Greed is Good: A Unifying Perspective on Guided Generation
Zander W. Blasingame, Chen Liu · 2025
Read Greed is Good: A Unifying Perspective on Guided Generation by Zander W. Blasingame, Chen Liu (2025) 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-2502-08006.txt.
Description
Zander Blasingame and Chen Liu unified two ways of steering image generators in this 2025 theory paper. Guided generation lets users push a diffusion model's output toward a target without retraining. Two method families dominate: one projects each step toward the target, the other backpropagates through the whole generation run. They looked like separate ideas.
The authors show the first family is a greedy shortcut of the second. That single lens clarifies how each method's gradients compare with the ideal ones. Building on the link, they offer a way to blend the two, trading computing cost against guidance accuracy. Tests on image restoration problems and molecule design confirmed the approach. The result gives practitioners one dial instead of two rival toolkits.
Catalog record
| CREATOR | Zander W. Blasingame, Chen Liu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 134 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-08006