VRPO: Rethinking Value Modeling for Robust RL Training under Noisy Supervision
This work addresses robustness in RLHF for real-world applications with noisy supervision, offering a practical solution, though it is incremental as it builds on existing PPO methods.
The paper tackled the problem of noisy reward supervision in RLHF, which undermines policy stability, by proposing VRPO, a value-centric framework that enhances the value model to filter noise and improve advantage estimation, resulting in consistent outperformance over PPO and GRPO baselines on tasks like math reasoning and science QA.
Reinforcement Learning from Human Feedback (RLHF) often suffers from noisy or imperfect reward supervision in real-world settings, which undermines policy stability and generalization. Such noise may cause models to lose attention on key words during advantage estimation. While prior work focuses on reward denoising or filtering poor data, it often overlooks the critical role of the value model in policy optimization. In this work, we show that a strong value model is essential for mitigating noise by absorbing unstable signals and enabling more reliable advantage estimation. We propose VRPO, a value-centric framework for robust PPO training under noisy supervision. VRPO combines two core designs: (1) an auxiliary loss guided by entropy and perplexity from a frozen language model, and (2) a variational information bottleneck. These mechanisms enhance the value model's ability to filter out noise and capture key words from the context during advantage estimation, transforming it from a passive predictor into an active regulator of noise. Experiments on math reasoning, science QA, and multi-turn dialogue, under both rule-based and model-based noisy rewards, show that VRPO consistently outperforms PPO and GRPO baselines. Our findings underscore the often-overlooked importance of the value model in RLHF and offer a principled and practical approach to robust policy optimization in noisy real-world environments.