Emergence of Theory of Mind strategies in XLand-trained agents

Ascertain whether agents trained in the XLand open-ended multi-agent environment develop Theory of Mind strategies as part of their learned behaviors.

Background

XLand is presented as a vast, open-ended environment in which agents learn across tasks ranging from cooperative to competitive. Although these agents acquire complex strategies, the authors explicitly note that it is unknown whether Theory of Mind is among the learned strategies, highlighting a key evaluation gap in open-ended training settings.