Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces MACA, a novel framework for multi-agent coordination that jointly learns a structural prior over agent participation and interactions alongside a policy-based orchestration mechanism. By framing coordination as posterior inference over the joint distribution of structure and orchestration, MACA achieves a balance between structural stability and dynamic adaptability in LLM-based multi-agent systems. Experiments demonstrate that MACA outperforms existing adaptive multi-agent baselines, achieving an average improvement of 8.42% in performance while reducing token usage by 43.19%.
LLM-based multi-agent systems can achieve better performance with significantly fewer tokens by jointly adapting the coordination structure and agent orchestration.
As large language model (LLM)-based multi-agent systems scale to handle increasingly complex tasks, balancing structural stability and dynamic adaptability becomes increasingly challenging. Existing systems typically adopt either structure-centric methods, committing to structures determined upfront that limit fine-grained control, or orchestration-centric methods, adapting decisions dynamically while leaving coordination structure implicit and unstable. To address this challenge, we revisit multi-agent coordination from a probabilistic perspective, casting it as posterior inference over the joint distribution of structure and orchestration. We introduce MACA, an automated coordination framework that learns a task- and budget-conditioned structural prior over agent participation and interactions. This prior guides a policy-based orchestration as an approximation to posterior inference, enabling efficient solutions with fine-grained control. Across benchmarks, MACA outperforms adaptive multi-agent baselines by an average of 8.42% while using 43.19% fewer tokens. Further investigation reveals that joint adaptation of structure and orchestration suppresses redundant interactions, converging coordination toward task-effective execution.