SiT-XL/2
PulseAugur coverage of SiT-XL/2 — every cluster mentioning SiT-XL/2 across labs, papers, and developer communities, ranked by signal.
-
TraceSampler method corrects generative sampling errors with minimal overhead
Researchers have developed TraceSampler, a novel post-training method designed to correct sampling errors in continuous-time generative models. This technique models endpoint error as a sample-specific prediction proble…
-
CoVA framework boosts diffusion transformer training with composed VAE features
Researchers have developed CoVA, a novel framework designed to enhance the training of diffusion transformers for generative learning. CoVA composes variational autoencoder (VAE) features to provide stage-wise guidance,…
-
New ZID metric offers improved evaluation for generative models
Researchers have introduced ZID (Z-resolved Integrated Diagnostic), a new evaluation metric for generative models that aims to improve upon existing metrics like Fréchet Inception Distance (FID) and Kernel Inception Dis…
-
New routing method boosts Diffusion Transformer training efficiency
Researchers have developed Diffusion-Adaptive Routing (DAR), a novel method to improve information flow in Diffusion Transformers (DiTs). By analyzing cross-layer information dynamics, they identified inefficiencies in …
-
Flow Matching research advances efficiency, control, and applications
Recent research explores advancements in Flow Matching, a generative modeling technique. Several papers introduce new methods to improve its efficiency, controllability, and applicability to diverse data types. Innovati…