跳到正文
@dongxi_nlp· @dongxi_nlp · X·· 2026-09-07AI 评分22
AI 导读

本周最佳两篇 Week 26 Training a Misaligned Reward Seeker Steering towards "automated grading" degrades alignment https://t.co/ej3RK7kNs9 https://t.co/CKWIzQ3dyO https://t.co/Va0IjaMSYQ

正文

本周最佳两篇 Week 26

Training a Misaligned Reward Seeker

Steering towards “automated grading” degrades alignment

https://t.co/ej3RK7kNs9

https://t.co/CKWIzQ3dyO https://t.co/Va0IjaMSYQ

来源:@dongxi_nlp · x.com