AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

SiT: Exploring Flow and Diffusion-based Generative Models with Scalable Interpolant Transformers

Nanye Ma, Mark Goldstein, Michael Albergo, Nicholas M Boffi, Eric Vanden-Eijnden, Saining Xie

ECCV 2024 · Image generation › Fast diffusion sampling · approach: Diffusion theory and design space · 372 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use velocity-based SiT recipe (continuous time, velocity prediction, Linear interpolant, SDE sampler) instead of DiT recipeDiT backbone, all sizes · ImageNet 256x256 (400K steps)FID-50K: 43.5 -> 33.0 (B), 19.5 -> 17.2 (XL)Improved
Switch training from discrete to continuous time (ablation)SiT-B · ImageNet 256x256FID-50KImproved
Predict velocity or use weighted score loss instead of plain score loss (ablation)SBDM-VP SiT-B · ImageNet 256x256FID-50KImproved
Replace SBDM-VP interpolant with GVP or Linear interpolant (ablation)SiT-B velocity model · ImageNet 256x256FID-50KImproved
Sample with SDE instead of ODE (ablation)SiT-B · ImageNet 256x256, 400KFID-50K: 34.84 -> 33.02Improved
Sample with SDE instead of ODE at low NFE (ablation)SiT · ImageNet 256x256FID vs NFEMixed
Use KL-optimal diffusion coefficient w^KL for GVP velocity model (ablation)SiT-B GVP velocity · ImageNet 256x256FID-50K: 33.4 (sigma_t) -> 32.9 (w^KL)Improved
Use cost-regularized w^{KL,eta} for Linear interpolant (ablation)SiT-B Linear velocity · ImageNet 256x256FID-50K: 33.6 (w^KL) -> 33.0Improved
Use regularized w^{KL,eta} for SBDM-VP (ablation)SiT-B SBDM-VP velocity · ImageNet 256x256FID-50K: 37.8 (w^KL) -> 41.1Hurt
Learn score instead of velocity (ablation)SiT-B Linear interpolant · ImageNet 256x256FID-50K: 33.6 -> 41.0 (w^KL)Hurt
Apply classifier-free guidance (cfg=1.5) to SiT-XL at 7M stepsSiT-XL · ImageNet 256x256FID-50K: 8.3 -> 2.06Improved
Increase SiT model size (S to XL) (ablation)SiT · ImageNet 256x256, 400KFID-50K: 57.6 -> 17.2Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy