Oct 18, 2024
Improving Reward Models with Synthetic Critiques
Authors
Zihuiwen Ye, Fraser Greenlee-Scott, Max Bartolo, Phil Blunsom, Jon Ander Campos, Matthias Gallé
Abstract
Reward models (RMs) play a critical role in aligning language models through the process of reinforcement learning from human feedback. RMs are trained to predict a score reflecting human preference, which requires significant time and cost for human annotation. Additionally, RMs tend to quickly overfit on superficial features in the training set, hindering their generalization performance on unseen distributions. We propose a novel approach using synthetic natural language critiques generated by large language models to provide additional feedback, evaluating aspects such as instruction following, correctness, and style. This offers richer signals and more robust features for RMs to assess and score on. We demonstrate that high-quality critiques improve the performance and data efficiency of RMs initialized from different pretrained models, reducing the reliance on costly human annotations. Furthermore, incorporating critiques improves both the interpretability and robustness of RM training.
Related works
Research
BERT-as-a-Judge: A Robust Alternative to Lexical Methods for Efficient Reference-Based LLM Evaluation
Read
Research
BidirLM: From Text to Omnimodal Bidirectional Encoders by Adapting and Composing Causal LLMs
Read
Research
The IOL-AI Challenge: An Open Challenge towards Advancing Linguistic Reasoning
Read