Metraipolis › archive › dc_3af064565bf2

dKV Forcing: Separating Generation and Memory Representations through a Differentiable KV Cache — arXiv:2611.01648

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.01648
License (as archived)
CC0-1.0
Archived
2026-11-03T10:27:43.433Z
SHA-256 of the archived text
6bdf273086acfa6cb564d8e885a2e47f3c1eeaf825d93ee59019be24748a8b14
Stable URL of this page
https://metraipolis.com/documento/dc_3af064565bf2

How to cite

dKV Forcing: Separating Generation and Memory Representations through a Differentiable KV Cache — arXiv:2611.01648 — arXiv (en), https://arxiv.org/abs/2611.01648 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_3af064565bf2 (Metraipolis archive dc_3af064565bf2, sha256 6bdf273086acfa6cb564d8e885a2e47f3c1eeaf825d93ee59019be24748a8b14)
BibTeX
@misc{dc3af064565bf2,
  title = {dKV Forcing: Separating Generation and Memory Representations through a Differentiable KV Cache — arXiv:2611.01648},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.01648},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_3af064565bf2; sha256 6bdf273086acfa6cb564d8e885a2e47f3c1eeaf825d93ee59019be24748a8b14},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.01648. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

A causal video generator produces two key--value (KV) states for every segment: the KV it uses while denoising that segment, and the KV it writes into the cache for later segments to read. In a pretrained generator the two already differ substantially, later segments attend to them differently, and replacing the written KV with the generation KV degrades the continuation. We propose dKV Forcing (dKV denotes differentiable KV) to learn a dedicated memory representation from its future use: a writer-specific low-rank residual on the shared backbone separates memory writing from generation and is trained through the future reads of what it writes. Learning a write from its future use needs gradients through the cache, which the mutable cache in common use cannot provide. dKV Forcing replaces the training-time cache by persistent differentiable pages with local rematerialization and, in one temporal traversal, computes exactly the gradient that a two-pass reconstruction of the same objective computes. One traversal computes the full memory gradient at $7\%$ generator-update overhead over detached-write training, with $28\%$ less resident memory and $20.8\%$ shorter training cycles than the two-pass reconstruction. Over Self Gradient Forcing (SGF), the same objective trained with the same traversal, writer adaptation raises VBench-Long aesthetic quality by $6.6$ points at 60\,s and $8.5$ points at 240\,s with the highest subject and background consistency, and rollout dropout of the residual keeps most of this gain while restoring the motion of training without write gradients.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2611.01648 — 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 2026-11-03T10:19:50.106Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).