Metraipolis › archive › dc_1801d33f4dcd
CORE: COverage CAlibration and Evicted-Mass REdistribution for KV Cache — arXiv:2610.02235
- Source
- arXiv
- Original
- https://arxiv.org/abs/2610.02235
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-05T22:49:50.225Z
- SHA-256 of the archived text
aa4461ab35007024d112d733949ed1003e449de8e2f347dcc6dfc0e7b81fe131- Stable URL of this page
https://metraipolis.com/documento/dc_1801d33f4dcd
How to cite
CORE: COverage CAlibration and Evicted-Mass REdistribution for KV Cache — arXiv:2610.02235 — arXiv, https://arxiv.org/abs/2610.02235 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_1801d33f4dcd (Metraipolis archive dc_1801d33f4dcd, sha256 aa4461ab35007024d112d733949ed1003e449de8e2f347dcc6dfc0e7b81fe131)
BibTeX
@misc{dc1801d33f4dcd,
title = {CORE: COverage CAlibration and Evicted-Mass REdistribution for KV Cache — arXiv:2610.02235},
howpublished = {arXiv},
url = {https://arxiv.org/abs/2610.02235},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_1801d33f4dcd; sha256 aa4461ab35007024d112d733949ed1003e449de8e2f347dcc6dfc0e7b81fe131},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.02235. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Long-context decoding is increasingly constrained by key--value (KV) cache memory and bandwidth. Existing fixed-budget compression methods typically separate retention from compensation, while a retention ranking specifies neither discarded attention mass nor the direction of induced output error. We start from an exact factorization: eviction error equals evicted attention mass times the directional gap between the evicted centroid and retained output, highlighting the importance of set-level coverage in retention and mass-preserving memory writing. We introduce CORE COverage Calibration and Evicted-Mass REdistribution for KV Cache, which distills an offline allocation combining query utility and log-determinant coverage into a lightweight cache-aware indexer. At inference, one calibrated distribution drives both channels: its Top-$B$ ordering retains complementary KV states, while its excluded allocation mass and conditional weights parameterize latent-memory writes without a separate write-weight predictor or online log-determinant evaluation. Our analysis provides a four-term pre-compensation error certificate, characterizes non-additive coverage interactions, and establishes mass-independent write stability with hierarchical bounds through recurrent and query-adaptive normalization. Across three backbones, CORE exceeds the strongest RULER baseline by up to 3.78 points at 90\% compression; LongBench and repeated-eviction evaluations further demonstrate strong effectiveness and decoding efficiency. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Hardware Architecture (cs.AR) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2610.02235 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-05T22:45:55.727Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).