1 paper
Tianqi Liu, Wei Xiong, Jie Ren +15
Reward models (RMs) play a pivotal role in aligning large language models (LLMs) with human preferences. However, traditional RM training, which relies on response pairs tied to sp…