AI Paper Atlas

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

Sparse CLIP: Co-Optimizing Interpretability and Performance in Contrastive Learning

Chuan Qin, Constantin Venhoff, Sonia Joseph, Fanyi Xiao, Stefan Scherer

ICLR 2026 · Vision-language and multimodal › CLIP and zero-shot · approach: CLIP training recipes and data curation · 1 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
train natively sparse CLIP (0.66% L0) vs denseViT-L/14 CLIP · zero-shot classification averageaccuracy: 75.1 -> 75.6Improved
train natively sparse CLIP vs denseViT-L/14 CLIP · zero-shot fine-grained averageaccuracy: 80.4 -> 81.0Improved
train natively sparse CLIP vs denseViT-L/14 CLIP · COCO BBox classificationAcc@1: 53.3 -> 55.5Improved
train natively sparse CLIP vs denseViT-L/14 CLIP · COCO retrievalIR@1: 45.5 -> 43.7Hurt
lower logit scale cap from 50 to 40 (Sparse+) (ablation)ViT-L/14 Sparse · zero-shot classification averageaccuracy: 75.6 -> 75.1No clear effect
lower logit scale cap from 50 to 40 (Sparse+) (ablation)ViT-L/14 Sparse · COCO retrievalIR@1: 43.7 -> 41.8Hurt
native sparse training vs post-hoc SAE (Prisma)OpenAI ViT-B/32 dense · zero-shot classification averageaccuracy: 57.5 -> 56.6Hurt
add non-negativity (ReLU) with dimension expansion (ablation)Sparse CLIP ViT-B/32 (15M pairs) · ImageNet-1k zero-shotaccuracyImproved
use ReLU instead of L1 loss or TopK for sparsity (ablation)Sparse CLIP ViT-B/32 (15M pairs) · ImageNet-1k zero-shotaccuracyImproved
reduce logit scale cap to 20 (ablation)Sparse CLIP ViT-B/32 (15M pairs) · ImageNet-1k zero-shotaccuracyHurt
use Sparse+ vision encoder in VLMViT-L/14 baseline VLM · AI2Daccuracy: 67.2 -> 70.6Improved
use Sparse+ vision encoder in VLMViT-L/14 baseline VLM · TextVQAaccuracy: 48.9 -> 48.5No 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