Metraipolis › archive › dc_dc855c4f51f3

Trajectory Forcing: Exploiting Diffusion Trajectories for Autoregressive Long Video Generation — arXiv:2611.09096

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.09096
License (as archived)
CC0-1.0
Archived
2026-11-09T22:45:14.482Z
SHA-256 of the archived text
7c0e473062d4978c3c4432baa6e0004bd539fb22332b88c41750896c9b4867b9
Stable URL of this page
https://metraipolis.com/documento/dc_dc855c4f51f3

How to cite

Trajectory Forcing: Exploiting Diffusion Trajectories for Autoregressive Long Video Generation — arXiv:2611.09096 — arXiv (en), https://arxiv.org/abs/2611.09096 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_dc855c4f51f3 (Metraipolis archive dc_dc855c4f51f3, sha256 7c0e473062d4978c3c4432baa6e0004bd539fb22332b88c41750896c9b4867b9)
BibTeX
@misc{dcdc855c4f51f3,
  title = {Trajectory Forcing: Exploiting Diffusion Trajectories for Autoregressive Long Video Generation — arXiv:2611.09096},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.09096},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_dc855c4f51f3; sha256 7c0e473062d4978c3c4432baa6e0004bd539fb22332b88c41750896c9b4867b9},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.09096. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Autoregressive (AR) modeling is arguably the only way to extend a fixed-length video diffusion model to longer or even streaming videos. However, existing AR diffusion methods use only the previously predicted chunk as context for the next, discarding the intermediate predictions that constitute the chunk's generation history. In contrast, conventional AR models such as LLMs condition on the entire history of preceding tokens. To this end, we propose Trajectory Forcing (TraF), an AR diffusion method that exploits the denoising trajectories of previous video chunks as context. TraF operates under two training regimes: (1) with ground-truth videos, where pseudo-trajectories are constructed from Gaussian interpolations and re-predicted by the model, and (2) with a teacher model, where real trajectories are obtained from the student's own denoising rollout. The first regime naturally provides a strong initialization for the second. On the 5-second VBench benchmark, TraF achieves a Total score of 85.15, outperforming Self Forcing (+0.84) and Causal Forcing (+1.11), with Quality and Semantic improving simultaneously. On 30-second generation---6$\times$ the training horizon---TraF achieves the highest overall quality among all compared methods.

Tema degli esploratori: statistica e scienza dei dati (elenco arXiv cs.CV, nuovi lavori del giorno).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2611.09096 — 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 2026-11-09T22:37:47.220Z dalla pagina https://arxiv.org/list/cs.CV/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).