Orchestrator: Active Inference for Multi-Agent Systems in Long-Horizon Tasks

Lukas Beckenbauer, Johannes-Lucas Löwe, Ge Zheng, Alexandra Brintrup

Keywords: Multi Agent Systems, Multi Agent Interactions, Active Inference, Agent Optimization

TL;DR: Orchestrator is a multi-agent framework that draws on attention-inspired architecture and active inference benchmarks, to improve multi-agent coordination in non-linear, long-horizon tasks and enable more efficient global task completion..

Abstract: Complex, non-linear tasks challenge LLM-enhanced multi-agent systems (MAS) due to partial observability and suboptimal coordination. We propose Orchestrator, a novel MAS framework that leverages attention-inspired self-emergent coordination and reflective benchmarking to optimize global task performance. Orchestrator introduces a monitoring mechanism to track agent-environment dynamics, using active inference benchmarks to optimize system behavior. By tracking agent-to-agent and agent-to-environment interaction, Orchestrator mitigates the effects of partial observability and enables agents to approximate global task solutions more efficiently. We evaluate the framework on a series of maze puzzles of increasing complexity, demonstrating its effectiveness in enhancing coordination and performance in dynamic, non-linear environments with long-horizon objectives.

Submission Number: 157