---
title: "Mage-Flow：用于图像生成与编辑的高效原生分辨率基础模型"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.19064v1"
permalink: "https://feed.astralor.com/items/itm_f02ef3d1984d1a6794d1af83"
published: "2026-07-21T12:55:34.000Z"
score: 80
channel: "ai"
tags: ["Diffusion Transformer","Mage-Flow","Mage-VAE","原生分辨率","图像生成","图像编辑"]
---

# Mage-Flow：用于图像生成与编辑的高效原生分辨率基础模型

## AI 点评

> 这篇论文让我眼前一亮，因为它没有盲目追求超大模型，而是通过精巧的协同设计在4B规模上实现了接近甚至超过更大模型的效果。对于想部署高分辨率图像生成应用的团队来说，Turbo变体的速度和低内存占用是直接可用的优势。尤其推荐给关注高效AI系统、边缘部署或交互式创意工具的读者。

## 速览

Mage-Flow是一个紧凑的4B参数生成堆栈，包含轻量级高保真潜在分词器Mage-VAE和原生分辨率多模态扩散变压器。通过锚点潜在正则化、原生分辨率打包和CUDA内核融合，训练吞吐量提升约2.5倍。该模型家族包括Base、RL对齐和Turbo变体，其中Turbo变体在单张A100上实现1024²分辨率图像生成仅0.59秒、编辑仅1.02秒，性能达到主流基准水平。

## 关键点

- Mage-VAE使用单步扩散式编解码和锚点潜在正则化，在保持重建质量的同时将分词成本降低一个数量级以上。
- 原生分辨率打包与栈级CUDA内核融合使端到端训练吞吐量提升约2.5倍。
- Diffusion-NFT改进提示遵循、文本渲染、美学质量和编辑保真度，少步蒸馏配合对抗感知引导生成4步Turbo模型。
- Turbo变体在单张A100上实现1024²分辨率图像生成0.59秒，编辑1.02秒，内存占用小。

## 全文

Large-scale visual generators are increasingly capable but costly to train, fine-tune, and deploy. We introduce Mage-Flow, a compact 4B-scale generative stack for efficient text-to-image generation and instruction-based image editing. The stack is built from two co-designed components: Mage-VAE, a lightweight high-fidelity latent tokenizer, and a Native-Resolution Multimodal Diffusion Transformer trained with rectified flow matching. Mage-VAE uses one-step diffusion-style encoding and decoding with anchor-latent regularization, preserving the reconstruction quality of strong public VAEs while reducing tokenization cost by more than an order of magnitude. Together with native-resolution packing and stack-level CUDA kernel fusion, the stack supports flexible-resolution training and improves end-to-end training throughput by about $2.5\times$. Built on this foundation, we develop a complete model family with Base, RL-aligned, and Turbo variants for both generation and editing. Diffusion-NFT improves prompt following, text rendering, aesthetic quality, and editing fidelity, while few-step distillation with adversarial perceptual guidance produces 4-step Turbo models for low-latency inference. Despite its compact scale, Mage-Flow and Mage-Flow-Edit achieves competitive performance across standard generation and editing benchmarks. More importantly, the Turbo variants make high-resolution generation and editing practical for interactive use: at $1024^2$ resolution on a single NVIDIA A100 GPU, Mage-Flow-Turbo generates an image in 0.59s, and Mage-Flow-Edit-Turbo edits an image in 1.02s, while maintaining a small memory footprint. These results show that careful tokenizer--backbone--system co-design can deliver strong high-resolution generation and editing within an efficient 4B model family.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_f02ef3d1984d1a6794d1af83