AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

PACEAttention: Principled and Adaptive Feature Compression-Expansion Grounded in the Geometry of $\text{MCR}^2$

Xiaojie Yu, Haibo Zhang, Jeremiah D. Deng, Lizhi Peng

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Sparse dictionary representations

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 softplus non-negativity constraint on expansion weights (ablation)PACENet-B · CIFAR-10accuracy vs rImproved
keep MLP in ToST (vs removing it) (ablation)ToST-S · CIFAR-10accuracy: 81.14 -> 84.50Improved
use PACEAttention (PACENet-S, 6M, r=10) instead of ToST-S without MLPToST-S-w/o MLP · CIFAR-10accuracy: 81.14 -> 81.67Improved
increase r from 10 to 40 (ablation)PACENet-S (6M) · CIFAR-10accuracy: 81.67 -> 82.17Improved
scale from PACENet-S (6M) to PACENet-B (20M) at r=40 (ablation)PACENet · CIFAR-10accuracy: 82.17 -> 83.85Improved
increase r beyond 20 (ablation)PACENet · CIFAR-10accuracyNo clear effect
scale PACENet from S to B+PACENet · ImageNet ReaLtop-1 accuracy: 79.0 -> 84.6Improved

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