mlsanityData to 7 Oct 2026

Is It Thinking or Cheating? Detecting Implicit Reward Hacking by Measuring Reasoning Effort

arXiv 2510.01367 ↗First posted 1 Oct 2025cs.AI cs.CLcited by 25

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

Abstract (arXiv)

Reward hacking, where a reasoning model exploits loopholes in a reward function to achieve high rewards without solving the intended task, poses a significant threat. This behavior may be explicit, i.e. verbalized in the model's chain-of-thought (CoT), or implicit, where the CoT appears benign thus bypasses CoT monitors. To detect implicit reward hacking, we propose TRACE (Truncated Reasoning AUC Evaluation). Our key observation is that hacking occurs when exploiting the loophole is easier than solving the actual task. This means that the model is using less 'effort' than required to achieve high reward. TRACE quantifies effort by measuring how early a model's reasoning becomes sufficient to obtain the reward. We progressively truncate a model's CoT at various lengths, force the model to answer, and estimate the expected reward at each cutoff. A hacking model, which takes a shortcut, will achieve a high expected reward with only a small fraction of its CoT, yielding a large area under the accuracy-vs-length curve. TRACE achieves over 65% gains over our strongest 72B CoT monitor in math reasoning, and over 30% gains over a 32B monitor in coding. We further show that TRACE can discover unknown loopholes during training. Overall, TRACE offers a scalable unsupervised approach for oversight where current monitoring methods prove ineffective.

Key facts. Cited by 25 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 19 of its 34 references that are in the map. First posted on arXiv 1 Oct 2025.

Builds on

19 of 34 references.

Closest ancestors · 7

Same direction, newest first

  1. 2025When Chain of Thought is Necessary, Language Models Struggle to Evade Monitors7 Jul
  2. 2025Reasoning Models Don't Always Say What They Think8 May
  3. 2025Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscation14 Mar
  4. 2025Chain-of-Thought Reasoning In The Wild Is Not Always Faithful11 Mar
  5. 2024Training Large Language Models to Reason in a Continuous Latent Space9 Dec
  6. 2023Preventing Language Models From Hiding Their Reasoning27 Oct
  7. 2023Measuring Faithfulness in Chain-of-Thought Reasoning17 Jul

Other references · 12

  1. 2022Training language models to follow instructions with human feedbackcited by 15,712
  2. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  3. 2024Qwen2.5 Technical Reportcited by 4,713
  4. 2021Measuring Coding Challenge Competence With APPScited by 879
  5. 2025ReTool: Reinforcement Learning for Strategic Tool Use in LLMscited by 377
  6. 2022Measuring Progress on Scalable Oversight for Large Language Modelscited by 212
  7. 2022Defining and Characterizing Reward Hackingcited by 184
  8. 2024Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Modelscited by 158
  9. 2024On Memorization of Large Language Models in Logical Reasoningcited by 114
  10. 2025Big-Math: A Large-Scale, High-Quality Math Dataset for Reinforcement Learning in Language Modelscited by 104
  11. 2024Back to Basics: Revisiting REINFORCE Style Optimization for Learning from Human Feedback in LLMscited by 103
  12. 2025Teaching Models to Verbalize Reward Hacking in Chain-of-Thought Reasoningcited by 19

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 25.

25 papers.

Sort:

By area:

  1. 2026Reasoning Theater: Disentangling Model Beliefs from Chain-of-Thoughtcited by 35
  2. 2026LLMs Gaming Verifiers: RLVR can Lead to Reward Hackingcited by 24
  3. 2026Reward Hacking in Rubric-Based Reinforcement Learningcited by 18
  4. 2026Counterfactual Simulation Training for Chain-of-Thought Faithfulnesscited by 15
  5. 2026Countdown-Code: A Testbed for Studying The Emergence and Generalization of Reward Hacking in RLVRcited by 15
  6. 2026Hack-Verifiable Environments: Towards Evaluating Reward Hacking at Scalecited by 9
  7. 2026Reproducing, Analyzing, and Detecting Reward Hacking in Rubric-Based Reinforcement Learningcited by 9
  8. 2026Large Reasoning Models Are (Not Yet) Multilingual Latent Reasonerscited by 8
  9. 2026Analyzing and Improving Chain-of-Thought Monitorability Through Information Theorycited by 7
  10. 2026Learning Next Action Predictors from Human-Computer Interactioncited by 7
  11. 2026Outcome Rewards Do Not Guarantee Verifiable or Causally Important Reasoningcited by 7
  12. 2026CoffeeBench: Benchmarking Long-Horizon LLM Agents in Heterogeneous Multi-Agent Economiescited by 7
13 more
  1. 2026From Rebound to Remedy: Understanding and Mitigating Reward Hacking via Representation Engineeringcited by 5
  2. 2026Detecting and Suppressing Reward Hacking with Gradient Fingerprintscited by 4
  3. 2026Do Coding Agents Deceive Us? Detecting and Preventing Cheating via Capped Evaluation with Randomized Testscited by 1
  4. 2026Self-Commitment Latency: A Reward-Free Probe for Prompted Implicit Hackingcited by 0
  5. 2026Context-Masked Truncated Reasoning Audits for Answer-Key Dependence in LLM Tutorscited by 0
  6. 2026Does Out-of-Sight Equal Out-of-Mind in CoT Monitorability?cited by 0
  7. 2026INTENT-AS-A-TOOL Makes it Easy to Track Agentic Misalignmentcited by 0
  8. 2026Detecting Hidden Behaviors in LLMs via Activation-matched Finetuningcited by 0
  9. 2026RLVR2: Reinforcement Learning with Verifiable Rubric-based Rankingcited by 0
  10. 2026TeacherGRPO: Closing the Capacity Gap in Reasoning Distillation via Teacher Alignmentcited by 0
  11. 2026When Confidence Rises Too Early: Detecting Shortcut Reasoning via Premature Answer Commitmentcited by 0
  12. 2026Verifier Errors in RLVR: Reward Hacking, Limits of Feedback, and Selective Controlcited by 0
  13. 2026CATCH: A Controllable Analysis Testbed for Reward Hacking in Coding RLcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Do Prompt-Elicited Trajectories Reflect Training-Time Reward Hacking? A Systematic Study on Monitoring Training-Time Reward Hacking in Code Generation
  2. 2026A Near-Zero Monitor Readout Is Not Evidence of Behavioral Control
  3. 2025The Ends Justify the Thoughts: RL-Induced Motivated Reasoning in LLM CoTs
  4. 2026Training on Documents About Monitoring Leads to CoT Obfuscation
  5. 2026A False Average: Pooled CoT-Monitor Accuracy Conceals a Reasoning-Dependent Fragility
  6. 2026Chain-of-thought obfuscation learned from output supervision can generalise to unseen tasks
  7. 2025Curing Miracle Steps in LLM Mathematical Reasoning with Rubric Rewards
  8. 2026hacktrace: behavior-supervised detection of reward hacking during code generation