AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Pay Attention to Attention Distribution: A New Local Lipschitz Bound for Transformers

Mikalai Yudzin, Sergei Kudriashov, Alexander Gaponov, Maxim Rakhuba

ECCV 2026 · Architectures and efficiency › Attention and transformers · approach: Attention head and block redesign · 12 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 JaSMin k=0 regularizerViT-B baseline · CIFAR-100 (ViT-B)PGD4 accuracy: 13.11 -> 18.90Improved
add JaSMin k=0 regularizerViT-B baseline · CIFAR-100 (ViT-B)Standard accuracy: 91.05 -> 88.83Hurt
add JaSMin k=0 regularizerViT-B baseline · CIFAR-100 (ViT-B)AA2 accuracy: 2.10 -> 3.62Improved
add JaSMin k=10 lambda=1e-3 regularizerViT-B baseline · CIFAR-100 (ViT-B)FGSM4 accuracy: 39.86 -> 42.34Improved
add JaSMin k=10 lambda=1e-3 regularizerViT-B baseline · CIFAR-100 (ViT-B)Standard accuracy: 91.05 -> 91.08No clear effect
add Specformer with all three weights 1e-2ViT-B baseline · CIFAR-100 (ViT-B)Standard accuracy: 91.05 -> 90.06Hurt
add Specformer weights (0,0,1e-2)ViT-B baseline · CIFAR-100 (ViT-B)PGD4 accuracy: 13.11 -> 12.56No clear effect
add JaSMin k=30 lambda=1e-2ViT-B baseline · Imagenette (ViT-B)PGD4 accuracy: 15.13 -> 20.67Improved
add JaSMin k=30 lambda=1e-2ViT-B baseline · Imagenette (ViT-B)Standard accuracy: 98.00 -> 97.07Hurt
add JaSMin k=70 lambda=1e-2ViT-B baseline · Imagenette (ViT-B)AA2 accuracyImproved
add Specformer (1e-2,0,0)ViT-B baseline · Imagenette (ViT-B)AA2 accuracyNo clear effect
use new bound instead of Gershgorin-based estimate for softmax Jacobian normsoftmax Jacobian estimate · ViT-L ImageNetmean relative error: 20.19 -> 0.035Improved

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