AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

PhenoLIP: Phenotype Guided Medical Vision–Language Pretraining

Cheng Liang, Chaoyi Wu, Weike Zhao, Ya Zhang, Yanfeng Wang, Weidi Xie

ECCV 2026 · Vision-language and multimodal › Human and embodied modeling · approach: Medical vision-language models · 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 knowledge distillation and data curation to CLIP+PubmedBert (ablation)PhenoLIP · PhenoBench I2TR@10: 49.53 -> 60.13Improved
initialize vision encoder from BiomedCLIP instead of CLIP (ablation)PhenoLIP · PhenoBench I2TR@10: 60.13 -> 63.30Improved
initialize vision encoder from BiomedCLIP instead of scratch/CLIP (no distill) (ablation)PhenoLIP · PhenoBench I2TR@10: 50.11 -> 53.42 (PMB text)Improved
use PubmedBert text init instead of the pretrained knowledge encoder, no distillation (ablation)PhenoLIP · PhenoBench I2TR@10: 50.11 -> 53.42Improved
add the knowledge distillation loss (ablation)PhenoLIP · PhenoBench retrievalR@10: not isolated in Table 5Improved
pretrain with PhenoKG phenotype knowledge vs BIOMEDICAPhenoLIP · PhenoBench zero-shot classificationaccuracy: 28.50 -> 36.56 (average)Improved
pretrain with PhenoKG on TissueMNISTPhenoLIP · Zero-shot Histologyaccuracy: 7.87 vs 5.68 PMC-CLIPNo clear effect
PhenoLIP vs BIOMEDICAPhenoLIP · PhenoBench I2P retrievalR@10: 8.12 -> 13.84Improved
PhenoLIP vs BiomedCLIPPhenoLIP · Face2Gene retrievalR@50: 49.84 -> 55.05Improved
PhenoLIP vs BiomedCLIPPhenoLIP · Linear probing 100% (avg)accuracy: 71.56 -> 73.76Improved

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