Metraipolis › archive › dc_2afb2e40bdc6
KV-COBRA: KV Cache Compression via Co-Optimized Bit-Rank Allocation — arXiv:2609.24298
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.24298
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-09T07:07:39.324Z
- SHA-256 of the archived text
a10d2334e7200b6a11ea8b342e0f933cb6779aa694df895d0b555b04dd0a0bc3- Stable URL of this page
https://metraipolis.com/documento/dc_2afb2e40bdc6
How to cite
KV-COBRA: KV Cache Compression via Co-Optimized Bit-Rank Allocation — arXiv:2609.24298 — arXiv (en), https://arxiv.org/abs/2609.24298 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2afb2e40bdc6 (Metraipolis archive dc_2afb2e40bdc6, sha256 a10d2334e7200b6a11ea8b342e0f933cb6779aa694df895d0b555b04dd0a0bc3)
BibTeX
@misc{dc2afb2e40bdc6,
title = {KV-COBRA: KV Cache Compression via Co-Optimized Bit-Rank Allocation — arXiv:2609.24298},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.24298},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2afb2e40bdc6; sha256 a10d2334e7200b6a11ea8b342e0f933cb6779aa694df895d0b555b04dd0a0bc3},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.24298. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
What limits KV-cache compression at extreme bit-rates? We argue that it is not the choice of compression scheme, but how its budget is allocated across attention heads. Existing methods apply rank and bit-width uniformly, ignoring that each head has a different optimal mix of rank truncation and quantization. We show that co-optimizing rank and bit-width per head, using only standard low-rank projection and scalar quantization, dominates uniform allocation, with the largest gains at low bit-rates. Our method, KV-COBRA (Co-Optimized Bit-Rank Allocation), formalizes this as a resource-allocation problem: it balances rank-truncation loss against quantization loss within each head, then redistributes budget across heads to minimize total distortion. A fused Hadamard rotation equalizes per-channel variance, and reordering the SVD basis by attention-KL importance makes the solver query-aware. The same allocator extends to joint $K{+}V$ compression. On perplexity, zero-shot, and long-context benchmarks from $0.5$ to $4$ bits per dimension (bpd), KV-COBRA shows the smallest accuracy degradation among evaluated methods at low bpd, with no per-token overhead.
Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-22 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.24298 — 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 2026-10-09T06:58:09.393Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-22?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).