https://github.com/XiaoYee/Awesome_Efficient_LRM_Reasoning
😎 A Survey of Efficient Reasoning for Large Reasoning Models: Language, Multimodality, and Beyond (From Shanghai AI Lab)
https://github.com/XiaoYee/Awesome_Efficient_LRM_Reasoning
List: Awesome_Efficient_LRM_Reasoning
budget-aware chain-of-thought cot efficient efficient-reasoning long-cot lrm o1 o3 r1 reasoning slow-fast
Last synced: about 2 months ago
JSON representation
😎 A Survey of Efficient Reasoning for Large Reasoning Models: Language, Multimodality, and Beyond (From Shanghai AI Lab)
- Host: GitHub
- URL: https://github.com/XiaoYee/Awesome_Efficient_LRM_Reasoning
- Owner: XiaoYee
- License: mit
- Created: 2025-03-25T11:49:36.000Z (2 months ago)
- Default Branch: main
- Last Pushed: 2025-04-06T06:22:06.000Z (about 2 months ago)
- Last Synced: 2025-04-06T06:27:41.218Z (about 2 months ago)
- Topics: budget-aware, chain-of-thought, cot, efficient, efficient-reasoning, long-cot, lrm, o1, o3, r1, reasoning, slow-fast
- Homepage: https://arxiv.org/abs/2503.21614
- Size: 245 KB
- Stars: 135
- Watchers: 4
- Forks: 2
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- StarryDivineSky - XiaoYee/Awesome_Efficient_LRM_Reasoning
- ultimate-awesome - Awesome_Efficient_LRM_Reasoning - 😎 A Survey of Efficient Reasoning for Large Reasoning Models: Language, Multimodality, and Beyond. (Other Lists / Julia Lists)
- Awesome-Efficient-Reasoning - XiaoYee/Awesome_Efficient_LRM_Reasoning
README
# A Survey of Efficient Reasoning for Large Reasoning Models: Language, Multimodality, and Beyond
[](https://arxiv.org/pdf/2503.21614) [](https://github.com/XiaoYee/Awesome_Efficient_LRM_Reasoning)
[](https://x.com/suzhaochen0110/status/1905461785693749709?s=46)
[](https://github.com/XiaoYee/Awesome_Efficient_LRM_Reasoning)
[](https://opensource.org/licenses/MIT)
---
## 🔔 News
- [2025-04] We add more Hybrid models (e.g Mamba-Transformer) in Efficient Reasoning during Pre-training. It is more efficient to infer.
- [2025-04] We add a new "Model Merge" category in Efficient Reasoning during Inference. It is feasible to be a promising direction.
- [2025-03] We released our survey "[A Survey of Efficient Reasoning for Large Reasoning Models: Language, Multimodality, and Beyond](https://arxiv.org/abs/2503.21614)". This is the **first survey** for efficient reasoning of **Large Reasoning Models**, covering both language and multimodality.
- [2025-03] We created this repository to maintain a paper list on Awesome-Efficient-LRM-Reasoning.
---
> If you find our survey useful for your research, please consider citing:
```
@article{qu2025survey,
title={A Survey of Efficient Reasoning for Large Reasoning Models: Language, Multimodality, and Beyond},
author={Qu, Xiaoye and Li, Yafu and Su, Zhaochen and Sun, Weigao and Yan, Jianhao and Liu, Dongrui and Cui, Ganqu and Liu, Daizong and Liang, Shuxian and He, Junxian and others},
journal={arXiv preprint arXiv:2503.21614},
year={2025}
}
```
---
## 🔥 Table of Contents
- [Awesome-Efficient-LRM-Reasoning](#awesome-efficient-lrm-reasoning)
- [👀 Introduction](#-introduction)
- [💭 Efficient Reasoning during Inference](#-efficient-reasoning-during-inference)
- [Length Budgeting](#length-budgeting)
- [System Switch](#system-switch)
- [Model Switch](#model-switch)
- [Model Merge](#model-merge)
- [Parallel Search](#parallel-search)
- [💫 Efficient Reasoning with SFT](#-efficient-reasoning-with-sft)
- [Reasoning Chain Compression](#reasoning-chain-compression)
- [Latent-Space SFT](#latent-space-sft)
- [🧩 Efficient Reasoning with Reinforcement Learning](#-efficient-reasoning-with-reinforcement-learning)
- [Efficient Reinforcement Learning with Length Reward](#efficient-reinforcement-learning-with-length-reward)
- [Efficient Reinforcement Learning without Length Reward](#efficient-reinforcement-learning-without-length-reward)
- [💬 Efficient Reasoning during Pre-training](#-efficient-reasoning-during-pre-training)
- [Pretraining with Latent Space](#pretraining-with-latent-space)
- [Subquadratic Attention](#subquadratic-attention)
- [Linearization](#linearization)
- [Efficient Reasoning with Subquadratic Attention](#efficient-reasoning-with-subquadratic-attention)
- [🔖 Future Directions](#-future-directions)
- [Efficient Multimodal Reasoning and Video Reasoning](#efficient-multimodal-reasoning-and-video-reasoning)
- [Efficient Test-time Scaling and Infinity Thinking](#efficient-test-time-scaling-and-infinity-thinking)
- [Efficient and Trustworthy Reasoning](#efficient-and-trustworthy-reasoning)
- [Building Efficient Reasoning Applications](#building-efficient-reasoning-applications)
- [Evaluation and Benchmark](#evaluation-and-benchmark)
---
## 📜Content
### 👀 Introduction
In the age of LRMs, we propose that "**Efficiency is the essence of intelligence.**"
Just as a wise human knows when to stop thinking and start deciding, a wise model should know when to halt unnecessary deliberation.
An intelligent model should manipulate the token economy, i.e., allocating tokens purposefully, skipping redundancy, and optimizing the path to a solution. Rather than naively traversing every possible reasoning path, it should emulate a master strategist, balancing cost and performance with elegant precision.
To summarize, this survey makes the following key contributions to the literature:
- Instead of offering a general overview of LRMs, we focus on the emerging and critical topic of **efficient reasoning** in LRMs, providing an in-depth and targeted analysis.
- We identify and characterize common patterns of reasoning inefficiency, and outline the current challenges that are unique to improving reasoning efficiency in large models.
- We provide a comprehensive review of recent advancements aimed at enhancing reasoning efficiency, structured across the end-to-end LRM development pipeline, from pretraining and supervised fine-tuning to reinforcement learning and inference.
---
## 🚀 Papers
### 💭 Efficient Reasoning during Inference
#### Length Budgeting
- [How Well do LLMs Compress Their Own Chain-of-Thought? A Token Complexity Approach](https://arxiv.org/abs/2503.01141) 
- [Sketch-of-Thought: Efficient LLM Reasoning with Adaptive Cognitive-Inspired Sketching](https://arxiv.org/abs/2503.05179) 
- [Chain of Draft: Thinking Faster by Writing Less](https://arxiv.org/abs/2502.18600) 
- [SafeChain: Safety of Language Models with Long Chain-of-Thought Reasoning Capabilities](https://arxiv.org/abs/2502.12025) 
- [s1: Simple test-time scaling](https://arxiv.org/abs/2501.19393) 
- [Token-budget-aware llm reasoning](https://arxiv.org/abs/2412.18547) 
- [Efficiently Serving LLM Reasoning Programs with Certaindex](https://arxiv.org/abs/2412.20993) 
- [Make every penny count: Difficulty-adaptive self-consistency for cost-efficient reasoning](https://arxiv.org/abs/2408.13457) 
- [Scaling llm test-time compute optimally can be more effective than scaling model parameters](https://arxiv.org/abs/2408.03314) 
- [Concise thoughts: Impact of output length on llm reasoning and cost](https://arxiv.org/abs/2407.19825) 
- [The impact of reasoning step length on large language models](https://arxiv.org/abs/2401.04925v3) 
- [The benefits of a concise chain of thought on problem-solving in large language models](https://arxiv.org/abs/2401.05618) 
- [Guiding language model reasoning with planning tokens](https://arxiv.org/abs/2310.05707) 
#### System Switch
- [Think More, Hallucinate Less: Mitigating Hallucinations via Dual Process of Fast and Slow Thinking](https://arxiv.org/abs/2501.01306) 
- [Dualformer: Controllable Fast and Slow Thinking by Learning with Randomized Reasoning Traces](https://arxiv.org/abs/2410.09918) 
- [Visual Agents as Fast and Slow Thinkers](https://arxiv.org/abs/2408.08862) 
- [System-1.x: Learning to Balance Fast and Slow Planning with Language Models](https://arxiv.org/abs/2407.14414) 
- [DynaThink: Fast or slow? A dynamic decision-making framework for large language models](https://arxiv.org/abs/2407.01009) 
#### Model Switch
- [MixLLM: Dynamic Routing in Mixed Large Language Models](https://arxiv.org/abs/2502.18482) 
- [Closer Look at Efficient Inference Methods: A Survey of Speculative Decoding](https://arxiv.org/abs/2411.13157) 
- [EAGLE-2: Faster Inference of Language Models with Dynamic Draft Trees](https://arxiv.org/abs/2406.16858) 
- [RouteLLM: Learning to Route LLMs with Preference Data](https://arxiv.org/abs/2406.18665) 
- [LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding](https://arxiv.org/abs/2404.16710) 
- [EAGLE: Speculative Sampling Requires Rethinking Feature Uncertainty](https://arxiv.org/abs/2401.15077) 
- [Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads](https://arxiv.org/abs/2401.10774) 
- [Routing to the Expert: Efficient Reward-guided Ensemble of Large Language Models](https://arxiv.org/abs/2311.08692) 
- [Speculative Decoding with Big Little Decoder](https://arxiv.org/abs/2302.07863) 
#### Model Merge
- [Unlocking efficient long-to-short llm reasoning with model merging](https://arxiv.org/abs/2503.20641) 
#### Parallel Search
- [Sampling-Efficient Test-Time Scaling: Self-Estimating the Best-of-N Sampling in Early Decoding](https://arxiv.org/abs/2503.01422) 
- [Efficient Test-Time Scaling via Self-Calibration](https://arxiv.org/abs/2503.00031) 
- [Meta-Reasoner: Dynamic Guidance for Optimized Inference-time Reasoning in Large Language Models](https://arxiv.org/abs/2502.19918) 
- [Test-Time Preference Optimization: On-the-Fly Alignment via Iterative Textual Feedback](https://arxiv.org/abs/2501.12895) 
- [Fast Best-of-N Decoding via Speculative Rejection](https://arxiv.org/abs/2410.20290) 
- [TreeBoN: Enhancing Inference-Time Alignment with Speculative Tree-Search and Best-of-N Sampling](https://arxiv.org/abs/2410.16033) 
- [Scaling llm test-time compute optimally can be more effective than scaling model parameters](https://arxiv.org/abs/2408.03314) 
### 💫 Efficient Reasoning with SFT
#### Reasoning Chain Compression
- [Self-Training Elicits Concise Reasoning in Large Language Models](https://arxiv.org/abs/2502.20122) 
- [TokenSkip: Controllable Chain-of-Thought Compression in LLMs](https://arxiv.org/abs/2502.12067) 
- [Stepwise Perplexity-Guided Refinement for Efficient Chain-of-Thought Reasoning in Large Language Models](https://arxiv.org/abs/2502.13260) 
- [C3oT: Generating Shorter Chain-of-Thought without Compromising Effectiveness](https://arxiv.org/abs/2412.11664) 
- [Can Language Models Learn to Skip Steps?](https://arxiv.org/abs/2411.01855) 
- [Distilling System 2 into System 1](https://arxiv.org/abs/2407.06023) 
#### Latent-Space SFT
- [From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step](https://arxiv.org/abs/2405.14838) 
- [CODI: Compressing Chain-of-Thought into Continuous Space via Self-Distillation](https://arxiv.org/abs/2502.21074) 
- [Token Assorted: Mixing Latent and Text Tokens for Improved Language Model Reasoning](https://arxiv.org/abs/2502.03275) 
- [SoftCoT: Soft Chain-of-Thought for Efficient Reasoning with LLMs](https://arxiv.org/abs/2502.12134) 
- [LightThinker: Thinking Step-by-Step Compression](https://arxiv.org/abs/2502.15589) 
- [Efficient Reasoning with Hidden Thinking](https://arxiv.org/abs/2501.19201) 
- [Training Large Language Models to Reason in a Continuous Latent Space](https://arxiv.org/abs/2412.06769) 
- [Compressed Chain of Thought: Efficient Reasoning Through Dense Representations](https://arxiv.org/abs/2412.13171) 
### 🧩 Efficient Reasoning with Reinforcement Learning
#### Efficient Reinforcement Learning with Length Reward
- [HAWKEYE: Efficient Reasoning with Model Collaboration](https://arxiv.org/pdf/2504.00424v1) 
- [ThinkPrune: Pruning Long Chain-of-Thought of LLMs via Reinforcement Learning](https://arxiv.org/abs/2504.01296) 
- [Think When You Need: Self-Adaptive Chain-of-Thought Learning](https://arxiv.org/abs/2504.03234) 
- [DAST: Difficulty-Adaptive Slow-Thinking for Large Reasoning Models](https://arxiv.org/abs/2503.04472) 
- [L1: Controlling How Long A Reasoning Model Thinks With Reinforcement Learning](https://www.arxiv.org/abs/2503.04697) 
- [Demystifying Long Chain-of-Thought Reasoning in LLMs](https://arxiv.org/abs/2502.03373) 
- [Training Language Models to Reason Efficiently](https://arxiv.org/abs/2502.04463) 
- [O1-Pruner: Length-Harmonizing Fine-Tuning for O1-Like Reasoning Pruning](https://arxiv.org/abs/2501.12570) 
- [Kimi k1.5: Scaling Reinforcement Learning with LLMs](https://arxiv.org/abs/2501.12599) 
#### Efficient Reinforcement Learning without Length Reward
- [Optimizing Test-Time Compute via Meta Reinforcement Fine-Tuning](https://arxiv.org/abs/2503.07572) 
- [Think Smarter not Harder: Adaptive Reasoning with Inference Aware Optimization](https://arxiv.org/abs/2501.17974) 
- [Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs](https://arxiv.org/abs/2412.21187) 
### 💬 Efficient Reasoning during Pre-training
#### Pretraining with Latent Space
- [LLM Pretraining with Continuous Concepts](https://arxiv.org/abs/2502.08524) 
- [Scalable Language Models with Posterior Inference of Latent Thought Vectors](https://arxiv.org/abs/2502.01567) 
- [Byte latent transformer: Patches scale better than tokens](https://arxiv.org/abs/2412.09871) 
- [Large Concept Models: Language Modeling in a Sentence Representation Space](https://arxiv.org/abs/2412.08821) 
#### Subquadratic Attention
- [RWKV-7 "Goose" with Expressive Dynamic State Evolution](https://arxiv.org/abs/2503.14456) 
- [LASP-2: Rethinking Sequence Parallelism for Linear Attention and Its Hybrid](https://arxiv.org/abs/2502.07563) 
- [Native sparse attention: Hardware-aligned and natively trainable sparse attention](https://arxiv.org/abs/2502.11089) 
- [MoBA: Mixture of Block Attention for Long-Context LLMs](https://arxiv.org/abs/2502.13189) 
- [MoM: Linear Sequence Modeling with Mixture-of-Memories](https://www.arxiv.org/abs/2502.13685) 
- [Gated Delta Networks: Improving Mamba2 with Delta Rule](https://arxiv.org/abs/2412.06464) 
- [Transformers are SSMs: Generalized models and efficient algorithms through structured state space duality](https://arxiv.org/abs/2405.21060) 
- [Various Lengths, Constant Speed: Efficient Language Modeling with Lightning Attention](https://arxiv.org/abs/2405.17381) 
- [Gated linear attention transformers with hardware-efficient training](https://arxiv.org/abs/2312.06635) 
#### Linearization
- [Liger: Linearizing Large Language Models to Gated Recurrent Structures](https://arxiv.org/abs/2503.01496) 
- [Llamba: Scaling Distilled Recurrent Models for Efficient Language Processing](https://arxiv.org/abs/2502.14458) 
- [LoLCATs: On Low-Rank Linearizing of Large Language Models](https://arxiv.org/abs/2410.10254) 
- [The Mamba in the Llama: Distilling and Accelerating Hybrid Models](https://arxiv.org/abs/2408.15237) 
#### Efficient Reasoning with Subquadratic Attention
- [Nemotron-H: A Family of Accurate and Efficient Hybrid Mamba-Transformer Models](https://arxiv.org/abs/2504.03624v1) 
- [Compositional Reasoning with Transformers, RNNs, and Chain of Thought](https://arxiv.org/abs/2503.01544) 
- [Cosmos-Reason1: From Physical Common Sense To Embodied Reasoning](https://arxiv.org/abs/2503.15558v1) 
- [Thinking Slow, Fast: Scaling Inference Compute with Distilled Reasoners](https://arxiv.org/abs/2502.20339) 
### 🔖 Future Directions
#### Efficient Multimodal Reasoning and Video Reasoning
- [Can Atomic Step Decomposition Enhance the Self-structured Reasoning of Multimodal Large Models?](https://arxiv.org/abs/2503.06252) 
- [Skywork R1V: Pioneering Multimodal Reasoning with Chain-of-Thought](https://huggingface.co/Skywork/Skywork-R1V-38B)
#### Efficient Test-time Scaling and Infinity Thinking
- [Efficient Test-Time Scaling via Self-Calibration](https://arxiv.org/abs/2503.00031) 
- [Dynamic self-consistency: Leveraging reasoning paths for efficient llm sampling](https://arxiv.org/abs/2408.17017) 
#### Efficient and Trustworthy Reasoning
- [X-Boundary: Establishing Exact Safety Boundary to Shield LLMs from Multi-Turn Jailbreaks without Compromising Usability](https://arxiv.org/abs/2502.09990) 
- [Deliberative alignment: Reasoning enables safer language models](https://arxiv.org/abs/2412.16339) 
#### Building Efficient Reasoning Applications
- [The Danger of Overthinking: Examining the Reasoning-Action Dilemma in Agentic Tasks](https://arxiv.org/abs/2502.08235) 
- [Chain-of-Retrieval Augmented Generation](https://arxiv.org/abs/2501.14342) 
#### Evaluation and Benchmark
- [DNA Bench: When Silence is Smarter -- Benchmarking Over-Reasoning in Reasoning LLMs](https://arxiv.org/abs/2503.15793) 
- [Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs](https://arxiv.org/abs/2412.21187) 
---
## Resources
**Reading lists related to Efficient Reasoning**
- [hemingkx/Awesome-Efficient-Reasoning](https://github.com/hemingkx/Awesome-Efficient-Reasoning)
- [Eclipsess/Awesome-Efficient-Reasoning-LLMs](https://github.com/Eclipsess/Awesome-Efficient-Reasoning-LLMs)
- [Hongcheng-Gao/Awesome-Long2short-on-LRMs](https://github.com/Hongcheng-Gao/Awesome-Long2short-on-LRMs)
- [DevoAllen/Awesome-Reasoning-Economy-Papers](https://github.com/DevoAllen/Awesome-Reasoning-Economy-Papers)
- [Blueyee/Efficient-CoT-LRMs](https://github.com/Blueyee/Efficient-CoT-LRMs)
- [EIT-NLP/Awesome-Latent-CoT](https://github.com/EIT-NLP/Awesome-Latent-CoT)
- [yzhangchuck/awesome-llm-reasoning-long2short-papers](https://github.com/yzhangchuck/awesome-llm-reasoning-long2short-papers)
## 🎉 Contribution
### Contributing to this paper list
⭐" **Join us in improving this repository!** If you know of any important works we've missed, please contribute. Your efforts are highly valued! "
### Contributors
---