The Promise of RL for Autoregressive Image Editing

TL;DR

This paper demonstrates that reinforcement learning combined with a large multimodal verifier significantly improves autoregressive text-guided image editing, achieving competitive results with less training data.

Contribution

It introduces EARL, a novel RL-based autoregressive model for image editing that outperforms baselines and advances multimodal editing capabilities.

Findings

RL with a multimodal verifier is most effective

EARL performs competitively with less training data

The approach advances autoregressive multimodal image editing

Abstract

While image generation techniques are now capable of producing high-quality images that respect prompts which span multiple sentences, the task of text-guided image editing remains a challenge. Even edit requests that consist of only a few words often fail to be executed correctly. We explore three strategies to enhance performance on a wide range of image editing tasks: supervised fine-tuning (SFT), reinforcement learning (RL), and Chain-of-Thought (CoT) reasoning. In order to study all these components in one consistent framework, we adopt an autoregressive multimodal model that processes textual and visual tokens in a unified manner. We find RL combined with a large multi-modal LLM verifier to be the most effective of these strategies. As a result, we release EARL: Editing with Autoregression and RL, a strong RL-based image editing model that performs competitively on a diverse range of edits compared to strong baselines, despite using much less training data. Thus, EARL pushes the frontier of autoregressive multimodal models on image editing. We release our code, training data, and trained models at https://github.com/mair-lab/EARL.