Metraipolis › archive › dc_2aa899a2cd2c

DART: Distillation-Aware Reparameterization for Training-Free LoRA Reuse in Few-Step Video Diffusion Models — arXiv:2609.20051

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.20051
License (as archived): CC0-1.0
Archived: 2026-10-10T03:21:23.637Z
SHA-256 of the archived text: fe0c300526fc8692b3585cfce4a0991d93ac968533d02da51118355f8d7acc42
Stable URL of this page: https://metraipolis.com/documento/dc_2aa899a2cd2c

How to cite

DART: Distillation-Aware Reparameterization for Training-Free LoRA Reuse in Few-Step Video Diffusion Models — arXiv:2609.20051 — arXiv (en), https://arxiv.org/abs/2609.20051 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2aa899a2cd2c (Metraipolis archive dc_2aa899a2cd2c, sha256 fe0c300526fc8692b3585cfce4a0991d93ac968533d02da51118355f8d7acc42)

BibTeX

@misc{dc2aa899a2cd2c,
  title = {DART: Distillation-Aware Reparameterization for Training-Free LoRA Reuse in Few-Step Video Diffusion Models — arXiv:2609.20051},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.20051},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2aa899a2cd2c; sha256 fe0c300526fc8692b3585cfce4a0991d93ac968533d02da51118355f8d7acc42},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.20051. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Few-step distillation reduces the inference cost of image-to-video generation, but directly reusing LoRA adapters trained for long denoising trajectories can weaken their intended effects and degrade video quality. We observe that adapters with similar measured static parameter geometry can behave differently under the shortened target schedule, motivating response-aware transfer. We propose DART, a training-free reparameterization method that transports source LoRAs into aligned coordinates through a low-rank distillation bridge. Paired forward evaluations measure channel-level incremental responses under the target schedule to fit coefficients that calibrate response direction, source-relative magnitude, and timestep allocation. Coordinate transport establishes update directions, while calibration adapts their contributions, requiring neither source training videos nor backpropagation. Experiments across multiple distilled I2V models demonstrate improved generation quality and aggregate functional retention over direct reuse. Further analyses show that coordinate transport complements response calibration, with adapter-level benefits encompassing both functional preservation and reduced negative transfer.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-18 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.20051 — 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 2026-10-10T03:04:13.690Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).