AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

DICE: Staleness-Centric Optimizations for Parallel Diffusion MoE Inference

Jiajun Luo, Lizhuo Luo, Jianru Xu, Jiajun Song, Rongwei Lu, Chen Tang, Zhi Wang

ICCV 2025 · Image generation › Fast diffusion sampling · approach: Training-free diffusion acceleration · 1 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
switch displaced to interweaved parallelism (1-step instead of 2-step staleness) (ablation)displaced expert parallelism · ImageNet 256x256 DiT-MoE-XL 50 stepsFID: 8.27 -> 6.97Improved
add selective synchronization of deep layers (ablation)Interweaved Parallelism · ImageNet 256x256 DiT-MoE-XL 50 stepsFID: 6.97 -> 5.74Improved
add selective synchronization of shallow layers (ablation)Interweaved Parallelism · ImageNet 256x256 DiT-MoE-XL 50 stepsFID: 6.97 -> 6.55Improved
add selective synchronization of staggered layers (ablation)Interweaved Parallelism · ImageNet 256x256 DiT-MoE-XL 50 stepsFID: 6.97 -> 5.95Improved
add conditional communication with reduced frequency for low-score tokens (ablation)Interweaved Parallelism · ImageNet 256x256 DiT-MoE-XL 50 stepsFID: 6.97 -> 7.24Hurt
reduce communication for high-score tokens instead of low-score (ablation)Conditional Communication · ImageNet 256x256 DiT-MoE-XL 50 stepsFID: 7.24 -> 7.51Hurt
reduce communication for random tokens instead of low-score (ablation)Conditional Communication · ImageNet 256x256 DiT-MoE-XL 50 stepsFID: 7.24 -> 7.37Hurt
use full DICE instead of displaced expert parallelism (ablation)DiT-MoE-XL · ImageNet 256x256 DiT-MoE-XL 50 stepsFID: 8.27 -> 6.11Improved
use DICE instead of DistriFusion (ablation)DiT-MoE-XL · ImageNet 256x256 DiT-MoE-XL 50 stepsFID: 7.79 -> 6.11Improved
use DICE instead of displaced expert parallelism at 10 steps (ablation)DiT-MoE-XL · ImageNet 256x256 DiT-MoE-XL 10 stepsFID: 27.61 -> 15.13Improved
use DICE instead of displaced expert parallelism at 20 steps (ablation)DiT-MoE-XL · ImageNet 256x256 DiT-MoE-XL 20 stepsFID: 15.27 -> 8.60Improved
DICE versus synchronous expert parallelism, speedup at some quality costDiT-MoE-XL · ImageNet 256x256 DiT-MoE-XL 50 stepsFID: 5.31 -> 6.11; speedup up to 1.26xMixed

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