mlsanityData to 7 Oct 2026

ThinkPrune: Pruning Long Chain-of-Thought of LLMs via Reinforcement Learning

arXiv 2504.01296 ↗First posted 2 Apr 2025cs.CLcited by 187

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

Abstract (arXiv)

We present ThinkPrune, a simple yet effective method for pruning the thinking length for long-thinking LLMs, which has been found to often produce inefficient and redundant thinking processes. Existing preliminary explorations of reducing thinking length primarily focus on forcing the thinking process to early exit, rather than adapting the LLM to optimize and consolidate the thinking process, and therefore the length-performance tradeoff observed so far is sub-optimal. To fill this gap, ThinkPrune offers a simple solution that continuously trains the long-thinking LLMs via reinforcement learning (RL) with an added token limit, beyond which any unfinished thoughts and answers will be discarded, resulting in a zero reward. To further preserve model performance, we introduce an iterative length pruning approach, where multiple rounds of RL are conducted, each with an increasingly more stringent token limit. We observed that ThinkPrune results in a remarkable performance-length tradeoff -- on the AIME24 dataset, the reasoning length of DeepSeek-R1-Distill-Qwen-1.5B can be reduced by half with only 2% drop in performance. We also observed that after pruning, the LLMs can bypass unnecessary steps while keeping the core reasoning process complete. Code is available at https://github.com/UCSB-NLP-Chang/ThinkPrune.

Key facts. Cited by 187 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 25 of its 33 references that are in the map. First posted on arXiv 2 Apr 2025.

Builds on

25 of 33 references.

Closest ancestors · 9

Same direction, newest first

  1. 2025Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models20 Mar
  2. 2025The First Few Tokens Are All You Need: An Efficient and Effective Unsupervised Prefix Fine-Tuning Method for Reasoning Models4 Mar
  3. 2025Revisiting the Test-Time Scaling of o1-like Models: Do they Truly Possess Test-Time Scaling Capabilities?17 Feb
  4. 2025Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach7 Feb
  5. 2025Competitive Programming with Large Reasoning Models3 Feb
  6. 2025s1: Simple test-time scaling31 Jan
  7. 2025Thoughts Are All Over the Place: On the Underthinking of o1-Like LLMs30 Jan
  8. 2024Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs30 Dec
  9. 2024V-STaR: Training Verifiers for Self-Taught Reasoners9 Feb

Other references · 16

  1. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  2. 2023Let's Verify Step by Stepcited by 3,598
  3. 2015Learning both Weights and Connections for Efficient Neural Networkscited by 1,998
  4. 2024HybridFlow: A Flexible and Efficient RLHF Frameworkcited by 1,980
  5. 2025Kimi k1.5: Scaling Reinforcement Learning with LLMscited by 1,193
  6. 2024OlympiadBench: A Challenging Benchmark for Promoting AGI with Olympiad-Level Bilingual Multimodal Scientific Problemscited by 1,099
  7. 2024Training Large Language Models to Reason in a Continuous Latent Spacecited by 763
  8. 2025SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wildcited by 505
  9. 2025LIMO: Less is More for Reasoningcited by 488
  10. 2025Process Reinforcement through Implicit Rewardscited by 478
  11. 2025Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Modelscited by 381
  12. 2025Cognitive Behaviors that Enable Self-Improving Reasoners, or, Four Habits of Highly Effective STaRscited by 373
4 more
  1. 2024Quiet-STaR: Language Models Can Teach Themselves to Think Before Speakingcited by 292
  2. 2025Training Language Models to Reason Efficientlycited by 242
  3. 2025CODI: Compressing Chain-of-Thought into Continuous Space via Self-Distillationcited by 216
  4. 2025OverThink: Slowdown Attacks on Reasoning LLMscited by 94

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 95; 2026: 92.

187 papers.

Sort:

By area:

  1. 2025Deep Think with Confidencecited by 199
  2. 2025The Landscape of Agentic Reinforcement Learning for LLMs: A Surveycited by 195
  3. 2025Efficient Reasoning Models: A Surveycited by 93
  4. 2025AdaCoT: Pareto-Optimal Adaptive Chain-of-Thought Triggering via Reinforcement Learningcited by 86
  5. 2025Sample More to Think Less: Group Filtered Policy Optimization for Concise Reasoningcited by 75
  6. 2025Learning When to Think: Shaping Adaptive Reasoning in R1-Style Models via Multi-Stage RLcited by 59
  7. 2025AdaThinkDrive: Adaptive Thinking via Reinforcement Learning for Autonomous Drivingcited by 57
  8. 2025Does Thinking More always Help? Mirage of Test-Time Scaling in Reasoning Modelscited by 52
  9. 2025Just Enough Thinking: Efficient Reasoning with Adaptive Length Penalties Reinforcement Learningcited by 51
  10. 2025AdaCtrl: Towards Adaptive and Controllable Reasoning via Difficulty-Aware Budgetingcited by 50
  11. 2025Token Reduction Should Go Beyond Efficiency in Generative Models -- From Vision, Language to Multimodalitycited by 49
  12. 2026CRISP: Compressed Reasoning via Iterative Self-Policy Distillationcited by 49
38 more
  1. 2025ARM: Adaptive Reasoning Modelcited by 38
  2. 2025Answer Convergence as a Signal for Early Stopping in Reasoningcited by 38
  3. 2025Optimizing Length Compression in Large Reasoning Modelscited by 37
  4. 2025Making Slow Thinking Faster: Compressing LLM Chain-of-Thought via Step Entropycited by 37
  5. 2025Don't Overthink It: A Survey of Efficient R1-style Large Reasoning Modelscited by 36
  6. 2025DLER: Doing Length pEnalty Right - Incentivizing More Intelligence per Token via Reinforcement Learningcited by 34
  7. 2025Steering LLM Thinking with Budget Guidancecited by 30
  8. 2025R1-Compress: Long Chain-of-Thought Compression via Chunk Compression and Searchcited by 29
  9. 2025PixelThink: Towards Efficient Chain-of-Pixel Reasoningcited by 28
  10. 2025Stable Reinforcement Learning for Efficient Reasoningcited by 27
  11. 2025Promoting Efficient Reasoning with Verifiable Stepwise Rewardcited by 27
  12. 2025BudgetThinker: Empowering Budget-aware LLM Reasoning with Control Tokenscited by 27
  13. 2025VeriThinker: Learning to Verify Makes Reasoning Model Efficientcited by 24
  14. 2025HardTests: Synthesizing High-Quality Test Cases for LLM Codingcited by 22
  15. 2025Activation Steering for Chain-of-Thought Compressioncited by 22
  16. 2025Incentivizing Dual Process Thinking for Efficient Large Language Model Reasoningcited by 21
  17. 2025Thinking Fast and Right: Balancing Accuracy and Reasoning Length with Adaptive Rewardscited by 20
  18. 2025Thought calibration: Efficient and confident test-time scalingcited by 20
  19. 2025KTAE: A Model-Free Algorithm to Key-Tokens Advantage Estimation in Mathematical Reasoningcited by 19
  20. 2025HAPO: Training Language Models to Reason Concisely via History-Aware Policy Optimizationcited by 18
  21. 2025OThink-R1: Intrinsic Fast/Slow Thinking Mode Switching for Over-Reasoning Mitigationcited by 17
  22. 2026Does Your Reasoning Model Implicitly Know When to Stop Thinking?cited by 17
  23. 2025Not All Thoughts are Generated Equal: Efficient LLM Reasoning via Multi-Turn Reinforcement Learningcited by 16
  24. 2025DRPO: Efficient Reasoning via Decoupled Reward Policy Optimizationcited by 16
  25. 2026MEMENTO: Teaching LLMs to Manage Their Own Contextcited by 16
  26. 2025How do Transformers Learn Implicit Reasoning?cited by 15
  27. 2025Lost at the Beginning of Reasoningcited by 15
  28. 2025TrimR: Verifier-based Training-Free Thinking Compression for Efficient Test-Time Scalingcited by 14
  29. 2025ConciseRL: Conciseness-Guided Reinforcement Learning for Efficient Reasoning Modelscited by 14
  30. 2025Making Small Language Models Efficient Reasoners: Intervention, Supervision, Reinforcementcited by 13
  31. 2025Reconsidering Overthinking: Penalizing Internal and External Redundancy in CoT Reasoningcited by 13
  32. 2025Leash: Adaptive Length Penalty and Reward Shaping for Efficient Large Reasoning Modelcited by 13
  33. 2026Self-Compression of Chain-of-Thought via Multi-Agent Reinforcement Learningcited by 13
  34. 2025Activation-Guided Consensus Merging for Large Language Modelscited by 12
  35. 2025AutoL2S: Auto Long-Short Reasoning for Efficient Large Language Modelscited by 12
  36. 2025SABER: Switchable and Balanced Training for Efficient LLM Reasoningcited by 12
  37. 2025The Markovian Thinker: Architecture-Agnostic Linear Scaling of Reasoningcited by 12
  38. 2025ReasonIF: Large Reasoning Models Fail to Follow Instructions During Reasoningcited by 12

50 most cited of 187 All 187 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025Thinkless: LLM Learns When to Think
  2. 2025Walk Before You Run! Concise LLM Reasoning via Reinforcement Learning
  3. 2026OS-Pruner: Pruning Chains-of-Thought of Reasoning Models via Optimal Stopping
  4. 2025When More is Less: Understanding Chain-of-Thought Length in LLMs
  5. 2025Demystifying Long Chain-of-Thought Reasoning in LLMs
  6. 2025RL of Thoughts: Navigating LLM Reasoning with Inference-time Reinforcement Learning
  7. 2025Thinking-Free Policy Initialization Makes Distilled Reasoning Models More Effective and Efficient Reasoners
  8. 2025Not All Tokens Matter: Towards Efficient LLM Reasoning via Token Significance in Reinforcement Learning