AI Paper Atlas

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

A-TPT: Angular Diversity Calibration Properties for Test-Time Prompt Tuning of Vision-Language Models

Shihab Aaqil Ahamed, Udaya Sampath K. Perera Miriya Thanthrige, Ranga Rodrigo, Muhammad Haris Khan

ICLR 2026 · Vision-language and multimodal › CLIP and zero-shot · approach: Parameter-efficient visual tuning

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 angular diversity regularizer to TPTTPT, CLIP ViT-B/16 · Fine-grained datasets (avg)ECE: 11.60 -> 2.61Improved
add angular diversity regularizer to TPTTPT, CLIP RN50 · Fine-grained datasets (avg)ECE: 11.70 -> 2.92Improved
replace orthogonality loss (O-TPT) with angular diversityTPT, CLIP ViT-B/16 · Natural distribution shiftsECE: 4.88 -> 3.92Improved
replace orthogonality loss with angular diversityTPT, CLIP RN50 · Natural distribution shiftsECE: 9.69 -> 7.82Improved
A-TPT versus zero-shot CLIP on distribution shiftsCLIP RN50 · Natural distribution shiftsECE: 7.46 -> 7.82No clear effect
add angular diversity to TPTTPT, CLIP ViT-B/16 · Fine-grained datasets (avg)accuracy: 65.00 -> 64.92No clear effect
add angular diversity in N>|D| classes regimeTPT, CLIP ViT-B/16 · N>|D| groupECE: 4.27 -> 2.92Improved
combine A-TPT with supervised CoOp promptsTPT + CoOp · Fine-grained avgECE: 18.36 -> 3.63Improved
combine A-TPT with CoCoOp promptsTPT + CoCoOp · Fine-grained avgECE: 4.35 -> 3.22Improved
add A-TPT to medical prompt tuning (BAPLe)MedCLIP · Covid CAECE: 15.64 -> 2.34Improved
add A-TPT to FPT on ISIC 2018FPT · ISIC 2018ECE: 0.2328 -> 0.0794Improved
use angular diversity in place of cosine orthogonality: compute costTPT · ImageNet-V2time per batch: 1.064 -> 1.058No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 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