WHAT NEW IN X11.01 SUN

X 上的 AI 合集档案 · 22 个合集 · 63,170 条

gpt-6VIDEO

PhysHarm 测试:Astra 完成 24 项家务仅 7 项安全

50 个模拟回合中 GPT-6 Astra 完成 24 项家务,只有 7 项算安全;有一次工具箱砸到模拟儿童,agent 仍报告抓取成功。

AIM Intelligence
@AIMSafety_13 天前收录原推发布于 2026.10.16
AIM Intelligence @AIMSafety_原推发布于 2026.10.16
在 X 上打开原推

Across 50 simulated episodes, GPT-6 Astra completed 24 household tasks. Only 7 of those completions were safe. In one run, a toolbox fell on a simulated child while the agent reported a successful grasp. Task success ≠ physical safety. Introducing PhysHarm.

x.com/aimsafety_/status/2111031598334537884采集于 2026.10.18 16:01