Metraipolis › archive › dc_45c813a100f2
SAKI-RoPE: Score-Aware Low-Rank Key Indexing with RoPE-Aware Covariance and Random-Matrix Noise Analysis — arXiv:2608.03228
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.03228
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-31T12:42:23.017Z
- SHA-256 of the archived text
248ac0da81be16e9e92dfe0a7ef6f54d4e3cef0dec66aba3ee8aa02bc9e99e15- Stable URL of this page
https://metraipolis.com/documento/dc_45c813a100f2
How to cite
SAKI-RoPE: Score-Aware Low-Rank Key Indexing with RoPE-Aware Covariance and Random-Matrix Noise Analysis — arXiv:2608.03228 — arXiv (en), https://arxiv.org/abs/2608.03228 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_45c813a100f2 (Metraipolis archive dc_45c813a100f2, sha256 248ac0da81be16e9e92dfe0a7ef6f54d4e3cef0dec66aba3ee8aa02bc9e99e15)
BibTeX
@misc{dc45c813a100f2,
title = {SAKI-RoPE: Score-Aware Low-Rank Key Indexing with RoPE-Aware Covariance and Random-Matrix Noise Analysis — arXiv:2608.03228},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.03228},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_45c813a100f2; sha256 248ac0da81be16e9e92dfe0a7ef6f54d4e3cef0dec66aba3ee8aa02bc9e99e15},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.03228. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Existing low-rank KV-cache indexes preserve model weights or key variance, not the attention scores that determine retrieval. We derive the expected score distortion of key-side rank-r compression, a two-sided covariance-weighted low-rank problem that uses both query and key statistics, and solve it exactly: the optimum is an optimal rank-r linear score map, computed in closed form from two d x d covariances per head, and its predicted per-head score-MSE reduction matches measurement (Pearson r = 0.997). The resulting training-free index, SAKI, beats key-PCA at every rank on LLaMA-3.1-8B, Qwen2.5-7B, Mistral-7B and Llama-3.2-3B, at every context length from 4K to 64K, and in every calibration-domain pair (0.748 -> 0.799 recall@64 at r = 32 on LLaMA-3.1-8B). Because RoPE rotates keys by their distance from the query, we make the closed form RoPE-aware by averaging the query covariance over rotations; this SAKI-RoPE, our main method, raises recall to 0.861, beats PCA on every head of all four models, and needs only 28-39% of PCA's rank to match it at r = 32 across the four models. Random-matrix theory of the covariance estimation noise shows that 512 calibration tokens suffice and that clipping the Marchenko-Pastur noise bulk helps SAKI when calibration is short. We make no perplexity claim: with the attention sink pinned, perplexity differences between indexes are small and depend on model and text, and we show that any low-rank index calibrated on a long prefix drops the attention-sink token, which collapses 32K perplexity unless the sink is pinned. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-05 in cs.LG). Categorie arXiv: Machine Learning (cs.LG) ; Applications (stat.AP). Pagina del lavoro: https://arxiv.org/abs/2608.03228 — 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 2026-10-31T12:31:30.345Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-08-05?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).