Primary sourceArtificial Analysis
View sourceEvent ·Multi-source
AA-PersonalityBench 发布:Grok 4.7 阿谀度最低,Meta 模型普遍偏高
Artificial Analysis 发布新基准 AA-PersonalityBench,用 Sycophantic(阿谀)人格维度给各家模型打分(0-100)。最老的三款模型 GPT-4o、Llama 4 Maverick、GPT-3.5 Turbo 在该维度排名最高,其中 Llama 4 Maverick 得 97 分。Meta 全部五款受测模型均达 65 分以上,Anthropic 的 Claude 系列全部在 55 分以下,Claude Fable 5.1 仅 4 分,Grok 4.7 是所有受测模型中最低的。
Current read
新基准专门测模型多爱顺着你说,Llama 4 Maverick 97 分、Grok 4.7 最低,Claude 全系都偏低,想挑一款不谄媚的模型可以看看这份排名。
11 sources51° AI heatLast updated 2026/10/21 17:35:18
Evidence
11 sourcesRelated source 1宝玉
View sourceRelated source 2ARC Prize
View sourceRelated source 3Teortaxes
View sourceRelated source 4Decoder
View sourceRelated source 5The Rundown AI
View sourceRelated source 6AWS Machine Learning Blog
View sourceRelated source 7rohanpaul_ai
View sourceRelated source 8The Information
View sourceRelated source 9kimmonismus
View sourceRelated source 10Andrew Curran
View sourceConflict check
Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.