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

2026-06-01

AI Summary

Existing diffusion Transformers struggle to effectively model complex, fine-grained details in image generation. To address this limitation, this work proposes a query masking mechanism that dynamically filters input queries to the feed-forward network (FFN) within the Diffusion Transformer architecture, retaining only the most critical queries to enhance semantic decoding of salient visual regions. By enabling query-level fine-grained control, the proposed method significantly improves both the fidelity of intricate details and the overall quality of images generated in text-to-image tasks.

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.

Problem

• fine-grained image generation
• diffusion transformer
• query tokens
• visual details
• token masking

Innovation

• Diffusion Transformer
• Query Masking
• Fine-grained Generation
• Feed-Forward Network
• Token Retrieval