Metraipolis › archive › dc_962bdf3ad3ec

In-Flight KV Cache with Clean Anchors for Faster Autoregressive Video Diffusion — arXiv:2609.32540

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.32540
License (as archived)
CC0-1.0
Archived
2026-10-07T21:00:22.952Z
SHA-256 of the archived text
644b21e2709d9bac2c2d23ade20916f825d46d6516ecc68de0306da911751ba0
Stable URL of this page
https://metraipolis.com/documento/dc_962bdf3ad3ec

How to cite

In-Flight KV Cache with Clean Anchors for Faster Autoregressive Video Diffusion — arXiv:2609.32540 — arXiv (en), https://arxiv.org/abs/2609.32540 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_962bdf3ad3ec (Metraipolis archive dc_962bdf3ad3ec, sha256 644b21e2709d9bac2c2d23ade20916f825d46d6516ecc68de0306da911751ba0)
BibTeX
@misc{dc962bdf3ad3ec,
  title = {In-Flight KV Cache with Clean Anchors for Faster Autoregressive Video Diffusion — arXiv:2609.32540},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.32540},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_962bdf3ad3ec; sha256 644b21e2709d9bac2c2d23ade20916f825d46d6516ecc68de0306da911751ba0},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.32540. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Few-step autoregressive video diffusion generates a long video by splitting the video into temporal chunks and generating chunk-by-chunk, each through a short sequence of denoising stages. To memorize chunks that are already generated, previous methods reconstruct a clean or less-noisy key--value (KV) cache by additional forwards to build the cache without advancing an output latent. However, every denoising forward itself already computes the in-flight KV of the current chunk. We introduce FlashForward, which directly reuses this cache to avoid the heavy cache-update-only model forwards. After the current chunk completes one denoising stage, its stage-specific cache is already available for the next chunk. Assigning one GPU to each stage therefore lets different chunks occupy different stages concurrently. This early availability has a quality cost: the resulting stage-matched history is noisy, causing appearance and motion drift among chunks. To complement it, FlashForward produces sparse auxiliary clean anchor latents before the corresponding region is generated so the generation trajectories can be stabilized by this two-sided conditioning. The two memories operate at different temporal scales: sparse clean anchor KV supplies coarse, long-range two-sided structural guidance, while dense stage-matched history preserves fine, recent evolution. With up to four GPUs, FlashForward runs $1.16$--$1.69\times$ faster than HiAR and $1.42$--$2.92\times$ faster than Self-Forcing for 16 FPS videos of 20 seconds or longer across 1.3B and 14B backbone scales at 480p and 720p. On VBench, for the 1.3B model at 480p, it achieves higher scores and remains stable at longer durations, demonstrating that FlashForward generates high-quality and temporally consistent videos across durations of 20s, 35s and 65s at a much faster generation speed.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multimedia (cs.MM).
Pagina del lavoro: https://arxiv.org/abs/2609.32540 — 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 2026-10-07T20:29:02.381Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).