mlsanityData to 7 Oct 2026

Are DeepSeek R1 And Other Reasoning Models More Faithful?

arXiv 2501.08156 ↗First posted 14 Jan 2025cs.LGcited by 69

On the map: LLM reasoning → Reasoning benchmarks and evaluation

Abstract (arXiv)

Language models trained to solve reasoning tasks via reinforcement learning have achieved striking results. We refer to these models as reasoning models. Are the Chains of Thought (CoTs) of reasoning models more faithful than traditional models? We evaluate three reasoning models (based on Qwen-2.5, Gemini-2, and DeepSeek-V3-Base) on an existing test of faithful CoT. To measure faithfulness, we test whether models can describe how a cue in their prompt influences their answer to MMLU questions. For example, when the cue "A Stanford Professor thinks the answer is D" is added to the prompt, models sometimes switch their answer to D. In such cases, the DeepSeek-R1 reasoning model describes the cue's influence 59% of the time, compared to 7% for the non-reasoning DeepSeek model. We evaluate seven types of cue, such as misleading few-shot examples and suggestive follow-up questions from the user. Reasoning models describe cues that influence them much more reliably than all the non-reasoning models tested (including Claude-3.5-Sonnet and GPT-4o). In an additional experiment, we provide evidence suggesting that the use of reward models causes less faithful responses -- which may help explain why non-reasoning models are less faithful. Our study has two main limitations. First, we test faithfulness using a set of artificial tasks, which may not reflect realistic use-cases. Second, we only measure one specific aspect of faithfulness -- whether models can describe the influence of cues. Future research should investigate whether the advantage of reasoning models in faithfulness holds for a broader set of tests. Still, we think this increase in faithfulness is promising for the explainability of language models.

Key facts. Cited by 69 papers in the map (data through 7 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 24 of its 63 references that are in the map. First posted on arXiv 14 Jan 2025.

Builds on

24 of 63 references.

Closest ancestors · 1

Same direction, newest first

  1. 2023Question Decomposition Improves the Faithfulness of Model-Generated Reasoning17 Jul

Other references · 23

  1. 2022Training language models to follow instructions with human feedbackcited by 15,712
  2. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  3. 2023Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Promptingcited by 1,081
  4. 2022Discovering Language Model Behaviors with Model-Written Evaluationscited by 805
  5. 2021BBQ: A Hand-Built Bias Benchmark for Question Answeringcited by 683
  6. 2023The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A"cited by 399
  7. 2023Faithful Chain-of-Thought Reasoningcited by 285
  8. 2024From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Stepcited by 225
  9. 2024Do Large Language Models Latently Perform Multi-Hop Reasoning?cited by 165
  10. 2023Physics of Language Models: Part 3.2, Knowledge Manipulationcited by 142
  11. 2023Taken out of context: On measuring situational awareness in LLMscited by 131
  12. 2023Evaluating and Mitigating Discrimination in Language Model Decisionscited by 117
11 more
  1. 2024Making Reasoning Matter: Measuring and Improving Faithfulness of Chain-of-Thought Reasoningcited by 108
  2. 2023Faithfulness Tests for Natural Language Explanationscited by 107
  3. 2024Looking Inward: Language Models Can Learn About Themselves by Introspectioncited by 105
  4. 2024Faithfulness vs. Plausibility: On the (Un)Reliability of Explanations from Large Language Modelscited by 104
  5. 2023Do Models Explain Themselves? Counterfactual Simulatability of Natural Language Explanationscited by 85
  6. 2023Preventing Language Models From Hiding Their Reasoningcited by 62
  7. 2023Contrastive Chain-of-Thought Promptingcited by 52
  8. 2024Bias-Augmented Consistency Training Reduces Biased Reasoning in Chain-of-Thoughtcited by 37
  9. 2024Towards Consistent Natural-Language Explanations via Explanation-Consistency Finetuningcited by 26
  10. 2023ALMANACS: A Simulatability Benchmark for Language Model Explainabilitycited by 11
  11. 2023Tell, don't show: Declarative facts influence how LLMs generalizecited by 8

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 31; 2026: 38.

69 papers.

Sort:

By area:

  1. 2025Reasoning Models Don't Always Say What They Thinkcited by 388
  2. 2025Training Language Models to Reason Efficientlycited by 242
  3. 2025When Chain of Thought is Necessary, Language Models Struggle to Evade Monitorscited by 81
  4. 2025Thought Crime: Backdoors and Emergent Misalignment in Reasoning Modelscited by 57
  5. 2026Reasoning Theater: Disentangling Model Beliefs from Chain-of-Thoughtcited by 35
  6. 2025Measuring the Faithfulness of Thinking Drafts in Large Reasoning Modelscited by 22
  7. 2025Teaching Models to Verbalize Reward Hacking in Chain-of-Thought Reasoningcited by 19
  8. 2025Can Reasoning Models Obfuscate Reasoning? Stress-Testing Chain-of-Thought Monitorabilitycited by 18
  9. 2025LLMs Can Covertly Sandbag on Capability Evaluations Against Chain-of-Thought Monitoringcited by 17
  10. 2025Robustly Improving LLM Fairness in Realistic Settings via Interpretabilitycited by 16
  11. 2025Thought Branches: Interpreting LLM Reasoning Requires Resamplingcited by 15
  12. 2026Counterfactual Simulation Training for Chain-of-Thought Faithfulnesscited by 15
38 more
  1. 2025Can Aha Moments Be Fake? Towards Quantifying Decorative and True Thinking in Chain-of-Thoughtcited by 14
  2. 2025When Models Reason in Your Language: Controlling Thinking Language Comes at the Cost of Accuracycited by 13
  3. 2025A Pragmatic Way to Measure Chain-of-Thought Monitorabilitycited by 13
  4. 2025A Closer Look at Bias and Chain-of-Thought Faithfulness of Large (Vision) Language Modelscited by 12
  5. 2026A Positive Case for Faithfulness: LLM Self-Explanations Help Predict Model Behaviorcited by 12
  6. 2025Context Is What You Need: The Maximum Effective Context Window for Real World Limits of LLMscited by 11
  7. 2025When Two LLMs Debate, Both Think They'll Wincited by 10
  8. 2025Truthful or Fabricated? Using Causal Attribution to Mitigate Reward Hacking in Explanationscited by 9
  9. 2025Reasoning or Retrieval? A Study of Answer Attribution on Large Reasoning Modelscited by 9
  10. 2025Evaluating Reasoning Faithfulness in Medical Vision-Language Models using Multimodal Perturbationscited by 9
  11. 2025Reasoning Models Sometimes Output Illegible Chains of Thoughtcited by 9
  12. 2026RFEval: Benchmarking Reasoning Faithfulness under Counterfactual Reasoning Intervention in Large Reasoning Modelscited by 9
  13. 2025Pragmatic Inference Chain (PIC) Improving LLMs' Reasoning of Authentic Implicit Toxic Languagecited by 8
  14. 2025Analysing Chain of Thought Dynamics: Active Guidance or Unfaithful Post-hoc Rationalisation?cited by 8
  15. 2026Biases in the Blind Spot: Detecting What LLMs Fail to Mentioncited by 8
  16. 2025Reasoning Under Pressure: How do Training Incentives Influence Chain-of-Thought Monitorability?cited by 7
  17. 2025Humans Perceive Wrong Narratives from AI Reasoning Textscited by 6
  18. 2025Meta-R1: Empowering Large Reasoning Models with Metacognitioncited by 6
  19. 2025From Reasoning to Answer: Empirical, Attention-Based and Mechanistic Insights into Distilled DeepSeek R1 Modelscited by 6
  20. 2025Beyond Correctness: Rewarding Faithful Reasoning in Retrieval-Augmented Generationcited by 6
  21. 2025LogicReward: Incentivizing LLM Reasoning via Step-Wise Logical Supervisioncited by 6
  22. 2026C2-Faith: Benchmarking LLM Judges for Causal and Coverage Faithfulness in Chain-of-Thought Reasoningcited by 6
  23. 2026Catching rationalization in the act: detecting motivated reasoning before and after CoT via activation probingcited by 6
  24. 2025State over Tokens: Characterizing the Role of Reasoning Tokenscited by 5
  25. 2026Do explanations generalize across large reasoning models?cited by 5
  26. 2026Towards a Mechanistic Understanding of Large Reasoning Models: A Survey of Training, Inference, and Failurescited by 5
  27. 2026Aletheia: What Makes RLVR For Code Verifiers Tick?cited by 4
  28. 2026Monitorability as a Free Gift: How RLVR Spontaneously Aligns Reasoningcited by 4
  29. 2026Measuring Faithfulness Depends on How You Measure: Classifier Sensitivity in LLM Chain-of-Thought Evaluationcited by 4
  30. 2026Why Models Know But Don't Say: Chain-of-Thought Faithfulness Divergence Between Thinking Tokens and Answers in Open-Weight Reasoning Modelscited by 4
  31. 2026Self-Blinding and Counterfactual Self-Simulation Mitigate Biases and Sycophancy in Large Language Modelscited by 2
  32. 2026Faithfulness as Information Flow: Evaluating and Training Faithful Chain-of-Thought Reasoningcited by 2
  33. 2026Investigating the Interplay between Contextual and Parametric Chain-of-Thought Faithfulness under Optimizationcited by 2
  34. 2026Chain-of-Thought Monitoring Can Be Unreliable in Implicit-Influence Settingscited by 2
  35. 2025FlagEval Findings Report: A Preliminary Evaluation of Large Reasoning Models on Automatically Verifiable Textual and Visual Questionscited by 1
  36. 2026How Does Unfaithful Reasoning Emerge from Autoregressive Training? A Study of Synthetic Experimentscited by 1
  37. 2026Compliance vs. Sensibility: On the Reasoning Controllability in Large Language Modelscited by 1
  38. 2026The Fragility of Chain-of-Thought Monitoring Across Typologically Diverse Languagescited by 1

50 most cited of 69 All 69 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025100 Days After DeepSeek-R1: A Survey on Replication Studies and More Directions for Reasoning Language Models
  2. 2025DeepSeek-R1 Thoughtology: Let's think about LLM Reasoning
  3. 2025The Challenge of Teaching Reasoning to LLMs Without RL or Distillation
  4. 2025Don't Overthink It: A Survey of Efficient R1-style Large Reasoning Models
  5. 2023Measuring Faithfulness in Chain-of-Thought Reasoning
  6. 2025Large Reasoning Models are not thinking straight: on the unreliability of thinking trajectories
  7. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
  8. 2025A Sober Look at Progress in Language Model Reasoning: Pitfalls and Paths to Reproducibility