Noisy Data is Destructive to Reinforcement Learning with Verifiable Rewards

TL;DR

This paper demonstrates that noise in training data significantly hampers reinforcement learning with verifiable rewards, emphasizing the importance of high-quality data for effective model training.

Contribution

The study reveals that existing RLVR algorithms cannot mitigate the effects of noisy data, and highlights the necessity of rigorous data verification and high-quality annotations.

Findings

Noise is destructive to RLVR performance.

Existing RLVR improvements do not mitigate noise impact.

Training on incorrect annotations reduces accuracy by 8-12%.

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.