AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Spatially-Adaptive Gradient Re-parameterization for 3D Large Kernel Optimization

Ho Hin Lee, Quan Liu, Shunxing Bao, Yuankai Huo, Bennett Allan Landman

ICML 2026 · Architectures and efficiency › CNN design · approach: Large and dynamic kernel design · 0 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
add LRBM learnable spatial bias to 3D UX-Net (k=7) (ablation)3D UX-Net (k=7) · AMOS CTaverage Dice: 0.890 -> 0.897Improved
use learnable LRBM generator instead of fixed distance prior (ablation)Rep3D · AMOS CT (from scratch)average Dice: 0.902 -> 0.910Improved
use learnable LRBM generator instead of fixed distance prior (ablation)Rep3D · AMOS MRI (from scratch)average Dice: 0.855 -> 0.864Improved
use learnable LRBM generator instead of fixed distance prior (ablation)Rep3D · KiTSmean Dice: 0.727 -> 0.736Improved
add spatial prior modulation (fixed prior) over vanilla plain large-kernel training (ablation)Rep3D · AMOS (60k iterations)validation Dice: 0.892 -> 0.902Improved
add learnable LRBM over vanilla plain large-kernel training (ablation)Rep3D · AMOS (60k iterations)validation Dice: 0.892 -> 0.910Improved
add learnable LRBM, training time cost (ablation)Rep3D · AMOStraining time (hrs): 17.3 -> 17.5No clear effect
use 3-layer instead of 2-layer generator (ablation)Rep3D LRBM · AMOS CTaverage Dice: 0.910 -> 0.899Hurt
use 1-layer instead of 2-layer generator (ablation)Rep3D LRBM · AMOS CTaverage Dice: 0.904 -> 0.910Hurt
increase generator kernel from 1x1x1 to 7x7x7 (ablation)Rep3D LRBM generator · AMOS CTaverage Dice: 0.905 -> 0.910Improved
generator kernel 5x5x5 instead of 3x3x3 (ablation)Rep3D LRBM generator · AMOS CTaverage Dice: 0.906 -> 0.903No clear effect
add a parallel 3x3x3 branch with higher learning rate (AdamW) (ablation)Rep3D 21x21x21 CSLA · AMOSmean Dice: 0.862 -> 0.889Improved

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