Tensor-Coord: Algebraic Decomposition of Joint Plan Tensors for Conflict-Free Multi-Agent LLM Planning
Researchers have developed Tensor-Coord, a novel framework utilizing multilinear algebra to represent and decompose joint plans generated by multiple large language models. This approach decomposes joint plans into tensors, allowing for the identification of coordination structures and conflict scores without domain-specific rules. Experiments demonstrated that Tensor-Coord can significantly improve the convergence rate of conflict-free plans in multi-agent scenarios, with success rates varying by the number of agents involved. AI
IMPACT Introduces a novel algebraic decomposition method to resolve coordination failures in multi-agent LLM planning.