Best-of-N Jailbreaking
"Best-of-N" 相关文章
Researchers at Google Deepmind Introduce BOND: A Novel RLHF Method that Fine-Tunes the Policy via Online Distillation of the Best-of-N Sampling Distribution
