#Llama

共 17 篇带此标签

17 articles Newest 2026-10-16
Archive · #Llama 11 articles
LLMs

Cache-to-Cache: LLMs Skip Text and Talk in KV-Cache

An NVIDIA paper proposes letting two LLMs exchange KV-caches directly instead of communicating through generated text, claiming an order-of-magnitude latency reduction for multi-model pipelines.

2026-09-19 Source 2026-09-18 5 min 327 views