AnchorKV compresses the KV cache 20x with near-lossless accuracy by representing each token as an anchor projection plus selectively allocated quantized residuals, beating eviction baselines at matched byte budgets on three model scales.
Croci and Bo Li and Pashmina Cameron and Martin Jaggi and Dan Alistarh and Torsten Hoefler and James Hensman , editor =
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
AnchorKV: Anchor-Residual KV Cache Compression
AnchorKV compresses the KV cache 20x with near-lossless accuracy by representing each token as an anchor projection plus selectively allocated quantized residuals, beating eviction baselines at matched byte budgets on three model scales.