AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Improving neural network representations using human similarity judgments

Lukas Muttenthaler, Lorenz Linhardt, Jonas Dippel, Robert A. Vandermeulen, Katherine Hermann, Andrew Kyle Lampinen, Simon Kornblith

NeurIPS 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 10 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
apply gLocal transformCLIP-ViT-L/14 (WIT) · Entity-13 5-shot5-shot accuracy: 65.34 -> 71.94Improved
apply gLocal transformCLIP-ViT-L/14 (WIT) · CIFAR100 anomaly detectionAUROC: 91.41 -> 97.19Improved
apply gLocal transformCLIP-ViT-L/14 (LAION-2B) · DTD 5-shot5-shot accuracy: 69.49 -> 68.44Hurt
apply gLocal transformImageNet models (AlexNet) · CIFAR100 5-shot5-shot accuracy: 28.70 -> 23.29Hurt
apply gLocal transformResNet-50 (ImageNet) · Entity-13 5-shot5-shot accuracy: 49.27 -> 59.18Improved
add local contrastive loss and shrinkage (gLocal) to naive alignment (ablation)CLIP-ViT-L/14 (WIT) · 5-shot FS (avg.)5-shot accuracy: 47.62 -> 69.91Improved
add local contrastive loss and shrinkage (gLocal) to naive alignment (ablation)CLIP-ViT-L/14 (WIT) · ImageNet nearest neighborsclosest neighbor preserved (%): 12.2 -> 76.3Improved
add local loss to global transform (ablation)CLIP ViT-L · ImageNet nearest neighborsclosest neighbor preserved (%): 73.7 -> 76.3Improved
gLocal instead of naive transform on human alignment (ablation)CLIP-ViT-L/14 (WIT) · THINGS odd-one-outodd-one-out accuracy (%): 60.13 -> 60.05No clear effect
apply gLocal on top of Tip-AdapterCLIP-ViT-L/14 (WIT) · CIFAR100 5-shot5-shot accuracy: 45.79 -> 64.51Improved
train on human-adversarial triplets (ablation)gLocal transform · downstream tasksaccuracyHurt

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