프로그래밍선

Saining Xie 님의 엄청난 논문이 나왔다VAE를 완전 안써버릴 대단한 논문.. RAEhttps://arxiv.org/abs/2510.11690 Diffusion Transformers with Representation AutoencodersLatent 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 continarx..

REPA 를 뛰어넘을 논문이 나왔다! REG ! Representation Entanglement for Generation:Training Diffusion Transformers Is Much Easier Than You Think[Submitted on 2 Jul 2025]https://arxiv.org/abs/2507.01467 Representation Entanglement for Generation:Training Diffusion Transformers Is Much Easier Than You ThinkREPA and its variants effectively mitigate training challenges in diffusion models by incorporating exter..