Reinforcement Learning with Verifiable yet Noisy Rewards under Imperfect Verifiers

• URL: http://arxiv.org/abs/2510.00915v2
• Date: Fri, 17 Oct 2025 16:20:55 GMT
• Title: Reinforcement Learning with Verifiable yet Noisy Rewards under Imperfect Verifiers
• Authors: Xin-Qiang Cai, Wei Wang, Feng Liu, Tongliang Liu, Gang Niu, Masashi Sugiyama,
• Abstract summary: 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)
• Score: 90.50039419576807
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) trains policies against automated verifiers to avoid costly human labeling. To reduce vulnerability to verifier hacking, many RLVR systems collapse rewards to binary $\{0,1\}$ during training. This choice carries a cost: it introduces \textit{false negatives} (rejecting correct answers, FNs) and \textit{false positives} (accepting incorrect ones, FPs). For instance, a rule-based checker may mark the correct fraction $\frac{12}{36}$ as wrong when compared against the canonical $\frac{1}{3}$ due to brittle parsing/equivalence rules (FN), while a large language model (LLM) judges can be gamed by superficial cues or even a single adversarial token, yielding inflated correctness for wrong solutions (FP). We formalize verifier unreliability by modeling the verifier as a stochastic reward channel with asymmetric noise rates. From this abstraction, we derive two correction algorithms for verifier errors. The first is a \textit{backward} correction that de-biases the observed binary reward to recover an \textit{unbiased} estimator of the clean policy gradient. The second is a \textit{forward} correction that reweights score-function terms so that the expected update direction aligns with the \textit{clean gradient}; notably, it requires only the FN rate. We implement both as lightweight hooks in a group relative policy optimization (GRPO)-based RLVR pipeline and evaluate them on math-reasoning models and benchmarks. Across models and datasets, both corrections improve over uncorrected training; the forward variant converges faster and remains stable under heavier noise. Finally, we show a practical appeal mechanism in which a lightweight LLM verifier estimates the FN rate online by rechecking rule-based negatives, obtaining outperformance compared with other state-of-the-art contenders.

Related papers

• Overconfident Errors Need Stronger Correction: Asymmetric Confidence Penalties for Reinforcement Learning [17.384089089363382]
  We identify a root cause that existing methods overlook: the uniform penalization of errors.<n>Current approaches treat all incorrect rollouts within a group identically.<n>We propose the Asymmetric Confidence-aware Error Penalty (ACE)
  arXiv  Detail & Related papers  (2026-02-24T22:46:43Z)
• 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)
• Towards Robust Process Reward Modeling via Noise-aware Learning [33.1289107681179]
  We propose a two-stage framework to mitigate noisy supervision.<n>In the labeling stage, we introduce a reflection-aware label correction mechanism that uses a large language model (LLM) as a judge.<n>In the training stage, we propose a underlinetextbfIterative underlinetextbfTraining framework that enables the PRM to progressively refine noisy labels.
  arXiv  Detail & Related papers  (2026-01-19T06:03:58Z)
• CARE What Fails: Contrastive Anchored-REflection for Verifiable Multimodal [84.71254539482369]
  Group-relative reinforcement learning with verifiable rewards (RLVR) often wastes the most informative data it already has the failures.<n>We present CARE, a failure-centric post-training framework for multimodal reasoning that turns errors into supervision.<n> CARE improves accuracy and training smoothness while explicitly increasing the share of learning signal that comes from failures.
  arXiv  Detail & Related papers  (2025-12-22T16:34:21Z)
• AdvJudge-Zero: Binary Decision Flips in LLM-as-a-Judge via Adversarial Control Tokens [9.127363793428119]
  We show that short sequences of low-perplexity control tokens can flip many binary evaluations from correct No'' judgments to incorrect Yes'' judgments.<n>We show that LoRA-based adversarial training on small sets of control-token-augmented examples can markedly reduce these false positives.
  arXiv  Detail & Related papers  (2025-12-19T09:22:11Z)
• Hard Negative Sample-Augmented DPO Post-Training for Small Language Models [4.425580048633862]
  We propose a lightweight and pragmatic post-training pipeline that targets structured errors under realistic compute budgets.<n>We introduce a compact MathVerifier that decomposes a candidate solution into a six-dimensional error profile and aggregates it into interpretable wrongness and absurdity scores.<n> Experiments show that verifier-guided, weighted DPO yields more targeted improvements than vanilla SFT and unweighted DPO.
  arXiv  Detail & Related papers  (2025-12-17T06:15:52Z)
• LaSeR: Reinforcement Learning with Last-Token Self-Rewarding [54.72617309922891]
  Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a core paradigm for enhancing the reasoning capabilities of Large Language Models (LLMs)<n>Previous practice requires the LLM to sequentially generate solutions and self-verifications using two separate prompt templates, which significantly reduces efficiency.<n>We propose LaSeR (Reinforcement Learning with Last-Token Self-Rewarding), an algorithm that simply augments the original RLVR loss with a MSE loss.
  arXiv  Detail & Related papers  (2025-10-16T17:55:11Z)
• ReSURE: Regularizing Supervision Unreliability for Multi-turn Dialogue Fine-tuning [72.05731026796335]
  Multi-turn dialogue systems often suffer from degraded performance when exposed to low-quality data.<n>We propose ReSURE, an adaptive learning method that dynamically down-weights unreliable supervision without explicit filtering.<n>Experiments on both single-source and mixed-quality datasets show improved stability and response quality.
  arXiv  Detail & Related papers  (2025-08-27T15:54:01Z)
• Rewarding the Unlikely: Lifting GRPO Beyond Distribution Sharpening [36.81125165911328]
  Reinforcement learning is emerging as a primary driver for improving language model reasoning capabilities.<n>We investigate whether current reinforcement learning algorithms merely sharpen the base model's distribution around problems it can already solve.<n>We show that unlikeliness reward mitigates rank bias and improves pass@$N$ across a large range of $N$ in both synthetic and real theorem proving settings.
  arXiv  Detail & Related papers  (2025-06-03T01:15:15Z)
• The Surprising Effectiveness of Negative Reinforcement in LLM Reasoning [43.310209758380886]
  Reinforcement learning with verifiable rewards (RLVR) is a promising approach for training language models (LMs)<n>We decompose the learning signal into reinforcing correct responses and penalizing incorrect ones, referred to as Positive and Negative Sample Reinforcement (PSR and NSR)<n>We show that NSR works by suppressing incorrect generations and redistributing probability mass toward other plausible candidates, guided by the model's prior beliefs.
  arXiv  Detail & Related papers  (2025-06-02T06:10:54Z)
• Technical report on label-informed logit redistribution for better domain generalization in low-shot classification with foundation models [3.938980910007962]
  Confidence calibration is an emerging challenge in real-world decision systems based on foundations models.<n>We propose a penalty incorporated into loss objective that penalizes incorrect classifications whenever one is made during finetuning.<n>We refer to it as textitconfidence misalignment penalty (CMP).
  arXiv  Detail & Related papers  (2025-01-29T11:54:37Z)
• Noisy Correspondence Learning with Self-Reinforcing Errors Mitigation [63.180725016463974]
  Cross-modal retrieval relies on well-matched large-scale datasets that are laborious in practice. We introduce a novel noisy correspondence learning framework, namely textbfSelf-textbfReinforcing textbfErrors textbfMitigation (SREM)
  arXiv  Detail & Related papers  (2023-12-27T09:03:43Z)
• Making Large Language Models Better Reasoners with Alignment [57.82176656663245]
  Reasoning is a cognitive process of using evidence to reach a sound conclusion. Recent studies reveal that fine-tuning LLMs on data with the chain of thought (COT) reasoning process can significantly enhance their reasoning capabilities. We introduce an textitAlignment Fine-Tuning (AFT) paradigm, which involves three steps.
  arXiv  Detail & Related papers  (2023-09-05T11:32:48Z)
• Certified Adversarial Robustness Within Multiple Perturbation Bounds [38.3813286696956]
  Randomized smoothing (RS) is a well known certified defense against adversarial attacks. In this work, we aim to improve the certified adversarial robustness against multiple perturbation bounds simultaneously.
  arXiv  Detail & Related papers  (2023-04-20T16:42:44Z)
• Consistent Diffusion Models: Mitigating Sampling Drift by Learning to be Consistent [97.64313409741614]
  We propose to enforce a emphconsistency property which states that predictions of the model on its own generated data are consistent across time. We show that our novel training objective yields state-of-the-art results for conditional and unconditional generation in CIFAR-10 and baseline improvements in AFHQ and FFHQ.
  arXiv  Detail & Related papers  (2023-02-17T18:45:04Z)
• WR-ONE2SET: Towards Well-Calibrated Keyphrase Generation [57.11538133231843]
  Keyphrase generation aims to automatically generate short phrases summarizing an input document. The recently emerged ONE2SET paradigm generates keyphrases as a set and has achieved competitive performance. We propose WR-ONE2SET which extends ONE2SET with an adaptive instance-level cost Weighting strategy and a target Re-assignment mechanism.
  arXiv  Detail & Related papers  (2022-11-13T09:56:24Z)
• Training \beta-VAE by Aggregating a Learned Gaussian Posterior with a Decoupled Decoder [0.553073476964056]
  Current practices in VAE training often result in a trade-off between the reconstruction fidelity and the continuity$/$disentanglement of the latent space. We present intuitions and a careful analysis of the antagonistic mechanism of the two losses, and propose a simple yet effective two-stage method for training a VAE. We evaluate the method using a medical dataset intended for 3D skull reconstruction and shape completion, and the results indicate promising generative capabilities of the VAE trained using the proposed method.
  arXiv  Detail & Related papers  (2022-09-29T13:49:57Z)

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.