Glance Forcing: Making Bidirectional Diffusion Autoregressive at a Glance

Zhuobai Dong 1, Junchao Yi 2, Hu Jian Guo 3, Linjie Li 4, Rui Zhao 5 † Alex Jinpeng Wang 6,
1WuHan University,    2University of Electronic Science and Technology of China,    3Sun Yat-sen University,
4University of Washington,    5National University of Singapore,    6Central South University
† Corresponding author
 
Paper   Code   Hugging Face Model
Click to zoom in

Overall framework of our Glance Forcing and the comparison with existing methods. (a) Causal Forcing (CF) fixes Self Forcing (SF)'s architecture gap but remains costly; our Glance Forcing (GF) is efficient, and superior. (b) Our GF reduces training cost by 11600×, requires only 1 extra sample curation, and achieves higher VBench scores than CF.

Qualitative Comparison

Synchronized video comparison across different methods

Self Forcing
Causal Forcing
Glance Forcing
Loading videos...
Loading video comparison...
All videos are synchronized in playback, seeking, and pause/play actions.
Previous Work: Glance (ECCV 2026)

Our previous work: Glance: Accelerating Diffusion Models with 1 Sample. Glance requires only one training sample to obtain Slow-LoRA and Fast-LoRA, providing plug-and-play acceleration of the base image generation model.

BibTeX 

@misc{dong2026glanceforcingmakingbidirectional,
      title={Glance Forcing: Making Bidirectional Diffusion Autoregressive at a Glance}, 
      author={Zhuobai Dong and Junchao Yi and Hu Jian Guo and Linjie Li and Rui Zhao and Alex Jinpeng Wang},
      year={2026},
      eprint={2610.20220},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2610.20220}, 
}
  @misc{dong2025glanceacceleratingdiffusionmodels,
  title={Glance: Accelerating Diffusion Models with 1 Sample},
  author={Zhuobai Dong and Rui Zhao and Songjie Wu and Junchao Yi and Linjie Li and Zhengyuan Yang and Lijuan Wang and Alex Jinpeng Wang},
  year={2025},
  eprint={2512.02899},
  archivePrefix={arXiv},
  primaryClass={cs.CV},
  url={https://arxiv.org/abs/2512.02899},
}