Metraipolis › archive › dc_ea5f41292c85

Conditional Residual Prediction: Improving Autoregressive Video Diffusion without a Bidirectional Teacher — arXiv:2610.11479

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.11479
License (as archived)
CC0-1.0
Archived
2026-10-09T05:57:50.166Z
SHA-256 of the archived text
3585e093e1b3a638d186dd31d9b8f8baedc36ee06bae8ea671ce99e6b8c21e2b
Stable URL of this page
https://metraipolis.com/documento/dc_ea5f41292c85

How to cite

Conditional Residual Prediction: Improving Autoregressive Video Diffusion without a Bidirectional Teacher — arXiv:2610.11479 — arXiv (en), https://arxiv.org/abs/2610.11479 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ea5f41292c85 (Metraipolis archive dc_ea5f41292c85, sha256 3585e093e1b3a638d186dd31d9b8f8baedc36ee06bae8ea671ce99e6b8c21e2b)
BibTeX
@misc{dcea5f41292c85,
  title = {Conditional Residual Prediction: Improving Autoregressive Video Diffusion without a Bidirectional Teacher — arXiv:2610.11479},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.11479},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ea5f41292c85; sha256 3585e093e1b3a638d186dd31d9b8f8baedc36ee06bae8ea671ce99e6b8c21e2b},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.11479. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Causal video diffusion models generate video autoregressively, which suits streaming, interactive, and long-video generation. Under standard training, however, they often yield lower generation quality than bidirectional models of the same size. Many existing approaches address this gap by initializing from or distilling a pretrained bidirectional teacher. We instead train a causal model from an image-model initialization, with no bidirectional video model at any stage. Because this path requires neither a large bidirectional teacher nor a complex distillation pipeline, it is simpler and more scalable. On this path, we find that a causal model trained on ground-truth history becomes strongly dependent on it, so that at inference errors in its own generated history propagate forward. We hypothesize that much of this dependence is unnecessary, because the current input already determines much of what the history provides. We propose Conditional Residual Prediction (CRP), a simple recipe for reducing a model's reliance on a condition: the model first predicts the target without the condition, and the condition may only add a residual on top of this prediction. Applied to history, CRP makes the model predict each chunk from the present as far as it can and use the past only for what the present cannot supply. In controlled experiments, CRP nearly closes the 6.14-point gap to a bidirectional model trained under the same setup. Scaling this recipe, we train Optica, a 2B-parameter causal video model that autoregressively generates 5-second 480p videos and reaches 82.78 on VBench with only about 15M training videos.

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); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.11479 — 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 2026-10-09T05:47:01.699Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).