RelayCaching: Accelerating LLM Collaboration via Decoding KV Cache Reuse

Geng, Yingsheng, Gao, Yuchong, Wu, Weihong · arXiv · 2026