1 paper
Zixuan Li
Hierarchical KV-cache systems can retain long-context LLM execution state beyond GPU memory, but retention capacity does not determine the local memory required to make that state…