CoT Faithfulness

CoT: Chain-of-Thought

Ask about this topic

Overview

Chain-of-thought faithfulness studies whether a model actually uses its visible reasoning to reach an answer, rather than producing a plausible explanation after deciding. Researchers test this with counterfactual hints, edits or deletions of reasoning steps, causal activation patching, and probes of hidden states across settings such as math, medical question answering, and vision. Training approaches include rewards for causal dependence and sufficient evidence, preference learning, and structural constraints on shortcuts. A central challenge is that accuracy and plausible explanations do not establish faithfulness, while measurements vary by task and method and often transfer poorly across models or settings.

Latest papers

CardsList

73 papers4 papers in the last four weeks, down 50% on the four weeks before. 0.0% of all new papers.

  1. Lie to Me: How Faithful Is Chain-of-Thought Reasoning in Reasoning Models?

    Mar 23, 2026Richard J. YoungCoT FaithfulnessFaithfulness of Language Model Explanations

    University of Nevada, Las Vegas, Department of Management, Entrepreneurship and Technology, Lee Business School, Las Vegas, NV, USA · DeepNeuro AI, Las Vegas, NV, USA

  2. Measuring Faithfulness Depends on How You Measure: Classifier Sensitivity in LLM Chain-of-Thought Evaluation

    Mar 20, 2026Richard J. YoungCoT FaithfulnessLLM Evaluation

    University of Nevada, Las Vegas, Department of Management, Entrepreneurship and Technology, Lee Business School, Las Vegas, NV, USA · DeepNeuro AI, Las Vegas, NV, USA

  3. ICE: Intervention-Consistent Explanation Evaluation with Statistical Grounding for LLMs

    Mar 19, 2026Abhinaba Basu, Pavan ChakrabortyExplanation EvaluationLLM Interpretability

    1Indian Institute of Information Technology, Allahabad (IIITA) · 2National Institute of Electronics and Information Technology (NIELIT)

  4. Catching rationalization in the act: detecting motivated reasoning before and after CoT via activation probing

    Mar 17, 2026Parsa Mirtaheri, Mikhail BelkinLLM InterpretabilityCoT Faithfulness

    UC San Diego

  5. Faithful or Just Plausible? Evaluating the Faithfulness of Closed-Source LLMs in Medical Reasoning

    Mar 14, 2026Halimat Afolabi, Zainab Afolabi, Elizabeth Friel +13Faithfulness of Language Model ExplanationsMedical LLMs

    Stratified Precision, United Kingdom · Harvard Medical School, United States · Imperial College London, United Kingdom +3

  6. C2-Faith: Benchmarking LLM Judges for Causal and Coverage Faithfulness in Chain-of-Thought Reasoning

    Mar 5, 2026Avni Mittal, Rauno ArikeCoT FaithfulnessLLM-as-a-Judge

  7. Counterfactual Simulation Training for Chain-of-Thought Faithfulness

    Feb 24, 2026Peter Hase, Christopher PottsCoT FaithfulnessCoT Monitoring

    Stanford University

  8. RFEval: Benchmarking Reasoning Faithfulness under Counterfactual Reasoning Intervention in Large Reasoning Models

    Feb 19, 2026Yunseok Han, Yejoon Lee, Jaeyoung DoCoT FaithfulnessLarge Reasoning Models

    AIDAS Laboratory, 1IPAI & 2ECE, Seoul National University

  9. Balancing Faithfulness and Performance in Reasoning via Multi-Listener Soft Execution

    Feb 18, 2026Nithin Sivakumaran, Shoubin Yu, Hyunji Lee +4CoT FaithfulnessFaithfulness of Language Model Explanations

    UNC Chapel Hill · Cisco · University of Texas at Austin

  10. Bypassing the Rationale: Causal Auditing of Implicit Reasoning in Language Models

    Feb 3, 2026Anish Sathyanarayanan, Aditya Nagarsekar, Aarush RathoreCoT FaithfulnessFaithfulness of Language Model Explanations

    BITS Pilani, K. K. Birla Goa Campus

  11. Do Language Models Reason Across Languages?

    Jan 10, 2026Yan Meng, Wafaa Mohammed, Christof MonzCoT FaithfulnessCross-Lingual Reasoning in Language Models

    Language Technology Lab · University of Amsterdam

  12. LIFT: Likelihood-Based Intervention Faithfulness Training

    Sep 10, 2025Anand Swaroop, Maheep Chaudhary, Kevin Zhu +2CoT FaithfulnessFaithfulness of Language Model Explanations

    Algoverse AI Research

  13. Truthful or Fabricated? Using Causal Attribution to Mitigate Reward Hacking in Explanations

    Apr 7, 2025Pedro Ferreira, Wilker Aziz, Ivan TitovCoT FaithfulnessReward Hacking

    Institute for Logic, Language and Computation (ILLC), University of Amsterdam · Institute for Language, Cognition and Computation (ILCC), University of Edinburgh