BeliefNest: A Joint Action Simulator for Embodied Agents with Theory of Mind
This addresses the problem of improving collaborative AI agents for researchers by providing a tool for Theory of Mind reasoning, though it appears incremental as it builds on existing concepts in embodied AI and simulation.
The paper tackles the problem of enabling embodied agents to perform collaborative tasks by introducing BeliefNest, an open-source simulator that dynamically constructs hierarchical simulators in Minecraft to represent nested belief states using Theory of Mind, and demonstrates through experiments that agents can infer others' beliefs and predict belief-based actions in false-belief tasks.
This paper introduces an open-source simulator, BeliefNest, designed to enable embodied agents to perform collaborative tasks by leveraging Theory of Mind. BeliefNest dynamically and hierarchically constructs simulators within a Minecraft environment, allowing agents to explicitly represent nested belief states about themselves and others. This enables agent control in open-domain tasks that require Theory of Mind reasoning. The simulator provides a prompt generation mechanism based on each belief state, facilitating the design and evaluation of methods for agent control utilizing large language models (LLMs). We demonstrate through experiments that agents can infer others' beliefs and predict their belief-based actions in false-belief tasks.