EditAR: Unified Conditional Generation with Autoregressive Models

TL;DR

EditAR introduces a unified autoregressive framework capable of handling multiple conditional image generation tasks by integrating images and instructions, achieving competitive results across benchmarks with improved task alignment.

Contribution

The paper presents a novel unified autoregressive model for diverse image editing tasks, incorporating knowledge distillation to improve text-image alignment.

Findings

Competitive performance on multiple benchmarks

Effective knowledge distillation enhances task alignment

Unified model simplifies multi-task image generation

Abstract

Recent progress in controllable image generation and editing is largely driven by diffusion-based methods. Although diffusion models perform exceptionally well in specific tasks with tailored designs, establishing a unified model is still challenging. In contrast, autoregressive models inherently feature a unified tokenized representation, which simplifies the creation of a single foundational model for various tasks. In this work, we propose EditAR, a single unified autoregressive framework for a variety of conditional image generation tasks, e.g., image editing, depth-to-image, edge-to-image, segmentation-to-image. The model takes both images and instructions as inputs, and predicts the edited images tokens in a vanilla next-token paradigm. To enhance the text-to-image alignment, we further propose to distill the knowledge from foundation models into the autoregressive modeling process. We evaluate its effectiveness across diverse tasks on established benchmarks, showing competitive performance to various state-of-the-art task-specific methods. Project page: https://jitengmu.github.io/EditAR/