AI Paper Atlas

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

Steerable Vision Transformers

Jona Ruthardt, Manu Gaur, Deva Ramanan, Makarand Tapaswi, Yuki Asano

ECCV 2026 · Vision-language and multimodal › CLIP and zero-shot · approach: Recaptioning and text-supervised pretraining

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
early fusion (gated cross-attention inside ViT) instead of late fusion (ablation)SteerViT · PODSPR-AUC: 36.6 -> 58.1Improved
early fusion instead of late fusion (ablation)SteerViT · COREacc@1: 93.3 -> 96.0Improved
early fusion instead of late fusion (ablation)SteerViT · FG-CLSaccuracy: 91.8 -> 87.7Hurt
add zero-initialized tanh gate (ablation)SteerViT · PODSPR-AUC: 47.1 -> 58.1Improved
add zero-initialized tanh gate (ablation)SteerViT · FG-CLSaccuracy: 83.5 -> 87.7Improved
add zero-initialized tanh gate (ablation)SteerViT · COREacc@1: 94.6 -> 96.0Improved
use two-layer MLP projector instead of linear (ablation)SteerViT · PODSPR-AUC: 56.4 -> 58.1Improved
use two-layer MLP projector instead of linear (ablation)SteerViT · FG-CLSaccuracy: 86.7 -> 87.7Improved
apply early fusion to SigLIP backbone vs late fusion (ablation)SigLIP · COREacc@1: 75.4 -> 91.3Improved
apply early fusion to MAE backbone vs late fusion (ablation)MAE · COREacc@1: 41.0 -> 74.9Improved
use detailed instance descriptions instead of supercategory prompts (ablation)SteerViT · PODS classificationPR-AUC: 27.9 -> 58.1Improved
scale cross-attention gates by 0.6 at inference (ablation)SteerViT (MAE) · representation qualityscore: 40 -> 50Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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