Noisy Data is Destructive to Reinforcement Learning with Verifiable Rewards

arXiv:2603.16140v2 Announce Type: replace Abstract: Reinforcement learning with verifiable rewards (RLVR) has driven recent capability advances of large language models across various domains. Recent studies suggest that improved RLVR algorithms allow models to learn effectively from incorrect annotations, achieving performance comparable to learning from clean data. In this work, we show that these findings are invalid because the claimed 100% noisy training data is "contaminated" with clean data. After rectifying the dataset with a rigorous re-verification pipeline, we demonstrate that noise is destructive to RLVR. We show that existing RLVR algorithm improvements fail to mitigate the impact of noise, achieving similar performance to that of the basic GRPO. Furthermore, we find that the model trained on truly incorrect annotations performs 8-10% worse than the model trained on clean data across mathematical reasoning benchmarks. Finally, we show that these findings hold for real-world noise in Text2SQL tasks, where training on real-world, human annotation errors cause 5-12% lower accuracy than clean data. Our results show that current RLVR methods cannot yet compensate for poor data quality. High-quality data remains essential.

Related

• Provable Benefits of RLVR over SFT for Reasoning Models: Learning to Backtrack Efficiently
• A Survey of Reinforcement Learning for Large Language Models under Data Scarcity: Challenges and Solutions
• MulFeRL: Enhancing Reinforcement Learning with Verbal Feedback in a Multi-turn Loop
• Contrastive Reinforced Policy Optimization via Privileged Self-Distillation

Source: arXiv cs.LG | 2026-07-31