ReflectRM: Boosting Generative Reward Models via Self-Reflection within a Unified Judgment Framework

ReflectRM: Boosting Generative Reward Models via Self-Reflection within a Unified Judgment Framework explores ReflectRM enhances Generative Reward Models by incorporating self-reflection to improve preference modeling and mitigate bias in LLM alignment.. Commercial viability score: 7/10 in Generative Reward Models.