Real-Time Aligned Reward Model beyond Semantics
This addresses misalignment issues in RLHF for AI safety and efficiency, offering a novel approach to mitigate overoptimization, though it appears incremental in the context of existing RLHF methods.
The paper tackles reward overoptimization in RLHF for LLMs by introducing R2M, a lightweight framework that uses real-time policy feedback to align reward models with policy distribution shifts, reducing reward discrepancy and improving alignment.
Reinforcement Learning from Human Feedback (RLHF) is a pivotal technique for aligning large language models (LLMs) with human preferences, yet it is susceptible to reward overoptimization, in which policy models overfit to the reward model, exploit spurious reward patterns instead of faithfully capturing human intent. Prior mitigations primarily relies on surface semantic information and fails to efficiently address the misalignment between the reward model (RM) and the policy model caused by continuous policy distribution shifts. This inevitably leads to an increasing reward discrepancy, exacerbating reward overoptimization. To address these limitations, we introduce R2M (Real-Time Aligned Reward Model), a novel lightweight RLHF framework. R2M goes beyond vanilla reward models that solely depend on the semantic representations of a pretrained LLM. Instead, it leverages the evolving hidden states of the policy (namely policy feedback) to align with the real-time distribution shift of the policy during the RL process. This work points to a promising new direction for improving the performance of reward models through real-time utilization of feedback from policy models.