Search papers, labs, and topics across Lattice.
DSS-GAN introduces a novel GAN architecture using Mamba as a hierarchical generator backbone for class-conditional image synthesis. The key innovation is Directional Latent Routing (DLR), which decomposes the latent vector into direction-specific subvectors conditioned on class embeddings to modulate Mamba scans. Experiments demonstrate that DSS-GAN achieves superior FID, KID, and precision-recall scores compared to StyleGAN2-ADA, and that the latent space exhibits specialization with direction-correlated changes in synthesized images.
Mamba, the darling of sequence modeling, now powers a GAN that beats StyleGAN2-ADA in image synthesis, thanks to a clever latent space routing trick.
We present DSS-GAN, the first generative adversarial network to employ Mamba as a hierarchical generator backbone for noise-to-image synthesis. The central contribution is Directional Latent Routing (DLR), a novel conditioning mechanism that decomposes the latent vector into direction-specific subvectors, each jointly projected with a class embedding to produce a feature-wise affine modulation of the corresponding Mamba scan. Unlike conventional class conditioning that injects a global signal, DLR couples class identity and latent structure along distinct spatial axes of the feature map, applied consistently across all generative scales. DSS-GAN achieves improved FID, KID, and precision-recall scores compared to StyleGAN2-ADA across multiple tested datasets. Analysis of the latent space reveals that directional subvectors exhibit measurable specialization: perturbations along individual components produce structured, direction-correlated changes in the synthesized image.
