Generator Matching: Generative modeling with arbitrary Markov processes
This work addresses the challenge of flexible generative modeling for AI researchers, offering a unified approach that is incremental in expanding existing methods.
The paper tackles the problem of generative modeling by introducing Generator Matching, a framework that uses arbitrary Markov processes to model data distributions, and demonstrates its effectiveness by unifying existing methods and expanding the design space to new processes like jump processes, with empirical validation showing improved performance in image and multimodal generation.
We introduce Generator Matching, a modality-agnostic framework for generative modeling using arbitrary Markov processes. Generators characterize the infinitesimal evolution of a Markov process, which we leverage for generative modeling in a similar vein to flow matching: we construct conditional generators which generate single data points, then learn to approximate the marginal generator which generates the full data distribution. We show that Generator Matching unifies various generative modeling methods, including diffusion models, flow matching and discrete diffusion models. Furthermore, it expands the design space to new and unexplored Markov processes such as jump processes. Finally, Generator Matching enables the construction of superpositions of Markov generative models and enables the construction of multimodal models in a rigorous manner. We empirically validate our method on image and multimodal generation, e.g. showing that superposition with a jump process improves performance.