AI Paper Atlas

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

Debiased Fine-Tuning for Vision-Language Models by Prompt Regularization

Beier Zhu, Yulei Niu, Saeil Lee, Minhoe Hur, Hanwang Zhang

AAAI 2023 · Vision-language and multimodal › CLIP and zero-shot · approach: Prompt learning for CLIP · 19 citations (OpenAlex)

Read the paper

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
Regularize fine-tuning with zero-shot prompt predictions using a sample-wise adaptive weightCLIP fine-tuning · BARaccuracy: 90.9 -> 92.1Improved
Regularize fine-tuning with zero-shot prompt predictions using a sample-wise adaptive weightCLIP fine-tuning · DomainNet OODaccuracy: 63.7 -> 64.4Improved
Regularize fine-tuning with zero-shot prompt predictions using a sample-wise adaptive weightCLIP fine-tuning · DomainNet IDaccuracy: 75.9 -> 75.6Hurt
Regularize fine-tuning with zero-shot prompt predictions using a sample-wise adaptive weightCLIP fine-tuning · DomainNetharmonic mean: 67.4 -> 69.2Improved
Regularize fine-tuning with zero-shot prompt predictions using a sample-wise adaptive weightCLIP fine-tuning · NICO Animal OODaccuracy: 92.5 -> 94.4Improved
Regularize fine-tuning with zero-shot prompt predictions using a sample-wise adaptive weightCLIP fine-tuning · NICO Vehicle OODaccuracy: 95.9 -> 94.1Hurt
Regularize fine-tuning with zero-shot prompt predictions using a sample-wise adaptive weightCLIP fine-tuning · PACSaverage accuracy: 94.7 -> 96.0Improved
Regularize fine-tuning with zero-shot prompt predictions using a sample-wise adaptive weightViLT fine-tuning · VQA-CPaccuracy: 46.34 -> 54.89Improved
Use sample-wise adaptive weighting instead of a constant distillation weight (ablation)ProReg fine-tuning · VQA-CPaccuracy: 53.15; -1.74% gap to ProReg fine-tuneImproved
Use sample-wise adaptive weighting instead of a constant distillation weight (ablation)ProReg fine-tuning · NICO Vehicleharmonic meanImproved
Ensemble zero-shot and conventional fine-tuning predictions instead of using ProReg (ablation)CLIP and ViLT predictions · VQA-CPaccuracy: 53.48% vs 54.89%Hurt
Ensemble zero-shot and conventional fine-tuning predictions instead of using ProReg (ablation)CLIP fine-tuning · NICO Vehicleharmonic meanHurt

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