Jiaxing Li1,2*,
Kai Zou1*,
Cindy Zhou1,3,
Kaichen Huang1,2,
Junyao Gao1,
Zile Wang1,
Yang Liu1,
Bin Liu1,
Bo An2,
Yangguang Li1†
1Riemann Dynamics 2Nanyang Technological University 3Wellington College, UK
*Equal contribution †Corresponding author
Project Page · Paper · Code
**DistillAlign** aligns and balances the **mode-covering** and **mode-seeking** objectives of the multi-stage video distillation pipeline — using only a **1.3B DMD teacher**, it already surpasses baselines refined with a **14B DMD teacher**. ## Checkpoints All released generators are **Wan2.1-1.3B students**; the size in each name refers to the **teacher** used during training. | Model | Checkpoint | Description | |---|---|---| | Initializer (1.3B teacher) | [distillalign_init_1p3b_teacher.pt](https://huggingface.co/LiJiaxing/DistillAlign/blob/main/checkpoints/distillalign_init_1p3b_teacher.pt) | Pre-DMD initializer, trained toward a Wan2.1-T2V-1.3B teacher | | Initializer (14B teacher) | [distillalign_init_14b_teacher.pt](https://huggingface.co/LiJiaxing/DistillAlign/blob/main/checkpoints/distillalign_init_14b_teacher.pt) | Pre-DMD initializer, trained toward a Wan2.1-T2V-14B teacher | | Distilled (1.3B teacher) | [distillalign_distill_1p3b_teacher.pt](https://huggingface.co/LiJiaxing/DistillAlign/blob/main/checkpoints/distillalign_distill_1p3b_teacher.pt) | Final joint-distilled generator, Wan2.1-T2V-1.3B DMD teacher | | Distilled (14B teacher) | [distillalign_distill_14b_teacher.pt](https://huggingface.co/LiJiaxing/DistillAlign/blob/main/checkpoints/distillalign_distill_14b_teacher.pt) | Final joint-distilled generator, Wan2.1-T2V-14B DMD teacher | Each checkpoint stores a plain `{"generator": state_dict}` and loads directly with the released inference and evaluation code. ## Distillation Data Prompts from VidProM, 25K clean-latent samples per teacher: | Data | Download | Description | |---|---|---| | Wan2.1-14B data | [DistillAlign_14B_25K](https://huggingface.co/datasets/LiJiaxing/DistillAlign_14B_25K) | 25K distillation samples generated by Wan2.1-T2V-14B | | Wan2.1-1.3B data | [DistillAlign_1p3b_25K](https://huggingface.co/datasets/LiJiaxing/DistillAlign_1p3b_25K) | 25K distillation samples generated by Wan2.1-T2V-1.3B | ## Evaluation Caches Teacher reference artifacts for the teacher-normalized distribution evaluation (16 prompts x 16 seeds, 25-step UniPC, timestep shift 8.0, CFG 5.0; V-JEPA2 ViT-H features, 2560-D): | Cache | Download | Description | |---|---|---| | Wan2.1-1.3B teacher reference features | [wan2.1_t2v_1.3b_reference_vjepa2.npz](https://huggingface.co/LiJiaxing/DistillAlign/blob/main/teacher_caches/wan2.1_t2v_1.3b_reference_vjepa2.npz) | 256 x 2560 features, ready for `--teacher-features` | | Wan2.1-14B teacher reference features | [wan2.1_t2v_14b_reference_vjepa2.npz](https://huggingface.co/LiJiaxing/DistillAlign/blob/main/teacher_caches/wan2.1_t2v_14b_reference_vjepa2.npz) | 256 x 2560 features, ready for `--teacher-features` | | Teacher reference videos | [teacher_caches/videos](https://huggingface.co/LiJiaxing/DistillAlign/tree/main/teacher_caches/videos) | 256 reference videos per teacher, for inspection and re-extraction | See [teacher_caches/README.md](https://huggingface.co/LiJiaxing/DistillAlign/blob/main/teacher_caches/README.md) for the exact protocol. ## Citation ```bibtex @misc{li2026distillalign, title = {DistillAlign: Coordinating Mode Covering and Mode Seeking in Autoregressive Video Distillation}, author = {Li, Jiaxing and Zou, Kai and Zhou, Cindy and Huang, Kaichen and Gao, Junyao and Wang, Zile and Liu, Yang and Liu, Bin and An, Bo and Li, Yangguang}, year = {2026}, eprint = {2607.26811}, archivePrefix = {arXiv}, primaryClass = {cs.CV}, url = {https://arxiv.org/abs/2607.26811} } ``` ## License The code and released weights follow the Apache License 2.0; upstream Wan2.1 model weights remain subject to their own licenses.