AI Paper Atlas

Research map / Label-efficient and robust learning / Test-time adaptation

Align Your Prompts: Test-Time Prompting with Distribution Alignment for Zero-Shot Generalization

Jameel Hassan Abdul Samadh, Hanan Gani, Noor Hazim Hussein, Muhammad Uzair Khattak, Muzammal Naseer, Fahad Khan, Salman Khan

NeurIPS 2023 · Label-efficient and robust learning › Test-time adaptation · approach: Test-time adaptation of vision-language models · 14 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
Add token distribution alignment to MaPLe+TPTMaPLe+TPT · ImageNet OOD variantsTop-1 accuracy (OOD average): 62.31 -> 63.55Improved
Add token distribution alignment to MaPLe+TPTMaPLe · ImageNet-SketchTop-1 accuracy: 48.16 -> 50.23Improved
Add token distribution alignment to MaPLe+TPTMaPLe · PUG-ImageNet (Worlds)Top-1 accuracy: 42.00 -> 43.24Improved
Add token distribution alignment to MaPLe+TPTMaPLe · Cross-dataset evaluationTop-1 accuracy (average): 66.50 -> 66.92Improved
Add entropy minimization alongside distribution alignment (ablation)PromptAlign · ImageNet-ATop-1 accuracy: 50.85 -> 59.37Improved
Use distribution alignment alone without entropy minimization (ablation)PromptAlign · ImageNet-ATop-1 accuracy: 50.90 -> 50.85No clear effect
Replace mean-and-variance alignment with fifth-order Central Moment Discrepancy (ablation)PromptAlign · Cross-dataset evaluationTop-1 accuracy (average): 66.92 -> 66.99No clear effect
Replace ImageNet proxy statistics with statistics from a LAION400M subset (ablation)PromptAlign · Cross-dataset evaluationTop-1 accuracy (average): 66.92 -> 67.17Improved
Estimate test statistics from a bag of same-class samples (N=5) rather than one test sample (ablation)PromptAlign · Cross-dataset evaluationTop-1 accuracy (average): 66.92 -> 69.59Improved
Replace source-statistics alignment with regularization to previous prompts (ablation)PromptAlign · Cross-dataset evaluationTop-1 accuracy (average): 66.92 -> 62.87Hurt
Apply alignment loss across visual encoder layers 1 to 3 (ablation)PromptAlign · ImageNet-ATop-1 accuracyImproved
Increase the number of augmented views (ablation)PromptAlign · ImageNet validation and evaluated datasetsTop-1 accuracyImproved

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