https://github.com/Eclipsess/Awesome-Efficient-Reasoning-LLMs
A Survey on Efficient Reasoning for LLMs
https://github.com/Eclipsess/Awesome-Efficient-Reasoning-LLMs
List: Awesome-Efficient-Reasoning-LLMs
efficiency large-language-models large-reasoning-models
Last synced: 19 days ago
JSON representation
A Survey on Efficient Reasoning for LLMs
- Host: GitHub
- URL: https://github.com/Eclipsess/Awesome-Efficient-Reasoning-LLMs
- Owner: Eclipsess
- Created: 2025-03-19T07:43:22.000Z (about 1 month ago)
- Default Branch: main
- Last Pushed: 2025-04-01T18:19:35.000Z (24 days ago)
- Last Synced: 2025-04-01T19:32:08.899Z (24 days ago)
- Topics: efficiency, large-language-models, large-reasoning-models
- Homepage:
- Size: 1.19 MB
- Stars: 230
- Watchers: 2
- Forks: 5
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
- ultimate-awesome - Awesome-Efficient-Reasoning-LLMs - A Survey on Efficient Reasoning for LLMs. (Other Lists / Julia Lists)
- Awesome_Efficient_LRM_Reasoning - Eclipsess/Awesome-Efficient-Reasoning-LLMs
- Awesome-Efficient-Reasoning - Eclipsess/Awesome-Efficient-Reasoning-LLMs
README
# Awesome-Efficient-Reasoning-LLMs
[](https://arxiv.org/abs/2503.16419)
## 📢 News
- **March 20, 2025**: We release the first survey for efficient reasoning of LLMs "[Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models](https://arxiv.org/abs/2503.16419)".
Feel free to cite, contribute, or open a pull request to add recent related papers!
In this paper, we present the first structured survey that systematically investigates and organizes the current progress in achieving **efficient reasoning in LLMs**.
## 📊 Taxonomy
Below is a taxonomy graph summarizing the current landscape of efficient reasoning research for LLMs:
---
## 📚 Table of Contents
- [Awesome-Efficient-Reasoning-LLM](#awesome-efficient-reasoning-llm)
- **Model-based Efficient Reasoning**
- [Section I: RL with Length Reward Design](#section-i-rl-with-length-reward-design)
- [Section II: SFT with Variable-Length CoT Data](#section-ii-sft-with-variable-length-cot-data)
- **Reasoning Output-based Efficient Reasoning**
- [Section III: Compressing Reasoning Steps into Fewer Latent Representation](#section-iii-compressing-reasoning-steps-into-fewer-latent-representation)
- [Section IV: Dynamic Reasoning Paradigm during Inference](#section-iv-dynamic-reasoning-paradigm-during-inference)
- **Input Prompt-based Efficient Reasoning**
- [Section V: Prompt-Guided Efficient Reasoning](#section-v-prompt-guided-efficient-reasoning)
- [Section VI: Prompts Attribute-Driven Reasoning Routing](#section-vi-prompts-attribute-driven-reasoning-routing)
- **Reasoning Abilities with Efficient Data and Small Language Models**
- [Section VII: Reasoning Abilities via Efficient Training Data and Model Compression](#section-vii-reasoning-abilities-via-efficient-training-data-and-model-compression)
- **Evaluation and Benchmark**
- [Section VIII: Evaluation and Benchmark](#section-viii-evaluation-and-benchmark)
---
## Section I: RL with Length Reward Design
* Demystifying Long Chain-of-Thought Reasoning in LLMs [[Paper]](https://arxiv.org/pdf/2502.03373) 
* O1-Pruner: Length-Harmonizing Fine-Tuning for O1-Like Reasoning Pruning [[Paper]](https://arxiv.org/pdf/2501.12570) 
* Kimi k1.5: Scaling Reinforcement Learning with LLMs [[Paper]](https://arxiv.org/pdf/2501.12599) 
* Training Language Models to Reason Efficiently [[Paper]](https://arxiv.org/pdf/2502.04463) 
* L1: Controlling How Long A Reasoning Model Thinks With Reinforcement Learning [[Paper]](https://www.arxiv.org/pdf/2503.04697) 
* DAST: Difficulty-Adaptive Slow-Thinking for Large Reasoning Models [[Paper]](https://arxiv.org/pdf/2503.04472) 
* Optimizing Test-Time Compute via Meta Reinforcement Fine-Tuning [[Paper]](https://arxiv.org/pdf/2503.07572) 
## Section II: SFT with Variable-Length CoT Data
* TokenSkip: Controllable Chain-of-Thought Compression in LLMs [[Paper]](https://arxiv.org/pdf/2502.12067) 
* C3oT: Generating Shorter Chain-of-Thought without Compromising Effectiveness [[Paper]](https://arxiv.org/pdf/2412.11664) 
* CoT-Valve: Length-Compressible Chain-of-Thought Tuning [[Paper]](https://arxiv.org/pdf/2502.09601) 
* Self-Training Elicits Concise Reasoning in Large Language Models [[Paper]](https://arxiv.org/pdf/2502.20122) 
* Distilling System 2 into System 1 [[Paper]](https://arxiv.org/pdf/2407.06023) 
* Can Language Models Learn to Skip Steps? [[Paper]](https://arxiv.org/pdf/2411.01855) 
* Stepwise Perplexity-Guided Refinement for Efficient Chain-of-Thought Reasoning in Large Language Models [[Paper]](https://arxiv.org/pdf/2502.13260) 
## Section III: Compressing Reasoning Steps into Fewer Latent Representation
* Training Large Language Models to Reason in a Continuous Latent Space [[Paper]](https://arxiv.org/pdf/2412.06769) 
* Compressed Chain of Thought: Efficient Reasoning through Dense Representations [[Paper]](https://arxiv.org/pdf/2412.13171) 
* Efficient Reasoning with Hidden Thinking (MLLM) [[Paper]](https://arxiv.org/pdf/2501.19201) 
* SoftCoT: Soft Chain-of-Thought for Efficient Reasoning with LLMs [[Paper]](https://arxiv.org/pdf/2502.12134) 
* Token Assorted: Mixing Latent and Text Tokens for Improved Language Model Reasoning [[Paper]](https://arxiv.org/pdf/2502.03275) 
* Reasoning with Latent Thoughts: On the Power of Looped Transformers [[Paper]](https://arxiv.org/pdf/2502.17416) 
* CODI: Compressing Chain-of-Thought into Continuous Space via Self-Distillation [[Paper]](https://arxiv.org/pdf/2502.21074) 
## Section IV: Dynamic Reasoning Paradigm during Inference
* Efficiently Serving LLM Reasoning Programs with Certaindex [[Paper]](https://arxiv.org/pdf/2412.20993) 
* When More is Less: Understanding Chain-of-Thought Length in LLMs [[Paper]](https://arxiv.org/pdf/2502.07266) 
* Sketch-of-Thought: Efficient LLM Reasoning with Adaptive Cognitive-Inspired Sketching [[Paper]](https://arxiv.org/pdf/2503.05179) 
* Reward-Guided Speculative Decoding for Efficient LLM Reasoning [[Paper]](https://arxiv.org/pdf/2501.19324) 
* Fast Best-of-N Decoding via Speculative Rejection [[Paper]](https://arxiv.org/pdf/2410.20290) 
* FastMCTS: A Simple Sampling Strategy for Data Synthesis [[Paper]](https://www.arxiv.org/pdf/2502.11476) 
* Dynamic Parallel Tree Search for Efficient LLM Reasoning [[Paper]](https://arxiv.org/pdf/2502.16235) 
* Sampling-Efficient Test-Time Scaling: Self-Estimating the Best-of-N Sampling in Early Decoding [[Paper]](https://arxiv.org/pdf/2503.01422) 
* LightThinker: Thinking Step-by-Step Compression (training LLMs to compress thoughts into gist tokens) [[Paper]](https://arxiv.org/pdf/2502.15589) 
* InftyThink: Breaking the Length Limits of Long-Context Reasoning in Large Language Models [[Paper]](https://www.arxiv.org/pdf/2503.06692) 
## Section V: Prompt-Guided Efficient Reasoning
* Token-Budget-Aware LLM Reasoning [[Paper]](https://arxiv.org/pdf/2412.18547) 
* Chain of Draft: Thinking Faster by Writing Less [[Paper]](https://arxiv.org/pdf/2502.18600) 
* How Well do LLMs Compress Their Own Chain-of-Thought? A Token Complexity Approach [[Paper]](https://arxiv.org/pdf/2503.01141) 
* The Benefits of a Concise Chain of Thought on Problem-Solving in Large Language Models [[Paper]](https://arxiv.org/pdf/2401.05618) 
## Section VI: Prompts Attribute-Driven Reasoning Routing
* Claude 3.7 Sonnet and Claude Code [[website]](https://www.anthropic.com/news/claude-3-7-sonnet) 
* Sketch-of-Thought: Efficient LLM Reasoning with Adaptive Cognitive-Inspired Sketching [[Paper]](https://arxiv.org/pdf/2503.05179) 
* Learning to Route LLMs with Confidence Tokens [[Paper]](https://arxiv.org/pdf/2410.13284) 
* Confident or Seek Stronger: Exploring Uncertainty-Based On-device LLM Routing From Benchmarking to Generalization [[Paper]](https://arxiv.org/pdf/2502.04428) 
* RouteLLM: Learning to Route LLMs with Preference Data [[Paper]](https://arxiv.org/pdf/2406.18665) 
## Section VII: Reasoning Abilities via Efficient Training Data and Model Compression
* LIMO: Less is More for Reasoning [[Paper]](https://arxiv.org/pdf/2502.03387) 
* s1: Simple test-time scaling [[Paper]](https://arxiv.org/pdf/2501.19393) 
* S2R: Teaching LLMs to Self-verify and Self-correct via Reinforcement Learning [[Paper]](https://arxiv.org/pdf/2502.12853) 
* Light-R1: Curriculum SFT, DPO and RL for Long COT from Scratch and Beyond [[Paper]](https://arxiv.org/pdf/2503.10460) 
* Small Models Struggle to Learn from Strong Reasoners [[Paper]](https://arxiv.org/pdf/2502.12143) 
* Towards Reasoning Ability of Small Language Models [[Paper]](https://arxiv.org/pdf/2502.11569) 
* Mixed Distillation Helps Smaller Language Models Reason Better [[Paper]](https://arxiv.org/pdf/2312.10730) 
* Small language models need strong verifiers to self-correct reasoning [[Paper]](https://arxiv.org/pdf/2404.17140) 
* Teaching Small Language Models Reasoning through Counterfactual Distillation [[Paper]](https://aclanthology.org/2024.emnlp-main.333.pdf) 
* Improving Mathematical Reasoning Capabilities of Small Language Models via Feedback-Driven Distillation [[Paper]](https://arxiv.org/pdf/2411.14698) 
* Probe then retrieve and reason: Distilling probing and reasoning capabilities into smaller language models [[Paper]](https://aclanthology.org/2024.lrec-main.1140.pdf) 
* Distilling Reasoning Ability from Large Language Models with Adaptive Thinking [[Paper]](https://arxiv.org/pdf/2404.09170) 
* SKIntern: Internalizing Symbolic Knowledge for Distilling Better CoT Capabilities into Small Language Models [[Paper]](https://arxiv.org/pdf/2409.13183) 
## Section VIII: Evaluation and Benchmark
* Can 1B LLM Surpass 405B LLM? Rethinking Compute-Optimal Test-Time Scaling [[Paper]](https://arxiv.org/pdf/2502.06703) 
* The Danger of Overthinking: Examining the Reasoning-Action Dilemma in Agentic Tasks [[Paper]](https://arxiv.org/pdf/2502.08235) 
* Inference-Time Computations for LLM Reasoning and Planning: A Benchmark and Insights [[Paper]](https://arxiv.org/pdf/2502.12521) 
## Citation
If you find this work useful, welcome to cite us.
```bib
@article{sui2025stop,
title={Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models},
author={Sui, Yang and Chuang, Yu-Neng and Wang, Guanchu and Zhang, Jiamu and Zhang, Tianyi and Yuan, Jiayi and Liu, Hongyi and Wen, Andrew and Chen, Hanjie and Hu, Xia and others},
journal={arXiv preprint arXiv:2503.16419},
year={2025}
}
```
## Acknowledgment
> 🧩 *Layout inspired by [zzli2022/Awesome-System2-Reasoning-LLM](https://github.com/zzli2022/Awesome-System2-Reasoning-LLM). Many thanks for the great structure!*