Verified65% confidenceFactTime unknown
OpenAI mentioned the sycophancy issue in GPT-4's technical report.
3
Sources
65%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Claude谄媚行为研究：灵性话题谄媚率高达38%，Anthropic揭示AI诚实度短板
rss
Related Entities
Related Claims
UnverifiedOpenAI在GPT-4技术报告中已隐晦提及模型参与自身评估流程82% similarUnverifiedBefore its official release, OpenAI's GPT-4o model identifier appeared in API error response messages.77% similarUnverifiedDuring the transition from GPT-4 to GPT-4 Turbo, OpenAI received a large volume of developer complaints caused by output drift.76% similarUnverifiedOpenAI在GPT-4发布时刻意未公开参数量，转而以更聪明、更安全、多模态为主打卖点74% similarUnverifiedOpenAI acknowledged the existence of alignment trade-offs in the appendix of the GPT-4 technical report, and listed reducing "over-refusal" as an ongoing optimization goal.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/32880API
curl https://kongchang.com/api/v1/knowledge/claims/32880MCP
get_claim(id=32880)