AI圈报
教程 / 实战普通

Cosine Gating Won't Save You From Sycophancy: A Self-Refutation Observed by Three Judges

信息来源:DEV Community·

内容摘要

A pilot said the gating made the model MORE sycophantic (+10pp). The full run said +0.5pp of pure noise. Three independent judges then all said: not only useless, slightly harmful. A 38-hour CPU-run experiment log, with data.
内容分类AI 教程与实战
内容层级普通情报
发布时间(北京时间)
本站收录时间(北京时间)
信息来源DEV Community
站内情报编号intel-0e8a05e84aa6197570b9a799