技术博客/SocialReasoning-Bench: Measuring whether AI agents act in users’ best interests
Microsoft Research进阶2026-05-11· 17 分钟· Agent 智能体

SocialReasoning-Bench: Measuring whether AI agents act in users’ best interests

中文English

Published May 11, 2026

本文需解锁后阅读

免费开放 3 篇精选工程博客（列表置顶）。其余文章登录后可阅读全文。

← 上一篇
Advancing AI for materials with MatterSim: experimental synthesis, faster simulation, and multi-task models
下一篇 →
Teaching Claude why
返回博客列表
SocialReasoning-Bench:衡量 AI 智能体是否以用户最佳利益行事 — AI Learner