Research map / Video / Video generation
Uniform Discrete Diffusion with Metric Path for Video Generation
Haoge Deng, Ting Pan, Fan Zhang, Yang Liu, Zhuoyan Luo, Yufeng Cui, Wenxuan Wang, Chunhua Shen, Shiguang Shan, Zhaoxiang Zhang, Xinlong Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add iterative refinement (uniform diffusion) vs masked diffusion sampling (ablation) | URSA discrete diffusion · GenEval at fewer steps | GenEval | Improved |
| add iterative refinement vs non-refinable sampling for video (ablation) | URSA · VBench video | VBench score | Improved |
| use linearized metric path with calibrated c and alpha (ablation) | URSA · image generation (256x256) | sample quality | Improved |
| increase model size 0.6B to 1.7B to 4B (ablation) | URSA · DPG-Bench, GenEval, VBench | semantic score | Improved |
| increase model size 0.6B to 1.7B to 4B (ablation) | URSA · text-to-image and text-to-video | generation quality | No clear effect |
| add timestep embedding or prompting (ablation) | URSA · GenEval | GenEval | No clear effect |
| use resolution-dependent timestep shifting (ablation) | URSA · VBench video | VBench score | Improved |
| use URSA vs discrete Emu3 on text-to-video | URSA · VBench | Total score: 81.0 -> 82.4 | Improved |
| URSA vs strongest continuous model Wan2.1 | URSA · VBench | Total score: 83.7 -> 82.4 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Lumos-1 — Lumos-1: On Autoregressive Video Generation with Discrete Diffusion from a Unified Model Perspective (ICLR 2026) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Grid Diffusion Models for Text-to-Video Generation (CVPR 2024 · similarity 0.86)
- Q-VDiT: Towards Accurate Quantization and Distillation of Video-Generation Diffusion Transformers (ICML 2025 · similarity 0.86)
- EIDT-V: Exploiting Intersections in Diffusion Trajectories for Model-Agnostic, Zero-Shot, Training-Free Text-to-Video Generation (CVPR 2025 · similarity 0.86)
- Tora: Trajectory-oriented Diffusion Transformer for Video Generation (CVPR 2025 · similarity 0.85)
- LongDiff: Training-Free Long Video Generation in One Go (CVPR 2025 · similarity 0.85)
- VideoMAR: Autoregressive Video Generation with Continuous Tokens (NeurIPS 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
