D-Flow: Differentiating through Flows for Controlled Generation

TL;DR

D-Flow introduces a method to control diffusion and flow-based generative models by differentiating through the flow, enabling effective inverse and conditional generation without retraining models.

Contribution

The paper presents D-Flow, a novel framework that controls generative models via differentiation through the flow, leveraging the prior implicitly for improved inverse and conditional generation.

Findings

Achieves state-of-the-art results in image and audio inverse problems.

Effective in conditional molecule generation.

Works with both linear and non-linear controlled generation tasks.

Abstract

Taming the generation outcome of state of the art Diffusion and Flow-Matching (FM) models without having to re-train a task-specific model unlocks a powerful tool for solving inverse problems, conditional generation, and controlled generation in general. In this work we introduce D-Flow, a simple framework for controlling the generation process by differentiating through the flow, optimizing for the source (noise) point. We motivate this framework by our key observation stating that for Diffusion/FM models trained with Gaussian probability paths, differentiating through the generation process projects gradient on the data manifold, implicitly injecting the prior into the optimization process. We validate our framework on linear and non-linear controlled generation problems including: image and audio inverse problems and conditional molecule generation reaching state of the art performance across all.