Balanced conic rectified flow

2025-10-29

AI Summary

Conical rectified flows suffer from heavy reliance on large-scale generated image pairs, high computational overhead, and susceptibility to biases in synthetic data. To address these limitations, this paper proposes a rectified flow generative model that integrates real images. The core innovation introduces, for the first time in rectified flows, an ODE path preservation mechanism leveraging real samples, coupled with a real–generated image collaborative reflow optimization framework that jointly incorporates contrastive learning and real-data regularization. This approach substantially reduces dependence on generated image pairs: on CIFAR-10, it achieves superior FID scores over baseline methods using only a minimal number of generated pairs. Moreover, it enhances trajectory linearity, distribution fidelity, and the stability and quality of both single-step and full-step generation.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Search and Optimization: Learning to Search

Application Category

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

Abstract

Rectified flow is a generative model that learns smooth transport mappings between two distributions through an ordinary differential equation (ODE). Unlike diffusion-based generative models, which require costly numerical integration of a generative ODE to sample images with state-of-the-art quality, rectified flow uses an iterative process called reflow to learn smooth and straight ODE paths. This allows for relatively simple and efficient generation of high-quality images. However, rectified flow still faces several challenges. 1) The reflow process requires a large number of generative pairs to preserve the target distribution, leading to significant computational costs. 2) Since the model is typically trained using only generated image pairs, its performance heavily depends on the 1-rectified flow model, causing it to become biased towards the generated data. In this work, we experimentally expose the limitations of the original rectified flow and propose a novel approach that incorporates real images into the training process. By preserving the ODE paths for real images, our method effectively reduces reliance on large amounts of generated data. Instead, we demonstrate that the reflow process can be conducted efficiently using a much smaller set of generated and real images. In CIFAR-10, we achieved significantly better FID scores, not only in one-step generation but also in full-step simulations, while using only of the generative pairs compared to the original method. Furthermore, our approach induces straighter paths and avoids saturation on generated images during reflow, leading to more robust ODE learning while preserving the distribution of real images.

Problem

• Reduces computational costs of rectified flow by using fewer generative pairs
• Incorporates real images to reduce bias from generated data dependency
• Achieves better image quality with straighter ODE paths and lower FID scores

Innovation

• Incorporates real images into training process
• Reduces reliance on large generated datasets
• Uses smaller set of real and generated images