AI Paper Atlas

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

Blind to Position, Biased in Language: Probing Mid-Layer Representational Bias in Vision-Language Encoders for Zero-Shot Language-Grounded Spatial Understanding

Na Min An, Inha Kang, Minhyun Lee, Hyunjung Shim

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
add B2G mid-layer P-Map rerankingGlobal-Local (Mask2Former) · Zero-shot RIS avg, DFN ViT-H/14mIoU: 44.64 -> 51.13Improved
add B2G mid-layer P-Map rerankingHybridGL (SAM) · Zero-shot RIS avg, CLIP ViT-B/16mIoU: 45.18 -> 46.24Improved
add B2G mid-layer P-Map rerankingGlobal-Local (Mask2Former) · Zero-shot RIS avg, SigLIP ViT-B/16mIoU: 46.62 -> 47.70Improved
choose visual layer dynamically (perm or cluster) instead of fixed mid-layer (ablation)B2G · Zero-shot RIS, CLIP ViT-B/16 Mask2FormermIoU: 50.20 vs 49.89 vs 50.13No clear effect
add B2G spatial guidanceGlobal-Local+CT+SG · RefCOCOg spatial subsetmIoU: +20.83% mIoUImproved
add B2G spatial guidanceGlobal-Local+CT+SG · RefCOCOg non-spatial subsetmIoU: +15.55% mIoUImproved
add mid-layer multilingual centroid injection instead of final-layer centroidSigLIP2 · Multilingual zero-shot RIS (9 benchmarks)mIoU: 42.42 -> 46.07 -> 49.62Improved
add mid-layer multilingual centroid injectionSigLIP2 · Multilingual zero-shot RIS (9 benchmarks)IoU@50: 45.13 -> 48.67 -> 53.41Improved
add B2G mid-layer centroidSigLIP2 · Multilingual retrieval, RefCOCO avgRecall@1: 46.66 -> 50.88Improved
add B2G mid-layer centroidSigLIP2 · Multilingual retrieval, RefCOCOg GermanRecall@1: 12.85 -> 26.40Improved
add negation of patch embeddings (ablation)B2G · RefCOCOg mIoUmIoU: 49.87 -> 54.92Improved
add clustering of P-Map (ablation)B2G · RefCOCOg mIoUmIoU: 50.36 -> 54.92Improved

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