mlsanityData to 7 Oct 2026

Reasoning Under Pressure: How do Training Incentives Influence Chain-of-Thought Monitorability?

arXiv 2512.00218 ↗First posted 28 Nov 2025cs.AI cs.CRcited by 7

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

Abstract (arXiv)

AI systems that output their reasoning in natural language offer an opportunity for safety -- we can monitor their chain of thought (CoT) for undesirable reasoning, such as the pursuit of harmful objectives. However, the extent to which CoT faithfully reflects the underlying reasoning process, and hence the extent to which it can be usefully monitored, may be influenced by certain aspects of training. We investigate how different training incentives, applied to a reasoning model, affect its monitorability. We introduce a novel methodology for measuring monitorability according to whether a monitor can predict a key latent variable using the model's reasoning. When controlling for accuracy, we do not find evidence for consistent effects from commonly used incentives (length penalties and KL regularisation), but we find that adversarial optimisation (penalising monitor accuracy) degrades monitor performance, while direct optimisation for monitorability does not reliably lead to improvements. Our code is available at https://github.com/QiyaoWei/reasoning-under-pressure.

Key facts. Cited by 7 papers in the map (data through 7 Oct 2026). Placed in Prompting and chain-of-thought reasoning, one of 17 directions in LLM reasoning. Builds on 27 of its 37 references that are in the map. First posted on arXiv 28 Nov 2025.

Builds on

27 of 37 references.

Closest ancestors · 8

Same direction, newest first

  1. 2025Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety15 Jul
  2. 2025When Chain of Thought is Necessary, Language Models Struggle to Evade Monitors7 Jul
  3. 2025CoT Red-Handed: Stress Testing Chain-of-Thought Monitoring29 May
  4. 2025Reasoning Models Don't Always Say What They Think8 May
  5. 2025Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscation14 Mar
  6. 2025Chain-of-Thought Reasoning In The Wild Is Not Always Faithful11 Mar
  7. 2023Measuring Faithfulness in Chain-of-Thought Reasoning17 Jul
  8. 2023Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting7 May

Other references · 19

  1. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  2. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,146
  3. 2023Let's Verify Step by Stepcited by 3,598
  4. 2019Fine-Tuning Language Models from Human Preferencescited by 2,124
  5. 2024OpenAI o1 System Cardcited by 1,220
  6. 2022Solving math word problems with process- and outcome-based feedbackcited by 871
  7. 2024Alignment faking in large language modelscited by 316
  8. 2025Towards Large Reasoning Models: A Survey of Reinforced Reasoning with Large Language Modelscited by 291
  9. 2025Reinforcement Learning for Reasoning in Large Language Models with One Training Examplecited by 288
  10. 2024Rewarding Progress: Scaling Automated Process Verifiers for LLM Reasoningcited by 267
  11. 2024Frontier Models are Capable of In-context Schemingcited by 252
  12. 2025Training Language Models to Reason Efficientlycited by 242
7 more
  1. 2025Don't Overthink it. Preferring Shorter Thinking Chains for Improved LLM Reasoningcited by 82
  2. 2025Stress Testing Deliberative Alignment for Anti-Scheming Trainingcited by 71
  3. 2025Are DeepSeek R1 And Other Reasoning Models More Faithful?cited by 69
  4. 2025SHADE-Arena: Evaluating Sabotage and Monitoring in LLM Agentscited by 58
  5. 2023Eliciting Latent Knowledge from Quirky Language Modelscited by 56
  6. 2025All Code, No Thought: Current Language Models Struggle to Reason in Ciphered Languagecited by 7
  7. 2023Benchmarks for Detecting Measurement Tamperingcited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 7.

7 papers.

  1. 2026Monitorability as a Free Gift: How RLVR Spontaneously Aligns Reasoningcited by 4
  2. 2026Aligned, Orthogonal or In-conflict: When can we safely optimize Chain-of-Thought?cited by 3
  3. 2026When Reasoning Traces Become Performative: Step-Level Evidence that Chain-of-Thought Is an Imperfect Oversight Channelcited by 2
  4. 2026Length Penalties Make Chain-of-Thought Less Monitorablecited by 1
  5. 2026Corrupt Plans, Clean Traces: Evading Chain-of-Thought Monitoring with Plan Injectioncited by 0
  6. 2026Monitor Jailbreaking: Evading Chain-of-Thought Monitoring Without Encoded Reasoningcited by 0
  7. 2026When Do Models Admit They Are Wrong? Failure Disclosure Is Unstable Under Reinforcement Learningcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Reasoning Models Sometimes Output Illegible Chains of Thought
  2. 2026Does Out-of-Sight Equal Out-of-Mind in CoT Monitorability?
  3. 2026Efficient Reasoning Training Does Not Always Harm CoT Faithfulness and Monitorability
  4. 2025Monitoring Monitorability
  5. 2025Investigating CoT Monitorability in Large Reasoning Models
  6. 2025The Ends Justify the Thoughts: RL-Induced Motivated Reasoning in LLM CoTs
  7. 2026Chain-of-Thought Monitoring Can Be Unreliable in Implicit-Influence Settings
  8. 2025Can Reasoning Models Obfuscate Reasoning? Stress-Testing Chain-of-Thought Monitorability