1 paper · 1 filter
Bole Ma, Jan Eitzinger, Harald Koestler
Modern KV cache management assumes the chatbot workload: prompts arrive once and the cache grows append-only, so prefix caching and forward-only eviction are correct by constructio…