AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

Alignment with human representations supports robust few-shot learning

Ilia Sucholutsky, Thomas L. Griffiths

NeurIPS 2023 · Adversarial robustness and security › Attacks on text and graphs · approach: Biologically inspired robust vision · 5 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
select the 5 most Pearson-aligned models instead of the 5 closest to the meanpretrained vision models · ImageNet-A top-1accuracy: 10.26 -> 59.93Improved
select the 5 least-aligned models instead of the 5 mid-aligned modelspretrained vision models · ImageNet-A top-1accuracy: 10.26 -> 30.89Improved
select the 5 most Pearson-aligned models instead of the 5 closest to the meanpretrained vision models · ImageNet-R top-1accuracy: 39.16 -> 61.57Improved
select the 5 most Pearson-aligned models instead of the 5 closest to the meanpretrained vision models · ImageNet-S top-1accuracy: 26.90 -> 48.06Improved
select the 5 most triplet-aligned models instead of the 5 closest to the meanpretrained vision models · ImageNet-A top-5accuracy: 46.63 -> 72.20Improved
use z^2 alignment instead of raw alignment to predict robustnessalignment metrics · ImageNet-AcorrelationMixed
measure alignment on dissimilar data (MNIST instead of CIFAR100)Pearson z^2 alignment · MNIST few-shotcorrelation with accuracyHurt
add ImageNet-21k/22k pre-training before ImageNet-1k fine-tuningvision models · alignment and few-shotalignment and accuracyImproved
apply knowledge distillationvision models · alignment and few-shotalignment and accuracyImproved
add batch normalization to VGGVGG · alignment and few-shotalignmentMixed
prune EfficientNetEfficientNet · alignment and few-shotalignmentMixed
increase ResNet depthResNet · alignment and few-shotalignmentMixed

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