Verified65% confidenceFactExact time
In 2023, multiple organizations including Anthropic and OpenAI published research reports confirming that sycophancy is prevalent in mainstream models
3
Sources
65%
Confidence
Long-term
Relevance
7/20/2026
First Seen
Sources
Related Claims
UnverifiedAnthropic、OpenAI等研究团队均已发表论文记录模型的奉承偏见现象75% similarUnverifiedAnthropic研究团队2023年发表了关于奉承性偏差的系统性分析72% similarUnverifiedIn a 2023 research paper, Anthropic formally named the phenomenon of models being overly compliant with users the "sycophancy problem" and identified it as a nearly inevitable systematic bias in RLHF training.67% similarUnverifiedIn 2024, Anthropic published a research paper specifically addressing the over-refusal problem, acknowledging an unresolved tension between safety and utility.66% similarUnverifiedIn 2024, multiple research institutions found that relying solely on RLHF may cause models to learn to please annotators rather than genuinely becoming more honest.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/572484API
curl https://kongchang.com/api/v1/knowledge/claims/572484MCP
get_claim(id=572484)