NormGym：通过社交互动评估智能体的规范能力
原标题：NormGym: Evaluating Normative Competence through Social Interaction
AI 摘要
研究团队提出 NormGym，一个用于评估智能体规范能力（normative competence）的多智能体交互基准：新加入的智能体需在完成自身任务的同时，从其他智能体的行为和公开制裁中推断出社区隐藏的任意规范，并在新情境中应用。实验显示最强 LLM 的任务与合规联合成功率仅 55.3%，模型能根据自身受到的制裁调整行为，却难以利用他人的行为和制裁推断规范，微调也未必能改善该能力。
正文节选
NormGym: Evaluating Normative Competence through Social Interaction **footnotetext: Equal advising. Abstract. Normative competence, the ability to infer socially enforced expectations from interaction and adjust behavior accordingly, is essential for agents operating in unfamiliar human communities. Yet existing evaluations largely test whether agents can recognize or apply norms they may already know, rather than whether they can acquire previously unknown norms from social interaction. We intr