mlsanityData to 14 Oct 2026

THOUGHTTERMINATOR: Benchmarking, Calibrating, and Mitigating Overthinking in Reasoning Models

arXiv 2504.13367 ↗First posted 17 Apr 2025cs.CLcited by 35

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

Abstract (arXiv)

Reasoning models have demonstrated impressive performance on difficult tasks that traditional language models struggle at. However, many are plagued with the problem of overthinking--generating large amounts of unnecessary tokens which don't improve accuracy on a question. We introduce approximate measures of problem-level difficulty and demonstrate that a clear relationship between problem difficulty and optimal token spend exists, and evaluate how well calibrated a variety of reasoning models are in terms of efficiently allocating the optimal token count. We find that in general, reasoning models are poorly calibrated, particularly on easy problems. To evaluate calibration on easy questions we introduce DUMB500, a dataset of extremely easy math, reasoning, code, and task problems, and jointly evaluate reasoning model on these simple examples and extremely difficult examples from existing frontier benchmarks on the same task domain. Finally, we introduce THOUGHTTERMINATOR, a training-free black box decoding technique that significantly improves reasoning model calibration.

Key facts. Cited by 35 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 7 of its 27 references that are in the map. First posted on arXiv 17 Apr 2025.

Builds on

7 of 27 references.

Closest ancestors · 1

Same direction, newest first

  1. 2024Efficiently Scaling LLM Reasoning with Certaindex30 Dec

Other references · 6

  1. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  2. 2024Qwen2.5 Technical Reportcited by 4,942
  3. 2024OpenAI o1 System Cardcited by 1,244
  4. 2024From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Stepcited by 235
  5. 2023DyVal: Dynamic Evaluation of Large Language Models for Reasoning Taskscited by 101
  6. 2024Can Language Models Learn to Skip Steps?cited by 59

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 22; 2026: 13.

35 papers.

Sort:

By area:

  1. 2025Don't Overthink it. Preferring Shorter Thinking Chains for Improved LLM Reasoningcited by 85
  2. 2025Budget-Aware Tool Use Enables Effective Agent Scalingcited by 44
  3. 2025ARM: Adaptive Reasoning Modelcited by 39
  4. 2025Mitigating Overthinking in Large Reasoning Models via Manifold Steeringcited by 33
  5. 2025AlphaOne: Reasoning Models Thinking Slow and Fast at Test Timecited by 29
  6. 2025BudgetThinker: Empowering Budget-aware LLM Reasoning with Control Tokenscited by 29
  7. 2025Shorten After You're Right: Lazy Length Penalties for Reasoning RLcited by 25
  8. 2025THINK-Bench: Evaluating Thinking Efficiency and Chain-of-Thought Quality of Large Reasoning Modelscited by 19
  9. 2025The Evolution of Thought: Tracking LLM Overthinking via Reasoning Dynamics Analysiscited by 18
  10. 2025Real-Time Reasoning Agents in Evolving Environmentscited by 15
  11. 2025Fractional Reasoning via Latent Steering Vectors Improves Inference Time Computecited by 14
  12. 2026Inference-Time Budget Control for LLM Search Agentscited by 13
23 more
  1. 2025OckBench: Measuring the Efficiency of LLM Reasoningcited by 11
  2. 2026Over-Searching in Search-Augmented Large Language Modelscited by 11
  3. 2025Do LLMs Really Need 10+ Thoughts for "Find the Time 1000 Days Later"? Towards Structural Understanding of LLM Overthinkingcited by 10
  4. 2025How Far Are We from Optimal Reasoning Efficiency?cited by 8
  5. 2026Neural Chain-of-Thought Search: Searching the Optimal Reasoning Path to Enhance Large Language Modelscited by 8
  6. 2025Asynchronous Reasoning: Training-Free Interactive Thinking LLMscited by 7
  7. 2025Do LLMs Overthink Basic Math Reasoning? Benchmarking the Accuracy-Efficiency Tradeoff in Language Modelscited by 5
  8. 2025Are Language Models Efficient Reasoners? A Perspective from Logic Programmingcited by 4
  9. 2026Intern-S2-Preview: Scientific Agentic Foundation Modelcited by 4
  10. 2025MixReasoning: Switching Modes to Thinkcited by 3
  11. 2025Optimizing Reasoning Efficiency through Prompt Difficulty Predictioncited by 3
  12. 2026Reasoning While Asking: Transforming Reasoning Large Language Models from Passive Solvers to Proactive Inquirerscited by 3
  13. 2026Reliable Chain-of-Thought via Prefix Consistencycited by 3
  14. 2025Improving the Distributional Alignment of LLMs using Supervisioncited by 2
  15. 2025Distractor Injection Attacks on Large Reasoning Models: Characterization and Defensecited by 2
  16. 2026From Passive Metric to Active Signal: The Evolving Role of Uncertainty Quantification in Large Language Modelscited by 2
  17. 2026AutoCRAT: Within-trajectory Joint Control of Stochasticity and Compute for LLM Reasoningcited by 2
  18. 2026Dynamic Rollout Editing for Reducing Overthinking in RL-Trained Reasoning Modelscited by 1
  19. 2025Step-Tagging: Toward controlling the generation of Language Reasoning Models through step monitoringcited by 0
  20. 2026Reasoning Traces Shape Outputs but Models Won't Say Socited by 0
  21. 2026TRACES: Tagging Reasoning Steps for Adaptive Cost-Efficient Early-Stoppingcited by 0
  22. 2026Solver-Hard Is Not Model-Hard: A Hardness-Controlled Diagnostic for LLM Constraint Reasoningcited by 0
  23. 2026DoubtLess: Training-Free Contrastive Decoding for Efficient Reasoningcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025DiffAdapt: Difficulty-Adaptive Reasoning for Token-Efficient LLM Inference
  2. 2025Thought calibration: Efficient and confident test-time scaling
  3. 2026Hint Tuning: Less Data Makes Better Reasoners
  4. 2025On the Bias of Next-Token Predictors Toward Systematically Inefficient Reasoning: A Shortest-Path Case Study
  5. 2025Plan and Budget: Effective and Efficient Test-Time Scaling on Reasoning Large Language Models
  6. 2024On Memorization of Large Language Models in Logical Reasoning
  7. 2025The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity
  8. 2025Long Is More Important Than Difficult for Training Reasoning Models