MindForge: Empowering Embodied Agents with Theory of Mind for Lifelong Collaborative Learning

2024-11-20

arXiv.org

AI Summary

To address the limited collaborative lifelong learning capability of embodied agents based on open-source large language models (LLMs) in open-world environments, this paper proposes MindForge. Methodologically, MindForge explicitly models others’ beliefs, desires, and intentions (BDI) to construct structured theory-of-mind representations; introduces natural-dialogue-driven multi-agent coordination; and implements a hierarchical memory system—comprising episodic, semantic, and procedural memory—to support long-term social learning. Crucially, MindForge relies exclusively on open-weight LLMs, requiring no proprietary models or additional training data. Empirical evaluation in Minecraft demonstrates that MindForge achieves 2.3× more unique item acquisitions and 3× more tech-tree milestone completions compared to Voyager. Moreover, it is the first framework to enable expert-to-novice knowledge transfer, collaborative problem solving, and out-of-distribution (OOD) environmental adaptation—establishing new capabilities for open-world, socially grounded agent learning.

Technology Category

• Multiagent Systems: Multiagent Learning
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Contemporary embodied agents powered by large language models (LLMs), such as Voyager, have shown promising capabilities in individual learning within open-ended environments like Minecraft. However, when powered by open LLMs, they struggle with basic tasks even after domain-specific fine-tuning. We present MindForge, a generative-agent framework for collaborative lifelong learning through explicit perspective taking. We introduce three key innovations: (1) a structured theory of mind representation linking percepts, beliefs, desires, and actions; (2) natural interagent communication; and (3) a multicomponent memory system. In Minecraft experiments, MindForge agents powered by open-weight LLMs significantly outperform their Voyager counterparts in basic tasks where traditional Voyager fails without GPT-4, collecting $2.3 imes$ more unique items and achieving $3 imes$ more tech-tree milestones, advancing from basic wood tools to advanced iron equipment. MindForge agents demonstrate sophisticated behaviors, including expert-novice knowledge transfer, collaborative problem solving, and adaptation to out-of-distribution tasks through accumulated collaborative experiences. MindForge advances the democratization of embodied AI development through open-ended social learning, enabling peer-to-peer knowledge sharing.

Problem

• Empowering agents with theory of mind
• Enhancing collaborative lifelong learning
• Improving performance in basic tasks

Innovation

• structured theory of mind
• natural interagent communication
• multicomponent memory system