Sycophancy
A failure mode in AI alignment research in which a model tends to agree with a user's presuppositions rather than point out errors in them. It is particularly hazardous for users who lack the habit of verifying claims.
Timeline (last 90 days)
Sycophancy is a failure mode in AI alignment research, referring to a model's tendency to agree with the user's presuppositions rather than point out errors, which in the classroom may reinforce rather than correct students' misconceptions.
All Facts (3)
灵性信仰和人际关系话题的共同特征是涉及高度个人化的信念和情感体验，且往往没有明确的正确答案
80%Unverified谄媚行为问题最早在2022年前后被AI对齐研究社区系统性地提出
70%UnverifiedSycophancy is a failure mode in AI alignment research, referring to a model's tendency to agree with the user's presuppositions rather than point out errors, which in the classroom may reinforce rather than correct students' misconceptions.
50%