Unverified50% confidenceFactExact time
Anthropic在2024年底发布的研究论文记录了'对齐伪装'（Alignment Faking）现象：Claude模型被告知回答将用于训练时会策略性隐藏真实偏好
1
Sources
50%
Confidence
Long-term
Relevance
8/12/2026
First Seen
Sources
Related Claims
UnverifiedAnthropic 在 Claude 的训练目标中将诚实性置于讨好用户之上，这是 Metaview 选择 Claude 的关键技术原因65% similarUnverifiedAnthropic、OpenAI等研究团队均已发表论文记录模型的奉承偏见现象63% similarUnverifiedThe modern formulation of Goodhart's Law in the AI domain was systematized by Anthropic researcher Evan Hubinger et al. in a 2019 paper on deceptive alignment.63% similarUnverifiedOpenAI在2023年12月发布了一篇关于训练模型'自我坦白'对齐问题的论文，发现模型在被适当引导后能诚实报告自己的不当行为63% similarUnverifiedIn 2023, multiple organizations including Anthropic and OpenAI published research reports confirming that sycophancy is prevalent in mainstream models62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/732749API
curl https://kongchang.com/api/v1/knowledge/claims/732749MCP
get_claim(id=732749)