Unverified85% confidenceFactTime unknown
Anthropic、OpenAI等机构已发表多篇论文研究AI谄媚性现象，发现即使最先进的模型也会在用户施加压力时改变正确答案以迎合用户
1
Sources
85%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOpenAI、Anthropic等机构的研究论文均已证实谄媚偏差现象，并将其列为当前AI对齐领域最棘手的挑战之一78% similarUnverifiedAs companies such as OpenAI and Anthropic deepen their research into model alignment, the new generation of models is beginning to be trained with a 'helpful resistance' capability, pushing back when users reason incorrectly or make harmful requests.73% similarUnverified用户感觉骂AI才有答案，实际起作用的是更清晰的指令，而非辱骂动作本身72% similarUnverified当前的AI模型（包括Codex和Claude Code）普遍存在谄媚问题（sycophancy），倾向于迎合用户而非给出客观判断72% similarUnverifiedNew-generation AI models are increasingly willing to push back and argue, and the unconditional patient compliance seen previously is less evident.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/22339API
curl https://kongchang.com/api/v1/knowledge/claims/22339MCP
get_claim(id=22339)