mlsanityData to 14 Oct 2026

Reasoning Models Better Express Their Confidence

arXiv 2505.14489 ↗First posted 20 May 2025cs.AI cs.CLcited by 65

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

Abstract (arXiv)

Despite their strengths, large language models (LLMs) often fail to communicate their confidence accurately, making it difficult to assess when they might be wrong and limiting their reliability. In this work, we demonstrate that reasoning models that engage in extended chain-of-thought (CoT) reasoning exhibit superior performance not only in problem-solving but also in accurately expressing their confidence. Specifically, we benchmark six reasoning models across six datasets and find that they achieve strictly better confidence calibration than their non-reasoning counterparts in 33 out of the 36 settings. Our detailed analysis reveals that these gains in calibration stem from the slow thinking behaviors of reasoning models (e.g., exploring alternative approaches and backtracking) which enable them to adjust their confidence dynamically throughout their CoT, making it progressively more accurate. In particular, we find that reasoning models become increasingly better calibrated as their CoT unfolds, a trend not observed in non-reasoning models. Moreover, removing slow thinking behaviors from the CoT leads to a significant drop in calibration. Lastly, we show that non-reasoning models also demonstrate enhanced calibration when simply guided to slow think via in-context learning, fully isolating slow thinking as the source of the calibration gains.

Key facts. Cited by 65 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 49 references that are in the map. First posted on arXiv 20 May 2025.

Builds on

33 of 49 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025Reasoning Models Know When They're Right: Probing Hidden States for Self-Verification7 Apr
  2. 2025s1: Simple test-time scaling31 Jan
  3. 2024Imitate, Explore, and Self-Improve: A Reproduction Report on Slow-thinking Reasoning Systems12 Dec
  4. 2022Language Models (Mostly) Know What They Know11 Jul

Other references · 29

  1. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  2. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  3. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  4. 2017On Calibration of Modern Neural Networkscited by 4,960
  5. 2017TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehensioncited by 2,955
  6. 2024MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmarkcited by 1,848
  7. 2023A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questionscited by 1,601
  8. 2024ChatGLM: A Family of Large Language Models from GLM-130B to GLM-4 All Toolscited by 1,121
  9. 2023Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMscited by 974
  10. 2023Just Ask for Calibration: Strategies for Eliciting Calibrated Confidence Scores from Language Models Fine-Tuned with Human Feedbackcited by 812
  11. 2023SelfCheckGPT: Zero-Resource Black-Box Hallucination Detection for Generative Large Language Modelscited by 810
  12. 2023Siren's Song in the AI Ocean: A Survey on Hallucination in Large Language Modelscited by 698
17 more
  1. 2023Active Retrieval Augmented Generationcited by 516
  2. 2019Measuring Calibration in Deep Learningcited by 446
  3. 2020AmbigQA: Answering Ambiguous Open-domain Questionscited by 390
  4. 2025Cognitive Behaviors that Enable Self-Improving Reasoners, or, Four Habits of Highly Effective STaRscited by 377
  5. 2024To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoningcited by 263
  6. 2023A Survey of Confidence Estimation and Calibration in Large Language Modelscited by 256
  7. 2023A Stitch in Time Saves Nine: Detecting and Mitigating Hallucinations of LLMs by Validating Low-Confidence Generationcited by 206
  8. 2023Med-HALT: Medical Domain Hallucination Test for Large Language Modelscited by 157
  9. 2024Language Model Cascades: Token-level uncertainty and beyondcited by 121
  10. 2023Decomposing Uncertainty for Large Language Models through Input Clarification Ensemblingcited by 114
  11. 2024Relying on the Unreliable: The Impact of Language Models' Reluctance to Express Uncertaintycited by 106
  12. 2024SaySelf: Teaching LLMs to Express Confidence with Self-Reflective Rationalescited by 105
  13. 2024On Verbalized Confidence Scores for LLMscited by 83
  14. 2023Llamas Know What GPTs Don't Show: Surrogate Models for Confidence Estimationcited by 39
  15. 2024Towards Uncertainty-Aware Language Agentcited by 32
  16. 2024Finetuning Language Models to Emit Linguistic Expressions of Uncertaintycited by 26
  17. 2024A Survey of Calibration Process for Black-Box LLMscited by 8

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 22; 2026: 43.

65 papers.

Sort:

By area:

  1. 2025Beyond Binary Rewards: Training LMs to Reason About Their Uncertaintycited by 117
  2. 2025Maximizing Confidence Alone Improves Reasoningcited by 100
  3. 2025Position: Uncertainty Quantification Needs Reassessment for Large-language Model Agentscited by 32
  4. 2026JEV-as-a-Judge: Accept When Confident, Escalate When Unsurecited by 30
  5. 2026How do LLMs Compute Verbal Confidencecited by 26
  6. 2025Trace Length is a Simple Uncertainty Signal in Reasoning Modelscited by 19
  7. 2025When Thinking LLMs Lie: Unveiling the Strategic Deception in Representations of Reasoning Modelscited by 18
  8. 2026The Confidence Dichotomy: Analyzing and Mitigating Miscalibration in Tool-Use Agentscited by 16
  9. 2025Anthropomimetic Uncertainty: What Verbalized Uncertainty in Language Models is Missingcited by 14
  10. 2026Decoupling Reasoning and Confidence: Resurrecting Calibration in Reinforcement Learning from Verifiable Rewardscited by 14
  11. 2025When Two LLMs Debate, Both Think They'll Wincited by 10
  12. 2025Can Reasoning Help Large Language Models Capture Human Annotator Disagreement?cited by 10
38 more
  1. 2026Causal Evidence that Language Models use Confidence to Drive Behaviorcited by 10
  2. 2025Verbalized Confidence Triggers Self-Verification: Emergent Behavior Without Explicit Reasoning Supervisioncited by 9
  3. 2025C2GSPG: Confidence-calibrated Group Sequence Policy Gradient towards Self-aware Reasoningcited by 8
  4. 2026Are LLM Decisions Faithful to Verbal Confidence?cited by 8
  5. 2026Reported Confidence in LLMs Tracks Commitment More Than Correctnesscited by 8
  6. 2026Probing the Trajectories of Reasoning Traces in Large Language Modelscited by 7
  7. 2026Decoupling the Effect of Chain-of-Thought Reasoning: A Human Label Variation Perspectivecited by 6
  8. 2026Same Answer, Different Confidence: Protocol Sensitivity in LLM Confidence Calibrationcited by 6
  9. 2026The Cost of Reasoning: Chain-of-Thought Induces Overconfidence in Vision-Language Modelscited by 5
  10. 2026Hallucinations Undermine Trust; Metacognition is a Way Forwardcited by 5
  11. 2026From Reasoning Traces to Reusable Modules: Understanding Compositional Generalization in Language Model Reasoningcited by 5
  12. 2025The Zero-Step Thinking: An Empirical Study of Mode Selection as Harder Early Exit in Reasoning Modelscited by 4
  13. 2025Re-FORC: Adaptive Reward Prediction for Efficient Chain-of-Thought Reasoningcited by 4
  14. 2026How Reliable are Confidence Estimators for Large Reasoning Models? A Systematic Benchmark on High-Stakes Domainscited by 4
  15. 2025Reasoning's Razor: Reasoning Improves Accuracy but Can Hurt Recall at Critical Operating Points in Safety and Hallucination Detectioncited by 3
  16. 2025Let the Model Distribute Its Doubt: Confidence Estimation through Verbalized Probability Distributioncited by 3
  17. 2026SELFDOUBT: Uncertainty Quantification for Reasoning LLMs via the Hedge-to-Verify Ratiocited by 3
  18. 2026Unsupervised Confidence Calibration for Reasoning LLMs from a Single Generationcited by 3
  19. 2026How LLMs Detect and Correct Their Own Errors: The Role of Internal Confidence Signalscited by 3
  20. 2026Scaling with Confidence: Calibrating Confidence of LLMs for Adaptive Test Time Scalingcited by 3
  21. 2025Anticipatory Evaluation of Language Models: Predicting LLM Benchmark Scores from Descriptionscited by 2
  22. 2025Expanding Computation Spaces of LLMs at Inference Timecited by 2
  23. 2025Beware of Reasoning Overconfidence: Pitfalls in the Reasoning Process for Multi-solution Taskscited by 2
  24. 2025Reliable LLM-Based Edge-Cloud-Expert Cascades for Telecom Knowledge Systemscited by 2
  25. 2026EpiCaR: Knowing What You Don't Know Matters for Better Reasoning in LLMscited by 2
  26. 2026Recursive Language Models Meet Uncertainty: The Surprising Effectiveness of Self-Reflective Program Search for Long Contextcited by 2
  27. 2026Large Language Models Decide Early and Explain Latercited by 2
  28. 2026Quantifying Faithful Confidence Expression in Large Reasoning Modelscited by 2
  29. 2026From Signals to Transfer: A Factorised Study of Probe-Based Uncertainty Estimation in Large Language Modelscited by 2
  30. 2026When Calibration Rankings Reverse: Accuracy-Controlled Evaluation for Fair Comparison of LLMscited by 2
  31. 2025Probing the "Psyche'' of Large Reasoning Models: Understanding Through a Human Lenscited by 1
  32. 2026Oops, Wait: Discourse Tokens Matter in Reasoning Modelcited by 1
  33. 2026Beyond In-Domain Detection: SpikeScore for Cross-Domain Hallucination Detectioncited by 1
  34. 2026How Uncertainty Estimation Scales with Sampling in Reasoning Modelscited by 1
  35. 2026Confidence Should Be Calibrated More Than One Turn Deepcited by 1
  36. 2026Unlearners Can Lie: Evaluating and Improving Honesty in LLM Unlearningcited by 1
  37. 2026Shared Doubt: Zero-Shot Cross-Lingual Confidence Estimation for Language Modelscited by 1
  38. 2026From Sequence to Structure: Relational Uncertainty Propagation for LLM Agentscited by 1

50 most cited of 65 All 65 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026Understanding and Mitigating Premature Confidence for Better LLM Reasoning
  2. 2026Calibration Drift Under Reasoning: How Chain-of-Thought Budgets Induce Overconfidence in Large Language Models
  3. 2025Thought calibration: Efficient and confident test-time scaling
  4. 2026CALIBER: Calibrating Confidence Before and After Reasoning in Language Models
  5. 2026Early Stopping for Large Reasoning Models via Confidence Dynamics
  6. 2025Thinking Out Loud: Do Reasoning Models Know When They're Right?
  7. 2026Post-Training Shifts Confidence: A Three-Stage Analysis of How SFT, RL, and OPD Shape CoT Calibration
  8. 2025Lost at the Beginning of Reasoning