arXiv:2511.10930v1 Announce Type: new 
Abstract: Biomedical text embeddings have primarily been developed using research literature from PubMed, yet clinical cardiology practice relies heavily on procedural knowledge and specialized terminology found in comprehensive textbooks rather than research abstracts. This research practice gap limits the effectiveness of existing embedding models for clinical applications incardiology. This study trained CardioEmbed, a domain-specialized embedding model based on Qwen3-Embedding-8B, using contrastive learning on a curated corpus of seven comprehensive cardiology textbooks totaling approximately 150,000 sentences after deduplication. The model employs InfoNCE loss with in-batch negatives and achieves 99.60% retrieval accuracy on cardiac-specific semantic retrieval tasks, a +15.94 percentage point improvement over MedTE, the current state-of-the-art medical embedding model. On MTEB medical benchmarks, the model obtained BIOSSES 0.77 Spearman and SciFact 0.61 NDCG@10, indicating competitive performance on related biomedical domains. Domain-specialized training on comprehensive clinical textbooks yields near-perfect cardiology retrieval (99.60% Acc@1), improving over MedTE by +15.94 percentage points.

تقدم الدراسة CardioEmbed، وهو نموذج متخصص في تضمين النصوص لطب القلب السريري. على عكس النماذج الحالية التي تعتمد على الأدبيات البحثية من PubMed، يتم تدريب CardioEmbed على مجموعة مختارة من سبعة كتب شاملة في cardiology، totaling approximately 150,000 sentences. هذا النهج يعالج الفجوة في تضمين النصوص الطبية الحيوية التي غالبًا ما تتجاهل المعرفة الإجرائية والمصطلحات المتخصصة الضرورية للممارسة السريرية. يحقق النموذج دقة استرجاع تبلغ 99.60% في المهام المحددة للقلب، متفوقًا بشكل كبير على النموذج الحالي الرائد، MedTE.

El estudio presenta CardioEmbed, un modelo de incrustación de texto especializado en cardiología clínica. A diferencia de los modelos existentes que se basan en la literatura de investigación de PubMed, CardioEmbed se entrena con un corpus curado de siete libros de texto de cardiología, que totalizan aproximadamente 150,000 oraciones. Este enfoque aborda la brecha en las incrustaciones de texto biomédico que a menudo pasan por alto el conocimiento procedimental y la terminología especializada esenciales para la práctica clínica. El modelo logra una precisión de recuperación del 99.60% en tarea…

L'étude présente CardioEmbed, un modèle d'embedding de texte spécialisé pour la cardiologie clinique. Contrairement aux modèles existants qui s'appuient sur la littérature de recherche de PubMed, CardioEmbed est formé sur un corpus sélectionné de sept manuels de cardiologie, totalisant environ 150 000 phrases. Cette approche comble le fossé dans les embeddings de texte biomédicaux qui négligent souvent les connaissances procédurales et la terminologie spécialisée essentielles à la pratique clinique. Le modèle atteint une précision de récupération de 99,60 % sur des tâches spécifiques à la card…

The study introduces CardioEmbed, a domain-specialized text embedding model designed for clinical cardiology. Unlike existing models that rely on research literature from PubMed, CardioEmbed is trained on a curated corpus of seven comprehensive cardiology textbooks, totaling around 150,000 sentences. This approach addresses the gap in biomedical text embeddings that often overlook procedural knowledge and specialized terminology essential for clinical practice. The model achieves a retrieval accuracy of 99.60% on cardiac-specific tasks, significantly outperforming the current state-of-the-art …

CardioEmbed: Domain-Specialized Text Embeddings for Clinical Cardiology

One More Thing in AI – Your Shortcut to AI Mastery

CardioEmbed: Domain-Specialized Text Embeddings for Clinical Cardiology

Was this article worth reading? Share it

One More Thing in AI

Vero Scribe

ClassX

IntelliQ

Vocablii

Meteoria

Ready to build your own newsroom?