FocusDiT: Masking Queries in Diffusion Transformers for Fine-grained Image Generation
2026-06-01 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors study Diffusion Transformer (DiT) models used for generating images, focusing on how certain parts called Feed-Forward Networks (FFNs) help decode visual details. They find that paying special attention to important tokens—pieces of information that represent complex image parts—can improve image quality. To do this, the authors propose FocusDiT, which masks out less important tokens so the model concentrates on key ones during decoding. Their experiments show that this selective attention helps the model create better detailed images.
Diffusion TransformerFeed-Forward NetworkQuery TokenMasking SchemeVisual TokenToken MaskingGenerative DiffusionText-to-Image Generation
Authors
Xueji Fang, Liyuan Ma, Jianhao Zeng, Jinjin Cao, Mingyuan Zhou, Guo-Jun Qi
Abstract
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.