On the Design of One-step Diffusion via Shortcutting Flow Paths

2025-12-03

AI Summary

Single-step diffusion models lack a unified theoretical foundation and modular, decoupled design principles. Method: This paper introduces the first systematic shortcut model design paradigm, establishing a general theoretical framework that rigorously derives and validates the validity of flow-path truncation modeling. It decouples three core components—noise scheduling, training objective, and network architecture—enabling modular innovation without relying on pretraining, knowledge distillation, or curriculum learning. The approach supports end-to-end single-step training and classifier-free guidance. Contribution/Results: On ImageNet-256×256, it achieves FID_50k = 2.85, setting a new state-of-the-art for single-step diffusion models. This work establishes a novel paradigm for efficient generative modeling that combines theoretical rigor with engineering flexibility.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Neural Architectures and Foundation Models
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Recent advances in few-step diffusion models have demonstrated their efficiency and effectiveness by shortcutting the probabilistic paths of diffusion models, especially in training one-step diffusion models from scratch (a.k.a. shortcut models). However, their theoretical derivation and practical implementation are often closely coupled, which obscures the design space. To address this, we propose a common design framework for representative shortcut models. This framework provides theoretical justification for their validity and disentangles concrete component-level choices, thereby enabling systematic identification of improvements. With our proposed improvements, the resulting one-step model achieves a new state-of-the-art FID50k of 2.85 on ImageNet-256x256 under the classifier-free guidance setting. Remarkably, the model requires no pre-training, distillation, or curriculum learning. We believe our work lowers the barrier to component-level innovation in shortcut models and facilitates principled exploration of their design space.

Problem

• Design framework for one-step diffusion models
• Improving image generation efficiency and quality
• Enabling systematic innovation in shortcut models

Innovation

• Framework for shortcutting diffusion model paths
• Component-level disentanglement for systematic improvements
• One-step generation without pre-training or distillation