# Selective Aggregation of Attention Maps Improves Diffusion-Based Visual Interpretation

Source: [Selective Aggregation of Attention Maps Improves Diffusion-Based Visual Interpretation](<https://www.sota2.com/research/paper/69d5e13c7d0172140ec75c7e>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-04-09T05:00:13.067Z

Authors: Jungwon Park, Jungmin Ko, Dongnam Byun, Wonjong Rhee

Published: 2026-04-07T14:09:00.000Z

## Abstract

Numerous studies on text-to-image (T2I) generative models have utilized cross-attention maps to boost application performance and interpret model behavior. However, the distinct characteristics of attention maps from different attention heads remain relatively underexplored. In this study, we show that selectively aggregating cross-attention maps from heads most relevant to a target concept can improve visual interpretability. Compared to the diffusion-based segmentation method DAAM, our approach achieves higher mean IoU scores. We also find that the most relevant heads capture concept-specific features more accurately than the least relevant ones, and that selective aggregation helps diagnose prompt misinterpretations. These findings suggest that attention head selection offers a promising direction for improving the interpretability and controllability of T2I generation.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Object Semantic Segmentation on Animals category 100 images generated from Stable Diffusion v1.4 (test)](<https://www.sota2.com/research/sota/object-semantic-segmentation-on-animals-category-100-images-generated-from-stable-diffusion-v1-4-test.md>): Object Semantic Segmentation; Animals category 100 images generated from Stable Diffusion v1.4 (test); mIoU: 77.65.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2604.05906>)

[PDF](<https://arxiv.org/pdf/2604.05906.pdf>)
