Editing Everything Everywhere All at Once

2026-06-30

AI Summary

This work addresses the challenge of semantic interference in concurrent multi-instruction image editing, which often leads to attribute leakage and difficulties in disentangling edits. The authors propose MICE, the first training-free and scalable approach for concurrent multi-instance editing. Built upon a multimodal diffusion Transformer, MICE leverages user-provided segmentation masks and, in a single forward pass, precisely associates instance-specific instructions, latent representations, and contextual interactions by modulating additive biases in joint attention. It further incorporates intra-instance enhancement, neighborhood suppression, and cross-instance irrelevant attention suppression mechanisms. Evaluated on LoMOE-Bench and the newly introduced MICE-Bench—featuring an average of 8.5 concurrent edits per image—MICE significantly outperforms existing methods in both visual fidelity and instruction adherence, effectively binding attributes while preserving global consistency.

Abstract

Editing multiple elements of an image in a single forward pass is a practical alternative to multi-turn image manipulation, offering improved efficiency and potentially better harmonization. However, when several instructions target different regions, semantic interference often leads to attribute leakage and poor edit disentanglement, especially as the number of edits increases. In this work, we propose MICE (Multi-Instance Concurrent Editing), a training-free strategy for scalable multi-instance image editing with Multimodal Diffusion Transformers. MICE modifies the additive bias of joint attention to regulate interactions between instance-specific edit instructions, latent, and context tokens identified via user-provided segmentation masks. Specifically, MICE allows intra-instance attention, penalizes interactions between neighboring region tokens, and suppresses unrelated cross-instance attention. As a result, our method enforces attribute binding while preserving global visual consistency. We evaluate MICE on LoMOE-Bench and introduce MICE-Bench, a more challenging benchmark with an average of 8.5 concurrent edits per image. The experiments demonstrate that our approach outperforms strong baselines and recent competitors in terms of visual quality preservation and faithfulness to the editing instructions.

Problem

• multi-instance editing
• semantic interference
• attribute leakage
• edit disentanglement
• concurrent image editing

Innovation

• Multi-Instance Editing
• Concurrent Image Manipulation
• Attention Bias Modulation
• Diffusion Transformers
• Edit Disentanglement