Reward Hacking in Rubric-Based Reinforcement Learning

• URL: http://arxiv.org/abs/2605.12474v1
• Date: Tue, 12 May 2026 17:54:25 GMT
• Title: Reward Hacking in Rubric-Based Reinforcement Learning
• Abstract summary: We study reward hacking in rubric-based RL, where a policy is optimized against a training verifier but evaluated against a cross-family panel of three frontier judges.<n>Our framework separates two sources of divergence: verifier failure and rubric-design limitations.<n>Across medical and science domains, weak verifiers produce large proxy-reward gains that do not transfer to the reference verifiers.
• Score: 23.418394508756464
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Reinforcement learning with verifiable rewards has enabled strong post-training gains in domains such as math and coding, though many open-ended settings rely on rubric-based rewards. We study reward hacking in rubric-based RL, where a policy is optimized against a training verifier but evaluated against a cross-family panel of three frontier judges, reducing dependence on any single evaluator. Our framework separates two sources of divergence: verifier failure, where the training verifier credits rubric criteria that reference verifiers reject, and rubric-design limitations, where even strong rubric-based verifiers favor responses that rubric-free judges rate worse overall. Across medical and science domains, weak verifiers produce large proxy-reward gains that do not transfer to the reference verifiers; exploitation grows over training and concentrates in recurring failures such as partial satisfaction of compound criteria, treating implicit content as explicit, and imprecise topical matching. Stronger verifiers substantially reduce, but do not eliminate, verifier exploitation. We also introduce a self-internalization gap, a verifier-free diagnostic based on policy log-probabilities, which tracks reference-verifier quality, detecting when the policy trained using the weak verifier stops improving. Finally, in our setting, stronger verification does not prevent reward hacking when the rubric leaves important failure modes unspecified: rubric-based verifiers prefer the RL checkpoint, while rubric-free judges prefer the base model. These disagreements coincide with gains concentrated in completeness and presence-based criteria, alongside declines in factual correctness, conciseness, relevance, and overall quality. Together, these results suggest that stronger verification reduces reward hacking, but does not by itself ensure that rubric gains correspond to broader quality gains.

Related papers

• AutoVerifier: Residual-Guided Non-Parametric Optimization for Reference-Based Answer Verification [5.990350746292239]
  AutoVerifier learns implicit assumptions from recurring verifier errors.<n>Experiments on four verifier benchmarks demonstrate that AutoVerifier outperforms state-of-the-art verifiers by a large margin.
  arXiv  Detail & Related papers  (2026-08-26T11:06:40Z)
• RecurSE: Bounded Recursive Self-Evaluation for LLM Rubric Judges [48.831446371402414]
  RecurSE pairs a trainable judge with a synchronized policy-copy checker that audits the judge's reasoning against meta-rubrics to supply a scalar process reward.<n>We show that RecurSE achieves consistent generalization gains across held-out medical, pairwise, summarization, and professional benchmarks.
  arXiv  Detail & Related papers  (2026-08-25T08:35:04Z)
• Rethinking Reward Supervision: Rubric-Conditioned Self-Distillation [60.55792673956761]
  We propose textbfRubric-Conditioned Self-Distillation, a framework that incorporates rubrics as structured, fine-grained feedback for on-policy self-distillation.<n>We evaluate on a diverse suite of science reasoning benchmarks and results show that rubric-conditioned self-distillation effectively converts rubric-level criteria into token-level guidance.
  arXiv  Detail & Related papers  (2026-06-17T17:54:04Z)
• VeriGate: Verifier-Gated Step-Level Supervision for GRPO [51.26100506256885]
  Group Relative Policy Optimization is an effective recipe for training reasoning models with verifier-based outcome rewards.<n>We present VeriGate, a verifier-gated extension of GRPO that addresses these limitations with three design choices.<n>We show that VeriGate improves average accuracy by about 20% and 12% for 1.5B and 7B models respectively.
  arXiv  Detail & Related papers  (2026-05-28T18:20:32Z)
• Soft-SVeRL: Self-Verified Reinforcement Learning with Soft Rewards [24.040621624203823]
  Reinforcement Learning from Verifiable Rewards (RLVR) has improved language models in domains such as mathematics and code.<n>We introduce Soft-RLVR, a framework for reinforcement learning from decomposed, learned verification signals.<n>We show that self-verification is prone to reward inflation from overly permissive self-judgments, and that explicit stabilization is needed to prevent this collapse.
  arXiv  Detail & Related papers  (2026-05-27T14:50:53Z)
• Cycle-Consistent Search: Question Reconstructability as a Proxy Reward for Search Agent Training [80.20022221643414]
  Cycle-Consistent Search is a gold-supervision-free framework for training search agents.<n>We show that CCS achieves performance comparable to supervised baselines.<n>These results suggest that CCS provides a scalable training paradigm for training search agents in settings where gold supervision is unavailable.
  arXiv  Detail & Related papers  (2026-04-14T17:00:18Z)
• What If Consensus Lies? Selective-Complementary Reinforcement Learning at Test Time [57.533031432715084]
  Test-Time Reinforcement Learning (TTRL) enables Large Language Models (LLMs) to enhance reasoning capabilities on unlabeled test streams.<n>Existing TTRL methods rely exclusively on positive pseudo-labeling strategies.<n>We propose SCRL, a robust test-time reinforcement learning framework that effectively mitigates label noise amplification.
  arXiv  Detail & Related papers  (2026-03-20T11:47:12Z)
• PRIME: A Process-Outcome Alignment Benchmark for Verifiable Reasoning in Mathematics and Engineering [71.15346406323827]
  We introduce PRIME, a benchmark for evaluating verifiers on Process-Outcome Alignment verification.<n>We find that current verifiers frequently fail to detect derivation flaws.<n>We propose a process-aware RLVR training paradigm utilizing verifiers selected via PRIME.
  arXiv  Detail & Related papers  (2026-02-12T04:45:01Z)
• Preventing the Collapse of Peer Review Requires Verification-First AI [49.995126139461085]
  We propose truth-coupling, i.e. how tightly venue scores track latent scientific truth.<n>We formalize two forces that drive a phase transition toward proxy-sovereign evaluation.
  arXiv  Detail & Related papers  (2026-01-23T17:17:32Z)
• Hybrid Reinforcement: When Reward Is Sparse, It's Better to Be Dense [36.71358559780692]
  HERO is a reinforcement learning framework that integrates verifier signals with reward-model scores in a structured way.<n> HERO consistently outperforms RM-only and verifier-only baselines, with strong gains on both verifiable and hard-to-verify tasks.
  arXiv  Detail & Related papers  (2025-10-08T17:09:41Z)
• Reinforcement Learning with Verifiable yet Noisy Rewards under Imperfect Verifiers [90.50039419576807]
  Reinforcement Learning with Verifiable Rewards (RLVR) trains policies against automated verifiers to avoid costly human labeling.<n>To reduce vulnerability to verifier hacking, many RLVR systems collapse rewards to binary $0,1$ during training.<n>This choice carries a cost: it introduces textitfalse negatives (rejecting correct answers, FNs) and textitfalse positives (accepting incorrect ones, FPs)
  arXiv  Detail & Related papers  (2025-10-01T13:56:44Z)
• From Accuracy to Robustness: A Study of Rule- and Model-based Verifiers in Mathematical Reasoning [41.02508512078575]
  We take mathematical reasoning as a case study and conduct a comprehensive analysis of various verifiers in both static evaluation and RL training scenarios.<n>First, we find that current open-source rule-based verifiers often fail to recognize equivalent answers presented in different formats, resulting in non-negligible false negative rates.<n>We investigate model-based verifiers as a potential solution to address these limitations.
  arXiv  Detail & Related papers  (2025-05-28T10:28:41Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.