Metraipolis › archive › dc_6d72c93fa487

Uncertainty DMD: Restoring Diversity in Few-Step Autoregressive Video Distillation — arXiv:2609.11265

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.11265
License (as archived)
CC0-1.0
Archived
2026-10-16T06:35:28.982Z
SHA-256 of the archived text
fb03206060eaa4457bb43dd2ee5a9b50969d46633ee15c060e7c23864a2b1b12
Stable URL of this page
https://metraipolis.com/documento/dc_6d72c93fa487

How to cite

Uncertainty DMD: Restoring Diversity in Few-Step Autoregressive Video Distillation — arXiv:2609.11265 — arXiv (en), https://arxiv.org/abs/2609.11265 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_6d72c93fa487 (Metraipolis archive dc_6d72c93fa487, sha256 fb03206060eaa4457bb43dd2ee5a9b50969d46633ee15c060e7c23864a2b1b12)
BibTeX
@misc{dc6d72c93fa487,
  title = {Uncertainty DMD: Restoring Diversity in Few-Step Autoregressive Video Distillation — arXiv:2609.11265},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.11265},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_6d72c93fa487; sha256 fb03206060eaa4457bb43dd2ee5a9b50969d46633ee15c060e7c23864a2b1b12},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.11265. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Few-step distillation improves the efficiency of autoregressive (AR) video generation, but often causes diversity collapse: under the same prompt, different noise samples tend to produce highly similar videos with weakened motion dynamics. We analyze this degradation in Distribution Matching Distillation (DMD)-distilled AR video generators and find that, in the autoregressive setting, it takes the form of a structured uncertainty collapse: the mode-seeking bias of DMD maps different noise samples to nearly identical first chunks, and the deterministic AR cache then propagates this collapsed state to all subsequent chunks, turning a local loss of stochasticity at the rollout root into a global suppression of temporal variation. Based on this analysis, we propose Uncertainty DMD, a simple uncertainty-injection framework that restores stochasticity at two key stages of AR generation: a timestep perturbation for the first chunk to increase first-chunk diversity, and a stochastic cache-writing mechanism for later chunks to preserve uncertainty in autoregressive conditioning. The method requires no architectural changes and introduces only lightweight perturbation operations. The same perturbation mechanisms are used during both training and inference. Experiments show that Uncertainty DMD consistently improves diversity and motion dynamics while maintaining comparable per-sample visual quality.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-11 in cs.CV).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2609.11265 — 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 2026-10-16T06:06:28.768Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-09-11?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).