mlsanityData to 7 Oct 2026

CoT-Valve: Length-Compressible Chain-of-Thought Tuning

arXiv 2502.09601 ↗First posted 13 Feb 2025cs.AI cs.CLcited by 167

On the map: LLM reasoning → Test-time compute and efficient thinking

Abstract (arXiv)

Chain-of-Thought significantly enhances a model's reasoning capability, but it also comes with a considerable increase in inference costs due to long chains. With the observation that the reasoning path can be easily compressed under easy tasks but struggle on hard tasks, we explore the feasibility of elastically controlling the length of reasoning paths with only one model, thereby reducing the inference overhead of reasoning models dynamically based on task difficulty. We introduce a new tuning and inference strategy named CoT-Valve, designed to allow models to generate reasoning chains of varying lengths. To achieve this, we propose to identify a direction in the parameter space that, when manipulated, can effectively control the length of generated CoT. Moreover, we show that this property is valuable for compressing the reasoning chain. We construct datasets with chains from long to short for the same questions and explore two enhanced strategies for CoT-Valve: (1) a precise length-compressible CoT tuning method, and (2) a progressive chain length compression approach. Our experiments show that CoT-Valve successfully enables controllability and compressibility of the chain and shows better performance than the prompt-based control. We applied this method to QwQ-32B-Preview, reducing reasoning chains on GSM8K from 741 to 225 tokens with a minor performance drop (95.07% to 94.92%) and on AIME from 6827 to 4629 tokens, with only one additional incorrect answer.

Key facts. Cited by 167 papers in the map (data through 7 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 27 of its 38 references that are in the map. First posted on arXiv 13 Feb 2025.

Builds on

27 of 38 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025O1-Pruner: Length-Harmonizing Fine-Tuning for O1-Like Reasoning Pruning22 Jan
  2. 2024Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs30 Dec
  3. 2024Compressed Chain of Thought: Efficient Reasoning Through Dense Representations17 Dec
  4. 2024C3oT: Generating Shorter Chain-of-Thought without Compromising Effectiveness16 Dec

Other references · 23

  1. 2024The Llama 3 Herd of Modelscited by 14,192
  2. 2021Training Verifiers to Solve Math Word Problemscited by 9,743
  3. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
  4. 2023Let's Verify Step by Stepcited by 3,598
  5. 2021WebGPT: Browser-assisted question-answering with human feedbackcited by 1,722
  6. 2024SimPO: Simple Preference Optimization with a Reference-Free Rewardcited by 951
  7. 2023Math-Shepherd: Verify and Reinforce LLMs Step-by-step without Human Annotationscited by 895
  8. 2024Training Large Language Models to Reason in a Continuous Latent Spacecited by 763
  9. 2024Phi-4 Technical Reportcited by 675
  10. 2019Linear Mode Connectivity and the Lottery Ticket Hypothesiscited by 627
  11. 2025LIMO: Less is More for Reasoningcited by 488
  12. 2024Improve Mathematical Reasoning in Language Models by Automated Process Supervisioncited by 426
11 more
  1. 2025The Lessons of Developing Process Reward Models in Mathematical Reasoningcited by 402
  2. 2025rStar-Math: Small LLMs Can Master Math Reasoning with Self-Evolved Deep Thinkingcited by 308
  3. 2024To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoningcited by 257
  4. 2024From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Stepcited by 225
  5. 2025Inference-Time Scaling for Diffusion Models beyond Scaling Denoising Stepscited by 225
  6. 2024Accessing GPT-4 level Mathematical Olympiad Solutions via Monte Carlo Tree Self-refine with LLaMa-3 8Bcited by 156
  7. 2024Distilling System 2 into System 1cited by 152
  8. 2024Can Language Models Learn to Skip Steps?cited by 59
  9. 2023Are Machine Rationales (Not) Useful to Humans? Measuring and Improving Human Utility of Free-Text Rationalescited by 37
  10. 2024Understanding Chain-of-Thought in LLMs through Information Theorycited by 36
  11. 2024Break the Chain: Large Language Models Can be Shortcut Reasonerscited by 32

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 104; 2026: 63.

167 papers.

Sort:

By area:

  1. 2025Stop Overthinking: A Survey on Efficient Reasoning for Large Language Modelscited by 456
  2. 2025Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Modelscited by 381
  3. 2025From System 1 to System 2: A Survey of Reasoning Large Language Modelscited by 319
  4. 2025Dynamic Early Exit in Reasoning Modelscited by 217
  5. 2025Reasoning Models Can Be Effective Without Thinkingcited by 182
  6. 2025DAST: Difficulty-Adaptive Slow-Thinking for Large Reasoning Modelscited by 154
  7. 2025A Survey of Efficient Reasoning for Large Reasoning Models: Language, Multimodality, and Beyondcited by 131
  8. 2025Safety Tax: Safety Alignment Makes Your Large Reasoning Models Less Reasonablecited by 120
  9. 2025Between Underthinking and Overthinking: An Empirical Study of Reasoning Length and correctness in LLMscited by 115
  10. 2025Thinkless: LLM Learns When to Thinkcited by 107
  11. 2025Efficient Reasoning Models: A Surveycited by 93
  12. 2025S-GRPO: Early Exit via Reinforcement Learning in Reasoning Modelscited by 87
38 more
  1. 2025AdaCoT: Pareto-Optimal Adaptive Chain-of-Thought Triggering via Reinforcement Learningcited by 86
  2. 2025Wait, We Don't Need to "Wait"! Removing Thinking Tokens Improves Reasoning Efficiencycited by 65
  3. 2025Learning When to Think: Shaping Adaptive Reasoning in R1-Style Models via Multi-Stage RLcited by 59
  4. 2025Unlocking Efficient Long-to-Short LLM Reasoning with Model Mergingcited by 58
  5. 2025Efficient Inference for Large Reasoning Models: A Surveycited by 58
  6. 2025InftyThink: Breaking the Length Limits of Long-Context Reasoning in Large Language Modelscited by 53
  7. 2025AdaCtrl: Towards Adaptive and Controllable Reasoning via Difficulty-Aware Budgetingcited by 50
  8. 2025SelfBudgeter: Adaptive Token Allocation for Efficient LLM Reasoningcited by 47
  9. 2025Implicit Reasoning in Large Language Models: A Comprehensive Surveycited by 42
  10. 2025Optimizing Length Compression in Large Reasoning Modelscited by 37
  11. 2025Making Slow Thinking Faster: Compressing LLM Chain-of-Thought via Step Entropycited by 37
  12. 2025Don't Overthink It: A Survey of Efficient R1-style Large Reasoning Modelscited by 36
  13. 2025DLER: Doing Length pEnalty Right - Incentivizing More Intelligence per Token via Reinforcement Learningcited by 34
  14. 2025ConCISE: Confidence-guided Compression in Step-by-step Efficient Reasoningcited by 33
  15. 2025Think or Not? Exploring Thinking Efficiency in Large Reasoning Models via an Information-Theoretic Lenscited by 32
  16. 2025Reasoning on a Budget: A Survey of Adaptive and Controllable Test-Time Compute in LLMscited by 31
  17. 2025DRP: Distilled Reasoning Pruning with Skill-aware Step Decomposition for Efficient Large Reasoning Modelscited by 29
  18. 2025R1-Compress: Long Chain-of-Thought Compression via Chunk Compression and Searchcited by 29
  19. 2025Entropy-based Exploration Conduction for Multi-step Reasoningcited by 28
  20. 2025PixelThink: Towards Efficient Chain-of-Pixel Reasoningcited by 28
  21. 2025Let LRMs Break Free from Overthinking via Self-Braking Tuningcited by 27
  22. 2025Promoting Efficient Reasoning with Verifiable Stepwise Rewardcited by 27
  23. 2025Shorten After You're Right: Lazy Length Penalties for Reasoning RLcited by 25
  24. 2025VeriThinker: Learning to Verify Makes Reasoning Model Efficientcited by 24
  25. 2025Thought Manipulation: External Thought Can Be Efficient for Large Reasoning Modelscited by 23
  26. 2025VerifyBench: A Systematic Benchmark for Evaluating Reasoning Verifiers Across Domainscited by 22
  27. 2025A Survey of Token Compression for Efficient Multimodal Large Language Modelscited by 22
  28. 2025A Comprehensive Survey on Trustworthiness in Reasoning with Large Language Modelscited by 22
  29. 2025Incentivizing Dual Process Thinking for Efficient Large Language Model Reasoningcited by 21
  30. 2025A Survey of Slow Thinking-based Reasoning LLMs using Reinforced Learning and Inference-time Scaling Lawcited by 18
  31. 2025Towards Concise and Adaptive Thinking in Large Reasoning Models: A Surveycited by 18
  32. 2025The Evolution of Thought: Tracking LLM Overthinking via Reasoning Dynamics Analysiscited by 18
  33. 2025OThink-R1: Intrinsic Fast/Slow Thinking Mode Switching for Over-Reasoning Mitigationcited by 17
  34. 2026Does Your Reasoning Model Implicitly Know When to Stop Thinking?cited by 17
  35. 2025Not All Thoughts are Generated Equal: Efficient LLM Reasoning via Multi-Turn Reinforcement Learningcited by 16
  36. 2025Causal Sufficiency and Necessity Improves Chain-of-Thought Reasoningcited by 16
  37. 2025ARES: Multimodal Adaptive Reasoning via Difficulty-Aware Token-Level Entropy Shapingcited by 16
  38. 2025Fast on the Easy, Deep on the Hard: Efficient Reasoning via Powered Length Penaltycited by 15

50 most cited of 167 All 167 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025Can Reasoning Path still be Effective as Input? Bridging Post-Reasoning to Chain-of-Thought Compression
  2. 2026Towards Efficient Large Language Reasoning Models via Extreme-Ratio Chain-of-Thought Compression
  3. 2025From Long to Lean: Performance-aware and Adaptive Chain-of-Thought Compression via Multi-round Refinement
  4. 2025Activation Steering for Chain-of-Thought Compression
  5. 2026Chain Of Thought Compression: A Theoretical Analysis
  6. 2025Adaptive Deep Reasoning: Triggering Deep Thinking When Needed
  7. 2026Zipping the Thought: When and How Compressed Reasoning Data Works in LLM Post-Training
  8. 2026TERMINATOR: Learning Optimal Exit Points for Early Stopping in Chain-of-Thought Reasoning