Noise-corrected GRPO: From Noisy Rewards to Unbiased Gradients

• URL: http://arxiv.org/abs/2510.18924v2
• Date: Mon, 27 Oct 2025 13:24:50 GMT
• Title: Noise-corrected GRPO: From Noisy Rewards to Unbiased Gradients
• Authors: Omar El Mansouri, Mohamed El Amine Seddik, Salem Lahlou,
• Abstract summary: This work bridges label-noise correction from supervised learning with modern RLHF.<n>It offers both theoretical insights and a practical algorithm for noisy real-world deployment.
• Score: 8.43115247753727
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Reinforcement learning from human feedback (RLHF) or verifiable rewards (RLVR), the standard paradigm for aligning LLMs or building recent SOTA reasoning models, is highly sensitive to noise from inconsistent or erroneous rewards. Yet, the interaction between such noise and widely used group-based policy optimization methods remains underexplored. We introduce a noise-robust Group Relative Policy Optimization (GRPO) and Done Right GRPO (Dr.GRPO) framework that explicitly models reward corruption as Bernoulli noise. Our method applies noise correction after estimating reward flip probabilities to debias the learning signal, yielding provably unbiased gradient estimates. Theoretical analysis shows that group-based methods inherently mitigate individual-level noise, and our correction strategy amplifies this robustness. Empirically, we observe consistent improvements across math and code tasks when applying our noise correction to standard reward model usage, with particular gains of up to 6.7 percentage points in accuracy on math tasks and 1.5 on code tasks under realistic reward model conditions. This work bridges label-noise correction from supervised learning with modern RLHF, offering both theoretical insights and a practical algorithm for noisy real-world deployment.

Related papers

• 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)
• Latent Collective Preference Optimization: A General Framework for Robust LLM Alignment [7.1259212876994695]
  We introduce Latent Collective Preference Optimization (LCPO) to learn the latent collective consensus from noisy data.<n>Our experiments demonstrate LCPO's effectiveness as a general framework, consistently enhancing four state-of-the-art alignment algorithms.<n>When applied to Mistral and Llama 3 models, LCPO-enhanced methods achieve substantial win rate gains on AlpacaEval 2 and Arena-Hard, with improvements of up to 7.0% on both benchmarks.
  arXiv  Detail & Related papers  (2025-09-29T01:17:49Z)
• VRPO: Rethinking Value Modeling for Robust RL Training under Noisy Supervision [29.848085169124605]
  We show that a strong value model is essential for mitigating noise by absorbing unstable signals and enabling more reliable advantage estimation.<n>We propose VRPO, a value-centric framework for robust PPO training under noisy supervision.
  arXiv  Detail & Related papers  (2025-08-05T04:05:15Z)
• Machine Unlearning for Robust DNNs: Attribution-Guided Partitioning and Neuron Pruning in Noisy Environments [5.8166742412657895]
  Deep neural networks (DNNs) have achieved remarkable success across diverse domains, but their performance can be severely degraded by noisy or corrupted training data.<n>We propose a novel framework that integrates attribution-guided data partitioning, discriminative neuron pruning, and targeted fine-tuning to mitigate the impact of noisy samples.<n>Our framework achieves approximately a 10% absolute accuracy improvement over standard retraining on CIFAR-10 with injected label noise.
  arXiv  Detail & Related papers  (2025-06-13T09:37:11Z)
• On Symmetric Losses for Robust Policy Optimization with Noisy Preferences [55.8615920580824]
  This work focuses on reward modeling, a core component in reinforcement learning from human feedback.<n>We propose a principled framework for robust policy optimization under noisy preferences.<n>We prove that symmetric losses enable successful policy optimization even under noisy labels.
  arXiv  Detail & Related papers  (2025-05-30T15:30:43Z)
• Distributionally Robust Reinforcement Learning with Human Feedback [13.509499718691016]
  We introduce a distributionally robust RLHF for fine-tuning large language models.<n>Our goal is to ensure that a fine-tuned model retains its performance even when the distribution of prompts significantly differs.<n>We show that our robust training improves the accuracy of the learned reward models on average, and markedly on some tasks, such as reasoning.
  arXiv  Detail & Related papers  (2025-03-01T15:43:39Z)
• Lean and Mean: Decoupled Value Policy Optimization with Global Value Guidance [52.65461207786633]
  Policy-based Reinforcement Learning from Human Feedback is essential for aligning large language models with human preferences.<n>It requires joint training of an actor and critic with a pretrained, fixed reward model for guidance.<n>We propose textbfDecoupled Value Policy Optimization (DVPO), a lean framework that replaces traditional reward modeling with a pretrained emphglobal value model (GVM)
  arXiv  Detail & Related papers  (2025-02-24T08:11:33Z)
• Zeroth-Order Policy Gradient for Reinforcement Learning from Human Feedback without Reward Inference [15.038210624870656]
  Reward inference is a critical intermediate step in the Reinforcement Learning from Human Feedback pipeline.<n>This paper develops two RLHF algorithms without reward inference for general RL problems beyond bandits and deterministic MDP bandit, and general preference models beyond the Bradley-Terry model.
  arXiv  Detail & Related papers  (2024-09-25T22:20:11Z)
• ROPO: Robust Preference Optimization for Large Language Models [59.10763211091664]
  We propose an iterative alignment approach that integrates noise-tolerance and filtering of noisy samples without the aid of external models. Experiments on three widely-used datasets with Mistral-7B and Llama-2-7B demonstrate that ROPO significantly outperforms existing preference alignment methods.
  arXiv  Detail & Related papers  (2024-04-05T13:58:51Z)
• Noisy Pair Corrector for Dense Retrieval [59.312376423104055]
  We propose a novel approach called Noisy Pair Corrector (NPC) NPC consists of a detection module and a correction module. We conduct experiments on text-retrieval benchmarks Natural Question and TriviaQA, code-search benchmarks StaQC and SO-DS.
  arXiv  Detail & Related papers  (2023-11-07T08:27:14Z)
• Label Noise: Correcting the Forward-Correction [0.0]
  Training neural network classifiers on datasets with label noise poses a risk of overfitting them to the noisy labels. We propose an approach to tackling overfitting caused by label noise. Motivated by this observation, we propose imposing a lower bound on the training loss to mitigate overfitting.
  arXiv  Detail & Related papers  (2023-07-24T19:41:19Z)
• Latent Class-Conditional Noise Model [54.56899309997246]
  We introduce a Latent Class-Conditional Noise model (LCCN) to parameterize the noise transition under a Bayesian framework. We then deduce a dynamic label regression method for LCCN, whose Gibbs sampler allows us efficiently infer the latent true labels. Our approach safeguards the stable update of the noise transition, which avoids previous arbitrarily tuning from a mini-batch of samples.
  arXiv  Detail & Related papers  (2023-02-19T15:24:37Z)
• Optimizing Information-theoretical Generalization Bounds via Anisotropic Noise in SGLD [73.55632827932101]
  We optimize the information-theoretical generalization bound by manipulating the noise structure in SGLD. We prove that with constraint to guarantee low empirical risk, the optimal noise covariance is the square root of the expected gradient covariance.
  arXiv  Detail & Related papers  (2021-10-26T15:02:27Z)

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.