跳到正文

Fable 5.2 登顶科研智能体基准

摘要

Fable 5.2 现已成为智能体科研的 SOTA Fable 5.2 在 Terminal-Bench-Science 0.1 上得分 74.6%,较 Fable 5.1(52.6%)大幅跃升,并领先于 Opus 5.5(58.7%)和 GPT-6 Astra(68.1%) 使用 Fable 5.2 的 Autoresearch 现已上线 https://openresearch.sh/

摘要由模型根据原文撰写,以原文为准。

正文 · 原文

Fable 5.2 is now SOTA for agentic scientific research

Fable 5.2 scores 74.6% on Terminal-Bench-Science 0.1, a large leap from Fable 5.1 (52.6%) and ahead of Opus 5.5 (58.7%) and GPT-6 Astra (68.1%)

Autoresearch with Fable 5.2 now on https://openresearch.sh/

来源:alphaXiv · x.com

引用原文

askalphaxiv. Fable 5.2 is now SOTA for agentic scientific research[EB/OL]. (2026-11-10)[访问日期]. https://x.com/askalphaxiv/status/2119881403966116069.