Metraipolis › archive › dc_6476faf93029
Close Enough is Nowhere Near: Fidelity-Consumability Decoupling in KV Caches — arXiv:2610.22016
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.22016
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T10:24:45.458Z
- SHA-256 of the archived text
22cb7ae1917bf8ec049062a21c2b9a801601758017fb466ca714c073ecb078ec- Stable URL of this page
https://metraipolis.com/documento/dc_6476faf93029
How to cite
Close Enough is Nowhere Near: Fidelity-Consumability Decoupling in KV Caches — arXiv:2610.22016 — arXiv (en), https://arxiv.org/abs/2610.22016 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_6476faf93029 (Metraipolis archive dc_6476faf93029, sha256 22cb7ae1917bf8ec049062a21c2b9a801601758017fb466ca714c073ecb078ec)
BibTeX
@misc{dc6476faf93029,
title = {Close Enough is Nowhere Near: Fidelity-Consumability Decoupling in KV Caches — arXiv:2610.22016},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.22016},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_6476faf93029; sha256 22cb7ae1917bf8ec049062a21c2b9a801601758017fb466ca714c073ecb078ec},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.22016. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Multi-agent LLM systems communicate in text, which is lossy and wasteful: the sender throws away the reading it already did. A growing literature instead transmits the KV cache, the natural carrier of that reading. We study the most appealing form of the idea: both models stay frozen and the sender's cache is substituted for the reader's with nothing trained. It does not work, and the obstruction belongs to the reader. A frozen transformer reads only caches close to the one it would have written itself, a region we call its tolerable set. We show that geometric similarity to that cache does not predict whether the reader can use it, and the failure is not gradual, so a more faithful substitute is not the remedy it appears to be. We explain this phenomenon with competitive attention. Experimentally, we test eight diverse substitution methods against two measured anchors: an uninformative cache baseline and the reader's own accuracy from having read the passage. All sit at the baseline. That turns a long tail of separate negative results into one outcome with a stated condition for overturning it. A different design does work: if the reader keeps its own cache and attends over the sender's as extra context, adapting only its queries, it beats its own baseline, and also clears the baseline with eight senders pooled where undirected attention over the same union falls below it. We offer this as a characterization on one probe, not a leaderboard claim. Our central contribution is a redirection: pursuing faithful substitute caches is the wrong target; adding a cache to what the reader already has is the promising one. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno). Categorie arXiv: Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2610.22016 — 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 2026-10-16T08:48:02.780Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).