A-SelecT: Automatic Timestep Selection for Diffusion Transformer Representation Learning
This work addresses the bottleneck of timestep selection in Diffusion Transformers for discriminative tasks, offering a more efficient and effective approach for representation learning.
A-SelecT automatically selects the most informative timestep for Diffusion Transformer representation learning, eliminating exhaustive search and improving efficiency. It achieves state-of-the-art results on classification and segmentation benchmarks, outperforming prior diffusion-based methods.
Diffusion models have significantly reshaped the field of generative artificial intelligence and are now increasingly explored for their capacity in discriminative representation learning. Diffusion Transformer (DiT) has recently gained attention as a promising alternative to conventional U-Net-based diffusion models, demonstrating a promising avenue for downstream discriminative tasks via generative pre-training. However, its current training efficiency and representational capacity remain largely constrained due to the inadequate timestep searching and insufficient exploitation of DiT-specific feature representations. In light of this view, we introduce Automatically Selected Timestep (A-SelecT) that dynamically pinpoints DiT's most information-rich timestep from the selected transformer feature in a single run, eliminating the need for both computationally intensive exhaustive timestep searching and suboptimal discriminative feature selection. Extensive experiments on classification and segmentation benchmarks demonstrate that DiT, empowered by A-SelecT, surpasses all prior diffusion-based attempts efficiently and effectively.