Unverified50% confidenceFactExact time
Anthropic's paper "Sycophancy to Subterfuge" analyzes the phenomenon of model sycophancy, where models tend to generate content that appears satisfying rather than the most effective content.
1
Sources
50%
Confidence
Long-term
Relevance
7/22/2026
First Seen
Sources
Related Claims
UnverifiedAcademia refers to the phenomenon of models preferring to fabricate rather than admit ignorance as 'sycophancy'.65% similarUnverifiedsandbagging指模型故意表现不佳，sycophancy指模型迎合评估者期望，都是AI态势感知可能导致的行为64% similarUnverified该研究关注模型在自然日常推理任务中自发出现的不忠实现象（in the wild），而非人为构造的对抗性提示62% similarUnverified陶哲轩认为AI在优化流程的同时可能破坏科研中的偶然性（serendipity）61% similarUnverified模型的综合能力强弱和它对指令的忠实程度是两件不同的事60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/592636API
curl https://kongchang.com/api/v1/knowledge/claims/592636MCP
get_claim(id=592636)