AI Paper Atlas

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

USE: A Unified Self-Ensembling Framework for Test-Time Prompt Tuning

Siru Jiang, Jian Liang, Ran He, Tieniu Tan

ICML 2026 · Label-efficient and robust learning › Test-time adaptation · approach: Test-time adaptation of vision-language models · 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
apply SE ensembling at inference instead of weak view only (ablation)TPT-style RCE tuning · ViT-B/16 avgaccuracy: 63.71 -> 64.77Improved
use SE at inference instead of uniform averaging (ablation)RCE-tuned prompt · ViT-B/16 avgaccuracy: 64.56 -> 64.77Improved
replace marginal entropy with RCE on SE pseudo label (with SE inference) (ablation)TPT · ViT-B/16 avgaccuracy: 64.43 -> 64.77Improved
RCE alone vs marginal entropy with standard inference (ablation)TPT · ViT-B/16 avgaccuracy: 63.60 -> 63.71No clear effect
add SE plug-in to C-TPTC-TPT · ImageNet + OOD (IN&O)accuracy: 61.09 -> 64.27Improved
add SE plug-in to TPTTPT · ViT-B/16 avgaccuracy: 63.60 -> 64.43Improved
add SE plug-in to STSSTS · ViT-B/16 IN&Oaccuracy: 63.89 -> 63.82No clear effect
SE as optimization-free method vs zero-shot CLIPCLIP · EuroSAT ViT-B/16accuracy: 42.05 -> 44.44Improved
ZERO marginalization vs zero-shot CLIPCLIP · EuroSAT ViT-B/16accuracy: 42.05 -> 37.19Hurt
use square-root rescaling of beta instead of linear (ablation)SE · ViT-B/16 avgaccuracy: 64.34 -> 64.19No clear effect
add SE with text prompt ensemblesCLIP · ViT-B/16 avgaccuracy: 62.87 -> 65.61Improved
add skip technique when weak and selected views agree (ablation)USE · efficiencyaccuracyNo 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