Duality Models: An Embarrassingly Simple One-step Generation Paradigm
Peng Sun, Xinyi Shang, Tao Lin, Zhiqiang Shen · 2026
Read Duality Models: An Embarrassingly Simple One-step Generation Paradigm by Peng Sun, Xinyi Shang, Tao Lin, Zhiqiang Shen (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-17682.txt.
Description
Peng Sun and team propose Duality Models, a one-step generation paradigm meant to end a costly training split. Consistency-based models such as Shortcut and MeanFlow use a target-aware design for the probability flow ODE, but the one-input, one-output setup forces much of the budget into the multi-step objective for stability. That leaves few-step generation undertrained.
Duality Models switch to one input, dual output. A shared backbone with dual heads predicts velocity and flow-map from a single input at once. Geometric constraints from the multi-step objective thus reach every sample, bounding few-step estimation with no split of objectives. On ImageNet at high resolution, a large diffusion transformer reaches a state-of-the-art FID score in just two steps, with far better stability and efficiency.
Catalog record
| CREATOR | Peng Sun, Xinyi Shang, Tao Lin, Zhiqiang Shen |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 59 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-17682