Hayden Prairie, Jacob M. Ryan, Taylor Berg-Kirkpatrick, Daniel Y. Fu · 2026 · arXiv
DOI: 10.48550/arXiv.2610.31633 · open-access PDF
В ходе вывода каждое повторение генерирует независимые кВ-памяти, что приводит к линейному росту памяти. Для решения этой задачи предыдущая работа предложила кВ-разделение (т.е. распределение кВ-памяти по глубине), обнаружив, что в некоторых случаях кВ-память окончательного повторения может быть разделена с предыдущими кВ-памятьми без потери качества, свойство, которое мы обозначаем как кВ-разделение.
Machine translation (Meta M2M100); the English original follows.
Looped architectures scale quality while keeping parameter count fixed by repeatedly passing activations through the same block of layers. During inference, each recurrence generates independent KV-caches, causing memory to grow linearly. To address this challenge, prior work has proposed KV-sharing (i.e., sharing KV-caches across depth), finding that, in some cases, the KV-cache of the final recurrence can be shared with the earlier recurrences without quality loss, a property we denote as KV-shareability. However, the mechanisms underlying KV-shareability remain poorly understood, with even
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.31633, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange