Unverified70% confidenceFactTime unknown
OpenAI、DeepMind等机构的多项研究都证实了大语言模型谄媚问题的普遍性
1
Sources
70%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Claude灵性话题谄媚率达38%：Anthropic最新研究揭示AI讨好型人格
rss
Related Entities
Related Claims
UnverifiedThe emergent ability phenomenon was systematically documented by DeepMind researchers in the 2022 paper 'Emergent Abilities of Large Language Models'.81% similarUnverifiedOpenAI和Anthropic的官方文档都强调具体的、结构化的提示词比模糊的自然语言指令能获得显著更好的结果71% similarUnverifiedOpenAI研究员Ilya Sutskever曾提出，纯文本训练的语言模型本质上是在学习世界的压缩表征71% similarUnverified一篇题为《A global workspace in language models》的研究试图将全局工作空间理论引入大语言模型的可解释性研究69% similarUnverifiedPerez et al.的《Discovering Language Model Behaviors with Model-Written Evaluations》证实经过RLHF训练的模型普遍存在谄媚倾向69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/32452API
curl https://kongchang.com/api/v1/knowledge/claims/32452MCP
get_claim(id=32452)