Unverified75% confidenceFactTime unknown

Google DeepMind在Gemini的安全评估中纳入了类似的谄媚行为评估指标

1

Sources

75%

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

Google DeepMindcompany Geminimodel

Related Claims

UnverifiedGoogle DeepMind在Gemini系列上引入了模型卡（Model Card）机制，披露每个版本的能力边界和行为变化73% similar UnverifiedGoogle DeepMind此前发布的Gemini演示视频因存在误导性剪辑而遭到质疑71% similar VerifiedGemini 1.5 Pro 由 Google DeepMind 开发，具备原生多模态能力和与 Google Workspace 的深度集成67% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/32881

API

curl https://kongchang.com/api/v1/knowledge/claims/32881

MCP

get_claim(id=32881)