AI Paper Atlas

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

An Inverse Scaling Law for CLIP Training

Xianhang Li, Zeyu Wang, Cihang Xie

NeurIPS 2023 · Vision-language and multimodal › CLIP and zero-shot · approach: Fine-grained improvements to CLIP · 8 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
use global average pooling instead of class token (ablation)CLIPA-L/16 (I17,T16) · ImageNet-1k zero-shottop-1: 66.3 -> 67.8Improved
add color jitter augmentation (ablation)CLIPA-L/16 (I17,T16) · ImageNet-1k zero-shottop-1: 66.3 -> 67.1Improved
combine GAP and color jitter (ablation)CLIPA-L/16 (I17,T16) · ImageNet-1k zero-shottop-1: 66.3 -> 68.1Improved
fine-tune with batch size 7k instead of 32k (ablation)CLIPA-L/16 (I17,T16) · ImageNet-1k zero-shottop-1: 68.1 -> 67.6No clear effect
fine-tuning batch size sweep 32k to 2k (ablation)CLIPA-L/16 (I17,T16) · ImageNet-1k zero-shottop-1: 68.1 -> 66.8Hurt
scale model from L/16 to H/14 (ablation)CLIPA · ImageNet-1k zero-shottop-1: 69.3 -> 72.8Improved
switch data from LAION-400M to LAION-2B (ablation)CLIPA-H/14 · ImageNet-1k zero-shottop-1: 72.8 -> 74.1Improved
increase pre-training schedule 5x (12.8B samples) (ablation)CLIPA-H/14 · ImageNet-1k zero-shottop-1: 74.1 -> 77.9Improved
25% random masking during fine-tuning (ablation)CLIPA-H/14 · ImageNet-1k zero-shottop-1: 77.9 -> 78.2Improved
75% masking during fine-tuning (block) (ablation)CLIPA-H/14 · ImageNet-1k zero-shottop-1: 77.9 -> 74.3Hurt
4x longer fine-tuning with 30% random masking (ablation)CLIPA-H/14 · ImageNet-1k zero-shottop-1: 78.0 -> 78.6Improved
progressive fine-tuning 224 then 336 vs direct 336 (ablation)CLIPA-H/14 · ImageNet-1k zero-shottop-1: 78.9 -> 79.1Improved

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