DetailFlow: 1D Coarse-to-Fine Autoregressive Image Generation via Next-Detail Prediction

2025-05-27

AI Summary

To address token redundancy, error accumulation, and low inference efficiency in autoregressive image generation models, this paper proposes DetailFlow—a one-dimensional coarse-to-fine autoregressive image generation framework. Its core innovation is a resolution-aware “next-detail prediction” mechanism that models images as progressively refined 1D token sequences, coupled with a parallel self-correcting sampling strategy to suppress error propagation. By integrating progressive degradation supervision with efficient 1D modeling, DetailFlow achieves a gFID of 2.96 on ImageNet 256×256 using only 128 tokens—outperforming VAR and FlexVAR. Moreover, it accelerates inference by approximately 2×, significantly reducing both computational overhead and token sequence complexity.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

This paper presents DetailFlow, a coarse-to-fine 1D autoregressive (AR) image generation method that models images through a novel next-detail prediction strategy. By learning a resolution-aware token sequence supervised with progressively degraded images, DetailFlow enables the generation process to start from the global structure and incrementally refine details. This coarse-to-fine 1D token sequence aligns well with the autoregressive inference mechanism, providing a more natural and efficient way for the AR model to generate complex visual content. Our compact 1D AR model achieves high-quality image synthesis with significantly fewer tokens than previous approaches, i.e. VAR/VQGAN. We further propose a parallel inference mechanism with self-correction that accelerates generation speed by approximately 8x while reducing accumulation sampling error inherent in teacher-forcing supervision. On the ImageNet 256x256 benchmark, our method achieves 2.96 gFID with 128 tokens, outperforming VAR (3.3 FID) and FlexVAR (3.05 FID), which both require 680 tokens in their AR models. Moreover, due to the significantly reduced token count and parallel inference mechanism, our method runs nearly 2x faster inference speed compared to VAR and FlexVAR. Extensive experimental results demonstrate DetailFlow's superior generation quality and efficiency compared to existing state-of-the-art methods.

Problem

• Develops a coarse-to-fine 1D autoregressive image generation method
• Reduces token count for efficient high-quality image synthesis
• Accelerates generation speed with parallel self-correcting inference

Innovation

• Coarse-to-fine 1D autoregressive image generation
• Next-detail prediction strategy
• Parallel inference with self-correction