分享

Certifiable Safe RLHF: Fixed-Penalty Constraint Optimization for Safer Language Models

326
查看论文
热度