Jieru Zhao

Associate Professor, Shanghai Jiao Tong University

1 paper at NeurIPS 2025

Homepage· OpenReview· Semantic Scholar· Google Scholar

Poster Session 1

1 paper

Wednesday, December 3, 2025 · 11:00 AM → 2:00 PM

Exhibit Hall C,D,E

Communication-Efficient Diffusion Denoising Parallelization via Reuse-then-Predict Mechanism

#4110 · Kunyun Wang, Bohan Li, Kai Yu, Minyi Guo, Jieru Zhao

We propose ParaStep, a step-wise parallelization method based on a reuse-then-predict mechanism that parallelizes diffusion inference by exploiting adjacent-step similarity, reducing latency with minimal quality loss.