken GPT-4 and LLaMA3-70B-Instruct. Our code and data can be found at https://github.com/RobustNLP/DeRTa. | Safety in LLMs |
- Video Diffusion Alignment via Reward Gradients - scale unsupervised data, it has become crucial to adapt these models to specific downstream tasks. Adapting these models via supervised fine-tuning requires collecting target datasets of videos, which is challenging and tedious. In this work, we utilize pre-trained reward models that are learned via preferences on top of powerful vision discriminative models to adapt video diffusion models. These models contain dense gradient information with respect to generated RGB pixels, which is critical to efficient learning in complex search spaces, such as videos. We show that backpropagating gradients from these reward models to a video diffusion model can allow for compute and sample efficient alignment of the video diffusion model. We show results across a variety of reward models and video diffusion models, demonstrating that our approach can learn much more efficiently in terms of reward queries and computation than prior gradient-free approaches. Our code, model weights,and more visualization are available at https://vader-vid.github.io. | Multimodal Alignment |
- Is Your Model Really A Good Math Reasoner? Evaluating Mathematical Reasoning with Checklist - world scenarios, has emerged as a critical issue. Current benchmarks predominantly concentrate on problem-solving capabilities, which presents a substantial risk of model overfitting and fails to accurately represent genuine mathematical reasoning abilities. In this paper, we argue that if a model really understands a problem, it should be robustly and readily applied across a diverse array of tasks. Motivated by this, we introduce MATHCHECK, a well-designed checklist for testing task generalization and reasoning robustness, as well as an automatic tool to generate checklists efficiently. MATHCHECK includes multiple mathematical reasoning tasks and robustness test types to facilitate a comprehensive evaluation of both mathematical reasoning ability and behavior testing. Utilizing MATHCHECK, we develop MATHCHECK-GSM and MATHCHECK-GEO to assess mathematical textual reasoning and multi-modal reasoning capabilities, respectively, serving as upgraded versions of benchmarks including GSM8k, GeoQA, UniGeo, and Geometry3K. We adopt MATHCHECK-GSM and MATHCHECK-GEO to evaluate over 20 LLMs and 11 MLLMs, assessing their comprehensive mathematical reasoning abilities. Our results demonstrate that while frontier LLMs like GPT-4o continue to excel in various abilities on the checklist, many other model families exhibit a significant decline. Further experiments indicate that, compared to traditional math benchmarks, MATHCHECK better reflects true mathematical abilities and represents mathematical intelligence more linearly, thereby supporting our design. On our MATHCHECK, we can easily conduct detailed behavior analysis to deeply investigate models. | Task Specific LLMs |
- PaliGemma: A versatile 3B VLM for transfer - Language Model (VLM) that is based on the SigLIP-So400m vision encoder and the Gemma-2B language model. It is trained to be a versatile and broadly knowledgeable base model that is effective to transfer. It achieves strong performance on a wide variety of open-world tasks. We evaluate PaliGemma on almost 40 diverse tasks including standard VLM benchmarks, but also more specialized tasks such as remote-sensing and segmentation. | Vision Language Models |
- On Leakage of Code Generation Evaluation Datasets - Generation |
- Internet of Agents: Weaving a Web of Heterogeneous Agents for Collaborative Intelligence - agent frameworks often struggle with integrating diverse capable third-party agents due to reliance on agents defined within their own ecosystems. They also face challenges in simulating distributed environments, as most frameworks are limited to single-device setups. Furthermore, these frameworks often rely on hard-coded communication pipelines, limiting their adaptability to dynamic task requirements. Inspired by the concept of the Internet, we propose the Internet of Agents (IoA), a novel framework that addresses these limitations by providing a flexible and scalable platform for LLM-based multi-agent collaboration. IoA introduces an agent integration protocol, an instant-messaging-like architecture design, and dynamic mechanisms for agent teaming and conversation flow control. Through extensive experiments on general assistant tasks, embodied AI tasks, and retrieval-augmented generation benchmarks, we demonstrate that IoA consistently outperforms state-of-the-art baselines, showcasing its ability to facilitate effective collaboration among heterogeneous agents. IoA represents a step towards linking diverse agents in an Internet-like environment, where agents can seamlessly collaborate to achieve greater intelligence and capabilities. Our codebase has been released at \url{https://github.com/OpenBMB/IoA}. | Multi-Agent Systems |
- RULE: Reliable Multimodal RAG for Factuality in Medical Vision Language Models - LVLMs) has enhanced medical diagnosis. However, current Med-LVLMs frequently encounter factual issues, often generating responses that do not align with established medical facts. Retrieval-Augmented Generation (RAG), which utilizes external knowledge, can improve the factual accuracy of these models but introduces two major challenges. First, limited retrieved contexts might not cover all necessary information, while excessive retrieval can introduce irrelevant and inaccurate references, interfering with the model's generation. Second, in cases where the model originally responds correctly, applying RAG can lead to an over-reliance on retrieved contexts, resulting in incorrect answers. To address these issues, we propose RULE, which consists of two components. First, we introduce a provably effective strategy for controlling factuality risk through the calibrated selection of the number of retrieved contexts. Second, based on samples where over-reliance on retrieved contexts led to errors, we curate a preference dataset to fine-tune the model, balancing its dependence on inherent knowledge and retrieved contexts for generation. We demonstrate the effectiveness of RULE on three medical VQA datasets, achieving an average improvement of 20.8% in factual accuracy. We publicly release our benchmark and code in https://github.com/richard-peng-xia/RULE. | Task-Specific LLMs |
- On scalable oversight with weak LLMs judging strong LLMs - answering, where the judge just answers outright without the AI. We use large language models (LLMs) as both AI agents and as stand-ins for human judges, taking the judge models to be weaker than agent models. We benchmark on a diverse range of asymmetries between judges and agents, extending previous work on a single extractive QA task with information asymmetry, to also include mathematics, coding, logic and multimodal reasoning asymmetries. We find that debate outperforms consultancy across all tasks when the consultant is randomly assigned to argue for the correct/incorrect answer. Comparing debate to direct question answering, the results depend on the type of task: in extractive QA tasks with information asymmetry debate outperforms direct question answering, but in other tasks without information asymmetry the results are mixed. Previous work assigned debaters/consultants an answer to argue for. When we allow them to instead choose which answer to argue for, we find judges are less frequently convinced by the wrong answer in debate than in consultancy. Further, we find that stronger debater models increase judge accuracy, though more modestly than in previous studies. | LLM Evaluation |
- Stark: Social Long-Term Multi-Modal Conversation with Persona Commonsense Knowledge - sharing behavior in singular sessions, leading to limited long-term social interaction, and (2) a lack of personalized image-sharing behavior. In this work, we introduce Stark, a large-scale long-term multi-modal conversation dataset that covers a wide range of social personas in a multi-modality format, time intervals, and images. To construct Stark automatically, we propose a novel multi-modal contextualization framework, Mcu, that generates long-term multi-modal dialogue distilled from ChatGPT and our proposed Plan-and-Execute image aligner. Using our Stark, we train a multi-modal conversation model, Ultron 7B, which demonstrates impressive visual imagination ability. Furthermore, we demonstrate the effectiveness of our dataset in human evaluation. We make our source code and dataset publicly available. | Multimodal Models |
- Evaluating Language Model Context Windows: A "Working Memory" Test and Inference-time Correction - world applications, often tasked with reasoning over large volumes of documents. An exciting development in this space is models boasting extended context capabilities, with some accommodating over 2 million tokens. Such long context model capabilities remain uncertain in production systems, motivating the need to benchmark their performance on real world use cases. We address this challenge by proposing SWiM, an evaluation framework that addresses the limitations of standard tests. Testing the framework on eight long context models, we find that even strong models such as GPT-4 and Claude 3 Opus degrade in performance when information is present in the middle of the context window (lost-in-the-middle effect). Next, in addition to our benchmark, we propose medoid voting, a simple, but effective training-free approach that helps alleviate this effect, by generating responses a few times, each time randomly permuting documents in the context, and selecting the medoid answer. We evaluate medoid voting on single document QA tasks, achieving up to a 24% lift in accuracy. Our code is available at https://github.com/snorkel-ai/long-context-eval. | LLM Context Evaluation |
- AgentInstruct: Toward Generative Teaching with Agentic Flows - training, specifically creating data by powerful models to teach a new skill or behavior to another model, we refer to this setting as Generative Teaching. We introduce AgentInstruct, an extensible agentic framework for automatically creating large amounts of diverse and high-quality synthetic data. AgentInstruct can create both the prompts and responses, using only raw data sources like text documents and code files as seeds. We demonstrate the utility of AgentInstruct by creating a post training dataset of 25M pairs to teach language models different skills, such as text editing, creative writing, tool usage, coding, reading comprehension, etc. The dataset can be used for instruction tuning of any base model. We post-train Mistral-7b with the data. When comparing the resulting model Orca-3 to Mistral-7b-Instruct (which uses the same base model), we observe significant improvements across many benchmarks. For example, 40% improvement on AGIEval, 19% improvement on MMLU, 54% improvement on GSM8K, 38% improvement on BBH and 45% improvement on AlpacaEval. Additionally, it consistently outperforms other models such as LLAMA-8B-instruct and GPT-3.5-turbo. | Agents |
- HEMM: Holistic Evaluation of Multimodal Foundation Models - world applications. However, it is challenging to characterize and study progress in multimodal foundation models, given the range of possible modeling decisions, tasks, and domains. In this paper, we introduce Holistic Evaluation of Multimodal Models (HEMM) to systematically evaluate the capabilities of multimodal foundation models across a set of 3 dimensions: basic skills, information flow, and real-world use cases. Basic multimodal skills are internal abilities required to solve problems, such as learning interactions across modalities, fine-grained alignment, multi-step reasoning, and the ability to handle external knowledge. Information flow studies how multimodal content changes during a task through querying, translation, editing, and fusion. Use cases span domain-specific challenges introduced in real-world multimedia, affective computing, natural sciences, healthcare, and human-computer interaction applications. Through comprehensive experiments across the 30 tasks in HEMM, we (1) identify key dataset dimensions (e.g., basic skills, information flows, and use cases) that pose challenges to today's models, and (2) distill performance trends regarding how different modeling dimensions (e.g., scale, pre-training data, multimodal alignment, pre-training, and instruction tuning objectives) influence performance. Our conclusions regarding challenging multimodal interactions, use cases, and tasks requiring reasoning and external knowledge, the benefits of data and model scale, and the impacts of instruction tuning yield actionable insights for future work in multimodal foundation models. | Multimodal Model Evaluation |
- InternLM-XComposer-2.5: A Versatile Large Vision Language Model Supporting Long-Contextual Input and Output - XComposer-2.5 (IXC-2.5), a versatile large-vision language model that supports long-contextual input and output. IXC-2.5 excels in various text-image comprehension and composition applications, achieving GPT-4V level capabilities with merely 7B LLM backend. Trained with 24K interleaved image-text contexts, it can seamlessly extend to 96K long contexts via RoPE extrapolation. This long-context capability allows IXC-2.5 to excel in tasks requiring extensive input and output contexts. Compared to its previous 2.0 version, InternLM-XComposer-2.5 features three major upgrades in vision-language comprehension: (1) Ultra-High Resolution Understanding, (2) Fine-Grained Video Understanding, and (3) Multi-Turn Multi-Image Dialogue. In addition to comprehension, IXC-2.5 extends to two compelling applications using extra LoRA parameters for text-image composition: (1) Crafting Webpages and (2) Composing High-Quality Text-Image Articles. IXC-2.5 has been evaluated on 28 benchmarks, outperforming existing open-source state-of-the-art models on 16 benchmarks. It also surpasses or competes closely with GPT-4V and Gemini Pro on 16 key tasks. The InternLM-XComposer-2.5 is publicly available at https://github.com/InternLM/InternLM-XComposer. | Foundation Model |
- A False Sense of Safety: Unsafe Information Leakage in 'Safe' AI Responses - tuning, are, and will remain, fundamentally insufficient for ensuring model safety. These defenses fail to address risks arising from dual-intent queries and the ability to composite innocuous outputs to achieve harmful goals. To address this critical gap, we introduce an information-theoretic threat model called inferential adversaries who exploit impermissible information leakage from model outputs to achieve malicious goals. We distinguish these from commonly studied security adversaries who only seek to force victim models to generate specific impermissible outputs. We demonstrate the feasibility of automating inferential adversaries through question decomposition and response aggregation. To provide safety guarantees, we define an information censorship criterion for censorship mechanisms, bounding the leakage of impermissible information. We propose a defense mechanism which ensures this bound and reveal an intrinsic safety-utility trade-off. Our work provides the first theoretically grounded understanding of the requirements for releasing safe LLMs and the utility costs involved. | Safety in LLMs |
- To Forget or Not? Towards Practical Knowledge Unlearning for Large Language Models
- Summary of a Haystack: A Challenge to Long-Context LLMs and RAG Systems - context tasks remains challenging, as tasks like Needle-in-a-Haystack lack complexity. In this work, we argue that summarization can play a central role in such evaluation. We design a procedure to synthesize Haystacks of documents, ensuring that specific \textit{insights} repeat across documents. The "Summary of a Haystack" (SummHay) task then requires a system to process the Haystack and generate, given a query, a summary that identifies the relevant insights and precisely cites the source documents. Since we have precise knowledge of what insights should appear in a haystack summary and what documents should be cited, we implement a highly reproducible automatic evaluation that can score summaries on two aspects - Coverage and Citation. We generate Haystacks in two domains (conversation, news), and perform a large-scale evaluation of 10 LLMs and corresponding 50 RAG systems. Our findings indicate that SummHay is an open challenge for current systems, as even systems provided with an Oracle signal of document relevance lag our estimate of human performance (56\%) by 10+ points on a Joint Score. Without a retriever, long-context LLMs like GPT-4o and Claude 3 Opus score below 20% on SummHay. We show SummHay can also be used to study enterprise RAG systems and position bias in long-context models. We hope future systems can equal and surpass human performance on SummHay. | LLM Context Length |
-
:star: Best Gen AI Papers List (May 2024)
- LLMs achieve adult human performance on higher-order theory of mind tasks - order theory of mind (ToM); the human ability to reason about multiple mental and emotional states in a recursive manner (e.g. I think that you believe that she knows). This paper builds on prior work by introducing a handwritten test suite – Multi-Order Theory of Mind Q&A – and using it to compare the performance of five LLMs to a newly gathered adult human benchmark. We find that GPT-4 and Flan-PaLM reach adult-level and near adult-level performance on ToM tasks overall, and that GPT-4 exceeds adult performance on 6th order inferences. Our results suggest that there is an interplay between model size and finetuning for the realisation of ToM abilities, and that the best-performing LLMs have developed a generalised capacity for ToM. Given the role that higher-order ToM plays in a wide range of cooperative and competitive human behaviours, these findings have significant implications for user-facing LLM applications. | Theory of Mind |
- JINA CLIP: Your CLIP Model Is Also Your Text Retriever - Image Pretraining (CLIP) is widely used to train models to align images and texts in a common embedding space by mapping them to fixed-sized vectors. These models are key to multimodal information retrieval and related tasks. However, CLIP models generally underperform in text-only tasks compared to specialized text models. This creates inefficiencies for information retrieval systems that keep separate embeddings and models for text-only and multimodal tasks. We propose a novel, multi-task contrastive training method to address this issue, which we use to train the jina-clip-v1 model to achieve the state-of-the-art performance on both text-image and text-text retrieval tasks. | Multimodal Models |
- Parrot: Efficient Serving of LLM-based Applications with Semantic Variable - based applications (a.k.a. AI agents or co-pilots), a new software paradigm that combines the strength of LLM and conventional software. Diverse LLM applications from different tenants could design complex workflows using multiple LLM requests to accomplish one task. However, they have to use the over-simplified request-level API provided by today’s public LLM services, losing essential application-level information. Public LLM services have to blindly optimize individual LLM requests, leading to sub-optimal end-to-end performance of LLM applications. This paper introduces Parrot, an LLM service system that focuses on the end-to-end experience of LLM-based applications. Parrot proposes Semantic Variable, a unified abstraction to expose application-level knowledge to public LLM services. A Semantic Variable annotates an input/output variable in the prompt of a request, and creates the data pipeline when connecting multiple LLM requests, providing a natural way to program LLM applications. Exposing Semantic Variables to the public LLM service allows it to perform conventional data flow analysis to uncover the correlation across multiple LLM requests. This correlation opens a brand-new optimization space for the end-to-end performance of LLMbased applications. Extensive evaluations demonstrate that Parrot can achieve up to an order-of-magnitude improvement for popular and practical use cases of LLM applications | LLM Agents |
- Perplexed by Perplexity: Perplexity-Based Data Pruning With Small Reference Models - scale text datasets that improve the performance of larger language models. While existing work has shown that pruning based on the perplexity of a larger model can yield high-quality data, we investigate whether smaller models can be used for perplexity-based pruning and how pruning is affected by the domain composition of the data being pruned. We demonstrate that for multiple dataset compositions, perplexity-based pruning of pretraining data can significantly improve downstream task performance: pruning based on perplexities computed with a 125 million parameter model improves the average performance on downstream tasks of a 3 billion parameter model by up to 2.04 and achieves up to a 1.45× reduction in pretraining steps to reach commensurate baseline performance. Furthermore, we demonstrate that such perplexity-based data pruning also yields downstream performance gains in the over-trained and data-constrained regimes. | Small Language Models |
- GNN-RAG: Graph Neural Retrieval for Large Language Model Reasoning - crafted factual knowledge in the form of triplets (head, relation, tail), which collectively form a graph. Question Answering over KGs (KGQA) is the task of answering natural questions grounding the reasoning to the information provided by the KG. Large Language Models (LLMs) are the state-of-the-art models for QA tasks due to their remarkable ability to understand natural language. On the other hand, Graph Neural Networks (GNNs) have been widely used for KGQA as they can handle the complex graph information stored in the KG. In this work, we introduce GNN-RAG, a novel method for combining language understanding abilities of LLMs with the reasoning abilities of GNNs in a retrieval-augmented generation (RAG) style. First, a GNN reasons over a dense KG subgraph to retrieve answer candidates for a given question. Second, the shortest paths in the KG that connect question entities and answer candidates are extracted to represent KG reasoning paths. The extracted paths are verbalized and given as input for LLM reasoning with RAG. In our GNN-RAG framework, the GNN acts as a dense subgraph reasoner to extract useful graph information, while the LLM leverages its natural language processing ability for ultimate KGQA. Furthermore, we develop a retrieval augmentation (RA) technique to further boost KGQA performance with GNN-RAG. Experimental results show that GNN-RAG achieves state-of-the-art performance in two widely used KGQA benchmarks (WebQSP and CWQ), outperforming or matching GPT-4 performance with a 7B tuned LLM. In addition, GNN-RAG excels on multi-hop and multi-entity questions outperforming competing approaches by 8.9–15.5% points at answer F1. We provide the code and KGQA results at https://github.com/cmavro/GNN-RAG. | RAG on Knowledge Graphs |
- Self-Exploring Language Models: Active Preference Elicitation for Online Alignment - aligned LLMs through an iterative process. However, achieving a globally accurate reward model requires systematic exploration to generate diverse responses that span the vast space of natural language. Random sampling from standard reward-maximizing LLMs alone is insufficient to fulfill this requirement. To address this issue, we propose a bilevel objective optimistically biased towards potentially high-reward responses to actively explore out-of-distribution regions. By solving the inner-level problem with the reparameterized reward function, the resulting algorithm, named Self-Exploring Language Models (SELM), eliminates the need for a separate RM and iteratively updates the LLM with a straightforward objective. Compared to Direct Preference Optimization (DPO), the SELM objective reduces indiscriminate favor of unseen extrapolations and enhances exploration efficiency. Our experimental results demonstrate that when finetuned on Zephyr-7B-SFT and Llama-3- 8B-Instruct models, SELM significantly boosts the performance on instructionfollowing benchmarks such as MT-Bench and AlpacaEval 2.0, as well as various standard academic benchmarks in different settings. Our code and models are available at https://github.com/shenao-zhang/SELM. | Alignment, Preference Optimization |
- OpenRLHF: An Easy-to-use, Scalable and High-performance RLHF Framework - tuning a single model, scaling reinforcement learning from human feedback (RLHF) for training large language models poses coordination challenges across four models. We present OpenRLHF, an open-source framework enabling efficient RLHF scaling. Unlike existing RLHF frameworks that co-locate four models on the same GPUs, OpenRLHF re-designs scheduling for the models beyond 70B parameters using Ray, vLLM, and DeepSpeed, leveraging improved resource utilization and diverse training approaches. Integrating seamlessly with Hugging Face, OpenRLHF provides an out-of-the-box solution with optimized algorithms and launch scripts, which ensures user-friendliness. OpenRLHF implements RLHF, DPO, rejection sampling, and other alignment techniques. Empowering state-of-the-art LLM development, OpenRLHF’s code is available at https://github.com/OpenLLMAI/OpenRLHF. | RLHF, Toolkit |
- LLAMA-NAS: EFFICIENT NEURAL ARCHITECTURE SEARCH FOR LARGE LANGUAGE MODELS - optimal network architectures based on LLaMA2-7B using one-shot NAS. In particular, we fine-tune LLaMA2-7B only once and then apply genetic algorithmbased search to find smaller, less computationally complex network architectures. We show that, for certain standard benchmark tasks, the pre-trained LLaMA2-7B network is unnecessarily large and complex. More specifically, we demonstrate a 1.5x reduction in model size and 1.3x speedup in throughput for certain tasks with negligible drop in accuracy. In addition to finding smaller, higherperforming network architectures, our method does so more effectively and efficiently than certain pruning or sparsification techniques. Finally, we demonstrate how quantization is complementary to our method and that the size and complexity of the networks we find can be further decreased using quantization. We believe that our work provides a way to automatically create LLMs which can be used on less expensive and more readily available hardware platforms. | Neural Architecture Search, Model Size Reduction |
- Don’t Forget to Connect! Improving RAG with Graph-based Reranking - RAG, a reranker based on graph neural networks (GNNs) between the retriever and reader in RAG. Our method combines both connections between documents and semantic information (via Abstract Meaning Representation graphs) to provide a context-informed ranker for RAG. G-RAG outperforms state-of-the-art approaches while having smaller computational footprint. Additionally, we assess the performance of PaLM 2 as a reranker and find it to significantly underperform G-RAG. This result emphasizes the importance of reranking for RAG even when using Large Language Models. | RAG for Reasoning |
- Meteor: Mamba-based Traversal of Rationale for Large Language and Vision Models - source LLVMs have curated high-quality visual instruction tuning datasets and utilized additional vision encoders or multiple computer vision models in order to narrow the performance gap with powerful closed-source LLVMs. These advancements are attributed to multifaceted information required for diverse capabilities, including fundamental image understanding, real-world knowledge about common-sense and non-object concepts (e.g., charts, diagrams, symbols, signs, and math problems), and step-by-step procedures for solving complex questions. Drawing from the multifaceted information, we present a new efficient LLVM, Mamba-based traversal of rationales ( Meteor), which leverages multifaceted rationale to enhance understanding and answering capabilities. To embed lengthy rationales containing abundant information, we employ the Mamba architecture, capable of processing sequential data with linear time complexity. We introduce a new concept of traversal of rationale that facilitates efficient embedding of rationale. Subsequently, the backbone multimodal language model (MLM) is trained to generate answers with the aid of rationale. Through these steps, Meteor achieves significant improvements in vision language performances across multiple evaluation benchmarks requiring diverse capabilities, without scaling up the model size or employing additional vision encoders and computer vision models. Code is available in https://github.com/ByungKwanLee/Meteor. | State Space Models, Multimodal Models |
- An Introduction to Vision-Language Modeling - level text description, the vision-language model (VLM) applications will significantly impact our relationship with technology. However, there are many challenges that need to be addressed to improve the reliability of those models. While language is discrete, vision evolves in a much higher dimensional space in which concepts cannot always be easily discretized. To better understand the mechanics behind mapping vision to language, we present this introduction to VLMs which we hope will help anyone who would like to enter the field. First, we introduce what VLMs are, how they work, and how to train them. Then, we present and discuss approaches to evaluate VLMs. Although this work primarily focuses on mapping images to language, we also discuss extending VLMs to videos. | Multimodal Models, Survey |
- Matryoshka Multimodal Models - linguistic reasoning. These models first embed images into a fixed large number of visual tokens and then feed them into a Large Language Model (LLM). However, this design causes an excessive number of tokens for dense visual scenarios such as high-resolution images and videos, leading to great inefficiency. While token pruning and merging methods exist, they produce a single-length output for each image and cannot afford flexibility in trading off information density v.s. efficiency. Inspired by the concept of Matryoshka Dolls, we propose M3 : Matryoshka Multimodal Models, which learns to represent visual content as nested sets of visual tokens that capture information across multiple coarse-to-fine granularities. Our approach offers several unique benefits for LMMs: (1) One can explicitly control the visual granularity per test instance during inference, e.g., adjusting the number of tokens used to represent an image based on the anticipated complexity or simplicity of the content; (2) M3 provides a framework for analyzing the granularity needed for existing datasets, where we find that COCO-style benchmarks only need around 9 visual tokens to obtain an accuracy similar to that of using all 576 tokens; (3) Our approach provides a foundation to explore the best trade-off between performance and visual token length at the sample level, where our investigation reveals that a large gap exists between the oracle upper bound and current fixed-scale representations. | Multimodal Models |
- Trans-LoRA: towards data-free Transferable Parameter Efficient Finetuning - rank adapters (LoRA) and their variants are popular parameter-efficient finetuning (PEFT) techniques that closely match full model fine-tune performance while requiring only a small number of additional parameters. These additional LoRA parameters are specific to the base model being adapted. When the base model needs to be deprecated and replaced with a new one, all the associated LoRA modules need to be re-trained. Such re-training requires access to the data used to train the LoRA for the original base model. This is especially problematic for commercial cloud applications where the LoRA modules and the base models are hosted by service providers who may not be allowed to host proprietary client task data. To address this challenge, we propose Trans-LoRA— a novel method for lossless, nearly data-free transfer of LoRAs across base models. Our approach relies on synthetic data to transfer LoRA modules. Using large language models, we design a synthetic data generator to approximate the data-generating process of the observed task data subset. Training on the resulting synthetic dataset transfers LoRA modules to new models. We show the effectiveness of our approach using both LLama and Gemma model families. Our approach achieves lossless (mostly improved) LoRA transfer between models within and across different base model families, and even between different PEFT methods, on a wide variety of tasks. | PEFT Methods, Fine-Tuning |
- Self-Play Preference Optimization for Language Model Alignment - Terry model fall short in capturing the intransitivity and irrationality in human preferences. Recent advancements suggest that directly working with preference probabilities can yield a more accurate reflection of human preferences, enabling more flexible and accurate language model alignment. In this paper, we propose a self-playbased method for language model alignment, which treats the problem as a constant-sum two-player game aimed at identifying the Nash equilibrium policy. Our approach, dubbed Self-Play Preference Optimization (SPPO), approximates the Nash equilibrium through iterative policy updates and enjoys a theoretical convergence guarantee. Our method can effectively increase the log-likelihood of the chosen response and decrease that of the rejected response, which cannot be trivially achieved by symmetric pairwise loss such as Direct Preference Optimization (DPO) and Identity Preference Optimization (IPO). In our experiments, using only 60k prompts (without responses) from the UltraFeedback dataset and without any prompt augmentation, by leveraging a pre-trained preference model PairRM with only 0.4B parameters, SPPO can obtain a model from fine-tuning Mistral-7B-Instruct-v0.2 that achieves the state-of-the-art lengthcontrolled win-rate of 28.53% against GPT-4-Turbo on AlpacaEval 2.0. It also outperforms the (iterative) DPO and IPO on MT-Bench and the Open LLM Leaderboard. Notably, the strong performance of SPPO is achieved without additional external supervision (e.g., responses, preferences, etc.) from GPT-4 or other stronger language models. | Alignment, Optimization |
- Not All Language Model Features Are Linear - dimensional representations of concepts (“features”) in activation space. In contrast, we explore whether some language model representations may be inherently multi-dimensional. We begin by developing a rigorous definition of irreducible multi-dimensional features based on whether they can be decomposed into either independent or non-co-occurring lower-dimensional features. Motivated by these definitions, we design a scalable method that uses sparse autoencoders to automatically find multi-dimensional features in GPT-2 and Mistral 7B. These auto-discovered features include strikingly interpretable examples, e.g. circular features representing days of the week and months of the year. We identify tasks where these exact circles are used to solve computational problems involving modular arithmetic in days of the week and months of the year. Finally, we provide evidence that these circular features are indeed the fundamental unit of computation in these tasks with intervention experiments on Mistral 7B and Llama 3 8B, and we find further circular representations by breaking down the hidden states for these tasks into interpretable components. | Linear Representation Analysis |
- AlignGPT: Multi-modal Large Language Models with Adaptive Alignment Capability - modal alignment. To attain this goal, current MLLMs typically follow a two-phase training paradigm: the pre-training phase and the instruction-tuning phase. Despite their success, there are shortcomings in the modeling of alignment capabilities within these models. Firstly, during the pre-training phase, the model usually assumes that all image-text pairs are uniformly aligned, but in fact the degree of alignment between different imagetext pairs is inconsistent. Secondly, the instructions currently used for finetuning incorporate a variety of tasks, different tasks’s instructions usually require different levels of alignment capabilities, but previous MLLMs overlook these differentiated alignment needs. To tackle these issues, we propose a new multimodal large language model AlignGPT. In the pre-training stage, instead of treating all imagetext pairs equally, we assign different levels of alignment capabilities to different image-text pairs. Then, in the instruction-tuning phase, we adaptively combine these different levels of alignment capabilities to meet the dynamic alignment needs of different instructions. Extensive experimental results show that our model achieves competitive performance on 12 benchmarks. | Alignment, Multimodal Model |
- HippoRAG: Neurobiologically Inspired Long-Term Memory for Large Language Models - changing natural environments, mammalian brains evolved to store large amounts of knowledge about the world and continually integrate new information while avoiding catastrophic forgetting. Despite the impressive accomplishments, large language models (LLMs), even with retrievalaugmented generation (RAG), still struggle to efficiently and effectively integrate a large amount of new experiences after pre-training. In this work, we introduce HippoRAG, a novel retrieval framework inspired by the hippocampal indexing theory of human long-term memory to enable deeper and more efficient knowledge integration over new experiences. HippoRAG synergistically orchestrates LLMs, knowledge graphs, and the Personalized PageRank algorithm to mimic the different roles of neocortex and hippocampus in human memory. We compare HippoRAG with existing RAG methods on multi-hop question answering and show that our method outperforms the state-of-the-art methods remarkably, by up to 20%. Singlestep retrieval with HippoRAG achieves comparable or better performance than iterative retrieval like IRCoT while being 10-30 times cheaper and 6-13 times faster, and integrating HippoRAG into IRCoT brings further substantial gains. Finally, we show that our method can tackle new types of scenarios that are out of reach of existing methods. | RAG Optimization |
- OmniGlue: Generalizable Feature Matching with Foundation Model Guidance - improving performance on conventional benchmarks. However, our investigation shows that despite these gains, their potential for real-world applications is restricted by their limited generalization capabilities to novel image domains. In this paper, we introduce OmniGlue, the first learnable image matcher that is designed with generalization as a core principle. OmniGlue leverages broad knowledge from a vision foundation model to guide the feature matching process, boosting generalization to domains not seen at training time. Additionally, we propose a novel keypoint position-guided attention mechanism which disentangles spatial and appearance information, leading to enhanced matching descriptors. We perform comprehensive experiments on a suite of 7 datasets with varied image domains, including scenelevel, object-centric and aerial images. OmniGlue’s novel components lead to relative gains on unseen domains of 20.9% with respect to a directly comparable reference model, while also outperforming the recent LightGlue method by 9.5% relatively. Code and model can be found at https: //hwjiang1510.github.io/OmniGlue. | Multimodal Models |
- MoRA: High-Rank Updating for Parameter-Efficient Fine-Tuning - rank adaptation (LoRA) is a popular parameter-efficient fine-tuning (PEFT) method for large language models (LLMs). In this paper, we analyze the impact of low-rank updating, as implemented in LoRA. Our findings suggest that the low-rank updating mechanism may limit the ability of LLMs to effectively learn and memorize new knowledge. Inspired by this observation, we propose a new method called MoRA, which employs a square matrix to achieve high-rank updating while maintaining the same number of trainable parameters. To achieve it, we introduce the corresponding non-parameter operators to reduce the input dimension and increase the output dimension for the square matrix. Furthermore, these operators ensure that the weight can be merged back into LLMs, which makes our method can be deployed like LoRA. We perform a comprehensive evaluation of our method across five tasks: instruction tuning, mathematical reasoning, continual pretraining, memory and pretraining. Our method outperforms LoRA on memoryintensive tasks and achieves comparable performance on other tasks. Our code will be available at https://github.com/kongds/MoRA. | PEFT Approaches, Fine-Tuning |
- Your Transformer is Secretly Linear - perfect linear relationship (Procrustes similarity score of 0.99). However, linearity decreases when the residual component is removed due to a consistently low output norm of the transformer layer. Our experiments show that removing or linearly approximating some of the most linear blocks of transformers does not affect significantly the loss or model performance. Moreover, in our pretraining experiments on smaller models we introduce a cosine-similarity-based regularization, aimed at reducing layer linearity. This regularization improves performance metrics on benchmarks like Tiny Stories and SuperGLUE and as well successfully decreases the linearity of the models. This study challenges the existing understanding of transformer architectures, suggesting that their operation may be more linear than previously assumed.1 | Transformer Analysis |
- Towards Modular LLMs by Building and Reusing a Library of LoRAs - efficient adaptations of a base large language model (LLM) calls for studying whether we can reuse such trained adapters to improve performance for new tasks. We study how to best build a library of adapters given multi-task data and devise techniques for both zero-shot and supervised task generalization through routing in such library. We benchmark existing approaches to build this library and introduce model-based clustering, MBC, a method that groups tasks based on the similarity of their adapter parameters, indirectly optimizing for transfer across the multi-task dataset. To re-use the library, we present a novel zero-shot routing mechanism, Arrow, which enables dynamic selection of the most relevant adapters for new inputs without the need for retraining. We experiment with several LLMs, such as Phi-2 and Mistral, on a wide array of held-out tasks, verifying that MBC-based adapters and Arrow routing lead to superior generalization to new tasks. We make steps towards creating modular, adaptable LLMs that can match or outperform traditional joint training. | PEFT Approaches, Fine-Tuning, Toolkit |
- Chameleon: Mixed-Modal Early-Fusion Foundation Models - fusion token-based mixed-modal models capable of understanding and generating images and text in any arbitrary sequence. We outline a stable training approach from inception, an alignment recipe, and an architectural parameterization tailored for the early-fusion, token-based, mixed-modal setting. The models are evaluated on a comprehensive range of tasks, including visual question answering, image captioning, text generation, image generation, and long-form mixed modal generation. Chameleon demonstrates broad and general capabilities, including state-of-the-art performance in image captioning tasks, outperforms Llama-2 in text-only tasks while being competitive with models such as Mixtral 8x7B and Gemini-Pro, and performs non-trivial image generation, all in a single model. It also matches or exceeds the performance of much larger models, including Gemini Pro and GPT-4V, according to human judgments on a new long-form mixed-modal generation evaluation, where either the prompt or outputs contain mixed sequences of both images and text. Chameleon marks a significant step forward in a unified modeling of full multimodal documents. | Multimodal Models, Foundation Model |
- Many-Shot In-Context Learning in Multimodal Foundation Models - known to be effective at few-shot in-context learning (ICL). Recent advancements in multimodal foundation models have enabled unprecedentedly long context windows, presenting an opportunity to explore their capability to perform ICL with many more demonstrating examples. In this work, we evaluate the performance of multimodal foundation models scaling from few-shot to many-shot ICL. We benchmark GPT-4o and Gemini 1.5 Pro across 10 datasets spanning multiple domains (natural imagery, medical imagery, remote sensing, and molecular imagery) and tasks (multi-class, multi-label, and fine-grained classification). We observe that many-shot ICL, including up to almost 2,000 multimodal demonstrating examples, leads to substantial improvements compared to few-shot (<100 examples) ICL across all of the datasets. Further, Gemini 1.5 Pro performance continues to improve log-linearly up to the maximum number of tested examples on many datasets. Given the high inference costs associated with the long prompts required for many-shot ICL, we also explore the impact of batching multiple queries in a single API call. We show that batching up to 50 queries can lead to performance improvements under zero-shot and many–shot ICL, with substantial gains in the zero-shot setting on multiple datasets, while drastically reducing per-query cost and latency. Finally, we measure ICL data efficiency of the models, or the rate at which the models learn from more demonstrating examples. We find that while GPT-4o and Gemini 1.5 Pro achieve similar zero-shot performance across the datasets, Gemini 1.5 Pro exhibits higher ICL data efficiency than GPT-4o on most datasets. Our results suggest that many-shot ICL could enable users to efficiently adapt multimodal foundation models to new applications and domains. Our codebase is publicly available at https://github.com/stanfordmlgroup/ManyICL. | ICL, Multimodal Models |
- LoRA Learns Less and Forgets Less - Rank Adaptation (LoRA) is a widely-used parameter-efficient finetuning method for large language models. LoRA saves memory by training only low rank perturbations to selected weight matrices. In this work, we compare the performance of LoRA and full finetuning on two target domains, programming and mathematics. We consider both the instruction finetuning (≈100K prompt-response pairs) and continued pretraining (≈10B unstructured tokens) data regimes. Our results show that, in most settings, LoRA substantially underperforms full finetuning. Nevertheless, LoRA exhibits a desirable form of regularization: it better maintains the base model’s performance on tasks outside the target domain. We show that LoRA provides stronger regularization compared to common techniques such as weight decay and dropout; it also helps maintain more diverse generations. We show that full finetuning learns perturbations with a rank that is 10-100X greater than typical LoRA configurations, possibly explaining some of the reported gaps. We conclude by proposing best practices for finetuning with LoRA. | PEFT Approaches, Fine-Tuning |
- Understanding the performance gap between online and offline alignment algorithms - policy sampling in RLHF. Within the context of reward over-optimization, we start with an opening set of experiments that demonstrate the clear advantage of online methods over offline methods. This prompts us to investigate the causes to the performance discrepancy through a series of carefully designed experimental ablations. We show empirically that hypotheses such as offline data coverage and data quality by itself cannot convincingly explain the performance difference. We also find that while offline algorithms train policy to become good at pairwise classification, it is worse at generations; in the meantime the policies trained by online algorithms are good at generations while worse at pairwise classification. This hints at a unique interplay between discriminative and generative capabilities, which is greatly impacted by the sampling process. Lastly, we observe that the performance discrepancy persists for both contrastive and non-contrastive loss functions, and appears not to be addressed by simply scaling up policy networks. Taken together, our study sheds light on the pivotal role of on-policy sampling in AI alignment, and hints at certain fundamental challenges of offline alignment algorithms. | Alignment |
- RLHF Workflow: From Reward Modeling to Online RLHF - source RLHF projects are still largely confined to the offline learning setting. In this technical report, we aim to fill in this gap and provide a detailed recipe that is easy to reproduce for online iterative RLHF. In particular, since online human feedback is usually infeasible for open-source communities with limited resources, we start by constructing preference models using a diverse set of open-source datasets and use the constructed proxy preference model to approximate human feedback. Then, we discuss the theoretical insights and algorithmic principles behind online iterative RLHF, followed by a detailed practical implementation. Our trained LLM, SFR-Iterative-DPO-LLaMA-3-8B-R, achieves impressive performance on LLM chatbot benchmarks, including AlpacaEval-2, Arena-Hard, and MT-Bench, as well as other academic benchmarks such as HumanEval and TruthfulQA. We have shown that supervised fine-tuning (SFT) and iterative RLHF can obtain state-of-the-art performance with fully open-source datasets. Further, we have made our models, curated datasets, and comprehensive step-by-step code guidebooks publicly available. Please refer to https://github.com/RLHFlow/RLHF-Reward-Modeling and https://github.com/RLHFlow/Online-RLHF for more detailed information. | Preference Optimization, RLHF |
- PROMETHEUS 2: An Open Source Language Model Specialized in Evaluating Other Language Models - 4 are often employed to assess the quality of responses from various LMs. However, concerns including transparency, controllability, and affordability strongly motivate the development of opensource LMs specialized in evaluations. On the other hand, existing open evaluator LMs exhibit critical shortcomings: 1) they issue scores that significantly diverge from those assigned by humans, and 2) they lack the flexibility to perform both direct assessment and pairwise ranking, the two most prevalent forms of assessment. Additionally, they do not possess the ability to evaluate based on custom evaluation criteria, focusing instead on general attributes like helpfulness and harmlessness. To address these issues, we introduce Prometheus 2, a more powerful evaluator LM than it’s predecessor that closely mirrors human and GPT-4 judgements. Moreover, it is capable of processing both direct assessment and pair-wise ranking formats grouped with a user-defined evaluation criteria. On four direct assessment benchmarks and four pairwise ranking benchmarks, PROMETHEUS 2 scores the highest correlation and agreement with humans and proprietary LM judges among all tested open evaluator LMs. Our models, code, and data are all publicly available 1 . | Evaluation, Agents |
- WILDCHAT: 1M CHATGPT INTERACTION LOGS IN THE WILD - 4 and ChatGPT are now serving millions of users. Despite their widespread use, there remains a lack of public datasets showcasing how these tools are used by a population of users in practice. To bridge this gap, we offered free access to ChatGPT for online users in exchange for their affirmative, consensual opt-in to anonymously collect their chat transcripts and request headers. From this, we compiled WILDCHAT, a corpus of 1 million user-ChatGPT conversations, which consists of over 2.5 million interaction turns. We compare WILDCHAT with other popular user-chatbot interaction datasets, and find that our dataset offers the most diverse user prompts, contains the largest number of languages, and presents the richest variety of potentially toxic use-cases for researchers to study. In addition to timestamped chat transcripts, we enrich the dataset with demographic data, including state, country, and hashed IP addresses, alongside request headers. This augmentation allows for more detailed analysis of user behaviors across different geographical regions and temporal dimensions. Finally, because it captures a broad range of use cases, we demonstrate the dataset’s potential utility in fine-tuning instruction-following models. WILDCHAT is released at https://wildchat.allen.ai under AI2 ImpACT Licenses1 . | Benchmark, Evaluation |
- STORYDIFFUSION: CONSISTENT SELF-ATTENTION FOR LONG-RANGE IMAGE AND VIDEO GENERATION - based generative models, maintaining consistent content across a series of generated images, especially those containing subjects and complex details, presents a significant challenge. In this paper, we propose a new way of self-attention calculation, termed Consistent Self-Attention, that significantly boosts the consistency between the generated images and augments prevalent pretrained diffusion-based text-to-image models in a zero-shot manner. To extend our method to long-range video generation, we further introduce a novel semantic space temporal motion prediction module, named Semantic Motion Predictor. It is trained to estimate the motion conditions between two provided images in the semantic spaces. This module converts the generated sequence of images into videos with smooth transitions and consistent subjects that are significantly more stable than the modules based on latent spaces only, especially in the context of long video generation. By merging these two novel components, our framework, referred to as StoryDiffusion, can describe a text-based story with consistent images or videos encompassing a rich variety of contents. The proposed StoryDiffusion encompasses pioneering explorations in visual story generation with the presentation of images and videos, which we hope could inspire more research from the aspect of architectural modifications. | Multimodal Models, Diffusion |
- FLAME : Factuality-Aware Alignment for Large Language Models - tune pre-trained large language models (LLMs) to follow natural language instructions and serve as helpful AI assistants. We have observed, however, that the conventional alignment process fails to enhance the factual accuracy of LLMs, and often leads to the generation of more false facts (i.e. hallucination). In this paper, we study how to make the LLM alignment process more factual, by first identifying factors that lead to hallucination in both alignment steps: supervised fine-tuning (SFT) and reinforcement learning (RL). In particular, we find that training the LLM on new knowledge or unfamiliar texts can encourage hallucination. This makes SFT less factual as it trains on human labeled data that may be novel to the LLM. Furthermore, reward functions used in standard RL can also encourage hallucination, because it guides the LLM to provide more helpful responses on a diverse set of instructions, often preferring longer and more detailed responses. Based on these observations, we propose factuality-aware alignment (FLAME ), comprised of factuality-aware SFT and factuality-aware RL through direct preference optimization. Experiments show that our proposed factuality-aware alignment guides LLMs to output more factual responses while maintaining instruction-following capability | Alignment, Factuality |
- NeMo-Aligner: Scalable Toolkit for Efficient Model Alignment - Aligner, a toolkit for model alignment that can efficiently scale to using hundreds of GPUs for training. NeMo-Aligner comes with highly optimized and scalable implementations for major paradigms of model alignment such as: Reinforcement Learning from Human Feedback (RLHF), Direct Preference Optimization (DPO), SteerLM, and Self-Play Fine-Tuning (SPIN). Additionally, our toolkit supports running most of the alignment techniques in a Parameter Efficient Fine-Tuning (PEFT) setting. NeMo-Aligner is designed for extensibility, allowing support for other alignment techniques with minimal effort. It is open-sourced with Apache 2.0 License and we invite community contributions at https://github.com/NVIDIA/NeMo-Aligner. | Alignment, Toolkit |
- Is Bigger Edit Batch Size Always Better? - An Empirical Study on Model Editing with Llama-3 - 3. We explore the efficacy of popular model editing techniques - ROME, MEMIT, and EMMET, which are designed for precise layer interventions. We identify the most effective layers for targeted edits through an evaluation that encompasses up to 4096 edits across three distinct strategies: sequential editing, batch editing, and a hybrid approach we call as sequential-batch editing. Our findings indicate that increasing edit batch-sizes may degrade model performance more significantly than using smaller edit batches sequentially for equal number of edits. With this, we argue that sequential model editing is an important component for scaling model editing methods and future research should focus on methods that combine both batched and sequential editing. This observation suggests a potential limitation in current model editing methods which push towards bigger edit batch sizes, and we hope it paves way for future investigations into optimizing batch sizes and model editing performance. | Model Editing |
- LoRA Land: 310 Fine-tuned LLMs that Rival GPT-4, A Technical Report - Tuning (PEFT) of Large Language Models (LLMs). LoRA reduces the number of trainable parameters and memory usage while achieving comparable performance to full fine-tuning. We aim to assess the viability of training and serving LLMs fine-tuned with LoRA in real-world applications. First, we measure the quality of LLMs fine-tuned with quantized low rank adapters across 10 base models and 31 tasks for a total of 310 models. We find that 4-bit LoRA fine-tuned models outperform base models by 34 points and GPT-4 by 10 points on average. Second, we investigate the most effective base models for fine-tuning and assess the correlative and predictive capacities of task complexity heuristics in forecasting the outcomes of fine-tuning. Finally, we evaluate the latency and concurrency capabilities of LoRAX, an open-source Multi-LoRA inference server that facilitates the deployment of multiple LoRA fine-tuned models on a single GPU using shared base model weights and dynamic adapter loading. LoRAX powers LoRA Land, a web application that hosts 25 LoRA fine-tuned Mistral-7B LLMs on a single NVIDIA A100 GPU with 80GB memory. LoRA Land highlights the quality and cost-effectiveness of employing multiple specialized LLMs over a single, general-purpose LLM. | PEFT Approaches, Fine-Tuning |
-
:star: Best GenAI Papers List (February 2024)
- Revisiting Feature Prediction for Learning Visual Representations from Video - JEPA, a collection of vision models trained solely on video data using a feature prediction objective, without relying on pretrained image encoders, text, negative examples, or reconstruction. Trained on 2 million videos, these models are evaluated on downstream image and video tasks, demonstrating versatile visual representations that excel in both motion and appearance-based tasks without requiring adaptation of model parameters. The largest model, a ViT-H/16 trained only on videos, achieves impressive performance on Kinetics-400, Something-Something-v2, and ImageNet1K datasets. | Multimodal LLMs |
- Sora: A Review on Background, Technology, Limitations, and Opportunities of Large Vision Models - to-video generative AI model launched by OpenAI. It examines Sora's evolution, underlying technologies, diverse applications across industries, and potential impact on creativity and productivity. Challenges like safety and bias in video generation are discussed, along with future directions for Sora and similar models, envisioning enhanced human-AI collaboration and innovation in video production. Note that this paper is not written by the creators of Sora, it is reverse engineered by a group of researchers. | Multimodal Models |
- OpenCodeInterpreter: Integrating Code Generation with Execution and Refinement - source code systems aimed at addressing the limitations of existing open-source models in code generation by incorporating execution capabilities and iterative refinement similar to advanced systems like the GPT-4 Code Interpreter. Leveraging the CodeFeedback dataset, which includes 68K multi-turn interactions, OpenCodeInterpreter integrates execution and human feedback for dynamic code refinement. Evaluation across key benchmarks demonstrates exceptional performance, with OpenCodeInterpreter33B achieving close accuracy to GPT-4 on HumanEval and MBPP benchmarks, effectively bridging the gap between open-source and proprietary code generation systems. | Task Specific LLMs, Evaluation |
- Evaluating Very Long-Term Conversational Memory of LLM Agents - human pipeline to generate high-quality, very long-term dialogues, spanning up to 35 sessions, using large language models and retrieval augmented generation techniques. The conversations are grounded on personas and temporal event graphs, with each agent capable of sharing and reacting to images. The resulting dataset, LOCOMO, comprises conversations with 300 turns on average. Evaluation benchmarks measure long-term memory in models, revealing challenges for LLMs in understanding lengthy conversations and comprehending long-range temporal dynamics. While strategies like long-context LLMs or RAG show improvements, models still lag behind human performance. | RAG, Benchmark, Long Context |
- When Scaling Meets LLM Finetuning: The Effect of Data, Model, and Finetuning Method - based multiplicative joint scaling law between finetuning data size and other factors, with LLM model scaling offering more benefits than pretraining data scaling. Additionally, the optimal finetuning method varies depending on the task and finetuning data. These findings aim to enhance understanding and development of LLM finetuning methods. | Fine-Tuning |
- The Era of 1-bit LLMs: - 1, 0, 1}, matching full-precision Transformer LLMs in both perplexity and end-task performance while offering significant cost-effectiveness. This 1.58-bit LLM sets a new standard for high-performance, cost-effective models and opens opportunities for new computation paradigms and hardware designs optimized for 1-bit LLMs. | Cost Effective LLMs |
- Rainbow Teaming: Open-Ended Generation of Diverse Adversarial Prompts - diversity problem, it uncovers vulnerabilities across various domains, including safety, question answering, and cybersecurity. Additionally, fine-tuning LLMs on synthetic data produced by Rainbow Teaming improves safety without compromising general capabilities, offering a path to open-ended self-improvement. | Red-Teaming |
- Do Large Language Models Latently Perform Multi-Hop Reasoning? - hop reasoning when processing complex prompts. By analyzing individual hops and their co-occurrence, the research examines how LLMs identify and utilize bridge entities to complete prompts. Results show strong evidence of latent multi-hop reasoning in certain relation types, with the reasoning pathway used in over 80% of prompts. However, the utilization varies contextually, and while evidence for the first hop is substantial, it's more moderate for the second hop. Additionally, there's a scaling trend with increasing model size for the first hop but not the second, indicating potential challenges and opportunities for future LLM development. | Evaluation |
- Revisiting REINFORCE Style Optimization for Learning from Human Feedback in LLMs - style optimization variants for RLHF, showing superior performance compared to PPO and other methods like DPO and RAFT. It suggests that adapting to LLM alignment characteristics allows for efficient online RL optimization. | Instruction Tuning |
- A Human-Inspired Reading Agent with Gist Memory of Very Long Contexts - document reading tasks demonstrates ReadAgent's superiority over baselines, enhancing performance while expanding the effective context window by 3 to 20 times. | LLM Agents |
- MobileLLM: Optimizing Sub-billion Parameter Language Models for On-Device Use Cases - of-the-art models. Additionally, MobileLLM-LS, incorporating block-wise weight sharing, further enhances accuracy with marginal latency overhead, highlighting the potential of small models for on-device use cases. | Smaller Models |
- Stable Diffusion 3 - to-image model, boasting significant improvements in multi-subject prompts, image quality, and spelling abilities. The waitlist for early access is now open, allowing users to contribute insights for enhancing performance and safety prior to its public release. Ranging from 800M to 8B parameters, the suite offers scalability options to cater to various creative needs.Emphasizing safe and responsible AI practices, [Stability.ai](http://Stability.ai) has implemented numerous safeguards and continues to collaborate with researchers and experts to ensure integrity throughout development and deployment. | Multimodal Models |
- Coercing Large Language Models (LLMs) to Do and Reveal (Almost) Anything - of-service, and data extraction. Controlled experiments reveal many attacks originate from pre-training with coding capabilities and the presence of "glitch" tokens in LLM vocabularies, emphasizing the need for security measures. | Red-Teaming |
- LongRoPE: Extending LLM Context Window Beyond 2 Million Tokens - trained LLMs to an impressive 2048k tokens, overcoming limitations of current extended context windows. Key innovations include exploiting non-uniformities in positional interpolation, a progressive extension strategy, and readjusting to recover short context window performance. Extensive experiments demonstrate the effectiveness of LongRoPE across various tasks, with models retaining the original architecture and minor modifications to positional embedding. | Long Context, Embedding |
- Gemma: Open Models Based on Gemini Research and Technology - of-the art open models built from the research and technology used to create Gemini models. Gemma models demonstrate strong performance across academic benchmarks for language understanding, reasoning, and safety. We release two sizes of models (2 billion and 7 billion parameters), and provide both pretrained and fine-tuned checkpoints. Gemma outperforms similarly sized open models on 11 out of 18 text-based tasks, and we present comprehensive evaluations of safety and responsibility aspects of the models, alongside a detailed description of model development. We believe the responsible release of LLMs is critical for improving the safety of frontier models, and for enabling the next wave of LLM innovations | Foundation LLMs |
- Large Language Models for Data Annotation: A Survey - 4, for automating data annotation, a labor-intensive process in machine learning. It offers insights into LLM-Based Data Annotation, Assessing LLM-generated Annotations, and Learning with LLM-generated annotations. The survey includes a taxonomy of methodologies, reviews learning strategies, and discusses challenges and limitations. Aimed at guiding researchers and practitioners, it aims to foster advancements in data annotation using the latest LLMs. | Task Specific LLMs |
- In Search of Needles in a 11M Haystack: Recurrent Memory Finds What LLMs Miss - tuning GPT-2 with recurrent memory augmentations enables it to handle tasks involving up to 11 × 10^6 elements. This achievement represents a substantial leap, demonstrating significant improvement in processing capabilities for long sequences and marking the longest input processed by any neural network model to date. | Benchmark, Long Context |
- Large Language Models: A Survey
- LongAgent: Scaling Language Models to 128k Context Through Multi-Agent Collaboration - agent collaboration to scale LLMs like LLaMA to process long texts up to 128K tokens. LongAgent utilizes a leader to interpret user intent and coordinate team members in acquiring information. To address hallucination-induced response inaccuracies, an inter-member communication mechanism resolves conflicts through information sharing. Experimental results demonstrate LongAgent's superiority over GPT-4 in tasks such as 128k-long text retrieval and multi-hop question answering. | LLM Agents |
- LoRA+: Efficient Low Rank Adaptation of Large Models - tuning in models with large width (embedding dimension) using Low Rank Adaptation (LoRA) due to updating adapter matrices A and B with the same learning rate. By setting different learning rates for A and B with a fixed ratio in a proposed algorithm called LoRA+, the suboptimality of LoRA can be corrected. Extensive experiments demonstrate that LoRA+ improves performance (1% − 2% improvements) and fine-tuning speed (up to ∼ 2X SpeedUp) at the same computational cost as LoRA. | PEFT |
- Generative Representational Instruction Tuning - of-the-art on the Massive Text Embedding Benchmark (MTEB) and outperforms all models of its size on generative tasks. Scaling up to GRITLM 8X7B further surpasses all open generative language models while remaining among the best embedding models. GRIT unifies generative and embedding training without performance loss, significantly speeding up RAG by over 60% for long documents. Models and code are available. | RAG, Instruction Tuning |
- Chain-of-Thought Reasoning Without Prompting - of-thought (CoT) reasoning paths. This method bypasses manual prompt engineering, assesses intrinsic reasoning abilities, and correlates CoT presence with higher confidence in decoded answers. Extensive empirical studies across benchmarks demonstrate significant performance improvement over standard greedy decoding. | Prompt Engineering |
- Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context - perfect recall across tasks, surpasses previous state-of-the-art models, and exhibits surprising translation abilities for rare languages like Kalamang. | Foundation LLMs |
- Revisiting Feature Prediction for Learning Visual Representations from Video - JEPA, a collection of vision models trained solely on video data using a feature prediction objective, without relying on pretrained image encoders, text, negative examples, or reconstruction. Trained on 2 million videos, these models are evaluated on downstream image and video tasks, demonstrating versatile visual representations that excel in both motion and appearance-based tasks without requiring adaptation of model parameters. The largest model, a ViT-H/16 trained only on videos, achieves impressive performance on Kinetics-400, Something-Something-v2, and ImageNet1K datasets. | Multimodal LLMs |
- World Model on Million-Length Video and Language with Ring Attention - form tasks. It curates a large dataset of diverse videos and books, trains transformers with RingAttention technique on long sequences, and gradually increases context size. Key contributions include training one of the largest context size transformers, overcoming vision-language training challenges, and open-sourcing optimized models capable of processing multimodal sequences over 1M tokens. This work enables training on massive datasets to develop understanding of both human knowledge and the multimodal world, paving the way for broader AI capabilities. | Multimodal LLMs |
- ChemLLM: A Chemical Large Language Model - based instruction construction method, ChemLLM transforms structured knowledge into plain dialogue for effective language model training. ChemLLM outperforms GPT-3.5 and GPT-4 on key chemistry tasks such as name conversion, molecular captioning, and reaction prediction, demonstrating exceptional adaptability to related mathematical and physical tasks. Moreover, ChemLLM showcases proficiency in specialized NLP tasks within chemistry, opening new avenues for exploration in chemical studies. | Task Specific LLMs |
- LLM Agents can Autonomously Hack Websites - 4, possess the capability to autonomously conduct website hacking tasks such as blind database schema extraction and SQL injections without prior knowledge of vulnerabilities. This ability, enabled by advanced models adept at tool usage and leveraging extended context, raises concerns about the potential offensive capabilities of LLM agents and prompts questions regarding their widespread deployment in cybersecurity contexts. | LLM Agents |
- AnyTool: Self-Reflective, Hierarchical Agents for Large-Scale API Calls - reflection mechanism. Powered by the function calling feature of GPT-4, AnyTool eliminates the need for external module training. Additionally, the paper revises the evaluation protocol to introduce AnyToolBench, demonstrating superior performance over strong baselines such as ToolLLM and a GPT-4 variant tailored for tool utilization across various datasets. The code is available at [https://github.com/dyabel/AnyTool](https://github.com/dyabel/AnyTool). | LLM Agents |
- Large Language Models as an Indirect Reasoner: Contrapositive and Contradiction for Automated Reasoning - of-Thought and Self-Consistency. By leveraging logic of contrapositives and contradictions, the IR method tackles tasks such as factual reasoning and mathematical proof. Experimental results on popular LLMs, including GPT-3.5-turbo and Gemini-pro, demonstrate a substantial improvement in accuracy for both factual reasoning and mathematical proof compared to traditional DR methods. Combining IR with DR further enhances performance, underscoring the effectiveness of the proposed strategy. | Prompt Engineering |
- Self-Discover: Large Language Models Self-Compose Reasoning Structures - DISCOVER, a framework for LLMs to autonomously identify task-specific reasoning structures, improving performance on challenging reasoning benchmarks like BigBench-Hard and MATH. By selecting and composing atomic reasoning modules during a self-discovery process, SELF-DISCOVER enhances reasoning abilities, surpassing models like GPT-4 and PaLM 2 by up to 32% compared to traditional methods like Chain of Thought (CoT). Notably, it outperforms inference-intensive methods like CoT-Self-Consistency by over 20%, with significantly lower inference compute requirements, while exhibiting universality across different LLM model families and echoing human reasoning patterns. | Prompt Engineering |
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models - Ultra and GPT-4. DeepSeekMath's success is attributed to leveraging web data via a sophisticated data selection pipeline and employing Group Relative Policy Optimization (GRPO) to enhance mathematical reasoning while optimizing memory usage. | Task Specific LLMs |
- Large Language Model for Table Processing: A Survey - centric tasks and the utilization of LLMs to automate them, including traditional areas like Table QA and fact verification, as well as newer aspects such as table manipulation and advanced data analysis. It delves into recent paradigms in LLM usage, focusing on instruction-tuning, prompting, and agent-based approaches. The paper also addresses challenges like private deployment, efficient inference, and the need for extensive benchmarks in table manipulation and advanced data analysis. | Task Specific LLMs |
- More Agents Is All You Need - and-voting method. This method is independent of existing complex enhancement techniques and its effectiveness correlates with task difficulty. Extensive experiments across various LLM benchmarks validate this finding and explore associated properties. The code for the experiments is publicly accessible on Git. | LLM Agents |
- OLMo: Accelerating the Science of Language Models - tuning, and evaluation, alongside a collaborative environment for researchers. The goal is to facilitate discoveries and advancements in LLM technology, making it more accessible to a wider audience. | Open-Source LLMs |
- Revisiting Feature Prediction for Learning Visual Representations from Video - JEPA, a collection of vision models trained solely on video data using a feature prediction objective, without relying on pretrained image encoders, text, negative examples, or reconstruction. Trained on 2 million videos, these models are evaluated on downstream image and video tasks, demonstrating versatile visual representations that excel in both motion and appearance-based tasks without requiring adaptation of model parameters. The largest model, a ViT-H/16 trained only on videos, achieves impressive performance on Kinetics-400, Something-Something-v2, and ImageNet1K datasets. | Multimodal LLMs |
-
:star: Best GenAI Papers List (March 2024)
- Don't Use Your Data All at Once: sDPO
- STaR-GATE: Teaching Language Models to Ask Clarifying Questions - GATE, a novel approach for enhancing language models' interaction skills by training them to ask clarifying questions. By employing a strategic teacher-student learning framework, STaR-GATE aims to improve the models' ability to clarify ambiguities in user queries, thereby enhancing communication effectiveness and accuracy in understanding and responding to complex requests | Prompt Engineering |
- GaLore: Memory-Efficient LLM Training by Gradient Low-Rank Projection - rank projection, significantly cutting down the memory used by optimizer states without sacrificing performance. It allows for the efficient training of large models on consumer-grade GPUs, marking a significant advancement in the accessibility of AI model training. | Memory Optimization |
- Gecko: Versatile Text Embeddings Distilled from Large Language Models - step distillation process that generates diverse, synthetic paired data, Gecko achieves superior retrieval performance. With a focus on compactness, it outperforms larger models and higher-dimensional embeddings on the Massive Text Embedding Benchmark (MTEB), demonstrating its efficacy and potential in improving information retrieval tasks. | LLM Embeddings |
- Grok-1.5 - related tasks, and long context understanding. With improvements in MATH, GSM8K, and HumanEval benchmarks, Grok-1.5 offers expanded memory capacity and exceptional retrieval capabilities. Built on a custom distributed training framework, it promises efficiency and reliability for large-scale language model research | Foundational LLM |
- Jamba: AI21's SSM-Transformer Model - Transformer model offering a 256K context window, aiming to balance the SSM model's efficiency with the Transformer's capability. It shows significant performance improvements across various benchmarks. Jamba is open-source under Apache 2.0, available on Hugging Face, and soon on NVIDIA's API catalog, marking a significant advancement in hybrid model architecture | Foundational LLM |
- Long-form factuality in large language models - generated content on open-ended topics. It introduces LongFact, a set of prompts for evaluating long-form factuality, and proposes the Search-Augmented Factuality Evaluator (SAFE) method. SAFE assesses the accuracy of facts in LLM responses through a multi-step reasoning process, comparing supported facts against Google Search results. The findings indicate LLMs' potential for superhuman factuality assessment, offering a cost-effective alternative to human annotation. | LLM Factuality |
- Mini-Gemini: Mining the Potential of Multi-modality Vision Language Models - Gemini presents a framework to enhance multi-modal Vision Language Models (VLMs) by improving visual tokens, constructing high-quality datasets, and guiding VLM-based generation for better performance. It uses an additional visual encoder for high-resolution refinement without increasing visual token count, aiming to enhance image understanding, reasoning, and simultaneous generation capabilities of VLMs. Mini-Gemini has shown leading performance in zero-shot benchmarks, surpassing developed private models. | Multimodal LLM |
- DBRX - of-the-art open large language model surpassing established models like GPT-3.5 and competing with Gemini 1.0 Pro. DBRX excels in programming and general LLM capabilities, featuring a fine-grained mixture-of-experts architecture for enhanced training and inference efficiency. It's 40% the size of Grok-1, offering faster inference and reduced compute requirements. The model is available on Hugging Face, emphasizing Databricks' commitment to open models and enabling customers to pretrain DBRX-class models with their infrastructure | Foundational LLM |
- RankPrompt: Step-by-Step Comparisons Make Language Models Better Reasoners - 4. Unlike existing solutions requiring human annotations or failing in inconsistent scenarios, RankPrompt enables LLMs to self-rank their responses by comparing diverse outputs. Experiments across 11 reasoning tasks demonstrate significant performance enhancements, with up to a 13% improvement. Moreover, RankPrompt aligns with human judgments 74% of the time in open-ended evaluations and exhibits robustness to response variations. This method proves effective in eliciting high-quality feedback from LLMs, offering promising avenues for advancing reasoning abilities. | Prompt Engineering |
- Mora: Enabling Generalist Video Generation via A Multi-Agent Framework - agent framework to address the gap in generalist video generation capabilities, aiming to match the performance of the pioneering model Sora. It leverages multiple visual AI agents to achieve text-to-video generation, image-to-video conversion, video extension, editing, connection, and digital world simulation, demonstrating close performance to Sora across various tasks but with a noticeable gap when assessed holistically. | Multimodal LLM |
- LLM2LLM: Boosting LLMs with Novel Iterative Data Enhancement - student LLM framework for improving performance in tasks with limited data. It involves fine-tuning a student LLM on initial seed data, identifying errors, and generating new data based on these errors using a teacher LLM. This iterative process significantly boosts LLM performance in low-data regimes across various datasets, demonstrating substantial improvements over traditional fine-tuning and other augmentation methods. | Data Augmentation |
- Adaptive-RAG: Learning to Adapt Retrieval-Augmented Large Language Models through Question Complexity - augmented LLMs with a complexity-level classifier. This approach aims to balance efficiency and accuracy in response generation across different query types, showing improvements over existing models and adaptive retrieval methods | RAG |
- Evaluating Frontier Models for Dangerous Capabilities - security, self-proliferation, and self-reasoning, using Gemini 1.0 models. While no strong dangerous capabilities were found, early warning signs were identified. The study aims to advance the science of evaluating such capabilities in AI models, preparing for future advancements. | LLM Attacks |
- Evolutionary Optimization of Model Merging Recipes - source models. It optimizes beyond individual model weights, facilitating cross-domain merging and achieving state-of-the-art performance, notably in Japanese language tasks. This approach introduces a new paradigm for automated model composition, offering efficient alternatives for foundation model development. | Model Merging |
- Agent-FLAN: Designing Data and Methods of Effective Agent Tuning for Large Language Models - effects of existing approaches. Introducing Agent-FLAN, a method for Fine-tuning LANguage models for Agents, the paper proposes a novel approach to address these challenges. By carefully redesigning the training corpus and incorporating negative samples, Agent-FLAN enables significant performance improvements, outperforming prior works by 3.5% across multiple evaluation datasets. Moreover, it mitigates hallucination issues and enhances LLMs' agent capabilities, even with scaled model sizes, while slightly improving their general capability. | Agents, Hallucination |
- Grok-1 - 1 is an autoregressive Transformer-based model designed for next-token prediction, fine-tuned with feedback from Grok-0 models and humans. Released in November 2023, it boasts a context length of 8,192 tokens and is geared towards various NLP tasks like question answering and coding assistance. However, while Grok-1 excels in information processing, human review is essential to ensure accuracy as it lacks independent web-search capabilities. Despite access to external sources, the model may still hallucinate. Trained on data up to Q3 2023 from the internet and AI Tutors, its performance was evaluated on reasoning tasks and foreign math questions, with ongoing testing involving early adopters for further refinement. | Foundational LLM |
- RAFT: Adapting Language Model to Domain Specific RAG - specific settings. RAFT leverages retrieval augmented fine-tuning to enable the model to effectively incorporate new knowledge into its reasoning process. By training the model to disregard irrelevant documents (distractor documents) and cite relevant sequences from retrieved documents, RAFT improves the model's ability to provide accurate and coherent responses. Experimental results across various datasets demonstrate the effectiveness of RAFT in domain-specific Retrieval Augmented Generation, offering a valuable post-training recipe for enhancing pre-trained LLMs in domain-specific contexts. | RAG, Fine-Tuning |
- Logits of API-Protected LLMs Leak Proprietary Information - vocabulary outputs. This includes efficiently discovering hidden model sizes, identifying different model updates, and estimating output layer parameters. Empirical investigations on OpenAI's gpt-3.5-turbo reveal its embedding size to be approximately 4,096. The paper concludes by discussing potential measures for LLM providers to mitigate such attacks and suggests viewing these capabilities as opportunities for enhanced transparency and accountability rather than vulnerabilities. | LLM Attacks, Privacy |
- Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking - STaR, a method aimed at enabling language models to learn to generate rationales to explain future text, thereby improving their predictive abilities. Building upon the Self-Taught Reasoner (STaR) framework, Quiet-STaR allows LMs to infer unstated rationales in arbitrary text. Key challenges addressed include computational costs, LM's initial unfamiliarity with generating internal thoughts, and predicting beyond individual tokens. The proposed method involves tokenwise parallel sampling, learnable tokens for indicating thought boundaries, and extended teacher-forcing techniques. Quiet-STaR leads to significant improvements in LM performance on tasks like GSM8K and CommonsenseQA without requiring fine-tuning, marking a step towards more general and scalable reasoning capabilities in LMs. | Prompt Engineering |
- MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training - performing Multimodal Large Language Models (MLLMs) and investigates the significance of various architecture components and data choices. Through meticulous ablations of the image encoder, vision language connector, and pre-training data options, several crucial design insights are uncovered. For instance, the careful integration of image-caption, interleaved image-text, and text-only data is shown to be essential for achieving state-of-the-art few-shot results across multiple benchmarks. Additionally, the impact of image resolution and token count in the image encoder is highlighted, while the vision-language connector design is found to be comparatively less critical. Scaling up the proposed approach results in MM1, a family of multimodal models with up to 30B parameters, including dense models and mixture-of-experts variants. MM1 achieves state-of-the-art pre-training metrics and competitive performance on various multimodal benchmarks, benefiting from enhanced in-context learning and multi-image reasoning capabilities enabled by large-scale pre-training. | Multimodal LLM |
- Knowledge Conflicts for LLMs: A Survey - memory, inter-context, and intra-memory conflicts, which can significantly impact LLM trustworthiness and performance, particularly in real-world scenarios with noise and misinformation. Through categorization, exploration of causes, observation of LLM behaviors, and review of existing solutions, the survey aims to provide insights into strategies for enhancing LLM robustness, serving as a valuable resource for advancing research in this domain. | LLM Robustness |
- MoAI: Mixture of All Intelligence for Large Language and Vision Models - shot visual language tasks, particularly in real-world scene understanding, without increasing model size or requiring additional visual instruction datasets. | Multimodal LLMs |
- Branch-Train-MiX: Mixing Expert LLMs into a Mixture-of-Experts LLM - Train-MiX (BTX), the approach starts with a seed model and branches to train experts in parallel, reducing communication costs. After training, BTX combines the experts' feedforward parameters into Mixture-of-Expert (MoE) layers, followed by an MoE-finetuning stage to learn token-level routing. BTX encompasses two special cases: Branch-Train-Merge, which lacks the MoE finetuning stage, and sparse upcycling, which skips asynchronous training. Results demonstrate that BTX offers the best accuracy-efficiency tradeoff compared to alternative methods. | MoEs, Foundational LLM |
- Stealing Part of a Production Language Model - stealing attack capable of extracting precise information from black-box production language models like OpenAI's ChatGPT or Google's PaLM-2. By leveraging typical API access, the attack can recover the embedding projection layer of a transformer model, including symmetries. Remarkably, the attack achieves this for under $20 USD, revealing hidden dimensions of 1024 and 2048 for OpenAI's Ada and Babbage models, respectively. Additionally, the exact hidden dimension size of the gpt-3.5-turbo model is recovered, with an estimated cost of under $2,000 in queries to retrieve the entire projection matrix. The paper concludes with discussions on potential defenses and mitigations, as well as implications for future work that could extend the attack. | LLM Attacks |
- Common 7B Language Models Already Possess Strong Math Capabilities - sized language models, specifically LLaMA-2, already exhibit strong mathematical abilities, challenging previous assumptions that such capabilities require very large models or extensive math-focused pre-training. By leveraging synthetic data and scaling strategies, the study significantly improves the model's math-solving accuracy, surpassing previous benchmarks and demonstrating that with appropriate training, even relatively small models can achieve remarkable math performance. | Domain Specific LLMs |
- Can Large Language Models Reason and Plan? - critique and correct their erroneous guesses, a capability humans occasionally demonstrate. This inquiry underscores the distinct nature of human cognitive processes compared to the computational mechanisms of LLMs, challenging the assumption of equivalent reasoning and self-correction abilities between the two. | Prompt Engineering |
- KnowAgent: Knowledge-Augmented Planning for LLM-Based Agents - in action knowledge, leading to planning hallucination. KnowAgent uses an action knowledge base and a self-learning strategy to guide planning trajectories, resulting in more accurate and efficient problem-solving across various domains. | Agents |
- AIOS: LLM Agent Operating System
- What Are Tools Anyway? A Survey from the Language Model Perspective
- ShortGPT: Layers in Large Language Models are More Redundant Than You Expect
-
:speaker: Announcements
-
:star: Best Gen AI Papers List (August 2024)
- Eagle: Exploring The Design Space for Multimodal LLMs with Mixture of Encoders - sensitive tasks, such as optical character recognition and document analysis. A number of recent MLLMs achieve this goal using a mixture of vision encoders. Despite their success, there is a lack of systematic comparisons and detailed ablation studies addressing critical aspects, such as expert selection and the integration of multiple vision experts. This study provides an extensive exploration of the design space for MLLMs using a mixture of vision encoders and resolutions. Our findings reveal several underlying principles common to various existing strategies, leading to a streamlined yet effective design approach. We discover that simply concatenating visual tokens from a set of complementary vision encoders is as effective as more complex mixing architectures or strategies. We additionally introduce Pre-Alignment to bridge the gap between vision-focused encoders and language tokens, enhancing model coherence. The resulting family of MLLMs, Eagle, surpasses other leading open-source models on major MLLM benchmarks. Models and code: https://github.com/NVlabs/Eagle |
- The Mamba in the Llama: Distilling and Accelerating Hybrid Models - scale Transformer models, we consider the challenge of converting these pretrained models for deployment. We demonstrate that it is feasible to distill large Transformers into linear RNNs by reusing the linear projection weights from attention layers with academic GPU resources. The resulting hybrid model, which incorporates a quarter of the attention layers, achieves performance comparable to the original Transformer in chat benchmarks and outperforms open-source hybrid Mamba models trained from scratch with trillions of tokens in both chat benchmarks and general benchmarks. Moreover, we introduce a hardware-aware speculative decoding algorithm that accelerates the inference speed of Mamba and hybrid models. Overall we show how, with limited computation resources, we can remove many of the original attention layers and generate from the resulting model more efficiently. Our top-performing model, distilled from Llama3-8B-Instruct, achieves a 29.61 length-controlled win rate on AlpacaEval 2 against GPT-4 and 7.35 on MT-Bench, surpassing the best instruction-tuned linear RNN model. |
- Text2SQL is Not Enough: Unifying AI and Databases with TAG - Augmented Generation (RAG) considers the limited subset of queries that can be answered with point lookups to one or a few data records within the database. We propose Table-Augmented Generation (TAG), a unified and general-purpose paradigm for answering natural language questions over databases. The TAG model represents a wide range of interactions between the LM and database that have been previously unexplored and creates exciting research opportunities for leveraging the world knowledge and reasoning capabilities of LMs over data. We systematically develop benchmarks to study the TAG problem and find that standard methods answer no more than 20% of queries correctly, confirming the need for further research in this area. We release code for the benchmark at https://github.com/TAG-Research/TAG-Bench. |
- Foundation Models for Music: A Survey - of-the-art (SOTA) pre-trained models and foundation models in music, spanning from representation learning, generative learning and multimodal learning. We first contextualise the significance of music in various industries and trace the evolution of AI in music. By delineating the modalities targeted by foundation models, we discover many of the music representations are underexplored in FM development. Then, emphasis is placed on the lack of versatility of previous methods on diverse music applications, along with the potential of FMs in music understanding, generation and medical application. By comprehensively exploring the details of the model pre-training paradigm, architectural choices, tokenisation, finetuning methodologies and controllability, we emphasise the important topics that should have been well explored, like instruction tuning and in-context learning, scaling law and emergent ability, as well as long-sequence modelling etc. A dedicated section presents insights into music agents, accompanied by a thorough analysis of datasets and evaluations essential for pre-training and downstream tasks. Finally, by underscoring the vital importance of ethical considerations, we advocate that following research on FM for music should focus more on such issues as interpretability, transparency, human responsibility, and copyright issues. The paper offers insights into future challenges and trends on FMs for music, aiming to shape the trajectory of human-AI collaboration in the music realm. |
- Controllable Text Generation for Large Language Models: A Survey - world applications, LLMs must meet increasingly complex requirements. Beyond avoiding misleading or inappropriate content, LLMs are also expected to cater to specific user needs, such as imitating particular writing styles or generating text with poetic richness. These varied demands have driven the development of Controllable Text Generation (CTG) techniques, which ensure that outputs adhere to predefined control conditions--such as safety, sentiment, thematic consistency, and linguistic style--while maintaining high standards of helpfulness, fluency, and diversity. This paper systematically reviews the latest advancements in CTG for LLMs, offering a comprehensive definition of its core concepts and clarifying the requirements for control conditions and text quality. We categorize CTG tasks into two primary types: content control and attribute control. The key methods are discussed, including model retraining, fine-tuning, reinforcement learning, prompt engineering, latent space manipulation, and decoding-time intervention. We analyze each method's characteristics, advantages, and limitations, providing nuanced insights for achieving generation control. Additionally, we review CTG evaluation methods, summarize its applications across domains, and address key challenges in current research, including reduced fluency and practicality. We also propose several appeals, such as placing greater emphasis on real-world applications in future research. This paper aims to offer valuable guidance to researchers and developers in the field. Our reference list and Chinese version are open-sourced at https://github.com/IAAR-Shanghai/CTGSurvey. |
-
:star: Best GenAI Papers List (April 2024)
- Make Your LLM Fully Utilize the Context - intensive (IN2) training to address the lost-in-the-middle challenge faced by contemporary LLMs. Leveraging a synthesized long-context question-answer dataset, IN2 training emphasizes fine-grained information awareness within long contexts. Applying this approach to Mistral-7B yields FILM-7B (FILl-in-the-Middle), which robustly retrieves information from various positions in a 32K context window. FILM-7B improves performance on real-world long-context tasks while maintaining comparable performance on short-context tasks | Context Length |
- SEED-Bench-2-Plus: Benchmarking Multimodal Large Language Models with Text-Rich Visual Comprehension - Bench-2-Plus, a benchmark specifically tailored for evaluating text-rich visual comprehension of Multimodal Large Language Models (MLLMs). With 2.3K multiple-choice questions covering Charts, Maps, and Webs, it aims to simulate real-world text-rich scenarios comprehensively. Evaluation involving 34 prominent MLLMs highlights current limitations in text-rich visual comprehension, emphasizing the need for further research and improvement in this area. SEED-Bench-2-Plus serves as a valuable addition to existing MLLM benchmarks, offering insightful observations and inspiring future developments in text-rich visual comprehension. | Multimodal LLMs |
- AURORA -M: The First Open Source Multilingual Language Model Red-teamed according to the U.S. Executive Order - M, a multilingual open-source language model trained on English, Finnish, Hindi, Japanese, Vietnamese, and code. It surpasses 2 trillion tokens in total training token count and is fine-tuned on human-reviewed safety instructions, aligning its development with the Biden-Harris Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence. AURORA -M is rigorously evaluated across various tasks and languages, demonstrating robustness against catastrophic forgetting and outperforming alternatives in multilingual settings, particularly in safety evaluations. | Domain-Specific LLMs |
- Multi-Head Mixture-of-Experts - Head Mixture-of-Experts (MH-MoE) to address issues in Sparse Mixtures of Experts (SMoE), specifically low expert activation and lack of fine-grained analytical capabilities. MH-MoE employs a multi-head mechanism to split tokens into sub-tokens, assigning them to diverse experts for parallel processing before reintegrating them. This approach enhances expert activation, deepening context understanding and alleviating overfitting. MH-MoE is easy to implement and integrates seamlessly with other SMoE models, as demonstrated across English-focused language modeling, Multi-lingual language modeling, and Masked multi-modality modeling tasks. | New Architecture |
- Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone - 3-mini, a compact 3.8 billion parameter language model trained on 3.3 trillion tokens, delivering competitive performance similar to larger models like Mixtral 8x7B and GPT-3.5. Achieving notable scores on benchmarks such as MMLU (69%) and MT-bench (8.38), phi-3-mini is designed for deployment on mobile devices. The innovation lies in its dataset, a scaled-up version of phi-2's, comprising heavily filtered web data and synthetic data. Additionally, initial parameter-scaling results with phi-3-small and phi-3-medium models trained on 4.8T tokens demonstrate further enhanced performance. | Foundational LLM |
- How Good Are Low-bit Quantized LLaMA3 Models? An Empirical Study - bit quantization, essential for resource-limited scenarios. Despite their impressive pre-training on over 15T tokens, LLaMA3 models exhibit notable degradation when quantized to low bit-width. Evaluating 10 quantization methods on 1-8 bits across diverse datasets, the study reveals significant performance gaps, especially in ultra-low bit-width scenarios, highlighting the need for future developments to bridge this gap for practical applications. | Quantization |
- FlowMind: Automatic Workflow Generation with LLMs - level workflow descriptions, allowing effective inspection and feedback. Evaluation on NCEN-QA dataset demonstrates FlowMind's success and the significance of its components in enhancing user interaction and workflow generation. | LLM Agents |
- SnapKV: LLM Knows What You are Looking for Before Generation - tuning-free approach to efficiently minimize Key-Value (KV) cache size in LLMs while maintaining comparable performance. SnapKV utilizes attention head-specific prompt features identified from an 'observation' window, automatically compressing KV caches by selecting clustered important positions. This significantly reduces computational overhead and memory footprint, achieving a 3.6x increase in generation speed and an 8.2x enhancement in memory efficiency compared to baseline models when processing long input sequences | Fine-Tuning |
- AutoCrawler: A Progressive Understanding Web Agent for Web Crawler Generation - stage framework merging LLMs with crawlers to enhance adaptability in web automation. Addressing limitations of traditional methods and standalone LLM-based agents, AutoCrawler employs a hierarchical HTML structure for progressive understanding through top-down and step-back operations. Comprehensive experiments validate the effectiveness of this approach in handling diverse and changing web environments efficiently | LLM Agents |
- Groma: Localized Visual Tokenization for Grounding Multimodal Large Language Models - grained visual perception capabilities, enabling region-level tasks like captioning and visual grounding. Groma employs a localized visual tokenization mechanism to decompose images into regions of interest, seamlessly integrating region tokens into user instructions and model responses. By curating a visually grounded instruction dataset, Groma outperforms MLLMs relying solely on language models or external modules for localization, demonstrating superior performance in standard referring and grounding benchmarks. | Multimodal LLMs |
- Toward Self-Improvement of LLMs via Imagination, Searching, and Criticizing - tuning. Introducing AlphaLLM, it integrates Monte Carlo Tree Search (MCTS) with LLMs to establish a self-improving loop. AlphaLLM includes components for prompt synthesis, an efficient MCTS approach for language tasks, and critic models for precise feedback. Experimental results on mathematical reasoning tasks demonstrate that AlphaLLM significantly enhances LLM performance without additional annotations, showcasing its potential for self-improvement in complex reasoning and planning tasks. | Instruction Tuning |
- Reuse Your Rewards: Reward Model Transfer for Zero-Shot Cross-Lingual Alignment - shot cross-lingual alignment of language models using reward models trained on preference data from one source language and applied to other target languages. Evaluations on summarization and open-ended dialog generation tasks consistently show the success of this method, with cross-lingually aligned models preferred by humans in over 70% of evaluation instances. Surprisingly, different-language reward models sometimes outperform same-language ones. The study also identifies best practices for alignment when language-specific data for supervised fine-tuning is unavailable. | Instruction Tuning |
- Introducing v0.5 of the AI Safety Benchmark from MLCommons - tuned language models. It introduces a principled approach, covering a single use case and personas, along with a taxonomy of 13 hazard categories and tests for 7 categories. Version 1.0 is planned for release by 2024, aiming to provide deeper insights into AI system safety. While v0.5 should not be used for safety assessment, it offers detailed documentation and tools for evaluation, including a grading system and an openly available platform called ModelBench. | Benchmark, Evaluation |
- Octopus v2: On-device language model for super agent - device language model with 2 billion parameters to surpass the performance of GPT-4 in both accuracy and latency, while reducing the context length by 95%. The method addresses concerns over privacy and cost associated with large-scale language models in cloud environments by enabling deployment on edge devices such as smartphones, cars, VR headsets, and personal computers. By enhancing latency and reducing inference costs, the method aligns with the performance requisites for real-world applications, making it suitable for deployment across a variety of edge devices in production environments. | Small LLMs |
- Learn Your Reference Model for Real Good Alignment - Leibler divergence between policies to prevent overfitting. Direct Preference Optimization (DPO) aims to eliminate the Reward Model but faces limitations. We propose Trust Region DPO (TR-DPO), updating the reference policy during training, which outperforms DPO by up to 19% on Anthropic HH and TLDR datasets, enhancing model quality across multiple parameters. | Prompt Engineering |
- Compression Represents Intelligence Linearly - sourced datasets and pipelines provided for future research in compression assessment. | Model Compression |
- Pre-training Small Base LMs with Fewer Tokens
- Scaling (Down) CLIP: A Comprehensive Analysis of Data, Architecture, and Training Strategies - Image Pre-training (CLIP) under limited computation budgets across data, architecture, and training strategies. It emphasizes the importance of high-quality data and suggests smaller ViT models for smaller datasets and larger ones for larger datasets with fixed compute. Additionally, it compares four training strategies, finding that CLIP+Data Augmentation achieves comparable results to CLIP using half the data, offering practical insights for CLIP training and deployment. | Vision Models |
- Megalodon: Efficient LLM Pretraining and Inference with Unlimited Context Length - norm with two-hop residual configuration. In a head-to-head comparison with Llama2, Megalodon demonstrates better efficiency than Transformers with 7 billion parameters and 2 trillion training tokens, achieving a training loss of 1.70, positioning between Llama2-7B (1.75) and 13B (1.67) | Context Length |
- RULER: What’s the Real Context Size of Your Long-Context Language Models? - in-a-haystack (NIAH) test, widely used to evaluate long-context language models, assesses the ability to retrieve information from long distractor texts. However, it only measures a superficial form of long-context understanding. To provide a more comprehensive evaluation, a new synthetic benchmark called RULER is introduced. RULER expands upon the NIAH test by incorporating variations with diverse types and quantities of needles and introduces new task categories like multi-hop tracing and aggregation to test behaviors beyond context searching. The evaluation of ten long-context LMs with 13 representative tasks in RULER reveals large performance drops as the context length increases, despite nearly perfect accuracy in the NIAH test. Only four models can maintain satisfactory performance at the length of 32K tokens. RULER is open-sourced to encourage comprehensive evaluation of long-context LMs. | Context Length |
- Social Skill Training with Large Language Models - disciplinary innovation to address the broader implications for workforce development and social equality.;Social Skill Training; LLMs | Alignment |
- Rho-1: Not All Tokens Are What You Need - training methods treat all tokens equally, but our research challenges this by showing that not all tokens are equally important. We introduce Rho-1, a new model that selectively trains on tokens aligned with the desired distribution, improving few-shot accuracy in math tasks by up to 30%. After fine-tuning, Rho-1 achieves state-of-the-art results on the MATH dataset with significantly fewer pretraining tokens compared to existing models. Moreover, pretraining Rho-1 on general tokens enhances performance across diverse tasks, boosting both efficiency and effectiveness in language model pre-training. | New Architecture |
- RecurrentGemma: Moving Past Transformers for Efficient Open Language Models - sized state, which reduces memory use and enables efficient inference on long sequences. We provide a pre-trained model with 2B non-embedding parameters, and an instruction tuned variant. Both models achieve comparable performance to Gemma-2B despite being trained on fewer tokens. | New Architecture |
- Ferret-v2: An Improved Baseline for Referring and Grounding with Large Language Models - v2, an upgraded version of Ferret that overcomes limitations in regional understanding within LLMs. Ferret-v2 introduces three key enhancements: (1) Any resolution grounding and referring for improved image processing at higher resolutions. (2) Multi-granularity visual encoding using the DINOv2 encoder to better capture diverse visual contexts. (3) A three-stage training paradigm, including high-resolution dense alignment, leading to substantial improvements over Ferret and other state-of-the-art methods in referring and grounding tasks. | Benchmark, Evaluation |
- JetMoE: Reaching Llama2 Performance with 0.1M Dollars - 8B, a cost-effective and high-performing Large Language Model trained with minimal resources. Its efficient architecture reduces computation significantly compared to previous models, while its transparency encourages collaboration and advancements in accessible LLM development. | Foundational LLM |
- Elephants Never Forget: Memorization and Learning of Tabular Data in Large Language Models - context statistical learning abilities of LLMs without fine-tuning, emphasizing the importance of evaluating whether an LLM has seen an evaluation dataset during pre-training. The paper introduces the tabmemcheck Python package for testing exposure to datasets. | Domain-Specific LLMs |
- Leave No Context Behind: Efficient Infinite Context Transformers with Infini-attention - based LLMs to infinitely long inputs with bounded memory and computation. A key component in the proposed approach is a new attention technique dubbed Infini-attention. The Infini-attention incorporates a compressive memory into the vanilla attention mechanism and builds both masked local attention and long-term linear attention mechanisms in a single Transformer block. The effectiveness of this approach is demonstrated on long-context language modeling benchmarks, 1M sequence length passkey context block retrieval, and 500K length book summarization tasks with 1B and 8B LLMs. The approach introduces minimal bounded memory parameters and enables fast streaming inference for LLMs. | Context Length |
- LLM2Vec: Large Language Models Are Secretly Powerful Text Encoders - only language models (LLMs) excel in NLP tasks but are underutilized for text embedding. This study introduces LLM2Vec, a method converting decoder-only LLMs into robust text encoders via bidirectional attention, masked token prediction, and contrastive learning. Applied to LLMs with 1.3B to 7B parameters, LLM2Vec surpasses encoder-only models on word-level tasks and achieves a new unsupervised state-of-the-art on the Massive Text Embeddings Benchmark (MTEB). Integration with supervised contrastive learning further boosts performance, demonstrating the potential to create universal text encoders from LLMs without costly adaptation or synthetic data. | New Architecture |
- Stream of Search (SoS): Learning to Search in Language - tuning allows models to solve 36% of previously unsolved problems. This approach enables language models to learn problem-solving strategies and potentially discover new ones. | Domain-Specific LLMs |
- Long-context LLMs Struggle with Long In-context Learning - context learning within the realm of extreme-label classification. The benchmark evaluates 13 long-context LLMs on datasets with input lengths ranging from 2K to 50K tokens and label ranges spanning 28 to 174 classes. While long-context LLMs perform relatively well on less challenging tasks with shorter demonstration lengths, they struggle on more difficult tasks, reaching close to zero accuracy on the most challenging task, Discovery with 174 labels. Further analysis reveals a gap in current LLM capabilities for processing and understanding long, context-rich sequences, indicating the need for improved long context understanding and reasoning abilities in future LLMs. | Context Length |
- ReFT: Representation Finetuning for Language Models - efficient fine-tuning (PEFT) methods for adapting large language models. ReFT methods operate on a frozen base model and learn task-specific interventions on hidden representations, aiming to edit representations rather than modifying weights. A strong instance of ReFT, called Low-rank Linear Subspace ReFT (LoReFT), is presented, which achieves 10-50 times more parameter efficiency than prior PEFTs. LoReFT is showcased on various evaluation tasks, delivering the best balance of efficiency and performance compared to existing methods. | Fine-Tuning, PEFT |
- Training LLMs over Neurally Compressed Text - suited for learning. To address this, the paper proposes Equal-Info Windows, a compression technique segmenting text into blocks that compress to the same bit length. This method enables effective learning over neurally compressed text, improving with scale and outperforming byte-level baselines on perplexity and inference speed benchmarks. The paper also provides suggestions for further improving high-compression tokenizers. | Model Compression |
- CODE EDITOR BENCH: EVALUATING CODE EDITING CAPABILITY OF LARGE LANGUAGE MODELS - world scenarios and practical aspects of software development by curating diverse coding challenges and scenarios from various sources. Evaluation of 19 LLMs reveals that closed-source models, particularly Gemini-Ultra and GPT-4, outperform open-source models in CodeEditorBench, highlighting differences in model performance based on problem types and prompt sensitivities. CodeEditorBench aims to catalyze advancements in LLMs by providing a robust platform for assessing code editing capabilities and will release all prompts and datasets to enable the community to expand the dataset and benchmark emerging LLMs. | Evaluation |
- GPT-4V Red-teamed under 11 Different Safety Policies - teaming experiments are conducted on 11 different LLMs and Multimodal Large Language Models (MLLMs), including both state-of-the-art proprietary and open-source models. Results reveal GPT4 and GPT-4V's superior robustness against jailbreak attacks compared to open-source models. Notably, Llama2 and Qwen-VL-Chat demonstrate higher robustness among open-source models. The transferability of visual jailbreak methods is found to be relatively limited compared to textual jailbreak methods. | Red Teaming |
- RALL-E: Robust Codec Language Modeling with Chain-of-Thought Prompting for Text-to-Speech Synthesis - E presents a robust language modeling approach for text-to-speech (TTS) synthesis, addressing issues of poor robustness in LLMs such as unstable prosody and high word error rate (WER). The method employs chain-of-thought (CoT) prompting to decompose the task into simpler steps, predicting prosody features of the input text and using them as intermediate conditions to predict speech tokens. Additionally, RALL-E utilizes predicted duration prompts to guide self-attention weights, improving focus on corresponding phonemes and prosody features. Objective and subjective evaluations demonstrate significant improvements in WER compared to baseline methods, showcasing RALL-E's effectiveness in synthesizing challenging sentences with reduced error rates. | Prompt Engineering |
- CantTalkAboutThis: Aligning Language Models to Stay on Topic in Dialogues - following tasks, including safety alignment | Alignment |
- On the Scalability of Diffusion-based Text-to-Image Generation - based text-to-image (T2I) models by conducting extensive ablations on scaling denoising backbones and training sets. The study explores various training settings and training costs to understand how to efficiently scale the model for better performance at reduced cost. The findings suggest that increasing the transformer blocks is more parameter-efficient for improving text-image alignment than increasing channel numbers. Additionally, the quality and diversity of the training set have a significant impact on text-image alignment performance and learning efficiency. Scaling functions are provided to predict text-image alignment performance based on model size, compute, and dataset size. | Multimodal LLMs |
- Direct Preference Optimization of Video Large Multimodal Models from Language Model Reward
- Advancing LLM Reasoning Generalists with Preference Trees - 7B and CodeLlama-70B, EURUS models achieve state-of-the-art results among open-source models on a diverse set of benchmarks covering mathematics, code generation, and logical reasoning problems. EURUS outperforms existing open-source models by margins more than 13.3% on challenging benchmarks like LeetCode and TheoremQA. The strong performance of EURUS is attributed to ULTRA INTERACT, a large-scale alignment dataset designed for complex reasoning tasks, and a novel reward modeling objective derived from preference learning techniques. | Domain-Specific LLMs |
- Mixture-of-Depths: Dynamically allocating compute in transformer-based language models
- LLaVA-Gemma: Accelerating Multimodal Foundation Models with a Compact Language Model - Gemma, a suite of multimodal foundation models trained using the LLaVA framework with the Gemma family of LLMs, particularly the 2B parameter Gemma model. The study evaluates the effect of ablating three design features: pretraining the connector, utilizing a more powerful image backbone, and increasing the size of the language backbone. While LLaVA-Gemma exhibits moderate performance on various evaluations, it fails to surpass current state-of-the-art models of comparable size. The paper releases training recipes, code, and weights for the LLaVA-Gemma models, facilitating further research in this area. | Multimodal LLMs |
- Octopus v4: Graph of language models - source language models optimized for specific tasks. Octopus v4 excels in directing user queries to the most appropriate model and reformulating queries for optimal performance, building upon previous iterations (v1, v2, and v3) with enhanced selection and parameter understanding. Additionally, it explores the use of graphs as a versatile data structure to coordinate multiple models effectively. | Foundational LLM |
- Better & Faster Large Language Models via Multi-token Prediction - token prediction on generative benchmarks like coding, showing notable gains in problem-solving tasks. Moreover, models trained with multi-token prediction demonstrate up to threefold faster inference speeds, even with large batch sizes, offering additional efficiency benefits. | New Architecture |
- Extending Llama-3's Context Ten-Fold Overnight - 3-8B-Instruct model's context length is extended from 8K to 80K through efficient QLoRA fine-tuning, requiring only 8 hours on a single 8xA800 GPU machine. This extension significantly enhances model performance across various evaluation tasks like NIHS and topic retrieval, while maintaining proficiency in short-context tasks. Surprisingly, the extension is achieved with just 3.5K synthetic training samples from GPT-4, showcasing the untapped potential of LLMs to extend context lengths. The team plans to release all associated resources publicly, including data, model, data generation pipeline, and training code. | Context Length |
- Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse Models - 4. The PoLL approach, composed of a larger number of smaller models, outperforms single large judges across three distinct settings and six datasets. It exhibits less intra-model bias and is over seven times less expensive, offering a cost-effective and more reliable evaluation method for LLMs. | Evaluation |
- AdvPrompter: Fast Adaptive Adversarial Prompting for LLMs - readable adversarial prompts, called AdvPrompter, to address jailbreaking attacks on LLMs. Unlike existing optimization-based approaches, AdvPrompter achieves adversarial prompt generation in seconds, 800 times faster, without requiring access to gradients from the TargetLLM. The method alternates between generating high-quality target adversarial suffixes and low-rank fine-tuning of AdvPrompter. Experimental results demonstrate state-of-the-art performance on the AdvBench dataset and transferability to closed-source black-box LLM APIs. | Adversarial Attacks, Evaluation |
- Capabilities of Gemini Models in Medicine - Gemini, a specialized multimodal model for medical tasks, surpasses GPT-4 on various benchmarks, achieving state-of-the-art results in medical text summarization and question answering. With its advanced long-context reasoning, it outperforms existing methods in tasks such as needle-in-a-haystack retrieval from medical records. While promising, further evaluation is needed before deployment in real-world medical applications. | Domain-Specific LLMs |
- How Far Are We to GPT-4V? Closing the Gap to Commercial Multimodal Models with Open-Source Suites - source multimodal large language model (MLLM), bridges the gap between open-source and proprietary commercial models in multimodal understanding. It introduces three improvements: a Strong Vision Encoder, Dynamic High-Resolution image processing supporting up to 4K resolution, and a High-Quality Bilingual Dataset. Evaluation across benchmarks demonstrates its effectiveness compared to both open-source and proprietary models. | Multimodal LLMs |
-
:star: Best Gen AI Papers List (Sept 2024)
- MM1.5: Methods, Analysis & Insights from Multimodal LLM Fine-tuning - rich image understanding, visual referring and grounding, and multi-image reasoning. Building upon the MM1 architecture, MM1.5 adopts a data-centric approach to model training, systematically exploring the impact of diverse data mixtures across the entire model training lifecycle. This includes high-quality OCR data and synthetic captions for continual pre-training, as well as an optimized visual instruction-tuning data mixture for supervised fine-tuning. Our models range from 1B to 30B parameters, encompassing both dense and mixture-of-experts (MoE) variants, and demonstrate that careful data curation and training strategies can yield strong performance even at small scales (1B and 3B). Additionally, we introduce two specialized variants: MM1.5-Video, designed for video understanding, and MM1.5-UI, tailored for mobile UI understanding. Through extensive empirical studies and ablations, we provide detailed insights into the training processes and decisions that inform our final designs, offering valuable guidance for future research in MLLM development. |
- MIO: A Foundation Model on Multimodal Tokens - to-end, autoregressive manner. While the emergence of large language models (LLMs) and multimodal large language models (MM-LLMs) propels advancements in artificial general intelligence through their versatile capabilities, they still lack true any-to-any understanding and generation. Recently, the release of GPT-4o has showcased the remarkable potential of any-to-any LLMs for complex real-world tasks, enabling omnidirectional input and output across images, speech, and text. However, it is closed-source and does not support the generation of multimodal interleaved sequences. To address this gap, we present MIO, which is trained on a mixture of discrete tokens across four modalities using causal multimodal modeling. MIO undergoes a four-stage training process: (1) alignment pre-training, (2) interleaved pre-training, (3) speech-enhanced pre-training, and (4) comprehensive supervised fine-tuning on diverse textual, visual, and speech tasks. Our experimental results indicate that MIO exhibits competitive, and in some cases superior, performance compared to previous dual-modal baselines, any-to-any model baselines, and even modality-specific baselines. Moreover, MIO demonstrates advanced capabilities inherent to its any-to-any feature, such as interleaved video-text generation, chain-of-visual-thought reasoning, visual guideline generation, instructional image editing, etc. |
- MaskLLM: Learnable Semi-Structured Sparsity for Large Language Models - structured (or ``N:M'') Sparsity in LLMs, aimed at reducing computational overhead during inference. Instead of developing a new importance criterion, MaskLLM explicitly models N:M patterns as a learnable distribution through Gumbel Softmax sampling. This approach facilitates end-to-end training on large-scale datasets and offers two notable advantages: 1) High-quality Masks - our method effectively scales to large datasets and learns accurate masks; 2) Transferability - the probabilistic modeling of mask distribution enables the transfer learning of sparsity across domains or tasks. We assessed MaskLLM using 2:4 sparsity on various LLMs, including LLaMA-2, Nemotron-4, and GPT-3, with sizes ranging from 843M to 15B parameters, and our empirical results show substantial improvements over state-of-the-art methods. For instance, leading approaches achieve a perplexity (PPL) of 10 or greater on Wikitext compared to the dense model's 5.12 PPL, but MaskLLM achieves a significantly lower 6.72 PPL solely by learning the masks with frozen weights. Furthermore, MaskLLM's learnable nature allows customized masks for lossless application of 2:4 sparsity to downstream tasks or domains. Code is available at \url{https://github.com/NVlabs/MaskLLM}. |
- Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Multimodal Models - weight models rely heavily on synthetic data from proprietary VLMs to achieve good performance, effectively distilling these closed models into open ones. As a result, the community is still missing foundational knowledge about how to build performant VLMs from scratch. We present Molmo, a new family of VLMs that are state-of-the-art in their class of openness. Our key innovation is a novel, highly detailed image caption dataset collected entirely from human annotators using speech-based descriptions. To enable a wide array of user interactions, we also introduce a diverse dataset mixture for fine-tuning that includes in-the-wild Q&A and innovative 2D pointing data. The success of our approach relies on careful choices for the model architecture details, a well-tuned training pipeline, and, most critically, the quality of our newly collected datasets, all of which will be released. The best-in-class 72B model within the Molmo family not only outperforms others in the class of open weight and data models but also compares favorably against proprietary systems like GPT-4o, Claude 3.5, and Gemini 1.5 on both academic benchmarks and human evaluation. We will be releasing all of our model weights, captioning and fine-tuning data, and source code in the near future. Select model weights, inference code, and demo are available at https://molmo.allenai.org. |
- VPTQ: Extreme Low-bit Vector Post-Training Quantization for Large Language Models - only quantization to extremely low-bit (even down to 2 bits). It reduces memory requirements, optimizes storage costs, and decreases memory bandwidth needs during inference. However, due to numerical representation limitations, traditional scalar-based weight quantization struggles to achieve such extreme low-bit. Recent research on Vector Quantization (VQ) for LLMs has demonstrated the potential for extremely low-bit model quantization by compressing vectors into indices using lookup tables. In this paper, we introduce Vector Post-Training Quantization (VPTQ) for extremely low-bit quantization of LLMs. We use Second-Order Optimization to formulate the LLM VQ problem and guide our quantization algorithm design by solving the optimization. We further refine the weights using Channel-Independent Second-Order Optimization for a granular VQ. In addition, by decomposing the optimization problem, we propose a brief and effective codebook initialization algorithm. We also extend VPTQ to support residual and outlier quantization, which enhances model accuracy and further compresses the model. Our experimental results show that VPTQ reduces model quantization perplexity by $0.01$-$0.34$ on LLaMA-2, $0.38$-$0.68$ on Mistral-7B, $4.41$-$7.34$ on LLaMA-3 over SOTA at 2-bit, with an average accuracy improvement of $0.79$-$1.5\%$ on LLaMA-2, $1\%$ on Mistral-7B, $11$-$22\%$ on LLaMA-3 on QA tasks on average. We only utilize $10.4$-$18.6\%$ of the quantization algorithm execution time, resulting in a $1.6$-$1.8\times$ increase in inference throughput compared to SOTA. |
- Time-MoE: Billion-Scale Time Series Foundation Models with Mixture of Experts - scale pre-training in language and vision domains, pre-trained time series models remain limited in scale and operate at a high cost, hindering the development of larger capable forecasting models in real-world applications. In response, we introduce Time-MoE, a scalable and unified architecture designed to pre-train larger, more capable forecasting foundation models while reducing inference costs. By leveraging a sparse mixture-of-experts (MoE) design, Time-MoE enhances computational efficiency by activating only a subset of networks for each prediction, reducing computational load while maintaining high model capacity. This allows Time-MoE to scale effectively without a corresponding increase in inference costs. Time-MoE comprises a family of decoder-only transformer models that operate in an auto-regressive manner and support flexible forecasting horizons with varying input context lengths. We pre-trained these models on our newly introduced large-scale data Time-300B, which spans over 9 domains and encompassing over 300 billion time points. For the first time, we scaled a time series foundation model up to 2.4 billion parameters, achieving significantly improved forecasting precision. Our results validate the applicability of scaling laws for training tokens and model size in the context of time series forecasting. Compared to dense models with the same number of activated parameters or equivalent computation budgets, our models consistently outperform them by large margin. These advancements position Time-MoE as a state-of-the-art solution for tackling real-world time series forecasting challenges with superior capability, efficiency, and flexibility. |
- A Preliminary Study of o1 in Medicine: Are We Closer to an AI Doctor? - of-thought technique using reinforcement learning strategies. While it has demonstrated surprisingly strong capabilities on various general language tasks, its performance in specialized fields such as medicine remains unknown. To this end, this report provides a comprehensive exploration of o1 on different medical scenarios, examining 3 key aspects: understanding, reasoning, and multilinguality. Specifically, our evaluation encompasses 6 tasks using data from 37 medical datasets, including two newly constructed and more challenging question-answering (QA) tasks based on professional medical quizzes from the New England Journal of Medicine (NEJM) and The Lancet. These datasets offer greater clinical relevance compared to standard medical QA benchmarks such as MedQA, translating more effectively into real-world clinical utility. Our analysis of o1 suggests that the enhanced reasoning ability of LLMs may (significantly) benefit their capability to understand various medical instructions and reason through complex clinical scenarios. Notably, o1 surpasses the previous GPT-4 in accuracy by an average of 6.2% and 6.6% across 19 datasets and two newly created complex QA scenarios. But meanwhile, we identify several weaknesses in both the model capability and the existing evaluation protocols, including hallucination, inconsistent multilingual ability, and discrepant metrics for evaluation. We release our raw data and model outputs at https://ucsc-vlaa.github.io/o1_medicine/ for future research. |
- Instruction Following without Instruction Tuning - response pairs. We discover two forms of adaptation (tuning) that are deficient compared to instruction tuning, yet still yield instruction following; we call this implicit instruction tuning. We first find that instruction-response pairs are not necessary: training solely on responses, without any corresponding instructions, yields instruction following. This suggests pretrained models have an instruction-response mapping which is revealed by teaching the model the desired distribution of responses. However, we then find it's not necessary to teach the desired distribution of responses: instruction-response training on narrow-domain data like poetry still leads to broad instruction-following behavior like recipe generation. In particular, when instructions are very different from those in the narrow finetuning domain, models' responses do not adhere to the style of the finetuning domain. To begin to explain implicit instruction tuning, we hypothesize that very simple changes to a language model's distribution yield instruction following. We support this by hand-writing a rule-based language model which yields instruction following in a product-of-experts with a pretrained model. The rules are to slowly increase the probability of ending the sequence, penalize repetition, and uniformly change 15 words' probabilities. In summary, adaptations made without being designed to yield instruction following can do so implicitly. |
- Imagine yourself: Tuning-Free Personalized Image Generation - to-image tasks. In this research, we introduce Imagine yourself, a state-of-the-art model designed for personalized image generation. Unlike conventional tuning-based personalization techniques, Imagine yourself operates as a tuning-free model, enabling all users to leverage a shared framework without individualized adjustments. Moreover, previous work met challenges balancing identity preservation, following complex prompts and preserving good visual quality, resulting in models having strong copy-paste effect of the reference images. Thus, they can hardly generate images following prompts that require significant changes to the reference image, \eg, changing facial expression, head and body poses, and the diversity of the generated images is low. To address these limitations, our proposed method introduces 1) a new synthetic paired data generation mechanism to encourage image diversity, 2) a fully parallel attention architecture with three text encoders and a fully trainable vision encoder to improve the text faithfulness, and 3) a novel coarse-to-fine multi-stage finetuning methodology that gradually pushes the boundary of visual quality. Our study demonstrates that Imagine yourself surpasses the state-of-the-art personalization model, exhibiting superior capabilities in identity preservation, visual quality, and text alignment. This model establishes a robust foundation for various personalization applications. Human evaluation results validate the model's SOTA superiority across all aspects (identity preservation, text faithfulness, and visual appeal) compared to the previous personalization models. |
- Training Language Models to Self-Correct via Reinforcement Learning - correction is a highly desirable capability of large language models (LLMs), yet it has consistently been found to be largely ineffective in modern LLMs. Current methods for training self-correction typically depend on either multiple models, a more advanced model, or additional forms of supervision. To address these shortcomings, we develop a multi-turn online reinforcement learning (RL) approach, SCoRe, that significantly improves an LLM's self-correction ability using entirely self-generated data. To build SCoRe, we first show that variants of supervised fine-tuning (SFT) on offline model-generated correction traces are often insufficient for instilling self-correction behavior. In particular, we observe that training via SFT falls prey to either a distribution mismatch between mistakes made by the data-collection policy and the model's own responses, or to behavior collapse, where learning implicitly prefers only a certain mode of correction behavior that is often not effective at self-correction on test problems. SCoRe addresses these challenges by training under the model's own distribution of self-generated correction traces and using appropriate regularization to steer the learning process into learning a self-correction behavior that is effective at test time as opposed to fitting high-reward responses for a given prompt. This regularization process includes an initial phase of multi-turn RL on a base model to generate a policy initialization that is less susceptible to collapse, followed by using a reward bonus to amplify self-correction. With Gemini 1.0 Pro and 1.5 Flash models, we find that SCoRe achieves state-of-the-art self-correction performance, improving the base models' self-correction by 15.6% and 9.1% respectively on MATH and HumanEval. |
- Qwen2.5-Coder Technical Report - Coder series, a significant upgrade from its predecessor, CodeQwen1.5. This series includes two models: Qwen2.5-Coder-1.5B and Qwen2.5-Coder-7B. As a code-specific model, Qwen2.5-Coder is built upon the Qwen2.5 architecture and continues pretrained on a vast corpus of over 5.5 trillion tokens. Through meticulous data cleaning, scalable synthetic data generation, and balanced data mixing, Qwen2.5-Coder demonstrates impressive code generation capabilities while retaining general versatility. The model has been evaluated on a wide range of code-related tasks, achieving state-of-the-art (SOTA) performance across more than 10 benchmarks, including code generation, completion, reasoning, and repair, consistently outperforming larger models of the same model size. We believe that the release of the Qwen2.5-Coder series will not only push the boundaries of research in code intelligence but also, through its permissive licensing, encourage broader adoption by developers in real-world applications. |
- Scaling Smart: Accelerating Large Language Model Pre-training with Small Model Initialization - training phase of language models often begins with randomly initialized parameters. With the current trends in scaling models, training their large number of parameters can be extremely slow and costly. In contrast, small language models are less expensive to train, but they often cannot achieve the accuracy of large models. In this paper, we explore an intriguing idea to connect these two different regimes: Can we develop a method to initialize large language models using smaller pre-trained models? Will such initialization bring any benefits in terms of training time and final accuracy? In this paper, we introduce HyperCloning, a method that can expand the parameters of a pre-trained language model to those of a larger model with increased hidden dimensions. Our method ensures that the larger model retains the functionality of the smaller model. As a result, the larger model already inherits the predictive power and accuracy of the smaller model before the training starts. We demonstrate that training such an initialized model results in significant savings in terms of GPU hours required for pre-training large language models. |
- A Controlled Study on Long Context Extension and Generalization in LLMs - context learning require language models that utilize full document contexts. Due to the implementation challenges associated with directly training long-context models, many methods have been proposed for extending models to handle long contexts. However, owing to differences in data and model classes, it has been challenging to compare these approaches, leading to uncertainty as to how to evaluate long-context performance and whether it differs from standard evaluation. We implement a controlled protocol for extension methods with a standardized evaluation, utilizing consistent base models and extension data. Our study yields several insights into long-context behavior. First, we reaffirm the critical role of perplexity as a general-purpose performance indicator even in longer-context tasks. Second, we find that current approximate attention methods systematically underperform across long-context tasks. Finally, we confirm that exact fine-tuning based methods are generally effective within the range of their extension, whereas extrapolation remains challenging. All codebases, models, and checkpoints will be made available open-source, promoting transparency and facilitating further research in this critical area of AI development. |
- LLMs + Persona-Plug = Personalized LLMs - tuning a unique personalized LLM for each user, which is too expensive for widespread application. Alternative approaches introduce personalization information in a plug-and-play manner by retrieving the user's relevant historical texts as demonstrations. However, this retrieval-based strategy may break the continuity of the user history and fail to capture the user's overall styles and patterns, hence leading to sub-optimal performance. To address these challenges, we propose a novel personalized LLM model, \ours{}. It constructs a user-specific embedding for each individual by modeling all her historical contexts through a lightweight plug-in user embedder module. By attaching this embedding to the task input, LLMs can better understand and capture user habits and preferences, thereby producing more personalized outputs without tuning their own parameters. Extensive experiments on various tasks in the language model personalization (LaMP) benchmark demonstrate that the proposed model significantly outperforms existing personalized LLM approaches. |
- NVLM: Open Frontier-Class Multimodal LLMs - class multimodal large language models (LLMs) that achieve state-of-the-art results on vision-language tasks, rivaling the leading proprietary models (e.g., GPT-4o) and open-access models (e.g., Llama 3-V 405B and InternVL 2). Remarkably, NVLM 1.0 shows improved text-only performance over its LLM backbone after multimodal training. In terms of model design, we perform a comprehensive comparison between decoder-only multimodal LLMs (e.g., LLaVA) and cross-attention-based models (e.g., Flamingo). Based on the strengths and weaknesses of both approaches, we propose a novel architecture that enhances both training efficiency and multimodal reasoning capabilities. Furthermore, we introduce a 1-D tile-tagging design for tile-based dynamic high-resolution images, which significantly boosts performance on multimodal reasoning and OCR-related tasks. Regarding training data, we meticulously curate and provide detailed information on our multimodal pretraining and supervised fine-tuning datasets. Our findings indicate that dataset quality and task diversity are more important than scale, even during the pretraining phase, across all architectures. Notably, we develop production-grade multimodality for the NVLM-1.0 models, enabling them to excel in vision-language tasks while maintaining and even improving text-only performance compared to their LLM backbones. To achieve this, we craft and integrate a high-quality text-only dataset into multimodal training, alongside a substantial amount of multimodal math and reasoning data, leading to enhanced math and coding capabilities across modalities. To advance research in the field, we are releasing the model weights and will open-source the code for the community: https://nvlm-project.github.io/. |
- Promptriever: Instruction-Trained Retrievers Can Be Prompted Like Language Models - tuned language models (LM) are able to respond to imperative commands, providing a more natural user interface compared to their base counterparts. In this work, we present Promptriever, the first retrieval model able to be prompted like an LM. To train Promptriever, we curate and release a new instance-level instruction training set from MS MARCO, spanning nearly 500k instances. Promptriever not only achieves strong performance on standard retrieval tasks, but also follows instructions. We observe: (1) large gains (reaching SoTA) on following detailed relevance instructions (+14.3 p-MRR / +3.1 nDCG on FollowIR), (2) significantly increased robustness to lexical choices/phrasing in the query+instruction (+12.9 Robustness@10 on InstructIR), and (3) the ability to perform hyperparameter search via prompting to reliably improve retrieval performance (+1.4 average increase on BEIR). Promptriever demonstrates that retrieval models can be controlled with prompts on a per-query basis, setting the stage for future work aligning LM prompting techniques with information retrieval. |
- A Comprehensive Evaluation of Quantized Instruction-Tuned Large Language Models: An Experimental Analysis up to 405B - scale models such as Llama 3.1 with up to 405B have not been thoroughly examined. This paper evaluates the performance of instruction-tuned LLMs across various quantization methods (GPTQ, AWQ, SmoothQuant, and FP8) on models ranging from 7B to 405B. Using 13 benchmarks, we assess performance across six task types: commonsense Q\&A, knowledge and language understanding, instruction following, hallucination detection, mathematics, and dialogue. Our key findings reveal that (1) quantizing a larger LLM to a similar size as a smaller FP16 LLM generally performs better across most benchmarks, except for hallucination detection and instruction following; (2) performance varies significantly with different quantization methods, model size, and bit-width, with weight-only methods often yielding better results in larger models; (3) task difficulty does not significantly impact accuracy degradation due to quantization; and (4) the MT-Bench evaluation method has limited discriminatory power among recent high-performing LLMs. |
- RetrievalAttention: Accelerating Long-Context LLM Inference via Vector Retrieval - based Large Language Models (LLMs) have become increasingly important. However, due to the quadratic time complexity of attention computation, scaling LLMs to longer contexts incurs extremely slow inference latency and high GPU memory consumption for caching key-value (KV) vectors. This paper proposes RetrievalAttention, a training-free approach to both accelerate attention computation and reduce GPU memory consumption. By leveraging the dynamic sparsity of attention mechanism, RetrievalAttention proposes to use approximate nearest neighbor search (ANNS) indexes for KV vectors in CPU memory and retrieves the most relevant ones with vector search during generation. Unfortunately, we observe that the off-the-shelf ANNS indexes are often ineffective for such retrieval tasks due to the out-of-distribution (OOD) between query vectors and key vectors in attention mechanism. RetrievalAttention addresses the OOD challenge by designing an attention-aware vector search algorithm that can adapt to the distribution of query vectors. Our evaluation shows that RetrievalAttention only needs to access 1--3% of data while maintaining high model accuracy. This leads to significant reduction in the inference cost of long-context LLMs with much lower GPU memory footprint. In particular, RetrievalAttention only needs a single NVIDIA RTX4090 (24GB) for serving 128K tokens in LLMs with 8B parameters, which is capable of generating one token in 0.188 seconds. |
- Kolmogorov-Arnold Transformer - layer perceptron (MLP) layers to mix the information between channels. In this paper, we introduce the Kolmogorov-Arnold Transformer (KAT), a novel architecture that replaces MLP layers with Kolmogorov-Arnold Network (KAN) layers to enhance the expressiveness and performance of the model. Integrating KANs into transformers, however, is no easy feat, especially when scaled up. Specifically, we identify three key challenges: (C1) Base function. The standard B-spline function used in KANs is not optimized for parallel computing on modern hardware, resulting in slower inference speeds. (C2) Parameter and Computation Inefficiency. KAN requires a unique function for each input-output pair, making the computation extremely large. (C3) Weight initialization. The initialization of weights in KANs is particularly challenging due to their learnable activation functions, which are critical for achieving convergence in deep neural networks. To overcome the aforementioned challenges, we propose three key solutions: (S1) Rational basis. We replace B-spline functions with rational functions to improve compatibility with modern GPUs. By implementing this in CUDA, we achieve faster computations. (S2) Group KAN. We share the activation weights through a group of neurons, to reduce the computational load without sacrificing performance. (S3) Variance-preserving initialization. We carefully initialize the activation weights to make sure that the activation variance is maintained across layers. With these designs, KAT scales effectively and readily outperforms traditional MLP-based transformers. |
- On the Diagram of Thought - regressive next-token prediction with role-specific tokens, DoT facilitates seamless transitions between proposing ideas and critically evaluating them, providing richer feedback than binary signals. Furthermore, we formalize the DoT framework using Topos Theory, providing a mathematical foundation that ensures logical consistency and soundness in the reasoning process. This approach enhances both the training and inference processes within a single LLM, eliminating the need for multiple models or external control mechanisms. DoT offers a conceptual framework for designing next-generation reasoning-specialized models, emphasizing training efficiency, robust reasoning capabilities, and theoretical grounding. The code is available at https://github.com/diagram-of-thought/diagram-of-thought. |
- DSBench: How Far Are Data Science Agents to Becoming Data Science Experts? - Language Models (LVLMs) have demonstrated impressive language/vision reasoning abilities, igniting the recent trend of building agents for targeted applications such as shopping assistants or AI software engineers. Recently, many data science benchmarks have been proposed to investigate their performance in the data science domain. However, existing data science benchmarks still fall short when compared to real-world data science applications due to their simplified settings. To bridge this gap, we introduce DSBench, a comprehensive benchmark designed to evaluate data science agents with realistic tasks. This benchmark includes 466 data analysis tasks and 74 data modeling tasks, sourced from Eloquence and Kaggle competitions. DSBench offers a realistic setting by encompassing long contexts, multimodal task backgrounds, reasoning with large data files and multi-table structures, and performing end-to-end data modeling tasks. Our evaluation of state-of-the-art LLMs, LVLMs, and agents shows that they struggle with most tasks, with the best agent solving only 34.12% of data analysis tasks and achieving a 34.74% Relative Performance Gap (RPG). These findings underscore the need for further advancements in developing more practical, intelligent, and autonomous data science agents. |
- PingPong: A Benchmark for Role-Playing Language Models with User Emulation and Multi-Model Evaluation - playing capabilities of language models. Our approach leverages language models themselves to emulate users in dynamic, multi-turn conversations and to assess the resulting dialogues. The framework consists of three main components: a player model assuming a specific character role, an interrogator model simulating user behavior, and a judge model evaluating conversation quality. We conducted experiments comparing automated evaluations with human annotations to validate our approach, demonstrating strong correlations across multiple criteria. This work provides a foundation for a robust and dynamic evaluation of model capabilities in interactive scenarios. |
- LLaMA-Omni: Seamless Speech Interaction with Large Language Models - 4o enable real-time interaction with large language models (LLMs) through speech, significantly enhancing user experience compared to traditional text-based interaction. However, there is still a lack of exploration on how to build speech interaction models based on open-source LLMs. To address this, we propose LLaMA-Omni, a novel model architecture designed for low-latency and high-quality speech interaction with LLMs. LLaMA-Omni integrates a pretrained speech encoder, a speech adaptor, an LLM, and a streaming speech decoder. It eliminates the need for speech transcription, and can simultaneously generate text and speech responses directly from speech instructions with extremely low latency. We build our model based on the latest Llama-3.1-8B-Instruct model. To align the model with speech interaction scenarios, we construct a dataset named InstructS2S-200K, which includes 200K speech instructions and corresponding speech responses. Experimental results show that compared to previous speech-language models, LLaMA-Omni provides better responses in both content and style, with a response latency as low as 226ms. Additionally, training LLaMA-Omni takes less than 3 days on just 4 GPUs, paving the way for the efficient development of speech-language models in the future. |
- Can Large Language Models Unlock Novel Scientific Research Ideas? - 2 and GPT-4 are more aligned with the author's perspective than GPT-3.5 and Gemini. We also found that Claude-2 generates more diverse future research ideas than GPT-4, GPT-3.5, and Gemini 1.0. We further performed a human evaluation of the novelty, relevancy, and feasibility of the generated future research ideas. This investigation offers insights into the evolving role of LLMs in idea generation, highlighting both its capability and limitations. Our work contributes to the ongoing efforts in evaluating and utilizing language models for generating future research ideas. We make our datasets and codes publicly available. |
- SongCreator: Lyrics-based Universal Song Generation - generation system designed to tackle this challenge. The model features two novel designs: a meticulously designed dual-sequence language model (DSLM) to capture the information of vocals and accompaniment for song generation, and an additional attention mask strategy for DSLM, which allows our model to understand, generate and edit songs, making it suitable for various song-related generation tasks. Extensive experiments demonstrate the effectiveness of SongCreator by achieving state-of-the-art or competitive performances on all eight tasks. Notably, it surpasses previous works by a large margin in lyrics-to-song and lyrics-to-vocals. Additionally, it is able to independently control the acoustic conditions of the vocals and accompaniment in the generated song through different prompts, exhibiting its potential applicability. Our samples are available at https://songcreator.github.io/. |
- HyperAgent: Generalist Software Engineering Agents to Solve Coding Tasks at Scale - to-end development tasks, these systems are typically designed for specific SE tasks. We introduce HyperAgent, a novel generalist multi-agent system designed to address a wide spectrum of SE tasks across different programming languages by mimicking human developers' workflows. Comprising four specialized agents - Planner, Navigator, Code Editor, and Executor. HyperAgent manages the full lifecycle of SE tasks, from initial conception to final verification. Through extensive evaluations, HyperAgent achieves state-of-the-art performance across diverse SE tasks: it attains a 25.01% success rate on SWE-Bench-Lite and 31.40% on SWE-Bench-Verified for GitHub issue resolution, surpassing existing methods. Furthermore, HyperAgent demonstrates SOTA performance in repository-level code generation (RepoExec), and in fault localization and program repair (Defects4J), often outperforming specialized systems. This work represents a significant advancement towards versatile, autonomous agents capable of handling complex, multi-step SE tasks across various domains and languages, potentially transforming AI-assisted software development practices. |
- MemoRAG: Moving towards Next-Gen RAG Via Memory-Inspired Knowledge Discovery - Augmented Generation (RAG) leverages retrieval tools to access external databases, thereby enhancing the generation quality of large language models (LLMs) through optimized context. However, the existing retrieval methods are constrained inherently, as they can only perform relevance matching between explicitly stated queries and well-formed knowledge, but unable to handle tasks involving ambiguous information needs or unstructured knowledge. Consequently, existing RAG systems are primarily effective for straightforward question-answering tasks. In this work, we propose MemoRAG, a novel retrieval-augmented generation paradigm empowered by long-term memory. MemoRAG adopts a dual-system architecture. On the one hand, it employs a light but long-range LLM to form the global memory of database. Once a task is presented, it generates draft answers, cluing the retrieval tools to locate useful information within the database. On the other hand, it leverages an expensive but expressive LLM, which generates the ultimate answer based on the retrieved information. Building on this general framework, we further optimize MemoRAG's performance by enhancing its cluing mechanism and memorization capacity. In our experiment, MemoRAG achieves superior performance across a variety of evaluation tasks, including both complex ones where conventional RAG fails and straightforward ones where RAG is commonly applied. |
- OneGen: Efficient One-Pass Unified Generation and Retrieval for LLMs - pass Generation and retrieval framework (OneGen), designed to improve LLMs' performance on tasks that require both generation and retrieval. The proposed framework bridges the traditionally separate training approaches for generation and retrieval by incorporating retrieval tokens generated autoregressively. This enables a single LLM to handle both tasks simultaneously in a unified forward pass. We conduct experiments on two distinct types of composite tasks, RAG and Entity Linking, to validate the pluggability, effectiveness, and efficiency of OneGen in training and inference. Furthermore, our results show that integrating generation and retrieval within the same context preserves the generative capabilities of LLMs while improving retrieval performance. To the best of our knowledge, OneGen is the first to enable LLMs to conduct vector retrieval during the generation. |
- Attention Heads of Large Language Models: A Survey - box systems. Consequently, the reasoning bottlenecks of LLMs are mainly influenced by their internal architecture. As a result, many researchers have begun exploring the potential internal mechanisms of LLMs, with most studies focusing on attention heads. Our survey aims to shed light on the internal reasoning processes of LLMs by concentrating on the underlying mechanisms of attention heads. We first distill the human thought process into a four-stage framework: Knowledge Recalling, In-Context Identification, Latent Reasoning, and Expression Preparation. Using this framework, we systematically review existing research to identify and categorize the functions of specific attention heads. Furthermore, we summarize the experimental methodologies used to discover these special heads, dividing them into two categories: Modeling-Free methods and Modeling-Required methods. Also, we outline relevant evaluation methods and benchmarks. Finally, we discuss the limitations of current research and propose several potential future directions. |
- How Do Your Code LLMs Perform? Empowering Code Instruction Tuning with High-Quality Data - known high-quality datasets perform poorly. This discovery reveals a new challenge: identifying which dataset genuinely qualify as high-quality code instruction data. To address this, we propose an efficient code data pruning strategy for selecting good samples. Our approach is based on three dimensions: instruction complexity, response quality, and instruction diversity. Based on our selected data, we present XCoder, a family of models finetuned from LLaMA3. Our experiments show XCoder achieves new state-of-the-art performance using fewer training data, which verify the effectiveness of our data strategy. Moreover, we perform a comprehensive analysis on the data composition and find existing code datasets have different characteristics according to their construction methods, which provide new insights for future code LLMs. Our models and dataset are released in https://github.com/banksy23/XCoder |
- Paper Copilot: A Self-Evolving and Efficient LLM System for Personalized Academic Assistance - to-date information efficiently. We present Paper Copilot, a self-evolving, efficient LLM system designed to assist researchers, based on thought-retrieval, user profile and high performance optimization. Specifically, Paper Copilot can offer personalized research services, maintaining a real-time updated database. Quantitative evaluation demonstrates that Paper Copilot saves 69.92\% of time after efficient deployment. This paper details the design and implementation of Paper Copilot, highlighting its contributions to personalized academic support and its potential to streamline the research process. |
- From MOOC to MAIC: Reshaping Online Teaching and Learning through LLM-driven Agents - empowered Course), a new form of online education that leverages LLM-driven multi-agent systems to construct an AI-augmented classroom, balancing scalability with adaptivity. Beyond exploring the conceptual framework and technical innovations, we conduct preliminary experiments at Tsinghua University, one of China's leading universities. Drawing from over 100,000 learning records of more than 500 students, we obtain a series of valuable observations and initial analyses. This project will continue to evolve, ultimately aiming to establish a comprehensive open platform that supports and unifies research, technology, and applications in exploring the possibilities of online education in the era of large model AI. We envision this platform as a collaborative hub, bringing together educators, researchers, and innovators to collectively explore the future of AI-driven online education. |
- Towards a Unified View of Preference Learning for Large Language Models: A Survey - explored, limiting the development of the preference alignment. In light of this, we break down the existing popular alignment strategies into different components and provide a unified framework to study the current alignment strategies, thereby establishing connections among them. In this survey, we decompose all the strategies in preference learning into four components: model, data, feedback, and algorithm. This unified view offers an in-depth understanding of existing alignment algorithms and also opens up possibilities to synergize the strengths of different strategies. Furthermore, we present detailed working examples of prevalent existing algorithms to facilitate a comprehensive understanding for the readers. Finally, based on our unified perspective, we explore the challenges and future research directions for aligning large language models with human preferences. |
- LongCite: Enabling LLMs to Generate Fine-grained Citations in Long-context QA - context large language models (LLMs) have demonstrated impressive capacities in answering user questions based on extensive text, the lack of citations in their responses makes user verification difficult, leading to concerns about their trustworthiness due to their potential hallucinations. In this work, we aim to enable long-context LLMs to generate responses with fine-grained sentence-level citations, improving their faithfulness and verifiability. We first introduce LongBench-Cite, an automated benchmark for assessing current LLMs' performance in Long-Context Question Answering with Citations (LQAC), revealing considerable room for improvement. To this end, we propose CoF (Coarse to Fine), a novel pipeline that utilizes off-the-shelf LLMs to automatically generate long-context QA instances with precise sentence-level citations, and leverage this pipeline to construct LongCite-45k, a large-scale SFT dataset for LQAC. Finally, we train LongCite-8B and LongCite-9B using the LongCite-45k dataset, successfully enabling their generation of accurate responses and fine-grained sentence-level citations in a single output. The evaluation results on LongBench-Cite show that our trained models achieve state-of-the-art citation quality, surpassing advanced proprietary models including GPT-4o. |
- LongLLaVA: Scaling Multi-modal LLMs to 1000 Images Efficiently via a Hybrid Architecture - context capabilities of Multi-modal Large Language Models~(MLLMs) is crucial for video understanding, high-resolution image understanding, and multi-modal agents. This involves a series of systematic optimizations, including model architecture, data construction and training strategy, particularly addressing challenges such as \textit{degraded performance with more images} and \textit{high computational costs}. In this paper, we adapt the model architecture to a hybrid of Mamba and Transformer blocks, approach data construction with both temporal and spatial dependencies among multiple images and employ a progressive training strategy. The released model \textbf{LongLLaVA}~(\textbf{Long}-Context \textbf{L}arge \textbf{L}anguage \textbf{a}nd \textbf{V}ision \textbf{A}ssistant) is the first hybrid MLLM, which achieved a better balance between efficiency and effectiveness. LongLLaVA not only achieves competitive results across various benchmarks, but also maintains high throughput and low memory consumption. Especially, it could process nearly a thousand images on a single A100 80GB GPU, showing promising application prospects for a wide range of tasks. |
- OLMoE: Open Mixture-of-Experts Language Models - of-the-art language model leveraging sparse Mixture-of-Experts (MoE). OLMoE-1B-7B has 7 billion (B) parameters but uses only 1B per input token. We pretrain it on 5 trillion tokens and further adapt it to create OLMoE-1B-7B-Instruct. Our models outperform all available models with similar active parameters, even surpassing larger ones like Llama2-13B-Chat and DeepSeekMoE-16B. We present various experiments on MoE training, analyze routing in our model showing high specialization, and open-source all aspects of our work: model weights, training data, code, and logs. |
- GenAgent: Build Collaborative AI Systems with Automated Workflow Generation -- Case Studies on ComfyUI - based framework that automatically generates complex workflows, offering greater flexibility and scalability compared to monolithic models. The core innovation of GenAgent lies in representing workflows with code, alongside constructing workflows with collaborative agents in a step-by-step manner. We implement GenAgent on the ComfyUI platform and propose a new benchmark, OpenComfy. The results demonstrate that GenAgent outperforms baseline approaches in both run-level and task-level evaluations, showing its capability to generate complex workflows with superior effectiveness and stability. |
- Building Math Agents with Multi-Turn Iterative Preference Learning - solving capabilities can be enhanced by integrating external tools, such as code interpreters, and employing multi-turn Chain-of-Thought (CoT) reasoning. While current methods focus on synthetic data generation and Supervised Fine-Tuning (SFT), this paper studies the complementary direct preference learning approach to further improve model performance. However, existing direct preference learning algorithms are originally designed for the single-turn chat task, and do not fully address the complexities of multi-turn reasoning and external tool integration required for tool-integrated mathematical reasoning tasks. To fill in this gap, we introduce a multi-turn direct preference learning framework, tailored for this context, that leverages feedback from code interpreters and optimizes trajectory-level preferences. This framework includes multi-turn DPO and multi-turn KTO as specific implementations. The effectiveness of our framework is validated through training of various language models using an augmented prompt set from the GSM8K and MATH datasets. Our results demonstrate substantial improvements: a supervised fine-tuned Gemma-1.1-it-7B model's performance increased from 77.5% to 83.9% on GSM8K and from 46.1% to 51.2% on MATH. Similarly, a Gemma-2-it-9B model improved from 84.1% to 86.3% on GSM8K and from 51.0% to 54.5% on MATH. |
- VideoLLaMB: Long-context Video Understanding with Recurrent Memory Bridges - scale video-language models have shown significant potential for real-time planning and detailed interactions. However, their high computational demands and the scarcity of annotated datasets limit their practicality for academic researchers. In this work, we introduce VideoLLaMB, a novel framework that utilizes temporal memory tokens within bridge layers to allow for the encoding of entire video sequences alongside historical visual data, effectively preserving semantic continuity and enhancing model performance across various tasks. This approach includes recurrent memory tokens and a SceneTilling algorithm, which segments videos into independent semantic units to preserve semantic integrity. Empirically, VideoLLaMB significantly outstrips existing video-language models, demonstrating a 5.5 points improvement over its competitors across three VideoQA benchmarks, and 2.06 points on egocentric planning. Comprehensive results on the MVBench show that VideoLLaMB-7B achieves markedly better results than previous 7B models of same LLM. Remarkably, it maintains robust performance as PLLaVA even as video length increases up to 8 times. Besides, the frame retrieval results on our specialized Needle in a Video Haystack (NIAVH) benchmark, further validate VideoLLaMB's prowess in accurately identifying specific frames within lengthy videos. Our SceneTilling algorithm also enables the generation of streaming video captions directly, without necessitating additional training. In terms of efficiency, VideoLLaMB, trained on 16 frames, supports up to 320 frames on a single Nvidia A100 GPU with linear GPU memory scaling, ensuring both high performance and cost-effectiveness, thereby setting a new foundation for long-form video-language models in both academic and practical applications. |
- ContextCite: Attributing Model Generation to Context - cite. |
- LongRecipe: Recipe for Efficient Long Context Generalization in Large Language Models - context tasks because of their limited effective context window size during pretraining, which restricts their ability to generalize over extended sequences. Meanwhile, extending the context window in LLMs through post-pretraining is highly resource-intensive. To address this, we introduce LongRecipe, an efficient training strategy for extending the context window of LLMs, including impactful token analysis, position index transformation, and training optimization strategies. It simulates long-sequence inputs while maintaining training efficiency and significantly improves the model's understanding of long-range dependencies. Experiments on three types of LLMs show that LongRecipe can utilize long sequences while requiring only 30% of the target context window size, and reduces computational training resource over 85% compared to full sequence training. Furthermore, LongRecipe also preserves the original LLM's capabilities in general tasks. Ultimately, we can extend the effective context window of open-source LLMs from 8k to 128k, achieving performance close to GPT-4 with just one day of dedicated training using a single GPU with 80G memory. Our code is released at https://github.com/zhiyuanhubj/LongRecipe. |
- SciLitLLM: How to Adapt LLMs for Scientific Literature Understanding - training (CPT) and supervised fine-tuning (SFT), to simultaneously infuse scientific domain knowledge and enhance instruction-following capabilities for domain-specific tasks.cIn this process, we identify two key challenges: (1) constructing high-quality CPT corpora, and (2) generating diverse SFT instructions. We address these challenges through a meticulous pipeline, including PDF text extraction, parsing content error correction, quality filtering, and synthetic instruction creation. Applying this strategy, we present a suite of LLMs: SciLitLLM, specialized in scientific literature understanding. These models demonstrate promising performance on scientific literature understanding benchmarks. Our contributions are threefold: (1) We present an effective framework that integrates CPT and SFT to adapt LLMs to scientific literature understanding, which can also be easily adapted to other domains. (2) We propose an LLM-based synthesis method to generate diverse and high-quality scientific instructions, resulting in a new instruction set -- SciLitIns -- for supervised fine-tuning in less-represented scientific domains. (3) SciLitLLM achieves promising performance improvements on scientific literature understanding benchmarks. |
- Mini-Omni: Language Models Can Hear, Talk While Thinking in Streaming - 4o, as a new milestone, has enabled real-time conversations with humans, demonstrating near-human natural fluency. Such human-computer interaction necessitates models with the capability to perform reasoning directly with the audio modality and generate output in streaming. However, this remains beyond the reach of current academic models, as they typically depend on extra TTS systems for speech synthesis, resulting in undesirable latency. This paper introduces the Mini-Omni, an audio-based end-to-end conversational model, capable of real-time speech interaction. To achieve this capability, we propose a text-instructed speech generation method, along with batch-parallel strategies during inference to further boost the performance. Our method also helps to retain the original model's language capabilities with minimal degradation, enabling other works to establish real-time interaction capabilities. We call this training method "Any Model Can Talk". We also introduce the VoiceAssistant-400K dataset to fine-tune models optimized for speech output. To our best knowledge, Mini-Omni is the first fully end-to-end, open-source model for real-time speech interaction, offering valuable potential for future research. |
- Jina-ColBERT-v2: A General-Purpose Multilingual Late Interaction Retriever - vector dense models, such as ColBERT, have proven highly effective in information retrieval. ColBERT's late interaction scoring approximates the joint query-document attention seen in cross-encoders while maintaining inference efficiency closer to traditional dense retrieval models, thanks to its bi-encoder architecture and recent optimizations in indexing and search. In this work we propose a number of incremental improvements to the ColBERT model architecture and training pipeline, using methods shown to work in the more mature single-vector embedding model training paradigm, particularly those that apply to heterogeneous multilingual data or boost efficiency with little tradeoff. Our new model, Jina-ColBERT-v2, demonstrates strong performance across a range of English and multilingual retrieval tasks. |
- CoRe: Context-Regularized Text Embedding Learning for Text-to-Image Personalization - to-image personalization have enabled high-quality and controllable image synthesis for user-provided concepts. However, existing methods still struggle to balance identity preservation with text alignment. Our approach is based on the fact that generating prompt-aligned images requires a precise semantic understanding of the prompt, which involves accurately processing the interactions between the new concept and its surrounding context tokens within the CLIP text encoder. To address this, we aim to embed the new concept properly into the input embedding space of the text encoder, allowing for seamless integration with existing tokens. We introduce Context Regularization (CoRe), which enhances the learning of the new concept's text embedding by regularizing its context tokens in the prompt. This is based on the insight that appropriate output vectors of the text encoder for the context tokens can only be achieved if the new concept's text embedding is correctly learned. CoRe can be applied to arbitrary prompts without requiring the generation of corresponding images, thus improving the generalization of the learned text embedding. Additionally, CoRe can serve as a test-time optimization technique to further enhance the generations for specific prompts. Comprehensive experiments demonstrate that our method outperforms several baseline methods in both identity preservation and text alignment. Code will be made publicly available. |
Programming Languages
Categories
:mortar_board: Courses
104
:star: Best GenAI Papers List (April 2024)
48
:star: Best Gen AI Papers List (Oct 2024)
46
:star: Best Gen AI Papers List (June 2024)
46
:star: Best Gen AI Papers List (Sept 2024)
45
:notebook: Code Notebooks
36
:star: Best GenAI Papers List (February 2024)
35
:star: Best Gen AI Papers List (July 2024)
35
:star: Best Gen AI Papers List (May 2024)
33
:star: Best GenAI Papers List (March 2024)
31
:star: Best Gen AI Papers List (August 2024)
5
:speaker: Announcements
1
Sub Categories
Keywords