aiqbench analysis
This paper addresses scalability challenges in Generative Adversarial Networks by integrating compact latent spaces with purely transformer-based architectures. The authors identify specific failure modes like early-layer underutilization and propose practical solutions such as lightweight intermediate supervision and width-aware learning rate adjustments. These modifications enable reliable training across various model sizes while achieving state-of-the-art performance with significantly fewer epochs than existing baselines.
Scalable GANs with Transformers
arXiv:2509.24935v3 Announce Type: replace-cross Abstract: Scalability has driven recent advances in generative modeling, yet its principles remain underexplored for adversarial learning. We investigate the scalability of Generative Adversarial Networks (GANs) through two design choices that have proven to be effective in other types of generative models: training in a compact Variational Autoencoder latent space and adopting purely transformer-based generators and discriminators. Training in latent space enables efficient computation while preserving perceptual fidelity, and this efficiency pairs naturally with plain transformers, whose performance scales with computational budget. Building on these choices, we analyze failure modes that emerge when naively scaling GANs. Specifically, we find issues as underutilization of early layers in the generator and optimization instability as the network scales. Accordingly, we provide simple and scale-friendly solutions as lightweight intermediate supervision and width-aware learning-rate adjustment. Our experiments show that GAT, a purely transformer-based and latent-space GANs, can be easily trained reliably across a wide range of capacities (S through XL). Moreover, GAT-XL/2 achieves state-of-the-art single-step, class-conditional generation performance (FID of 2.18) on ImageNet-256 in just 60 epochs, 4x fewer epochs than strong baselines. Project page: https://hse1032.github.io/GAT.
Abstract:Scalability has driven recent advances in generative modeling, yet its principles remain underexplored for adversarial learning. We investigate the scalability of Generative Adversarial Networks (GANs) through two design choices that have proven to be effective in other types of generative models: training in a compact Variational Autoencoder latent space and adopting purely transformer-based generators and discriminators. Training in latent space enables efficient computation while preserving perceptual fidelity, and this efficiency pairs naturally with plain transformers, whose performance scales with computational budget. Building on these choices, we analyze failure modes that emerge when naively scaling GANs. Specifically, we find issues as underutilization of early layers in the generator and optimization instability as the network scales. Accordingly, we provide simple and scale-friendly solutions as lightweight intermediate supervision and width-aware learning-rate adjustment. Our experiments show that GAT, a purely transformer-based and latent-space GANs, can be easily trained reliably across a wide range of capacities (S through XL). Moreover, GAT-XL/2 achieves state-of-the-art single-step, class-conditional generation performance (FID of 2.18) on ImageNet-256 in just 60 epochs, 4x fewer epochs than strong baselines. Project page: this https URL.
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
arXiv:2509.24935 [cs.CV]
(or arXiv:2509.24935v3 [cs.CV] for this version)
Full text under the publisher's open licence · 202 words
Related coverage
This is SParTA: A modular, efficient and scalable transformer architecture for jet-tagging
arXiv · Oct 21, 2026
Global Communication or Graph-Specific Memory?
arXiv · Oct 6, 2026
Scaling Transformer Energy-based Models For High Quality Image Synthesis
arXiv · Oct 27, 2026
Image-Conditional Diffusion Transformer for Underwater Image Enhancement
arXiv · Aug 25, 2026