AI Paper Atlas

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

RILS: Masked Visual Reconstruction in Language Semantic Space

Shusheng Yang, Yixiao Ge, Kun Yi, Dian Li, Ying Shan, Xiaohu Qie, Xinggang Wang

CVPR 2023 · Vision-language and multimodal › CLIP and zero-shot · approach: Recaptioning and text-supervised pretraining · 9 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
perform masked reconstruction using language-semantic text prototypes instead of raw pixels or learnable vision-space prototypes (ablation)RILS · ImageNet-1K, L-10Mzero-shot accuracy: 34.2 -> 37.5; 34.8 -> 37.5Improved
combine MIM and image-text contrastive learning in an end-to-end language-space reconstruction framework rather than two-stage trainingRILS · ImageNet-1K, L-10Mlinear probing accuracy: 64.8 -> 68.5Improved
reduce the mask ratio to 60% (ablation)RILS · ImageNet-1K, L-10Mlinear probing accuracy: 68.5 -> 69.3Improved
reduce the mask ratio to 60% (ablation)RILS · ImageNet-1K, L-10Mzero-shot accuracy: 37.5 -> 35.9Hurt
increase the mask ratio to 90% (ablation)RILS · ImageNet-1K, L-10Mlinear probing accuracy: 68.5 -> 67.6Hurt
increase the mask ratio to 90% (ablation)RILS · ImageNet-1K, L-10Mfine-tuning accuracy: 82.7 -> 82.3Hurt
increase decoder depth from 1 to 4 blocks (ablation)RILS · ImageNet-1K, L-10Mlinear probing accuracy: 68.5 -> 69.6Improved
increase decoder depth from 1 to 4 blocks (ablation)RILS · ImageNet-1K, L-10Mfine-tuning accuracy: 82.7 -> 82.5Hurt
increase decoder depth from 1 to 4 blocks (ablation)RILS · ImageNet-1K, L-10Mzero-shot accuracy: 37.5 -> 37.3No clear effect
increase contrastive-to-reconstruction loss ratio from 1:1 to 2:1 (ablation)RILS · ImageNet-1K, L-10Mzero-shot accuracy: 36.4 -> 37.5Improved
increase contrastive-to-reconstruction loss ratio from 1:2 to 2:1 (ablation)RILS · ImageNet-1K, L-10Mzero-shot accuracy: 35.9 -> 37.5Improved
scale pre-training data from LAION-10M to LAION-100MRILS · ImageNet-1Kzero-shot accuracy: 37.5 -> 50.6Improved

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