Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wang, Zitong, Shen, Zijun, Xu, Haohao, Luo, Zhengjie, Wu, Weibin
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2603.10210
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915851755585536
author Wang, Zitong
Shen, Zijun
Xu, Haohao
Luo, Zhengjie
Wu, Weibin
author_facet Wang, Zitong
Shen, Zijun
Xu, Haohao
Luo, Zhengjie
Wu, Weibin
contents While Diffusion Models excel in text-to-image synthesis, they often suffer from concept omission when synthesizing complex multi-instance scenes. Existing training-free methods attempt to resolve this by rescaling attention maps, which merely exacerbates unstructured noise without establishing coherent semantic representations. To address this, we propose Delta-K, a backbone-agnostic and plug-and-play inference framework that tackles omission by operating directly in the shared cross-attention Key space. Specifically, with Vision-language model, we extract a differential key $ΔK$ that encodes the semantic signature of missing concepts. This signal is then injected during the early semantic planning stage of the diffusion process. Governed by a dynamically optimized scheduling mechanism, Delta-K grounds diffuse noise into stable structural anchors while preserving existing concepts. Extensive experiments demonstrate the generality of our approach: Delta-K consistently improves compositional alignment across both modern DiT models and classical U-Net architectures, without requiring spatial masks, additional training, or architectural modifications.
format Preprint
id arxiv_https___arxiv_org_abs_2603_10210
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Delta-K: Boosting Multi-Instance Generation via Cross-Attention Augmentation
Wang, Zitong
Shen, Zijun
Xu, Haohao
Luo, Zhengjie
Wu, Weibin
Computer Vision and Pattern Recognition
Artificial Intelligence
While Diffusion Models excel in text-to-image synthesis, they often suffer from concept omission when synthesizing complex multi-instance scenes. Existing training-free methods attempt to resolve this by rescaling attention maps, which merely exacerbates unstructured noise without establishing coherent semantic representations. To address this, we propose Delta-K, a backbone-agnostic and plug-and-play inference framework that tackles omission by operating directly in the shared cross-attention Key space. Specifically, with Vision-language model, we extract a differential key $ΔK$ that encodes the semantic signature of missing concepts. This signal is then injected during the early semantic planning stage of the diffusion process. Governed by a dynamically optimized scheduling mechanism, Delta-K grounds diffuse noise into stable structural anchors while preserving existing concepts. Extensive experiments demonstrate the generality of our approach: Delta-K consistently improves compositional alignment across both modern DiT models and classical U-Net architectures, without requiring spatial masks, additional training, or architectural modifications.
title Delta-K: Boosting Multi-Instance Generation via Cross-Attention Augmentation
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2603.10210