AI Paper Atlas

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

Generate, Transduct, Adapt: Iterative Transduction with VLMs

Oindrila Saha, Logan Lawrence, Grant Van Horn, Subhransu Maji

ICCV 2025 · Vision-language and multimodal › CLIP and zero-shot · approach: Adapting VLMs for zero-shot classification · 2 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 Adapt (fine-tune CLIP) after dynamic-attribute transduction (ablation)GTA-CLIP (ViT-B/16) · 5-dataset averagetop-1 accuracy: 65.56 -> 67.52Improved
add dynamic (confusion-driven) attributes instead of static ones, with transduct and adapt (ablation)GTA-CLIP (ViT-B/16) · 5-dataset averagetop-1 accuracy: 66.76 -> 67.52Improved
use dynamic instead of static attributes without Adapt (ablation)GTA-CLIP (ViT-B/16) · 5-dataset averagetop-1 accuracy: 65.60 -> 65.56No clear effect
add Transduct to inductive CLIP (ablation)GTA-CLIP (ViT-B/16) · 5-dataset averagetop-1 accuracy: 60.56 -> 64.26Improved
add Adapt to static-attribute transduction (ablation)GTA-CLIP (ViT-B/16) · 5-dataset averagetop-1 accuracy: 65.60 -> 66.76Improved
add static attributes to transduction (ablation)GTA-CLIP (ViT-B/16) · 5-dataset averagetop-1 accuracy: 64.26 -> 65.60Improved
add static attributes to inductive CLIP only (ablation)CLIP (ViT-B/16) · 5-dataset averagetop-1 accuracy: 60.56 -> 61.59Improved
replace TransCLIP-ZS with GTA-CLIPGTA-CLIP · 12-dataset average, zero-shot (ViT-B/16)top-1 accuracy: 69.80 -> 73.81Improved
replace TransCLIP-ZS with GTA-CLIP on UCF101 with ViT-L/14GTA-CLIP · UCF101 (ViT-L/14)top-1 accuracy: 81.73 -> 81.28Hurt
replace TransCLIP-FS with GTA-CLIP at 4 shotsGTA-CLIP · 12-dataset average, 4-shot (ViT-B/16)top-1 accuracy: 75.22 -> 79.08Improved
replace TransCLIP-FS with GTA-CLIP at 1 shotGTA-CLIP · 12-dataset average, 1-shot (ViT-B/16)top-1 accuracy: 72.08 -> 75.49Improved
replace TransCLIP-ZS with GTA-CLIP using seen-class pretrained CLIPGTA-CLIP · Cars unseen classestop-1 accuracy: 78.04 -> 82.17Improved

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