Mask Forcing: Improving Autoregressive Video Diffusion Distillation via Dual-Noise Masking Rollout
- 类型：arxiv
- 标识：2609.09123
- 链接：https://arxiv.org/abs/2609.09123
- 主分类：multimodal
- 形态：method
- TLDR：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 pr
- 待LLM分类：否
- 来源文件：
- /inbox/tom/_candidates/2026-09-09-agent-rag-longcontext-candidates.json