\textsc{MasFACT}: Continual Multi-Agent Topology Learning via Geometry-Aware Posterior Transfer

TL;DR

MasFACT introduces a geometry-aware framework for continual multi-agent topology learning, effectively preserving and transferring collaboration structures across evolving tasks to prevent topology forgetting.

Contribution

It proposes a novel posterior transfer method using Fused Gromov-Wasserstein transport and PAC-Bayes adaptation to maintain effective communication topologies over multiple tasks.

Findings

MasFACT improves average accuracy in continual settings.

It reduces topology forgetting compared to baselines.

The method is compatible with various topology generators.

Abstract

Multi-agent systems (MAS) powered by large language models (LLMs) have emerged as a powerful paradigm for complex problem solving, where performance critically depends on the underlying inter-agent communication topology. However, existing topology generation methods mainly optimize for isolated tasks, while real-world deployments involve streams of evolving tasks, requiring previously effective collaboration patterns to be retained and reused rather than rediscovered or overwritten. We identify a previously underexplored failure mode, \emph{topology forgetting}, in which adapting to new tasks shifts the topology generator away from communication structures required by earlier tasks. This issue stems from cross-task misalignment in both agent-level functional semantics and relational communication structures. To address this challenge, we propose \textbf{\textsc{MasFACT}}, a geometry-aware posterior transfer framework that preserves and reuses historical collaboration knowledge as transferable topology priors. We transfer these priors across task-specific agent spaces through Fused Gromov-Wasserstein optimal transport and perform PAC-Bayes-guided conservative posterior adaptation to balance task-specific plasticity with structural stability. Experiments across class-, domain-, and task-level continual settings demonstrate that \textsc{MasFACT} consistently improves average accuracy while reducing topology forgetting compared to strong topology generation and replay-based baselines, and can be seamlessly integrated with different MAS topology generators.