FocusDiT: Masking Queries in Diffusion Transformers for Fine-grained Image Generation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Fang, Xueji, Ma, Liyuan, Zeng, Jianhao, Cao, Jinjin, Zhou, Mingyuan, Qi, Guo-Jun
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866916074285432832
author Fang, Xueji
Ma, Liyuan
Zeng, Jianhao
Cao, Jinjin
Zhou, Mingyuan
Qi, Guo-Jun
author_facet Fang, Xueji
Ma, Liyuan
Zeng, Jianhao
Cao, Jinjin
Zhou, Mingyuan
Qi, Guo-Jun
contents Diffusion transformer (DiT) has been widely adopted in the generative diffusion field, advancing the denoising of query tokens through attention and Feed-Forward (\text{FFN}) layers. FFN actually acts as the key-value vocabulary for decoding visual contents where the value embeds the visual semantical knowledge. We present that focusing on critical query tokens corresponding to more complex details and encouraging the model to improve these tokens is essential for fine-grained visual generation. To this end, we propose FocusDiT, which applies a Masking scheme to focus on critical query tokens that are exclusively fed into FFN. The masked queries can retrieve visual tokens from the FFN vocabularies, and use them to decode their visual details. Extensive text-to-image experiments validate the effectiveness of token masking in enhancing generative performance.
format Preprint
id arxiv_https___arxiv_org_abs_2606_02090
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle FocusDiT: Masking Queries in Diffusion Transformers for Fine-grained Image Generation
Fang, Xueji
Ma, Liyuan
Zeng, Jianhao
Cao, Jinjin
Zhou, Mingyuan
Qi, Guo-Jun
Computer Vision and Pattern Recognition
Diffusion transformer (DiT) has been widely adopted in the generative diffusion field, advancing the denoising of query tokens through attention and Feed-Forward (\text{FFN}) layers. FFN actually acts as the key-value vocabulary for decoding visual contents where the value embeds the visual semantical knowledge. We present that focusing on critical query tokens corresponding to more complex details and encouraging the model to improve these tokens is essential for fine-grained visual generation. To this end, we propose FocusDiT, which applies a Masking scheme to focus on critical query tokens that are exclusively fed into FFN. The masked queries can retrieve visual tokens from the FFN vocabularies, and use them to decode their visual details. Extensive text-to-image experiments validate the effectiveness of token masking in enhancing generative performance.
title FocusDiT: Masking Queries in Diffusion Transformers for Fine-grained Image Generation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2606.02090