AI Paper Atlas

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

MED-LCDS: Multi-Expert-Domain CLIP Classification via Logit Calibration

Zheng Zeng, Deepak Sridhar, Nuno Vasconcelos

ECCV 2026 · Vision-language and multimodal › CLIP and zero-shot · approach: Attribute-based zero-shot learning

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 domain supervision and logit scaling to MoLEMoLE · Cross-domain, base splitmean accuracy (%): 71.80 -> 85.51Improved
add domain supervision to softmax MoE gating (ablation)MoLE · Cross-domain, all splitmean accuracy (%): 65.85 -> 71.61Improved
add domain-wise logit scaling on top of domain supervision (ablation)MED-DS · Cross-domain, all splitmean accuracy (%): 71.61 -> 71.80Improved
add domain-wise logit scaling on top of domain supervision (ablation)MED-DS · Cross-domain, base splitmean accuracy (%): 85.43 -> 85.51No clear effect
use softmax instead of sigmoid gating (ablation)MED-DSLC · Cross-domain, all splitmean accuracy (%): 69.97 -> 71.80Improved
add logit scaling under extreme target-domain imbalance (k=2) (ablation)MED-DS · target-domain imbalance sweeptarget-domain accuracy: +2.2 pointsImproved
add logit scaling under target-domain imbalance (k=4) (ablation)MED-DS · target-domain imbalance sweeptarget-domain accuracy: +1.3 pointsImproved
one-shot training of gating and temperatures only (ablation)MED-DSLC · Cross-domain, all splitmean accuracy (%): 71.80 -> 71.71No clear effect
use SigLIP backbone instead of CLIP (ablation)MED-DSLC · Cross-domain, all splitmean accuracy (%): 71.80 -> 62.27Hurt
increase number of classes per domain from k=5 to 40 (First-k) (ablation)MED-DSLC · All split First-kmean accuracy (%): 91.37 -> 79.39Hurt
add MED-DSLC over MoLE with more classes (First-k, k=40) (ablation)MoLE · All split First-kmean accuracy (%): 66.99 -> 79.39Improved
use Single-LoRA with 10x parameters instead of expert LoRAsSingle-LoRA · Cross-domain, base splitmean accuracy (%): 66.22 -> 42.35Hurt

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