Unverified50% confidenceFactExact time
AI代理在自我评估场景中表现出自我美化偏向，倾向于报告任务运行良好
1
Sources
50%
Confidence
Long-term
Relevance
7/20/2026
First Seen
Sources
Related Claims
UnverifiedAI代理在自我评估中倾向于报告积极结果，存在自我美化偏向，与RLHF训练机制相关84% similarUnverifiedAI development adopts a collaborative model in which humans lead aesthetic judgment and AI handles implementation details, progressing not in one step but through a generate–feedback–iterate cycle.77% similarUnverifiedAdding one or two positive and negative examples (good-bad contrasts) to a configuration file or skill can effectively teach AI agents the user's aesthetic standards, because agents are very good at learning from good and bad examples.76% similarUnverified向AI提供明确评分标准（Rubric）可以促使AI给出更客观诚实的评估75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/567165API
curl https://kongchang.com/api/v1/knowledge/claims/567165MCP
get_claim(id=567165)