AI RESEARCH
When Less Latent Leads to Better Relay: Information-Preserving Compression for Latent Multi-Agent LLM Collaboration
arXiv CS.LG
•
ArXi:2604.13349v1 Announce Type: new Communication in Large Language Model (LLM)-based multi-agent systems is moving beyond discrete tokens to preserve richer context. Recent work such as LatentMAS enables agents to exchange latent messages through full key-value (KV) caches. However, full KV relay incurs high memory and communication cost. We adapt eviction-style KV compression to this setting and