mlsanityData to 7 Oct 2026

Self-Training Elicits Concise Reasoning in Large Language Models

arXiv 2502.20122 ↗First posted 27 Feb 2025cs.CL cs.AI cs.LGcited by 97

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

also in 1 more
Abstract (arXiv)

Chain-of-thought (CoT) reasoning has enabled large language models (LLMs) to utilize additional computation through intermediate tokens to solve complex tasks. However, we posit that typical reasoning traces contain many redundant tokens, incurring extraneous inference costs. Upon examination of the output distribution of current LLMs, we find evidence on their latent ability to reason more concisely, relative to their default behavior. To elicit this capability, we propose simple fine-tuning methods which leverage self-generated concise reasoning paths obtained by best-of-N sampling and few-shot conditioning, in task-specific settings. Our combined method achieves a 30% reduction in output tokens on average, across five model families on GSM8K and MATH, while maintaining average accuracy. By exploiting the fundamental stochasticity and in-context learning capabilities of LLMs, our self-training approach robustly elicits concise reasoning on a wide range of models, including those with extensive post-training. Code is available at https://github.com/TergelMunkhbat/concise-reasoning

Key facts. Cited by 97 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 30 of its 45 references that are in the map. First posted on arXiv 27 Feb 2025.

Builds on

30 of 45 references.

Closest ancestors · 5

Same direction, newest first

  1. 2025Demystifying Long Chain-of-Thought Reasoning in LLMs5 Feb
  2. 2025O1-Pruner: Length-Harmonizing Fine-Tuning for O1-Like Reasoning Pruning22 Jan
  3. 2024Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs30 Dec
  4. 2024Training Language Models to Self-Correct via Reinforcement Learning19 Sep
  5. 2024Concise Thoughts: Impact of Output Length on LLM Reasoning and Cost29 Jul

Other references · 25

  1. 2020Language Models are Few-Shot Learnerscited by 32,199
  2. 2024The Llama 3 Herd of Modelscited by 14,192
  3. 2021Training Verifiers to Solve Math Word Problemscited by 9,743
  4. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  5. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
  6. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,203
  7. 2022Large Language Models are Zero-Shot Reasonerscited by 4,323
  8. 2023Let's Verify Step by Stepcited by 3,598
  9. 2019The Curious Case of Neural Text Degenerationcited by 2,881
  10. 2024Gemma 2: Improving Open Language Models at a Practical Sizecited by 1,928
  11. 2024MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmarkcited by 1,758
  12. 2023LIMA: Less Is More for Alignmentcited by 1,080
13 more
  1. 2024Qwen2.5-Math Technical Report: Toward Mathematical Expert Model via Self-Improvementcited by 1,057
  2. 2021Show Your Work: Scratchpads for Intermediate Computation with Language Modelscited by 871
  3. 2024Taming Throughput-Latency Tradeoff in LLM Inference with Sarathi-Servecited by 507
  4. 2022Complexity-Based Prompting for Multi-Step Reasoningcited by 387
  5. 2024Step-DPO: Step-wise Preference Optimization for Long-chain Reasoning of LLMscited by 265
  6. 2024To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoningcited by 257
  7. 2024The Impact of Reasoning Step Length on Large Language Modelscited by 152
  8. 2024Physics of Language Models: Part 2.2, How to Learn From Mistakes on Grade-School Math Problemscited by 49
  9. 2024Block Transformer: Global-to-Local Language Modeling for Fast Inferencecited by 32
  10. 2024Over-Reasoning and Redundant Calculation of Large Language Modelscited by 27
  11. 2024On the Representational Capacity of Neural Language Models with Chain-of-Thought Reasoningcited by 26
  12. 2024I Learn Better If You Speak My Language: Understanding the Superior Performance of Fine-Tuning Large Language Models with LLM-Generated Responsescited by 13
  13. 2024Chain-of-Probe: Examining the Necessity and Accuracy of CoT Step-by-Stepcited by 8

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 67; 2026: 30.

97 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. 2025Dynamic Early Exit in Reasoning Modelscited by 217
  4. 2025Reasoning Models Can Be Effective Without Thinkingcited by 182
  5. 2025Reasoning Models Know When They're Right: Probing Hidden States for Self-Verificationcited by 154
  6. 2025A Survey of Efficient Reasoning for Large Reasoning Models: Language, Multimodality, and Beyondcited by 131
  7. 2025A Survey of Context Engineering for Large Language Modelscited by 120
  8. 2025Between Underthinking and Overthinking: An Empirical Study of Reasoning Length and correctness in LLMscited by 115
  9. 2025Efficient Reasoning Models: A Surveycited by 93
  10. 2025S-GRPO: Early Exit via Reinforcement Learning in Reasoning Modelscited by 87
  11. 2025AdaCoT: Pareto-Optimal Adaptive Chain-of-Thought Triggering via Reinforcement Learningcited by 86
  12. 2025Unlocking Efficient Long-to-Short LLM Reasoning with Model Mergingcited by 58
38 more
  1. 2025Efficient Inference for Large Reasoning Models: A Surveycited by 58
  2. 2025AdaCtrl: Towards Adaptive and Controllable Reasoning via Difficulty-Aware Budgetingcited by 50
  3. 2025SelfBudgeter: Adaptive Token Allocation for Efficient LLM Reasoningcited by 47
  4. 2025Answer Convergence as a Signal for Early Stopping in Reasoningcited by 38
  5. 2025Don't Overthink It: A Survey of Efficient R1-style Large Reasoning Modelscited by 36
  6. 2025ConCISE: Confidence-guided Compression in Step-by-step Efficient Reasoningcited by 33
  7. 2025Reasoning on a Budget: A Survey of Adaptive and Controllable Test-Time Compute in LLMscited by 31
  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. 2025Let LRMs Break Free from Overthinking via Self-Braking Tuningcited by 27
  11. 2025VeriThinker: Learning to Verify Makes Reasoning Model Efficientcited by 24
  12. 2025A Comprehensive Survey on Trustworthiness in Reasoning with Large Language Modelscited by 22
  13. 2025Incentivizing Dual Process Thinking for Efficient Large Language Model Reasoningcited by 21
  14. 2025Thinking Fast and Right: Balancing Accuracy and Reasoning Length with Adaptive Rewardscited by 20
  15. 2026Does Your Reasoning Model Implicitly Know When to Stop Thinking?cited by 17
  16. 2025Not All Thoughts are Generated Equal: Efficient LLM Reasoning via Multi-Turn Reinforcement Learningcited by 16
  17. 2025Reasoning Path Compression: Compressing Generation Trajectories for Efficient LLM Reasoningcited by 16
  18. 2025Causal Sufficiency and Necessity Improves Chain-of-Thought Reasoningcited by 16
  19. 2025DRPO: Efficient Reasoning via Decoupled Reward Policy Optimizationcited by 16
  20. 2025Fast on the Easy, Deep on the Hard: Efficient Reasoning via Powered Length Penaltycited by 15
  21. 2025Generative AI Act II: Test Time Scaling Drives Cognition Engineeringcited by 14
  22. 2025TrimR: Verifier-based Training-Free Thinking Compression for Efficient Test-Time Scalingcited by 14
  23. 2025Activation-Guided Consensus Merging for Large Language Modelscited by 12
  24. 2025Sample Complexity and Representation Ability of Test-time Scaling Paradigmscited by 12
  25. 2025SynAdapt: Learning Adaptive Reasoning in Large Language Models via Synthetic Continuous Chain-of-Thoughtcited by 12
  26. 2025Reasoning Path Divergence: A New Metric and Curation Strategy to Unlock LLM Diverse Thinkingcited by 10
  27. 2025The Overthinker's DIET: Cutting Token Calories with DIfficulty-AwarE Trainingcited by 9
  28. 2025ConciseHint: Boosting Efficient Reasoning via Continuous Concise Hints during Generationcited by 9
  29. 2025SpecExit: Accelerating Large Reasoning Model via Speculative Exitcited by 9
  30. 2025Can LLMs Reason Abstractly Over Math Word Problems Without CoT? Disentangling Abstract Formulation From Arithmetic Computationcited by 8
  31. 2025How Far Are We from Optimal Reasoning Efficiency?cited by 8
  32. 2025Exploring and Exploiting the Inherent Efficiency within Large Reasoning Models for Self-Guided Efficiency Enhancementcited by 8
  33. 2026VTC-R1: Vision-Text Compression for Efficient Long-Context Reasoningcited by 8
  34. 2026FoE: Forest of Errors Makes the First Solution the Best in Large Reasoning Modelscited by 8
  35. 2025Don't "Overthink" Passage Reranking: Is Reasoning Truly Necessary?cited by 7
  36. 2026Neural Chain-of-Thought Search: Searching the Optimal Reasoning Path to Enhance Large Language Modelscited by 7
  37. 2025Reasoning Efficiently Through Adaptive Chain-of-Thought Compression: A Self-Optimizing Frameworkcited by 6
  38. 2025Are Large Reasoning Models Interruptible?cited by 6

50 most cited of 97 All 97 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025Lost at the Beginning of Reasoning
  2. 2025Think, Prune, Train, Improve: Scaling Reasoning without Scaling Models
  3. 2025Advancing Reasoning in Large Language Models: Promising Methods and Approaches
  4. 2026Learning from Partial Chain-of-Thought via Truncated-Reasoning Self-Distillation
  5. 2026Enhancing reasoning accuracy in large language models during inference time
  6. 2026Correct, Concise and Complete: Multi-stage Training For Adaptive Reasoning
  7. 2025LLMs Can Easily Learn to Reason from Demonstrations Structure, not content, is what matters!
  8. 2025Small Models Struggle to Learn from Strong Reasoners