Latent compression speeds up multi-agent AI collaboration fourfold
LatCom: Cross-Agent Latent Compression for Efficient Multi-Agent Collaboration
Computation and Language
Summary
Multi-agent AI systems often share information internally using compressed messages to work together efficiently. The authors found that compressing each agent’s message separately misses opportunities to reduce overlap and wastes space. Their LatCom method compresses all messages together into a fixed, efficient format that keeps the important information for the task. This approach speeds up the system, lowers memory and communication costs, and still keeps accuracy steady.
What this means in practice
- •For multi-agent system developers: Increase processing speed and reduce communication costs by applying efficient joint latent compression for multiple collaborating AI agents.
- •For distributed ai engineers: Enable scalable multi-agent collaboration with lower memory and token usage by integrating cross-agent latent fusion techniques.
Authors
Shinan Zhang, Tao Zhang, Qihui Zhu, Mengjie Zhang, Dong Jin, Yunpeng Hou, Shuangwu Chen, Xiaobin Tan, Quan Zheng, Jian Yang
Abstract
LLM-based multi-agent systems (MAS) increasingly use latent collaboration to avoid the information loss and repeated encoding-decoding overhead of natural-language communication. However, directly forwarding all sender latents makes the receiver-side context scale with both the number of agents and the reasoning length, increasing computation, memory usage, and collaboration latency. A natural solution is latent compression. But we find that cross-agent redundancy remains unresolved in existing latent compression approaches, which typically compress each sender independently and then concatenate the results. We propose LatCom, a cross-agent latent compression framework for efficient multi-agent latent collaboration. LatCom maps multiple sender latents into a fixed number of receiver-readable and task-relevant slots. Rather than reconstructing all sender hidden states, it optimizes the compressed latents for receiver-side task utility. LatCom trains the compressor in two stages: single-sender readability learning first establishes a latent interface interpretable by the frozen receiver, and multi-sender fusion learning then trains the compressor to fuse complementary evidence and remove redundancy across agents. Experiments on multiple benchmarks with Qwen3-4B show that LatCom achieves an average 2.46x inference speed-up over LatentMAS and reduces output token usage by 70.3% while maintaining comparable average accuracy.