AI Paper Atlas

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

Controlling Embedding Spaces with Text-Conditioned Transformations

Joseph Fioresi, Fabian Caba, Pankaj Nathani, Shah Mubarak, Kushal Kafle

ECCV 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
use affine transform instead of translation-only (ablation)text-conditioned hypernetwork · Clevr-4 query retrievalmAP: 35.4 -> 77.6Improved
use affine transform instead of translation-only (ablation)text-conditioned hypernetwork · Stanford-40 query retrievalmAP: 43.8 -> 84.4Improved
use affine transform instead of translation-only (ablation)text-conditioned hypernetwork · Stanford-40 cluster accuracyaccuracy: 55.6 -> 79.4Improved
add multimodal guidance loss (ablation)hypernetwork · text-based classification ShotBenchaccuracy: 31.3 -> 51.6Improved
add multimodal guidance loss (ablation)hypernetwork · Stanford40 attribute-aware retrievalmAP: 87.3 -> 91.2Improved
add multimodal guidance loss (ablation)hypernetwork · ShotBench multi-clusteringclustering accuracy: 53.2 -> 51.6Hurt
blend action and location projected queries (alpha=0.5)query transformation · Stanford40 joint-label retrievalmAP (Both): 70.0 -> 80.9Improved
train with one color held out and test on ithypernetwork · Clevr4 color retrievalmAP: 22.9 -> 58.0Improved
raise multimodal loss weight from 0.1 to 1 (ablation)hypernetwork · Clevr-4 retrieval and clusteringmAP / clustering accuracy: retrieval 73.4 -> 73.0, clustering 79.6 -> 81.5Mixed
remove multimodal loss (omega 0) versus 0.1 (ablation)hypernetwork · Clevr-4 retrievalmAP: 72.3 -> 73.4Improved
apply color-aware query transform on top of SEARLE-XL rerankingSEARLE-XL · FashionIQ averageR@10: 23.5 -> 24.6Improved
apply learned transformation to text embeddings with sentiment conditioningE5-large-v2 · SST-5zero-shot accuracy: 29.6 -> 54.3Improved

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