Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces the Interval Denoiser, a novel framework for latent-free image generation that leverages an exact analytical mapping derived from flow matching ODEs to enhance few-step sampling efficiency. By operating directly on pixel data and ensuring unbiased gradient evaluations, the model effectively trains on a low-dimensional manifold, which simplifies the regression process. The results demonstrate significant performance improvements, achieving a state-of-the-art FID of 4.55 in one step and 3.98 in two steps on the ImageNet dataset, highlighting the framework's potential for high-quality image synthesis with minimal computational overhead.
Achieving an FID of 3.98 in just two steps, the Interval Denoiser redefines the efficiency of latent-free image generation.
Modern diffusion and flow-based models are increasingly moving toward few-step, latent-free generation to bypass the computational overhead of multi-step sampling and the reconstruction bottlenecks of external autoencoders. We propose the Interval Denoiser, a theoretically rigorous framework for latent-free generation. Derived directly from the flow matching ODE, it establishes an exact analytical mapping for intermediate trajectory states. Unlike prior formulations, our prediction is shown to reside on a low-dimensional manifold across any time interval, making the regression tractable for a network operating directly on pixels. Furthermore, by avoiding empirical algebraic substitutions, our formulation correctly isolates the pure time derivative to prevent biased gradient evaluations and ensure exact first-order optimization. By analyzing this objective, we equip our framework with residual clipping and a time-sampling curriculum, enabling effective long-interval training and improving few-step performance. Trained from scratch on ImageNet 256x256, our model achieves an FID of 4.55 in one step (1-NFE) and 3.98 in two steps (2-NFE) without perceptual losses.