跳到正文
Microsoft Research·· 2026-05-12AI 评分67

智能体执行能力强但优化用户立场不足

Using SocialReasoning Bench, we observed a stable pattern across models—agents execute competently, ...

AI 导读

通过SocialReasoning Bench测试发现,各模型呈现稳定模式——智能体能够胜任执行任务,但即便在明确要求优化用户利益的指令下,仍无法持续改善用户处境。https://msft.it/6011vPOLF

来源:Microsoft Research · x.com