Diffusion Transformers with Representation Autoencoders episode artwork

EPISODE · Oct 15, 2025 · 24 MIN

Diffusion Transformers with Representation Autoencoders

from Daily Paper Cast · host Jingwen Liang, Gengyu Wang

🤗 Upvotes: 93 | cs.CV, cs.LG Authors: Boyang Zheng, Nanye Ma, Shengbang Tong, Saining Xie Title: Diffusion Transformers with Representation Autoencoders Arxiv: http://arxiv.org/abs/2510.11690v1 Abstract: Latent generative modeling, where a pretrained autoencoder maps pixels into a latent space for the diffusion process, has become the standard strategy for Diffusion Transformers (DiT); however, the autoencoder component has barely evolved. Most DiTs continue to rely on the original VAE encoder, which introduces several limitations: outdated backbones that compromise architectural simplicity, low-dimensional latent spaces that restrict information capacity, and weak representations that result from purely reconstruction-based training and ultimately limit generative quality. In this work, we explore replacing the VAE with pretrained representation encoders (e.g., DINO, SigLIP, MAE) paired with trained decoders, forming what we term Representation Autoencoders (RAEs). These models provide both high-quality reconstructions and semantically rich latent spaces, while allowing for a scalable transformer-based architecture. Since these latent spaces are typically high-dimensional, a key challenge is enabling diffusion transformers to operate effectively within them. We analyze the sources of this difficulty, propose theoretically motivated solutions, and validate them empirically. Our approach achieves faster convergence without auxiliary representation alignment losses. Using a DiT variant equipped with a lightweight, wide DDT head, we achieve strong image generation results on ImageNet: 1.51 FID at 256x256 (no guidance) and 1.13 at both 256x256 and 512x512 (with guidance). RAE offers clear advantages and should be the new default for diffusion transformer training.

Episode metadata supplied by the publisher feed · Published Oct 15, 2025

🤗 Upvotes: 93 | cs.CV, cs.LG Authors: Boyang Zheng, Nanye Ma, Shengbang Tong, Saining Xie Title: Diffusion Transformers with Representation Autoencoders Arxiv: http://arxiv.org/abs/2510.11690v1 Abstract: Latent generative modeling, where a pretrained autoencoder maps pixels into a latent space for the diffusion process, has become the standard strategy for Diffusion Transformers (DiT); however, the autoencoder component has barely evolved. Most DiTs continue to rely on the original VAE encoder, which introduces several limitations: outdated backbones that compromise architectural simplicity, low-dimensional latent spaces that restrict information capacity, and weak representations that result from purely reconstruction-based training and ultimately limit generative quality. In this work, we explore replacing the VAE with pretrained representation encoders (e.g., DINO, SigLIP, MAE) paired with trained decoders, forming what we term Representation Autoencoders (RAEs). These models provide both high-quality reconstructions and semantically rich latent spaces, while allowing for a scalable transformer-based architecture. Since these latent spaces are typically high-dimensional, a key challenge is enabling diffusion transformers to operate effectively within them. We analyze the sources of this difficulty, propose theoretically motivated solutions, and validate them empirically. Our approach achieves faster convergence without auxiliary representation alignment losses. Using a DiT variant equipped with a lightweight, wide DDT head, we achieve strong image generation results on ImageNet: 1.51 FID at 256x256 (no guidance) and 1.13 at both 256x256 and 512x512 (with guidance). RAE offers clear advantages and should be the new default for diffusion transformer training.

PodParley-generated summary based on available episode metadata and transcript content.

NOW PLAYING

Diffusion Transformers with Representation Autoencoders

0:00 24:28

No transcript for this episode yet

We transcribe on demand. Request one and we'll notify you when it's ready — usually under 10 minutes.

No similar episodes found.

Frequently Asked Questions

How long is this episode of Daily Paper Cast?

This episode is 24 minutes long.

When was this Daily Paper Cast episode published?

This episode was published on October 15, 2025.

What is this episode about?

🤗 Upvotes: 93 | cs.CV, cs.LG Authors: Boyang Zheng, Nanye Ma, Shengbang Tong, Saining Xie Title: Diffusion Transformers with Representation Autoencoders Arxiv: ...

Can I download this Daily Paper Cast episode?

Yes, you can download this episode by clicking the download button on the episode player, or subscribe to the podcast in your preferred podcast app for automatic downloads.
URL copied to clipboard!