Latent Channels Outperform Text for Multi-Agent Communication
July 17, 2026
LLM agents communicating via sparse latent channels achieve 99.4% probe accuracy at 28-fold compression, significantly outperforming text-based communication at 80.4% accuracy. This indicates agents can transfer complex information more effectively through internal representations than through natural language.
HOW THIS AFFECTS YOU
●
builderYou can achieve higher efficiency in multi-agent systems by using latent space communication instead of text.
●
researcherThis provides a pathway to studying information density in model internal states.