Mask Forcing: Improving Autoregressive Video Diffusion Distillation via Dual-Noise Masking Rollout

Hugging Face Daily Papers Papers

Summary

Mask Forcing mitigates mode collapse in autoregressive video diffusion distillation by injecting masked cleaner signals during self-rollout, improving visual quality without extra training data.

Autoregressive (AR) video diffusion models have shown great potential in real-time video generation. Recent methods distill pretrained bidirectional video diffusion models into causal AR students through Distribution Matching Distillation (DMD), but the generated videos often suffer from over-saturation and over-smoothing issues, resulting in limited visual quality and realism. The key contributing factor is the mode-seeking behavior of the reverse KL objective in DMD, which can cause the student distribution to collapse onto only a few modes of the teacher distribution. To address this, we propose Mask Forcing, a Dual-Noise Masking Rollout strategy that perturbs the AR student self-rollout to mitigate mode collapse induced by reverse-KL mode seeking. The core idea is to inject cleaner signals into noisy rollout inputs via random masks along spatial and temporal axes during the self-rollout process of AR diffusion distillation. Such perturbations encourage the student rollouts to explore more regions of the teacher distribution, allowing DMD to provide learning signals beyond the modes already covered by the student. Moreover, the cleaner tokens act as denoising guidance for other noisier tokens, improving the intermediate rollout predictions and reducing error accumulation. Extensive experiments demonstrate that our method improves multiple AR video diffusion distillation methods with higher visual quality efficiently, without incorporating real video data or additional post-training stages.
Original Article
View Cached Full Text

Cached at: 09/09/26, 08:33 AM

Paper page - Mask Forcing: Improving Autoregressive Video Diffusion Distillation via Dual-Noise Masking Rollout

Source: https://huggingface.co/papers/2609.09123

Abstract

Mask Forcing mitigates mode collapse in distilled autoregressive video diffusion by injecting masked cleaner signals during self-rollout, improving visual quality without extra training data.

Autoregressive (AR) video diffusion models have shown great potential in real-time video generation. Recent methods distill pretrained bidirectional video diffusion models into causal AR students throughDistribution Matching Distillation(DMD), but the generated videos often suffer from over-saturation and over-smoothing issues, resulting in limited visual quality and realism. The key contributing factor is the mode-seeking behavior of thereverse KLobjective in DMD, which can cause the student distribution to collapse onto only a few modes of the teacher distribution. To address this, we proposeMask Forcing, aDual-Noise Masking Rolloutstrategy that perturbs the AR studentself-rolloutto mitigatemode collapseinduced by reverse-KL mode seeking. The core idea is to inject cleaner signals into noisy rollout inputs via random masks along spatial and temporal axes during theself-rolloutprocess of AR diffusion distillation. Such perturbations encourage the student rollouts to explore more regions of the teacher distribution, allowing DMD to provide learning signals beyond the modes already covered by the student. Moreover, the cleaner tokens act asdenoising guidancefor other noisier tokens, improving the intermediate rollout predictions and reducing error accumulation. Extensive experiments demonstrate that our method improves multiple AR video diffusion distillation methods with higher visual quality efficiently, without incorporating real video data or additional post-training stages.

View arXiv pageView PDFProject pageGitHub11Add to collection

Get this paper in your agent:

hf papers read 2609\.09123

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2609.09123 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2609.09123 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2609.09123 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles