MetaMind: General and Cognitive World Models in Multi-Agent Systems by Meta-Theory of Mind

Meta researchers introduced MetaMind, a cognitive world model for multi-agent systems that enables agents to understand and predict other agents' behaviors without centralized supervision or communication. The system uses a meta-theory of mind framework allowing agents to reason about goals and beliefs of others through self-reflective learning and analogical reasoning.

• →MetaMind enables multi-agent systems to operate without centralized supervision or explicit communication between agents.
• →The system uses bidirectional inference loops for agents to learn metacognitive abilities in a self-supervised manner.
• →Agents can reason about other agents' goals and beliefs from limited observable behavior trajectories in zero-shot scenarios.
• →The framework generalizes metacognitive abilities from first-person to third-person through analogical reasoning.
• →Simulation results show superior performance compared to baselines in few-shot multi-agent generalization tasks.