AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

HierarchicalPrune: Position-Aware Compression for Large-Scale Diffusion Models

Young D. Kwon, Rui Li, Sijia Li, Da Li, Sourav Bhattacharya, Stylianos I. Venieris

AAAI 2026 · Image generation › Fast diffusion sampling · approach: Training-free diffusion acceleration · 2 citations (OpenAlex)

Read the paper

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
add positional weight preservation (PWP) to HPP, 20% pruning (ablation)HierarchicalPrune on SD3.5 Large Turbo · GenEvalGenEval score: 0.03 -> 0.71Improved
add PWP to HPP, 20% pruning (ablation)HierarchicalPrune on SD3.5 Large Turbo · HPSv2HPSv2 score: 11.08 -> 28.97Improved
add PWP to HPP, 30% pruning (ablation)HierarchicalPrune on SD3.5 Large Turbo · GenEvalGenEval score: 0.0 -> 0.46Improved
add sensitivity-guided distillation (SGDistill) to HPP+PWP, 30% pruning (ablation)HierarchicalPrune on SD3.5 Large Turbo · GenEvalGenEval score: 0.46 -> 0.64Improved
add SGDistill to HPP+PWP, 30% pruning (ablation)HierarchicalPrune on SD3.5 Large Turbo · HPSv2HPSv2 score: 21.74 -> 27.29Improved
add W4A16 quantisation at 20% pruning (ablation)HierarchicalPrune on SD3.5 Large Turbo · GenEvalGenEval score: 0.71 -> 0.69Hurt
add W4A16 quantisation at 30% pruning (ablation)HierarchicalPrune on SD3.5 Large Turbo · HPSv2HPSv2 score: 27.29 -> 26.29Hurt
use HierarchicalPrune (HPP+PWP+Q) instead of KOALASD3.5 Large Turbo · HPSv2HPSv2 score: 19.99 -> 28.15Improved
use HierarchicalPrune instead of BK-SDM on FLUX.1 SchnellFLUX.1 Schnell · GenEvalGenEval score: 0.45 -> 0.64Improved
use HierarchicalPrune instead of KOALA/BK-SDM on SD3.5 for latencySD3.5 Large Turbo · Per-step latency (A6000)latency (ms): 642 ms -> 593 msImproved
use HierarchicalPrune instead of KOALA/BK-SDM on FLUX.1 for latencyFLUX.1 Schnell · Per-step latency (A6000)latency (ms): 432 ms -> 469 msHurt

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