mlsanityData to 14 Oct 2026

Stop When Enough: Adaptive Early-Stopping for Chain-of-Thought Reasoning

arXiv 2510.10103 ↗First posted 11 Oct 2025cs.CLcited by 16

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

Abstract (arXiv)

Chain-of-Thought (CoT) reasoning has driven recent gains of large language models (LLMs) on reasoning-intensive tasks by externalizing intermediate steps. However, excessive or redundant reasoning -- so-called overthinking -- can increase inference costs and lead LLMs toward incorrect conclusions. In this paper, we present REFRAIN (underlineREFlective-underlineRedundancy for underlineAdaptive underlineINference), a training-free framework that adaptively determines when to stop reasoning to mitigate overthinking. REFRAIN integrates a two-stage stop discriminator to identify reflective yet redundant reasoning and a sliding-window Upper Confidence Bound (SW-UCB) multi-armed bandit controller to dynamically adjust stopping thresholds according to problem difficulty without supervision or fine-tuning. Across four representative benchmarks and two model families, REFRAIN reduces token usage by 20-55% while maintaining or improving accuracy compared to standard CoT prompting. Extensive ablation and robustness analyses demonstrate its stability across models, scorers, and prompt variations. In summary, our findings highlight when-to-stop as a new and practical axis of test-time scaling -- enabling models to reason not just more, but just enough.

Key facts. Cited by 16 papers in the map (data through 14 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 33 of its 41 references that are in the map. First posted on arXiv 11 Oct 2025.

Builds on

33 of 41 references.

Closest ancestors · 20

Same direction, newest first

  1. 2025Deep Think with Confidence21 Aug
  2. 2025gpt-oss-120b & gpt-oss-20b Model Card8 Aug
  3. 2025Efficient Reasoning for Large Reasoning Language Models via Certainty-Guided Reflection Suppression7 Aug
  4. 2025Inverse Scaling in Test-Time Compute19 Jul
  5. 2025SmartThinker: Learning to Compress and Preserve Reasoning by Step-Level Length Control6 Jul
  6. 2025AALC: Large Language Model Efficient Reasoning via Adaptive Accuracy-Length Control25 Jun
  7. 2025AlphaOne: Reasoning Models Thinking Slow and Fast at Test Time30 May
  8. 2025SoftCoT++: Test-Time Scaling with Soft Chain-of-Thought Reasoning16 May
  9. 2025ConCISE: Confidence-guided Compression in Step-by-step Efficient Reasoning8 May
  10. 2025Dynamic Early Exit in Reasoning Models22 Apr
  11. 2025Reasoning Models Can Be Effective Without Thinking14 Apr
  12. 2025Speculative Thinking: Enhancing Small-Model Reasoning with Large Model Guidance at Inference Time12 Apr
8 more
  1. 2025Missing Premise exacerbates Overthinking: Are Reasoning Models losing Critical Thinking Skill?9 Apr
  2. 2025Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models20 Mar
  3. 2025CoT-Valve: Length-Compressible Chain-of-Thought Tuning13 Feb
  4. 2025The Danger of Overthinking: Examining the Reasoning-Action Dilemma in Agentic Tasks12 Feb
  5. 2025Demystifying Long Chain-of-Thought Reasoning in LLMs5 Feb
  6. 2025s1: Simple test-time scaling31 Jan
  7. 2024Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs30 Dec
  8. 2024Escape Sky-high Cost: Early-stopping Self-Consistency for Multi-step Reasoning19 Jan

Other references · 13

  1. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  2. 2025Qwen3 Technical Reportcited by 8,820
  3. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  4. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
  5. 2019HuggingFace's Transformers: State-of-the-art Natural Language Processingcited by 2,218
  6. 2018CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledgecited by 1,822
  7. 2025The Lessons of Developing Process Reward Models in Mathematical Reasoningcited by 415
  8. 2025From System 1 to System 2: A Survey of Reasoning Large Language Modelscited by 320
  9. 2025Stepwise Perplexity-Guided Refinement for Efficient Chain-of-Thought Reasoning in Large Language Modelscited by 51
  10. 2025Is Chain-of-Thought Reasoning of LLMs a Mirage? A Data Distribution Lenscited by 43
  11. 2025DRP: Distilled Reasoning Pruning with Skill-aware Step Decomposition for Efficient Large Reasoning Modelscited by 30
  12. 2024Over-Reasoning and Redundant Calculation of Large Language Modelscited by 27
1 more
  1. 2024Large Language Models Struggle with Unreasonability in Math Problemscited by 18

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 4; 2026: 12.

16 papers.

Sort:

By area:

  1. 2025Mull-Tokens: Modality-Agnostic Latent Thinkingcited by 19
  2. 2025Asynchronous Reasoning: Training-Free Interactive Thinking LLMscited by 7
  3. 2026Probing the Trajectories of Reasoning Traces in Large Language Modelscited by 7
  4. 2026Stop When Reasoning Converges: Semantic-Preserving Early Exit for Reasoning Modelscited by 7
  5. 2026Towards a Mechanistic Understanding of Large Reasoning Models: A Survey of Training, Inference, and Failurescited by 6
  6. 2026From Chatbot to Digital Colleague: The Paradigm Shift Toward Persistent Autonomous AIcited by 6
  7. 2025Re-FORC: Adaptive Reward Prediction for Efficient Chain-of-Thought Reasoningcited by 4
  8. 2026Online Reasoning Calibration: Test-Time Training Enables Generalizable Conformal LLM Reasoningcited by 4
  9. 2025When to Think and When to Look: Uncertainty-Guided Lookbackcited by 3
  10. 2026interwhen: A Generalizable Framework for Steering Reasoning Models with Test-time Verificationcited by 2
  11. 2026Dynamic Rollout Editing for Reducing Overthinking in RL-Trained Reasoning Modelscited by 1
  12. 2026Shorter Reasoning, Earlier Answers? An Evaluation of Reasoning Interfacescited by 1
4 more
  1. 2026Optimal Stopping of Self-Refining Foundation Modelscited by 0
  2. 2026Divergent-Convergent Reasoning: Scaling Test-Time Compute through Structured Solution Synthesiscited by 0
  3. 2026Reason What Matters: Retrieval-Grounded Reasoning for Universal Multimodal Embeddingscited by 0
  4. 2026Budget Boundary Effects in Test-Time Mathematical Reasoningcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026OS-Pruner: Pruning Chains-of-Thought of Reasoning Models via Optimal Stopping
  2. 2026Correct, Concise and Complete: Multi-stage Training For Adaptive Reasoning
  3. 2025Lost at the Beginning of Reasoning
  4. 2025Early Stopping Chain-of-thoughts in Large Language Models
  5. 2025Logit-Entropy Adaptive Stopping Heuristic for Efficient Chain-of-Thought Reasoning
  6. 2026TERMINATOR: Learning Optimal Exit Points for Early Stopping in Chain-of-Thought Reasoning
  7. 2025Answer Convergence as a Signal for Early Stopping in Reasoning
  8. 2026Mitigating Overthinking in Large Reasoning Language Models via Reasoning Path Deviation Monitoring