2 papers
cs.LG2026
VestigeKV: The NoPE-MLA KV Cache Carries Its Own Sparse-Attention Signal in a Vestigial Branch
WenJie Fan
A long-lived KV cache must be compressed before the queries that will read it exist. Selection by observed attention collapses there: on a NoPE-MLA model, H2O and SnapKV retrieve 0…
cs.DC2026
Q-First: Most of Attention Needs Only the Query in Disaggregated LLM Decoding
WenJie Fan
Disaggregated LLM serving puts the KV-cache sweep on memory-optimised hardware and the projections and feed-forward on compute-optimised hardware, then inherits from the decoder bl…