AI Paper Atlas

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

Uncertainty-guided Compositional Alignment with Part-to-Whole Semantic Representativeness in Hyperbolic Vision-Language Models

Hayeon Kim, Ji Ha Jang, Junghun James Kim, Se Young Chun

CVPR 2026 · Vision-language and multimodal › CLIP and zero-shot · approach: Open-vocabulary alignment from 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
add uncertainty calibration in the entailment loss (ablation)UNCHA · zero-shot classification (general datasets), ViT-S/16accuracy: 64.57 -> 68.98Improved
add uncertainty-aware temperature scaling in global-local contrastive loss (ablation)UNCHA · zero-shot classification (general datasets), ViT-S/16accuracy: 65.14 -> 68.98Improved
add entropy regularizer to uncertainty calibration (ablation)UNCHA · zero-shot classification (general datasets), ViT-S/16accuracy: 65.61 -> 68.98Improved
add uncertainty calibration in the entailment loss (ablation)UNCHA · image retrieval (Table 4), ViT-S/16retrieval score: 69.68 -> 73.90Improved
add entropy regularizer to uncertainty calibration (ablation)UNCHA · misc. datasets classification, ViT-S/16accuracy: 24.78 -> 27.55Improved
uncertainty-guided part-to-whole alignment vs HyCoCLIPhyperbolic VLM (ViT-B/16) · ImageNet zero-shottop-1 accuracy: 45.8 -> 48.8Improved
uncertainty-guided part-to-whole alignment vs HyCoCLIPhyperbolic VLM (ViT-B/16) · fine-grained datasets (CUB, Cars, Aircraft, Pets, Flowers, DTD)top-1 accuracy: CUB 16.4 -> 14.8, Cars 11.6 -> 14.0Mixed
uncertainty-guided part-to-whole alignment vs HyCoCLIPhyperbolic VLM (ViT-S/16) · EuroSAT zero-shottop-1 accuracy: 38.3 -> 36.2Hurt
uncertainty-guided alignment vs HyCoCLIP on retrievalhyperbolic VLM (ViT-B/16) · Flickr text retrieval R@1R@1: 92.6 -> 91.4Hurt
uncertainty-guided alignment vs HyCoCLIP on hierarchy metricshyperbolic VLM (ViT-B/16) · ImageNet hierarchical classificationTIE: 3.17 -> 2.94Improved
uncertainty-guided alignment vs HyCoCLIP on multi-object sceneshyperbolic VLM (ViT-B/16) · ComCo 5 objectsmAP: 72.90 -> 81.18Improved
uncertainty-guided alignment vs HyCoCLIP on hard negativeshyperbolic VLM (ViT-B/16) · Densely Captioned Images, All Hard Negsaccuracy: 53.33 -> 56.51Improved

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