AI Paper Atlas

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

Test-Time Prompt Tuning for Zero-Shot Generalization in Vision-Language Models

Manli Shu, Weili Nie, De-An Huang, Zhiding Yu, Tom Goldstein, Anima Anandkumar, Chaowei Xiao

NeurIPS 2022 · Vision-language and multimodal › CLIP and zero-shot · approach: Prompt learning for CLIP · 111 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
Tune the prompt on a single test image using confidence-selected augmented viewsCLIP-RN50 · ImageNet variants (OOD Average)Top-1 accuracy: 40.69 -> 43.89Improved
Tune the prompt on a single test image using confidence-selected augmented viewsCLIP-ViT-B/16 · ImageNet variants (OOD Average)Top-1 accuracy: 57.2 -> 60.81Improved
Initialize test-time tuning from a CoOp promptCLIP-RN50 · ImageNet variants (OOD Average)Top-1 accuracy: 42.43 -> 45.75Improved
Tune prompts per test sample rather than use an ImageNet-trained promptCLIP-RN50 · ImageNet to fine-grained datasets (Average)Top-1 accuracy: 57.23 -> 57.66Improved
Tune prompt and class tokens on the support images of each reasoning taskCLIP-RN50 · Bongard-HOI (Average)accuracy: 62.46 -> 66.59Improved
Add confidence selection to entropy-minimizing TPT (ablation)CLIP-RN50 · ImageNet variants (OOD Average)Top-1 accuracy: 42.28 -> 43.89 (+1.61)Improved
Use top-10% confident views rather than a less selective cutoff (ablation)TPT · ImageNet variantsTop-1 accuracyImproved
Optimize the text prompt instead of other CLIP parameter groups (ablation)CLIP test-time optimization · ImageNet variantsTop-1 accuracyImproved
Optimize the visual encoder instead of the text prompt (ablation)CLIP test-time optimization · ImageNet variantsTop-1 accuracyHurt
Increase augmented views from 8 to 64 (ablation)TPT · Natural distribution shiftsAverage accuracy: over 2%Improved
Increase optimization steps from 1 to 2 (ablation)TPT · Natural distribution shiftsAverage accuracy: 0.4%Improved
Increase optimization beyond 2 steps (ablation)TPT · Natural distribution shiftsAverage accuracyNo clear effect

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