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

arXiv: https://arxiv.org/abs/2504.13367 · HTML: https://mlsanity.com/paper/2504.13367/

## 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.

## Placed in

- LLM reasoning / [Test-time compute and efficient thinking](https://mlsanity.com/lr/test-time-compute-and-efficient-thinking/)

## Builds on (most cited first, 7 of 7)

- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [OpenAI o1 System Card](https://mlsanity.com/paper/2412.16720/) (2024)
- [From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step](https://mlsanity.com/paper/2405.14838/) (2024)
- [DyVal: Dynamic Evaluation of Large Language Models for Reasoning Tasks](https://mlsanity.com/paper/2309.17167/) (2023)
- [Can Language Models Learn to Skip Steps?](https://mlsanity.com/paper/2411.01855/) (2024)
- [Efficiently Scaling LLM Reasoning with Certaindex](https://mlsanity.com/paper/2412.20993/) (2024)

## Built on by (most cited first, 30 of 35)

- [Don't Overthink it. Preferring Shorter Thinking Chains for Improved LLM Reasoning](https://mlsanity.com/paper/2505.17813/) (2025)
- [Budget-Aware Tool Use Enables Effective Agent Scaling](https://mlsanity.com/paper/2511.17006/) (2025)
- [ARM: Adaptive Reasoning Model](https://mlsanity.com/paper/2505.20258/) (2025)
- [Mitigating Overthinking in Large Reasoning Models via Manifold Steering](https://mlsanity.com/paper/2505.22411/) (2025)
- [AlphaOne: Reasoning Models Thinking Slow and Fast at Test Time](https://mlsanity.com/paper/2505.24863/) (2025)
- [BudgetThinker: Empowering Budget-aware LLM Reasoning with Control Tokens](https://mlsanity.com/paper/2508.17196/) (2025)
- [Shorten After You're Right: Lazy Length Penalties for Reasoning RL](https://mlsanity.com/paper/2505.12284/) (2025)
- [THINK-Bench: Evaluating Thinking Efficiency and Chain-of-Thought Quality of Large Reasoning Models](https://mlsanity.com/paper/2505.22113/) (2025)
- [The Evolution of Thought: Tracking LLM Overthinking via Reasoning Dynamics Analysis](https://mlsanity.com/paper/2508.17627/) (2025)
- [Real-Time Reasoning Agents in Evolving Environments](https://mlsanity.com/paper/2511.04898/) (2025)
- [Fractional Reasoning via Latent Steering Vectors Improves Inference Time Compute](https://mlsanity.com/paper/2506.15882/) (2025)
- [Inference-Time Budget Control for LLM Search Agents](https://mlsanity.com/paper/2605.05701/) (2026)
- [OckBench: Measuring the Efficiency of LLM Reasoning](https://mlsanity.com/paper/2511.05722/) (2025)
- [Over-Searching in Search-Augmented Large Language Models](https://mlsanity.com/paper/2601.05503/) (2026)
- [Do LLMs Really Need 10+ Thoughts for "Find the Time 1000 Days Later"? Towards Structural Understanding of LLM Overthinking](https://mlsanity.com/paper/2510.07880/) (2025)
- [How Far Are We from Optimal Reasoning Efficiency?](https://mlsanity.com/paper/2506.07104/) (2025)
- [Neural Chain-of-Thought Search: Searching the Optimal Reasoning Path to Enhance Large Language Models](https://mlsanity.com/paper/2601.11340/) (2026)
- [Asynchronous Reasoning: Training-Free Interactive Thinking LLMs](https://mlsanity.com/paper/2512.10931/) (2025)
- [Do LLMs Overthink Basic Math Reasoning? Benchmarking the Accuracy-Efficiency Tradeoff in Language Models](https://mlsanity.com/paper/2507.04023/) (2025)
- [Are Language Models Efficient Reasoners? A Perspective from Logic Programming](https://mlsanity.com/paper/2510.25626/) (2025)
- [Intern-S2-Preview: Scientific Agentic Foundation Model](https://mlsanity.com/paper/2608.13505/) (2026)
- [MixReasoning: Switching Modes to Think](https://mlsanity.com/paper/2510.06052/) (2025)
- [Optimizing Reasoning Efficiency through Prompt Difficulty Prediction](https://mlsanity.com/paper/2511.03808/) (2025)
- [Reasoning While Asking: Transforming Reasoning Large Language Models from Passive Solvers to Proactive Inquirers](https://mlsanity.com/paper/2601.22139/) (2026)
- [Reliable Chain-of-Thought via Prefix Consistency](https://mlsanity.com/paper/2605.07654/) (2026)
- [Improving the Distributional Alignment of LLMs using Supervision](https://mlsanity.com/paper/2507.00439/) (2025)
- [Distractor Injection Attacks on Large Reasoning Models: Characterization and Defense](https://mlsanity.com/paper/2510.16259/) (2025)
- [From Passive Metric to Active Signal: The Evolving Role of Uncertainty Quantification in Large Language Models](https://mlsanity.com/paper/2601.15690/) (2026)
- [AutoCRAT: Within-trajectory Joint Control of Stochasticity and Compute for LLM Reasoning](https://mlsanity.com/paper/2608.29988/) (2026)
- [Dynamic Rollout Editing for Reducing Overthinking in RL-Trained Reasoning Models](https://mlsanity.com/paper/2606.17890/) (2026)

## 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.