Reproduce “CacheWeave: Learning to Recompute for Efficient Position-Independent KV Caching” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yueyang Pan, Qianchao Zhu, Yuliang Liu, Teng Ma, Sanidhya Kashyap
CacheWeave adapts language models to position-independent caching and learns which cached blocks to recompute with preceding context restored. At a 5% recomputation budget, it beats CacheBlend and ProphetKV by averages of 7.2 and 3.5 percentage points across evaluated model-task pairs. Its mean relative accuracy loss versus full prefill is 1.6%, with a worst case of 8.3%, conditional on evaluated settings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.