AI Paper Atlas

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

Prototype-Based Test-Time Adaptation of Vision-Language Models

Zhaohong Huang, Yuxin Zhang, Wenjing Liu, Fei Chao, Rongrong Ji

ICML 2026 · Vision-language and multimodal › CLIP and zero-shot · approach: Test-time adaptation of VLMs · 0 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 adaptive decay rate beta instead of raw confidence scores for prototype updates (ablation)PTA · ImageNetaccuracy (%): 0.10 -> 70.28Improved
use adaptive decay rate beta instead of raw confidence scores (ablation)PTA · EuroSATaccuracy (%): 11.11 -> 61.57Improved
add the interpolation weight w (0 -> 0.01) between visual and text-anchored prototypes (ablation)PTA · ImageNetaccuracy (%): 0.10 -> 70.28Improved
increase w from 0.01 to 0.1 (ablation)PTA · ImageNetaccuracy (%): 70.28 -> 68.00Hurt
increase w from 0.0001 to 0.001 to 0.01 (ablation)PTA · Cross-Domainaccuracy (%): 69.35 -> 69.35 -> 69.38No clear effect
add PTA to GS-BiasGS-Bias · Cross-dataset (11 incl. ImageNet)accuracy (%): 67.35 -> 69.61Improved
add PTA to MTAMTA · Cross-dataset (11 incl. ImageNet)accuracy (%): 66.33 -> 69.33Improved
add PTA to TDATDA · Cross-dataset (11 incl. ImageNet)accuracy (%): 68.07 -> 69.58Improved
add PTA to TDA on EuroSATTDA · EuroSATaccuracy (%): 62.51 -> 61.88Hurt
place confidently misclassified samples first in the stream (poison size All vs Normal) (ablation)PTA · ImageNetaccuracy (%): 70.3 -> 70.3No clear effect
use PTA instead of TDA under low non-iid streams (gamma=0.1)TDA · non-iid average, 11 datasetsaccuracy (%): 67.7 -> 67.8No clear effect
use PTA instead of TDA for speedTDA · ImageNet efficiencyFPS: 44.6 -> 81.8Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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