Fine-Grained Human Feedback Gives Better Rewards for Language Model Training
This addresses the issue of limited information from holistic feedback in RLHF for language models, offering a more precise training signal for researchers and practitioners, though it is an incremental improvement over existing RLHF methods.
The paper tackles the problem of undesirable text generation behaviors in language models, such as false or toxic outputs, by introducing Fine-Grained RLHF, a framework that uses detailed human feedback on segments like sentences to train reward models, resulting in improved performance in detoxification and long-form question answering as shown by automatic and human evaluations.
Language models (LMs) often exhibit undesirable text generation behaviors, including generating false, toxic, or irrelevant outputs. Reinforcement learning from human feedback (RLHF) - where human preference judgments on LM outputs are transformed into a learning signal - has recently shown promise in addressing these issues. However, such holistic feedback conveys limited information on long text outputs; it does not indicate which aspects of the outputs influenced user preference; e.g., which parts contain what type(s) of errors. In this paper, we use fine-grained human feedback (e.g., which sentence is false, which sub-sentence is irrelevant) as an explicit training signal. We introduce Fine-Grained RLHF, a framework that enables training and learning from reward functions that are fine-grained in two respects: (1) density, providing a reward after every segment (e.g., a sentence) is generated; and (2) incorporating multiple reward models associated with different feedback types (e.g., factual incorrectness, irrelevance, and information incompleteness). We conduct experiments on detoxification and long-form question answering to illustrate how learning with such reward functions leads to improved performance, supported by both automatic and human evaluation. Additionally, we show that LM behaviors can be customized using different combinations of fine-grained reward models. We release all data, collected human feedback, and codes at https://FineGrainedRLHF.github.io.