AI Paper Atlas

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

Condensed Test-Time Adaptation of VLMs for Action Recognition

Wenxuan Ge, Hongyu Qu, Rui Yan, Guo-Sen Xie, Yazhou Yao, Xiangbo Shu, Jinhui Tang

CVPR 2026 · Vision-language and multimodal › CLIP and zero-shot · approach: Test-time adaptation of VLMs

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 PSPS (probabilistic semantic patch selection) (ablation)CONDA / ViCLIP · Kinetics-600top-1 accuracy: 70.9 -> 71.2Improved
add ATC (adaptive tube construction) (ablation)CONDA / ViCLIP · COINtop-1 accuracy: 65.6 -> 66.4Improved
combine PSPS and ATC (ablation)CONDA / ViCLIP · COINtop-1 accuracy: 65.6 -> 67.4Improved
use probabilistic selection instead of top-k hard selection (ablation)CONDA · Kinetics-600top-1 accuracy: 71.6 -> 72.1Improved
use both gradients and attention maps in semantic patch activation (ablation)CONDA · COINtop-1 accuracy: 66.7 -> 67.4Improved
add spatial expansion in tubes (ablation)CONDA · COINtop-1 accuracy: 66.3 -> 67.4Improved
use random temporal expansion (ablation)CONDA · Kinetics-600top-1 accuracy: 71.6 -> 71.5No clear effect
use adaptive instead of fixed temporal expansion (ablation)CONDA · COINtop-1 accuracy: 66.9 -> 67.4Improved
vary number of selected patches k from 5 to 20 (ablation)CONDA · Kinetics-600top-1 accuracy: 71.8 -> 72.1No clear effect
add CONDA to TDA-style cache on ViCLIPViCLIP · Kinetics-600top-1 accuracy: 70.9 -> 72.1Improved
add CONDA to vanilla CLIP ViT-L/14CLIP ViT-L/14 · Kinetics-600top-1 accuracy: 76.4 -> 76.9Improved
use CONDA instead of DPE on egocentric EGTEAEgoVLP · EGTEAtop-1 accuracy: 19.7 -> 22.7Improved

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