AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

Multi-Prompt Alignment for Multi-Source Unsupervised Domain Adaptation

Haoran Chen, Xintong Han, Zuxuan Wu, Yu-Gang Jiang

NeurIPS 2023 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Feature alignment for unsupervised adaptation · 15 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 pair-specific prompts, denoise them with an autoencoder, and align reconstructed promptsMPA · ImageCLEF Avgtop-1 accuracy (%): 89.4 -> 91.7Improved
Use pair-specific prompts, denoise them with an autoencoder, and align reconstructed promptsMPA · Office-Home Avgtop-1 accuracy (%): 74.1 -> 75.4Improved
Use pair-specific prompts, denoise them with an autoencoder, and align reconstructed promptsMPA · DomainNet Avgtop-1 accuracy (%): 52.0 -> 54.1Improved
Tune the learned latent subspace instead of retraining MPA for a later targetMPA continuous adaptation · DomainNet Avg, continuous adaptationtop-1 accuracy (%): 53.9 -> 53.6Hurt
Replace MFSAN's ImageNet-pretrained backbone with CLIP's image encoder (ablation)MFSAN · ImageCLEF Avgtop-1 accuracy (%): 89.4 -> 89.1Hurt
Use domain-specific manual prompt templates instead of the simple class-name prompt (ablation)MPA · Office-Home Avgtop-1 accuracy (%): 75.4 -> 74.1; 75.4 -> 75.2Hurt
Raise the pseudo-label confidence threshold from 0.3 to 0.8 (ablation)MPA · Office-Home Avgtop-1 accuracy (%): 75.4 -> 74.7Hurt
Increase prompt token lengths from M1=M2=8 to M1=M2=20 (ablation)MPA · Office-Home Avgtop-1 accuracy (%): 75.1 -> 75.6Improved
Remove L1 prompt-alignment loss (ablation)MPA · Office-Home Avgtop-1 accuracy (%): 75.4 -> 74.7Hurt
Remove L1 prompt-alignment loss (ablation)MPA · Office-Home Cliparttop-1 accuracy (%): 54.9 -> 53.4Hurt
Remove autoencoder reconstruction loss (ablation)MPA · Office-Home Avgtop-1 accuracy (%): 75.4 -> 74.9Hurt
Remove domain-specific prompt tokens (ablation)MPA · Office-Home Avgtop-1 accuracy (%): 75.4 -> 74.4Hurt

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