Stanford's Level-of-Token Diffusion Speeds Up Generation Up to 4.6x

Stanford's Gordon Wetzstein group proposes Level-of-Token (LoT) Diffusion, which allocates fine-grained tokens adaptively where detail is needed. With minimal changes to pretrained DiTs, it speeds up image generation up to 4.6x and accelerates video generation.

2026-10-07 ~ 2026-10-07 · 4 related posts

• Stanford's Level-of-Token Diffusion allocates fine tokens only where detail matters — GordonWetzstein · 2026-10-07
• Stanford's Level-of-Token Diffusion Cuts Video Gen Tokens, Speeds Up Generation Up to 4.6x — GordonWetzstein · 2026-10-07
• Level-of-Token DiT Lets Pretrained Diffusion Models Use Arbitrary-Size Token Grids — GordonWetzstein · 2026-10-07
• LoT: Adaptive Token Layout Speeds Up Image Generation Up to 4.6x — GordonWetzstein · 2026-10-07