论文32

Carbon 在 CritPt 物理评测中表现最佳

原标题:Carbon performs best on CritPt (a graduate-level physics evaluation) but loses out in agentic research and knowledge work. This is driven...

Artificial Analysis@ArtificialAnlys发布时间 (北京时间)

Carbon 在 CritPt(研究生水平的物理评测)上表现最佳,但在智能体研究和知识工作方面落败。这是因为相比 Neon 和 Hydrogen,它在更高推理层级上的提升有限。

中文译文尚未完成,先为你展示原文。

Carbon performs best on CritPt (a graduate-level physics evaluation) but loses out in agentic research and knowledge work. This is driven by its limited improvement at higher reasoning levels compared to Neon and Hydrogen.

推理能力评测基准#评测/基准#推理