# Noise-corrected GRPO: From Noisy Rewards to Unbiased Gradients

arXiv: https://arxiv.org/abs/2510.18924 · HTML: https://mlsanity.com/paper/2510.18924/

## Key facts

- Cited by 13 papers in the map (data through 14 Oct 2026).
- Placed in Reinforcement learning for reasoning, one of 17 directions in LLM reasoning.
- Builds on 23 of its 28 references that are in the map.
- First posted on arXiv 21 Oct 2025.

## Placed in

- LLM reasoning / [Reinforcement learning for reasoning](https://mlsanity.com/lr/reinforcement-learning-for-reasoning/)
- LLM alignment / RLHF / preference optimisation / [Process reward models for reasoning](https://mlsanity.com/alignment/process-reward-models-for-reasoning/)

## Builds on (most cited first, 23 of 23)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Measuring Mathematical Problem Solving With the MATH Dataset](https://mlsanity.com/paper/2103.03874/) (2021)
- [DAPO: An Open-Source LLM Reinforcement Learning System at Scale](https://mlsanity.com/paper/2503.14476/) (2025)
- [Fine-Tuning Language Models from Human Preferences](https://mlsanity.com/paper/1909.08593/) (2019)
- [Understanding R1-Zero-Like Training: A Critical Perspective](https://mlsanity.com/paper/2503.20783/) (2025)
- [Tulu 3: Pushing Frontiers in Open Language Model Post-Training](https://mlsanity.com/paper/2411.15124/) (2024)
- [Measuring Coding Challenge Competence With APPS](https://mlsanity.com/paper/2105.09938/) (2021)
- [REINFORCE++: Stabilizing Critic-Free Policy Optimization with Global Advantage Normalization](https://mlsanity.com/paper/2501.03262/) (2025)
- [Reinforcement Learning with Verifiable Rewards: GRPO's Effective Loss, Dynamics, and Success Amplification](https://mlsanity.com/paper/2503.06639/) (2025)
- [One Token to Fool LLM-as-a-Judge](https://mlsanity.com/paper/2507.08794/) (2025)
- [Diverging Preferences: When do Annotators Disagree and do Models Know?](https://mlsanity.com/paper/2410.14632/) (2024)
- [Revisiting Group Relative Policy Optimization: Insights into On-Policy and Off-Policy Training](https://mlsanity.com/paper/2505.22257/) (2025)
- [Reward-Robust RLHF in LLMs](https://mlsanity.com/paper/2409.15360/) (2024)
- [Robust Reinforcement Learning from Corrupted Human Feedback](https://mlsanity.com/paper/2406.15568/) (2024)
- [VerifyBench: A Systematic Benchmark for Evaluating Reasoning Verifiers Across Domains](https://mlsanity.com/paper/2507.09884/) (2025)
- [A Baseline Analysis of Reward Models' Ability To Accurately Analyze Foundation Models Under Distribution Shift](https://mlsanity.com/paper/2311.14743/) (2023)
- [Adversarial Training of Reward Models](https://mlsanity.com/paper/2504.06141/) (2025)
- [DisCO: Reinforcing Large Reasoning Models with Discriminative Constrained Optimization](https://mlsanity.com/paper/2505.12366/) (2025)
- [Reward Model Interpretability via Optimal and Pessimal Tokens](https://mlsanity.com/paper/2506.07326/) (2025)

## Built on by (most cited first, 13 of 13)

- [An Imperfect Verifier is Good Enough: Learning with Noisy Rewards](https://mlsanity.com/paper/2604.07666/) (2026)
- [Noisy Data is Destructive to Reinforcement Learning with Verifiable Rewards](https://mlsanity.com/paper/2603.16140/) (2026)
- [Are Verifier Errors Independent Within a GRPO Group? Evidence from Qwen2.5 Rollouts](https://mlsanity.com/paper/2609.06386/) (2026)
- [MemReward: Graph-Based Experience Memory for LLM Reward Prediction with Limited Labels](https://mlsanity.com/paper/2603.19310/) (2026)
- [Advantage Scale Calibration Imbalance in Group-Relative Optimization under Low-Variance Rewards: Diagnosis and Bounded Recovery](https://mlsanity.com/paper/2609.19164/) (2026)
- [Constraint-Rectified Training for Efficient Chain-of-Thought](https://mlsanity.com/paper/2602.12526/) (2026)
- [Aligning to Illusions: Choice Blindness in Human and AI Feedback](https://mlsanity.com/paper/2603.08412/) (2026)
- [Multi-Stakeholder LLM Alignment: Decomposing Estimation from Aggregation](https://mlsanity.com/paper/2605.26878/) (2026)
- [It Takes One to Bias Them All: Breaking Bad with One-Shot GRPO](https://mlsanity.com/paper/2606.10931/) (2026)
- [Marginally Correct Tool Caches Can Reverse Group-Normalized Policy Updates](https://mlsanity.com/paper/2609.26866/) (2026)
- [A Cheap Verifier is Good Enough: LLM Post-training is Robust to Erroneous Rewards](https://mlsanity.com/paper/2609.33467/) (2026)
- [Many Right Answers, One Training Example: Removing Responder Noise from GRPO for LLM Recommenders](https://mlsanity.com/paper/2610.17771/) (2026)
- [Checking the Checkers: A Synthesis of Verifier Noise and Robustness in RLVR](https://mlsanity.com/paper/2610.17824/) (2026)

## Abstract (arXiv)

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.