mlsanityData to 14 Oct 2026

Detecting Unfaithful Chain-of-Thought via Circuit-Guided Internal-External Discrepancy

arXiv 2605.25603 ↗First posted 25 May 2026cs.AIcited by 3

On the map: LLM reasoning → Prompting and chain-of-thought reasoning

Abstract (arXiv)

Chain-of-thought (CoT) reasoning improves the problem-solving ability of large language models (LLMs), but generated reasoning traces may not faithfully reflect the model's actual decision process. Existing CoT unfaithfulness detectors mainly rely on external signals from generated rationales, such as textual plausibility or answer consistency, while overlooking evidence from the model's internal computation. Although recent circuit tracing methods provide a way to obtain model-internal evidence by tracing how information flows through model components during reasoning, constructing full reasoning circuits for long CoTs is costly and difficult to scale. To address these challenges, we propose Circuit-guided Internal-External Discrepancy Scorer (CIE-Scorer), a framework for instance-level CoT unfaithfulness detection. The key idea is that faithful reasoning traces should align with the model's computational process, whereas unfaithful traces may diverge from it. CIE-Scorer efficiently traces compact sentence-level circuits from informative reasoning tokens, constructs internal and external reasoning graphs, and measures their discrepancy using Fused Gromov--Wasserstein distance. Experiments on four datasets from FaithCoT-Bench show that CIE-Scorer achieves state-of-the-art performance while reducing the cost of circuit construction, demonstrating the effectiveness of combining mechanistic interpretability signals with external reasoning traces for CoT unfaithfulness detection.

Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Placed in Prompting and chain-of-thought reasoning, one of 17 directions in LLM reasoning. Builds on 30 of its 54 references that are in the map. First posted on arXiv 25 May 2026.

Builds on

30 of 54 references.

Closest ancestors · 11

Same direction, newest first

  1. 2025Verifying Chain-of-Thought Reasoning via Its Computational Graph10 Oct
  2. 2025FaithCoT-Bench: Benchmarking Instance-Level Faithfulness of Chain-of-Thought Reasoning5 Oct
  3. 2025Is Chain-of-Thought Reasoning of LLMs a Mirage? A Data Distribution Lens2 Aug
  4. 2025Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Models12 Mar
  5. 2025Chain-of-Thought Reasoning In The Wild Is Not Always Faithful11 Mar
  6. 2025Stepwise Perplexity-Guided Refinement for Efficient Chain-of-Thought Reasoning in Large Language Models18 Feb
  7. 2024Towards Better Chain-of-Thought: A Reflection on Effectiveness and Faithfulness29 May
  8. 2024Dissociation of Faithful and Unfaithful Reasoning in LLMs23 May
  9. 2024Making Reasoning Matter: Measuring and Improving Faithfulness of Chain-of-Thought Reasoning21 Feb
  10. 2023Measuring Faithfulness in Chain-of-Thought Reasoning17 Jul
  11. 2023Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting7 May

Other references · 19

  1. 2021TruthfulQA: Measuring How Models Mimic Human Falsehoodscited by 2,975
  2. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,420
  3. 2017Program Induction by Rationale Generation : Learning to Solve and Explain Algebraic Word Problemscited by 650
  4. 2025Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoningcited by 559
  5. 2025Humanity's Last Examcited by 491
  6. 2024Safety Alignment Should Be Made More Than Just a Few Tokens Deepcited by 472
  7. 2025Demystifying Long Chain-of-Thought Reasoning in LLMscited by 326
  8. 2025Reinforcement Learning with Verifiable Rewards Implicitly Incentivizes Correct Reasoning in Base LLMscited by 289
  9. 2025ThinkPrune: Pruning Long Chain-of-Thought of LLMs via Reinforcement Learningcited by 192
  10. 2025A Comprehensive Survey in LLM(-Agent) Full Stack Safety: Data, Training and Deploymentcited by 116
  11. 2024Faithfulness vs. Plausibility: On the (Un)Reliability of Explanations from Large Language Modelscited by 105
  12. 2024The BiGGen Bench: A Principled Benchmark for Fine-grained Evaluation of Language Models with Language Modelscited by 93
7 more
  1. 2020LogiQA: A Challenge Dataset for Machine Reading Comprehension with Logical Reasoningcited by 69
  2. 2025Walk the Talk? Measuring the Faithfulness of Large Language Model Explanationscited by 45
  3. 2025On the Trustworthiness of Generative Foundation Models: Guideline, Assessment, and Perspectivecited by 41
  4. 2025Measuring the Faithfulness of Thinking Drafts in Large Reasoning Modelscited by 23
  5. 2025How Well Can Reasoning Models Identify and Recover from Unhelpful Thoughts?cited by 11
  6. 2026Do LLM Self-Explanations Help Users Predict Model Behavior? Evaluating Counterfactual Simulatability with Pragmatic Perturbationscited by 7
  7. 2025Agree, Disagree, Explain: Decomposing Human Label Variation in NLI through the Lens of Explanationscited by 5

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026Two Regimes of Chain-of-Thought Unfaithfulness: Metric-Based Detection Fails Where Models Are Wrongcited by 1
  2. 2026Context-Masked Truncated Reasoning Audits for Answer-Key Dependence in LLM Tutorscited by 0
  3. 2026From Concept Alignment to Causal Grounding: An Intervention Test of Chain-of-Thought Faithfulnesscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026The Tell-Tale Trace: Detecting Reasoning Failures in LLMs Using Chain-of-Thought Dynamics
  2. 2025Deep Hidden Cognition Facilitates Reliable Chain-of-Thought Reasoning
  3. 2026Mechanistic Evidence for Faithfulness Decay in Chain-of-Thought Reasoning
  4. 2026Faithfulness as Information Flow: Evaluating and Training Faithful Chain-of-Thought Reasoning
  5. 2026C2-Faith: Benchmarking LLM Judges for Causal and Coverage Faithfulness in Chain-of-Thought Reasoning
  6. 2026The Fragility of Chain-of-Thought Monitoring Across Typologically Diverse Languages
  7. 2026Stateful Reasoning via Insight Replay
  8. 2026Detecting Hidden Chain-of-Thought in Large Language Models with Linguistic, Behavioral, and Mechanistic Indicators