AI Paper Atlas

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

Batch-Conditioned Semantic Anchors for Robust Transductive Adaptation of Vision--Language Models

Mohammed Rahman Sherif Khan Mohammad, Ardhendu Behera, Sandip Pradhan, Swagat Kumar, Amr Ahmed

NeurIPS 2026 · Vision-language and multimodal › CLIP and zero-shot · approach: Prompt learning for CLIP · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes batch-conditioned semantic anchors for transductive test-time adaptation of CLIP-style vision-language models, balancing batch exploitation against preserving text semantics.

Problem

Transductive adaptation of vision-language models on unlabeled test batches faces a trade-off: aggressive methods drift under sparse batches, prior shift, or limited label-space coverage, while conservative methods keep CLIP text semantics but under-adapt.

Approach

Makes the semantic anchors used to refine predictions conditioned on the unlabeled test batch rather than fixed text embeddings, so adaptation can use batch structure while staying anchored to CLIP text semantics.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use batch-aware transported anchor instead of fixed CLIP text anchor (ablation)JPTA · Ablation mean (Aircraft, EuroSAT, Flowers)top-1 accuracy (%): 50.23 -> 60.90Improved
blend transported anchor with original text anchor rather than transported-only (ablation)JPTA · Ablation mean (Aircraft, EuroSAT, Flowers)top-1 accuracy (%): 57.26 -> 60.90Improved
add batch-marginal prior-aware refinement (ablation)JPTA · Ablation mean (Aircraft, EuroSAT, Flowers)top-1 accuracy (%): 54.66 -> 60.90Improved
add regime-adaptive trust control (ablation)JPTA · Ablation mean (Aircraft, EuroSAT, Flowers)top-1 accuracy (%): 54.50 -> 60.90Improved
use adaptive pseudocounts for anchored statistics instead of fixed schedule (ablation)JPTA · Ablation mean (Aircraft, EuroSAT, Flowers)top-1 accuracy (%): 57.79 -> 60.90Improved
free transduction (TransCLIP) on sparse low-coverage batchesTransCLIP · B=64 Very Low, 11-dataset avgtop-1 accuracy (%): 65.2 (CLIP) -> 38.9Hurt
replace fixed anchoring (StatA) with JPTA transported anchor in sparse batchesStatA · B=64 Very Low, 11-dataset avgtop-1 accuracy (%): 70.4 -> 82.4Improved
replace StatA with JPTA in large high-coverage batchesStatA · B=1000 High (25-50), per datasettop-1 accuracy (%): 69.7 -> 73.1 avgMixed
replace StatA with JPTA in all-class evaluationStatA · All-class, 11-dataset avgtop-1 accuracy (%): 69.9 -> 73.3Improved
replace StatA with JPTA in online correlated streamsStatA · Online stream Separatetop-1 accuracy (%): 69.1 -> 75.1Improved
transfer JPTA config unchanged to SigLIP2 backboneJPTA · SigLIP2 So400m/14-384, 11 datasets x 3 regimesconditions improved over zero-shot: 29/33 improvedMixed
random hyperparameter perturbations of regularization weights (0.5x-2x) (ablation)JPTA · Very Low, High, All-Class macromacro accuracy difference (pp): worst impact -1.33Hurt

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