1 paper · 1 filter
Jiyao Liu, Qi Zhang, Yaoyi Jia +2
Multi-agent LLM systems split work across models, so answering often requires knowledge that sits in another agent's context: a Sharer has encoded information that a Receiver needs…