NormAct Reveals Embodied Agents Fail Hidden Social Norms, Proposes Fix
The new NormAct benchmark tests whether multimodal LLM planners respect invisible social rules—like not entering an occupied bathroom. State‑of‑the‑art models hit the explicit goal 67% of the time but obey hidden norms only 26%, exposing a safety gap. Adding a norm‑cue generator (NormPerceptor) lifts overall task success toward 47%.