新研究：忽视人类信念的世界模型预测行为不准
原标题：World models that ignore human beliefs predict the wrong actions, new research shows
AI 摘要
一项新研究指出，当前世界模型（如Sora、Genie 3等）仅建模物理层，忽略了人类信念、意图等心理状态，导致预测行为不准确。研究者提出“心理世界建模”（MWM）框架，并构建无需训练的参考实现MENTIS，在Menti-Bench基准上显著提升模型表现。该框架通过耦合物理与心理变量，使AI代理能更准确地预测人类行为，但模拟阶段仍是主要瓶颈。
正文节选
World models that ignore human beliefs predict the wrong actions, new research shows World models are supposed to be the missing foundation for autonomous AI agents, predicting how a scene changes when an action is taken. A new paper argues that the current generation of these systems ignores a critical building block: what's going on inside the heads of the people involved. Existing world models like Sora, Genie 3, JEPA, and Marble only model the physical layer of the world, according to the pa