TIDE : Temporal-Aware Sparse Autoencoders for Interpretable Diffusion Transformers in Image Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Huang, Victor Shea-Jay, Zhuo, Le, Xin, Yi, Wang, Zhaokai, Wang, Fu-Yun, Wang, Yuchi, Zhang, Renrui, Gao, Peng, Li, Hongsheng
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909734155583488
author Huang, Victor Shea-Jay
Zhuo, Le
Xin, Yi
Wang, Zhaokai
Wang, Fu-Yun
Wang, Yuchi
Zhang, Renrui
Gao, Peng
Li, Hongsheng
author_facet Huang, Victor Shea-Jay
Zhuo, Le
Xin, Yi
Wang, Zhaokai
Wang, Fu-Yun
Wang, Yuchi
Zhang, Renrui
Gao, Peng
Li, Hongsheng
contents Diffusion Transformers (DiTs) are a powerful yet underexplored class of generative models compared to U-Net-based diffusion architectures. We propose TIDE-Temporal-aware sparse autoencoders for Interpretable Diffusion transformErs-a framework designed to extract sparse, interpretable activation features across timesteps in DiTs. TIDE effectively captures temporally-varying representations and reveals that DiTs naturally learn hierarchical semantics (e.g., 3D structure, object class, and fine-grained concepts) during large-scale pretraining. Experiments show that TIDE enhances interpretability and controllability while maintaining reasonable generation quality, enabling applications such as safe image editing and style transfer.
format Preprint
id arxiv_https___arxiv_org_abs_2503_07050
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle TIDE : Temporal-Aware Sparse Autoencoders for Interpretable Diffusion Transformers in Image Generation
Huang, Victor Shea-Jay
Zhuo, Le
Xin, Yi
Wang, Zhaokai
Wang, Fu-Yun
Wang, Yuchi
Zhang, Renrui
Gao, Peng
Li, Hongsheng
Computer Vision and Pattern Recognition
Artificial Intelligence
Multimedia
Diffusion Transformers (DiTs) are a powerful yet underexplored class of generative models compared to U-Net-based diffusion architectures. We propose TIDE-Temporal-aware sparse autoencoders for Interpretable Diffusion transformErs-a framework designed to extract sparse, interpretable activation features across timesteps in DiTs. TIDE effectively captures temporally-varying representations and reveals that DiTs naturally learn hierarchical semantics (e.g., 3D structure, object class, and fine-grained concepts) during large-scale pretraining. Experiments show that TIDE enhances interpretability and controllability while maintaining reasonable generation quality, enabling applications such as safe image editing and style transfer.
title TIDE : Temporal-Aware Sparse Autoencoders for Interpretable Diffusion Transformers in Image Generation
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Multimedia
url https://arxiv.org/abs/2503.07050