mlsanityData to 14 Oct 2026

Self-Commitment Latency: A Reward-Free Probe for Prompted Implicit Hacking

arXiv 2606.05625 ↗First posted 4 Jun 2026cs.AI cs.LGcited by 1

On the map: LLM reasoning → Reinforcement learning for reasoning

also in 1 more
Abstract (arXiv)

Implicit reward hacking is hard to audit when a language model's chain of thought appears benign: a final answer may be anchored by a prompt shortcut while the written reasoning still resembles ordinary problem solving. Verifier-based probes expose such behavior by measuring how early truncated reasoning contexts obtain high reward, but require a task-specific reward signal. This paper proposes a weaker-input alternative, self-commitment latency, which measures how early a prompted reasoning context commits to the model's own final answer. We evaluate the probe in a controlled paired GSM8K setting using Qwen2.5-3B-Instruct-4bit, comparing ordinary prompts with prompts that include an answer hint. Hinted contexts commit substantially earlier and with lower uncertainty than honest contexts. The primary latency metric, first-commitment latency at threshold 0.8, reaches AUROC 0.878; supporting whole-curve summaries reach AUROC 0.926 for commitment range and 0.904 for mean uncommitted mass. The signal is stronger when both prompt conditions answer correctly and remains stable across thresholds. These results show that shortcut-available reasoning contexts can leave an early behavioral commitment signature detectable without a reward model, external judge, or trained classifier.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Reinforcement learning for reasoning, one of 17 directions in LLM reasoning. Builds on 27 of its 28 references that are in the map. First posted on arXiv 4 Jun 2026.

Builds on

27 of 28 references.

Closest ancestors · 3

Same direction, newest first

  1. 2023Let's Verify Step by Step31 May
  2. 2022Constitutional AI: Harmlessness from AI Feedback15 Dec
  3. 2022Solving math word problems with process- and outcome-based feedback25 Nov

Other references · 24

  1. 2022Training language models to follow instructions with human feedbackcited by 15,971
  2. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  3. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  4. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,695
  5. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,708
  6. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
  7. 2022Large Language Models are Zero-Shot Reasonerscited by 4,374
  8. 2017Deep reinforcement learning from human preferencescited by 3,912
  9. 2020Learning to summarize from human feedbackcited by 2,588
  10. 2022Language Models (Mostly) Know What They Knowcited by 1,856
  11. 2016Concrete Problems in AI Safetycited by 1,642
  12. 2022Least-to-Most Prompting Enables Complex Reasoning in Large Language Modelscited by 1,407
12 more
  1. 2022Solving Quantitative Reasoning Problems with Language Modelscited by 1,392
  2. 2023Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Promptingcited by 1,122
  3. 2022Scaling Laws for Reward Model Overoptimizationcited by 1,095
  4. 2023Measuring Faithfulness in Chain-of-Thought Reasoningcited by 703
  5. 2022PAL: Program-aided Language Modelscited by 652
  6. 2020Towards Faithfully Interpretable NLP Systems: How should we define and evaluate faithfulness?cited by 599
  7. 2018Scalable agent alignment via reward modeling: a research directioncited by 447
  8. 2023Faithful Chain-of-Thought Reasoningcited by 288
  9. 2019Risks from Learned Optimization in Advanced Machine Learning Systemscited by 258
  10. 2024Scaling Laws for Reward Model Overoptimization in Direct Alignment Algorithmscited by 143
  11. 2019Reward Tampering Problems and Solutions in Reinforcement Learning: A Causal Influence Diagram Perspectivecited by 108
  12. 2025Is It Thinking or Cheating? Detecting Implicit Reward Hacking by Measuring Reasoning Effortcited by 27

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026A Near-Zero Monitor Readout Is Not Evidence of Behavioral Controlcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Diagnosing and Correcting Reward Hacking in Verifiable Instruction Following
  2. 2026Optimizing the Score, Losing Sight of the Task: Reward Hacking Across Weights, Selection, and Prompts
  3. 2026More Convincing, Not More Correct: Self-Play Reward Hacking of Reference-Free LLM Judges
  4. 2026Monitoring Emergent Reward Hacking During Generation via Internal Activations
  5. 2025Teaching Models to Verbalize Reward Hacking in Chain-of-Thought Reasoning
  6. 2026The Commit-Abstain Circuit: Why Language Models Hallucinate Instead of Abstaining
  7. 2025Rethinking Reward Models for Multi-Domain Test-Time Scaling
  8. 2025Test-time Prompt Intervention