RED: Unleashing Token-Level Rewards from Holistic Feedback via Reward Redistribution

ACL ARR 2025 May Submission381 Authors

Abstract: Reinforcement learning from human feedback (RLHF) offers a promising approach to aligning large language models (LLMs) with human preferences. Typically, a reward model is trained or supplied to act as a proxy for humans in evaluating generated responses during the reinforcement training phase. However, current reward models operate as sequence-to-one models, allocating a single, sparse, and delayed reward to an entire output sequence. This approach may overlook the significant contributions of individual tokens toward the desired outcome. To this end, we propose a more fine-grained, token-level guidance approach for RL training. Specifically, we introduce RED, a novel REward reDistribition method that evaluates and assigns specific credit to each token using an off-the-shelf reward model. Utilizing these fine-grained rewards enhances the model's understanding of language nuances, leading to more precise performance improvements. Notably, our method does not require modifying the reward model or introducing additional training steps, thereby incurring minimal computational costs. Experimental results across diverse datasets and tasks demonstrate the superiority of our approach.

Paper Type: Long

Research Area: Machine Learning for NLP

Research Area Keywords: reward redistribution; reinforcement learning from human feedback

Contribution Types: Theory

Languages Studied: English

Keywords: RLHF, Large Language Model, Reward Redistribution

Submission Number: 381