ASPLOS 2026
Meta Info
Homepage: https://www.asplos-conference.org/asplos2026/
Acceptance Rate
Spring: 9.6% (= 20 / 208)
Papers
Diffusion Models
Video DiT Training
DSV: Exploiting Dynamic Sparsity to Accelerate Large-Scale Video DiT Training [arXiv]
CUHK & StepFun
Leverage the dynamic attention sparsity.
Adopt a hybrid sparsity-aware context parallelism that re-balances the skewed workload across attention heads and blocks due to sparsity heterogeneity.
Acronyms
DiT: Diffusion Transformer
Last updated
Was this helpful?