Architectural Design, Not Only Model Intelligence, Governs Multi-Agent LLM Performance
An architectural taxonomy is introduced that decomposes multi-agent LLM frameworks along five dimensions: orchestration, memory, planning interfaces, specialization, and communication topology, and a controlled empirical study is conducted, fixing the underlying LLM and varying only architectural design choices.
Abdelghny Orogat, Ana Rostam, Essam Mansour
· 4 citations