AI Paper Atlas

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

Learning Visual Proxy for Compositional Zero-Shot Learning

Shiyu Zhang, Cheng Yan, Yang Liu, Chenchen Jing, Lei Zhou, Wenjun Wang

ICCV 2025 · Vision-language and multimodal › CLIP and zero-shot · approach: Attribute-based zero-shot learning · 3 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 visual proxy alongside text prototype (ablation)VP-CMJL · UT-Zappos closed-worldAUC: 37.8 -> 47.9Improved
add visual proxy alongside text prototype (ablation)VP-CMJL · MIT-States closed-worldAUC: 20.8 -> 23.3Improved
add text prototype alongside visual proxy (ablation)VP-CMJL · UT-Zappos closed-worldAUC: 42.1 -> 47.9Improved
add text prototype alongside visual proxy (ablation)VP-CMJL · MIT-States closed-worldAUC: 20.7 -> 23.3Improved
keep visual proxy branch at inference (ablation)VP-CMJL · UT-Zappos closed-worldAUC: 44.2 -> 47.9Improved
keep text prototype branch at inference (ablation)VP-CMJL · MIT-States closed-worldAUC: 20.4 -> 23.3Improved
keep visual proxy branch at inference (ablation)VP-CMJL · MIT-States closed-worldAUC: 23.0 -> 23.3No clear effect
use CA for text prototype and MLP for visual proxy decoupling instead of CA/CA (ablation)VP-CMJL · UT-Zappos closed-worldAUC: 42.0 -> 47.9Improved
use CA/MLP instead of MLP/MLP decoupling (ablation)VP-CMJL · MIT-States closed-worldAUC: 21.8 -> 23.3Improved
use CA/MLP instead of MLP/CA decoupling (ablation)VP-CMJL · UT-Zappos closed-worldAUC: 44.2 -> 47.9Improved
use VP-CMJL instead of TroikaVP-CMJL · C-GQA closed-worldAUC: 12.4 -> 16.3Improved
use VP-CMJL instead of CDS-CZSLVP-CMJL · MIT-States open-worldAUC: 8.5 -> 8.3Hurt

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