phenomenon · first seen 5 Oct 2026
chain-of-thought faithfulness
Also: CoT reliance, CoT faithfulness, reasoning faithfulness, reasoning unfaithfulness
The degree to which a model's stated reasoning actually reflects the factors that determined its output.
Mentions per week
PapersCodeDiscussion
Earliest sources
- SpatialChain: A Benchmark for Auditing Spatial Reasoning Faithfulness in VLMsPapers · 5 Oct 2026 · Rafael Teixeira Sousa, Vinícius Paulo Lopes de Oliveira, Elisa Ayumi Masasi de…
- Cited but Not Consulted: A Counterfactual Audit of Legal Chain-of-Thought FaithfulnessPapers · 8 Oct 2026 · Saisab Sadhu, Shreeyans Arora, Pratinav Seth
- Chain of Thought Monitoring Can Fail Even When CoT Is NecessaryPapers · 13 Oct 2026 · Lennart Bürger, Piotr Komorowski, Moritz Weckbecker, Evgenii Kortukov, Aakriti…
- The Reliance Hypothesis: Chain-of-Thought Faithfulness Rises as Language Resource Level FallsPapers · 16 Oct 2026 · Lucas Resck, Jingyi Sun, Qianli Wang, Isabelle Augenstein, Anna Korhonen
- Verbalising Model's Computation for Chain-of-Thought Faithfulness DetectionPapers · 16 Oct 2026 · Yujing Wang, Yulan He, Wenya Wang, Hanqi Yan
- CoTs remain monitorable when being unfaithful requires computation: Replicating and Extending the Necessity Effect Across Eleven ModelsPapers · 17 Oct 2026 · Arav Dhoot, Yixiong Hao, Zephaniah Roe
Latest sources
- Chain of Faith: Policy-Internal Rewards for Reasoning ReliancePapers · 2 Nov 2026 · Fanyu meng, Zhaodan Kong, Xin Liu
- Chain-of-Thought Oversight Requires Separate Evidence for Faithfulness and MonitorabilityPapers · 1 Nov 2026 · Sichao Li, Sai Ma, Xiyang Hu, Chudi Zhong, Tongliang Liu
- CRoS: Causality Reward on Scrambled CoTs Improves Accuracy and Reasoning FaithfulnessPapers · 30 Oct 2026 · Xiao Yu, Justin Chih-Yao Chen, Zaid Khan, Hyunji Lee, Mohit Bansal
- Does Chain-of-Thought Reliance Follow a Scaling Law?Larger Models Follow Their Steps More, Independent of Post-TrainingPapers · 28 Oct 2026 · Esther Xin
- Correctness Conditioning as Selection in Chain-of-Thought Faithfulness EvaluationPapers · 27 Oct 2026 · Yueer Pan, Yuexi Pan
- Natural Language Autoencoders Identify Without ExplainingPapers · 23 Oct 2026 · Kristina Šekrst, Barbara Geld
- LLMs Can Tackle High Energy Theoretical Physics, yet Gaps Remain: Part I, Quantum Field TheoryPapers · 23 Oct 2026 · Tong Liu, Bohan Li, Gongrui Cheng, Zixiong Yu, Yulan He, Volker Tresp
- CoTs remain monitorable when being unfaithful requires computation: Replicating and Extending the Necessity Effect Across Eleven ModelsPapers · 17 Oct 2026 · Arav Dhoot, Yixiong Hao, Zephaniah Roe
- Verbalising Model's Computation for Chain-of-Thought Faithfulness DetectionPapers · 16 Oct 2026 · Yujing Wang, Yulan He, Wenya Wang, Hanqi Yan
- The Reliance Hypothesis: Chain-of-Thought Faithfulness Rises as Language Resource Level FallsPapers · 16 Oct 2026 · Lucas Resck, Jingyi Sun, Qianli Wang, Isabelle Augenstein, Anna Korhonen
- Chain of Thought Monitoring Can Fail Even When CoT Is NecessaryPapers · 13 Oct 2026 · Lennart Bürger, Piotr Komorowski, Moritz Weckbecker, Evgenii Kortukov, Aakriti…
- Cited but Not Consulted: A Counterfactual Audit of Legal Chain-of-Thought FaithfulnessPapers · 8 Oct 2026 · Saisab Sadhu, Shreeyans Arora, Pratinav Seth