Single-Step Bidirectional Unpaired Image Translation Using Implicit Bridge Consistency Distillation

ICLR 2026 Conference Submission7212 Authors

Keywords: image translation, consistency distillation, unpaired, single-step, diffusion models

TL;DR: IBCD enables single-step bidirectional unpaired image translation by carefully reformulating consistency distillation with a diffusion implicit bridge, enhanced by distribution matching and adaptive weighting, achieving state-of-the-art results.

Abstract: Unpaired image-to-image translation has seen significant progress since the introduction of CycleGAN. However, methods based on diffusion models or Schrödinger bridges have yet to be widely adopted in real-world applications due to their iterative sampling nature. To address this challenge, we propose a novel framework, Implicit Bridge Consistency Distillation (IBCD), which enables single-step bidirectional unpaired translation without using adversarial loss. IBCD reformulates consistency distillation by using a diffusion implicit bridge model that connects PF-ODE trajectories between distributions, with a novel design parametrization to enable effective translation in a single step. Additionally, we introduce two key improvements: 1) distribution matching for consistency distillation and 2) adaptive weighting method based on distillation difficulty. Experimental results demonstrate that IBCD achieves state-of-the-art performance on benchmark datasets in a single generation step.

Primary Area: generative models

Submission Number: 7212