Reproduce “DR. WELL: Dynamic Reasoning and Learning with Symbolic World Model for Embodied LLM-Based Multi-Agent Collaboration” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Narjes Nourzad, Hanqing Yang, Shiyu Chen, Carlee Joe-Wong

Researchers at Carnegie Mellon University developed DR. WELL, a decentralized neurosymbolic framework that integrates large language models with a dynamic symbolic world model for multi-agent collaboration. The framework notably improved task completion rates and coordination efficiency in a cooperative block-pushing environment, allowing agents to learn and adapt collective strategies over time.

Key insights

  • Channeling LLM outputs through a symbolic planning layer and a dynamic world model effectively mitigates the fragility and generalization issues associated with direct LLM control in embodied multi-agent systems.
  • Structured, two-round negotiation protocols enable efficient decentralized task allocation and consensus among agents with partial information, promoting effective division of labor without excessive communication.
  • A continually updated shared symbolic world model, integrating episodic experiences and performance statistics, provides a robust memory and learning scaffold for agents to adapt and refine their collective strategies over time.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.