AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

Parts of Speech–Grounded Subspaces in Vision-Language Models

James Oldfield, Christos Tzelepis, Yannis Panagakis, Mihalis Nicolaou, Ioannis Patras

NeurIPS 2023 · Vision-language and multimodal › CLIP and zero-shot · approach: Fine-grained improvements to CLIP · 2 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
project image embedding onto noun submanifoldCLIP ViT-B-32 · ImageNettop-1 zero-shot accuracy: 56.50 -> 57.80Improved
project image embedding onto noun submanifoldCLIP ViT-B-32 · UT Zappostop-1 zero-shot accuracy: 82.20 -> 87.70Improved
project image embedding onto noun submanifoldCLIP ViT-B-32 · FER2013top-1 zero-shot accuracy: 37.80 -> 44.70Improved
project image embedding onto noun submanifoldCLIP ViT-B-32 · Flowers102top-1 zero-shot accuracy: 62.50 -> 61.00Hurt
project image embedding onto noun submanifoldCLIP ViT-B-16 · MNISTtop-1 zero-shot accuracy: 48.00 -> 55.00Improved
project image embedding onto noun submanifoldCLIP ViT-B-16 · UT Zappostop-1 zero-shot accuracy: 89.10 -> 88.10Hurt
project image embedding onto noun submanifoldCLIP ViT-L-14 · CIFAR10top-1 zero-shot accuracy: 91.70 -> 90.60Hurt
project image embedding onto noun submanifoldCLIP ViT-L-14 · FER2013top-1 zero-shot accuracy: 35.70 -> 46.40Improved
project image embedding onto noun submanifoldCLIP ViT-L-14 · MNISTtop-1 zero-shot accuracy: 59.70 -> 58.90Hurt
use geometry-aware submanifold instead of Euclidean PoS PGA subspace (ablation)noun projection · ImageNettop-1 zero-shot accuracy: 56.70 -> 57.80Improved
use Euclidean PoS PCA subspace instead of CLIP baseline (ablation)CLIP ViT-B-32 · ImageNettop-1 zero-shot accuracy: 56.50 -> 56.50No clear effect
lambda 0 vs 0.5 (no negative PoS guidance) (ablation)PoS subspace · class invariance (qualitative)top-1 zero-shot accuracyMixed

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