GAS: Improving Discretization of Diffusion ODEs via Generalized Adversarial Solver
Aleksandr Oganov, Ilya Bykov, Eva Neudachina, Mishan Aliev, Alexander Tolmachev, Alexander Sidorov, Aleksandr Zuev, Andrey Okhotin, Denis Rakitin, Aibek Alanov
2025-10-22
Summary
This paper focuses on making diffusion models, which are really good at creating things like images, faster and more efficient without losing quality.
What's the problem?
Diffusion models are powerful but slow because they require many steps to generate a final result. Some recent attempts to speed things up use optimization techniques to simplify the process, but these methods can be complicated to set up and sometimes don't capture all the fine details in the generated output, leading to blurry or artifact-filled images.
What's the solution?
The researchers introduced a new method called the Generalized Solver. It's a simpler way to speed up the generation process that doesn't need complex training procedures. They also combined this with a technique called adversarial training, which essentially pits two parts of the model against each other to refine the details and reduce unwanted imperfections in the final image.
Why it matters?
This work is important because it offers a more straightforward and effective way to accelerate diffusion models. By improving speed and detail, it makes these powerful image generation tools more practical for a wider range of applications without sacrificing the quality of the results.
Abstract
While diffusion models achieve state-of-the-art generation quality, they still suffer from computationally expensive sampling. Recent works address this issue with gradient-based optimization methods that distill a few-step ODE diffusion solver from the full sampling process, reducing the number of function evaluations from dozens to just a few. However, these approaches often rely on intricate training techniques and do not explicitly focus on preserving fine-grained details. In this paper, we introduce the Generalized Solver: a simple parameterization of the ODE sampler that does not require additional training tricks and improves quality over existing approaches. We further combine the original distillation loss with adversarial training, which mitigates artifacts and enhances detail fidelity. We call the resulting method the Generalized Adversarial Solver and demonstrate its superior performance compared to existing solver training methods under similar resource constraints. Code is available at https://github.com/3145tttt/GAS.