When Evidence Shapes Collaboration: Knowledge-Conditioned Topology Generation for Multi-Agent Systems

DOI10.48550/arXiv.2608.27984MaRDI QIDQ7285180arXivFDO

Authors Shanqing Xu, Yangxiao Jiang, Jiarun Fan, Mingcong Xu, Yanxi Guo, Jiwen Feng, Mengchen Qian, Wei Chen, Xiaojin Zhang

Publication date 28 August 2026

Published in arXiv

Tldr We introduce K-GAT, a neuro-symbolic framework that incorporates external evidence into autoregressive graph generation for dynamically generated collaboration topologies in Multi-Agent Systems (MAS). By integrating knowledge-conditioned structure learning, K-GAT outperforms existing methods on expert-level benchmarks by +15.7% in accuracy while consuming less than half the computational resources.

Full work available at URL https://doi.org/10.48550/arXiv.2608.27984

arXiv classification

cs.AI

Cited in

(1)

• CollabFlow: Recursive Self-Improvement of Agent Collaboration

This page was built for publication: When Evidence Shapes Collaboration: Knowledge-Conditioned Topology Generation for Multi-Agent Systems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7285180)