B-DENSE: Branching For Dense Ensemble Network Supervision Efficiency

TL;DR

B-DENSE introduces a multi-branch trajectory alignment framework for diffusion model distillation, enabling dense supervision of intermediate steps, which improves image generation quality and reduces discretization errors.

Contribution

It proposes a novel multi-branch architecture that aligns intermediate trajectories in diffusion models, enhancing distillation efficiency and generative performance.

Findings

Improved image generation quality over baseline methods.

Effective dense supervision of intermediate steps.

Reduced discretization errors in diffusion model distillation.

Abstract

Inspired by non-equilibrium thermodynamics, diffusion models have achieved state-of-the-art performance in generative modeling. However, their iterative sampling nature results in high inference latency. While recent distillation techniques accelerate sampling, they discard intermediate trajectory steps. This sparse supervision leads to a loss of structural information and introduces significant discretization errors. To mitigate this, we propose B-DENSE, a novel framework that leverages multi-branch trajectory alignment. We modify the student architecture to output $K$-fold expanded channels, where each subset corresponds to a specific branch representing a discrete intermediate step in the teacher's trajectory. By training these branches to simultaneously map to the entire sequence of the teacher's target timesteps, we enforce dense intermediate trajectory alignment. Consequently, the student model learns to navigate the solution space from the earliest stages of training, demonstrating superior image generation quality compared to baseline distillation frameworks.