BindEdit: Taming Attention Leakage for Precise Multi-Object Image Editing

2026-06-17

AI Summary

Existing image editing methods often suffer from attention leakage in multi-object scenes, leading to semantic ambiguity, object duplication, or incomplete edits. This work is the first to explicitly distinguish two types of attention leakage—Edit-Token Leakage and Source Dominance—and introduces a novel attention regularization framework operating within a single diffusion trajectory. The proposed approach binds target text tokens to their corresponding spatial regions, jointly regularizes cross- and self-attention mechanisms, rebalances cross-attention weights, and incorporates a region fidelity term to amplify target influence while suppressing source interference. Evaluated on a newly curated multi-object editing benchmark, the method significantly outperforms current state-of-the-art techniques, achieving robust, precise, and high-quality editing results across both single- and multi-object tasks.

Abstract

Real image editing enables precise manipulation of visual content, yet existing methods often fail in complex multi-object scenarios, causing semantic blending, object duplication, or incomplete edits. We attribute these failures to attention leakage, where signals across spatial regions and text tokens become entangled during the denoising process. Specifically, we identify two distinct forms of leakage: Edit-Token Leakage, where ambiguous token-region alignment leads to object blending, and Source Dominance Leakage, where tokens of unchanged source objects overwhelm the attention intended for target entities. To resolve these leakages, we propose \textbf{BindEdit}, which enforces attention-level constraints within a single diffusion trajectory. To suppress Edit-Token Leakage, BindEdit jointly regularizes cross- and self-attention so that each target token group is bound to its corresponding spatial region while maintaining instance-level separation. To suppress Source Dominance Leakage, a cross-attention re-balancing mechanism amplifies target token influence and attenuates residual source semantics within editable regions. Moreover, a region fidelity term ensures that each target concept is expressed coherently across the entire editing mask. Additionally, we propose a comprehensive multi-object benchmark encompassing diverse object counts and categories. Extensive experiments demonstrate that BindEdit consistently outperforms existing methods within a single diffusion trajectory, maintaining robust performance across both single- and multi-object editing scenarios.

Problem

• multi-object image editing
• attention leakage
• semantic blending
• object duplication
• incomplete edits

Innovation

• attention leakage
• multi-object image editing
• diffusion models
• cross-attention regularization
• BindEdit