Scholar Feed

KV-cache compression

Host CPU decompression

Superseded baseline#153 of 234 most-superseded

Cited as a baseline — critiqued by newer work, not yet beaten on a benchmark here

1 papers critique it · 0 beat it on benchmarks

What papers say

Verbatim critique sentences, each from a paper that cites Host CPU decompression as a baseline.

Offloading decompression to the host CPUs seems like a feasible workaround, but host CPUs in serving clusters are already burdened with other tasks like vector search and data preprocessing, and they are inefficient at a variety of decompression algorithms.
— ShadowServe: Interference-Free KV Cache Fetching for Distributed Prefix Caching

What to use instead

Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.