AI RESEARCH
Auditing Cascading Risks in Multi-Agent Systems via Semantic-Geometric Co-evolution
arXiv CS.AI
•
ArXi:2603.13325v1 Announce Type: cross Large Language model (LLM)-based Multi-Agent Systems (MAS) are prone to cascading risks, where early-stage interactions remain semantically fluent and policy-compliant, yet the underlying interaction dynamics begin to distort in ways that amplify latent instability or misalignment. Traditional auditing methods that focus on per-message semantic content are inherently reactive and lagging, failing to capture these early structural precursors.