Artificial Analysis· ArtificialAnlys·· 7 小时前AI 评分18
三模型并列65分，Elephant等尝试率各异
All three models tied at 65 hallucinate the least on AA-Omniscience-Decision questions. Elephant 9B attempts 18% of questions, SubQ Snap ...
AI 导读
三个模型在 AA-Omniscience-Decision 问题上幻觉最少，均得 65 分。Elephant 9B 尝试了 18% 的问题，SubQ Snap 1.1 为 32%，d1-ultra 为 42%
来源：Artificial Analysis · x.com