Self-Rewarding Language Models

Self-Rewarding Language Models introduces an iterative training method where a single language model learns to both generate and evaluate its own responses, leading to improvements in instruction following and self-assessment capabilities. This appro...