AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Activation with Intrinsic-Extrinsic Consensus

Tian Qiu, Zunlei Feng, Yang Gao, Bingde Hu, Yi Gao, Mingli Song

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Mechanistic circuit interpretability

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
use AIEC activation instead of ReLUViT-Tiny · CIFAR-100top-1 accuracy: 65.7 -> 71.1Improved
use AIEC activation instead of ReLUTNT-Small · ImageNet-1Ktop-1 accuracy: 73.4 -> 78.1Improved
add gatekeeping loss (auxiliary loss alone) (ablation)AIEC on ViT-Tiny · CIFAR-100top-1 accuracy: 66.7 -> 71.1Improved
use intrinsic-extrinsic consensus rather than intrinsic only (ablation)AIEC on ViT-Tiny · CIFAR-100top-1 accuracy: 69.7 -> 71.1Improved
use intrinsic-extrinsic consensus rather than extrinsic only (ablation)AIEC on ViT-Tiny · CIFAR-100top-1 accuracy: 68.5 -> 71.1Improved
use consensus (intersection) rather than union of assessments (ablation)AIEC on ViT-Tiny · CIFAR-100top-1 accuracy: 69.6 -> 71.1Improved
suppress only irrelevant channels instead of all channels (ablation)AIEC on ViT-Tiny · CIFAR-100top-1 accuracy: 65.8 -> 71.1Improved
use AIEC instead of L0 regularizationViT-Tiny · CIFAR-100top-1 accuracy: 67.2 -> 71.1Improved
use AIEC instead of SENetResNet-18 · CIFAR-100top-1 accuracy: 75.6 -> 77.0Improved
use AIEC for graph anomaly detection instead of ReLUGraphSAGE · T-FinanceAUC: 84.5 -> 90.9Improved
use AIEC for graph anomaly detection instead of ReLUGCN · Weibo21AUC: 98.5 -> 98.6No clear effect
use AIEC instead of GELUViT-Tiny training · ImageNet 224x224 batch 1024training GPU memory: 28.3 -> 32.5 GiBNo clear effect

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