Event ·Multi-source

六款前沿模型实测 LIBERO-10 机器人操作基准,GPT-6 Astra 以 44% 成功率居首

一项研究构建了统一闭环操作框架,在不做任务微调的前提下,让 GPT-6 Astra、Claude Opus 5.5、Gemini-4 Argon 等 6 款前沿模型直接控制机器人完成 LIBERO-10 基准任务。测试覆盖 3 档推理强度、共 18 种配置和 1,800 次评估,GPT-6 Astra 以 44% 成功率排名第一,Gemini-4 Argon 为 43%。提高推理强度总体能提升操作表现,但各模型受益程度不同且推理成本更高。模型在物体放置类任务上表现较好,在精确空间排布和多步骤交互上仍有明显短板。

Current read

研究者让 GPT-6、Claude Opus 5.5 这些模型直接操控真实机器人跑 LIBERO-10,1,800 次测试排出了名次,最强也只有 44% 成功率,想了解大模型动手能力差在哪的可以看。

8 sources44° AI heatLast updated 2026/10/21 17:11:52

Evidence

8 sources
Primary sourcearXiv cs.AI
View source
Related source 1lmarena.ai
View source
Related source 2Artificial Analysis
View source
Related source 3Bindu Reddy
View source
Related source 5TestingCatalog
View source
Related source 6rohanpaul_ai
View source

Conflict 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.