SciFive HyDE Generator (Cardiology Fine Tuned)
Model Description
This model is a fine-tuned version of razent/SciFive-base-Pubmed_PMC for Hypothetical Document Generation (HyDE) in biomedical Retrieval-Augmented Generation (RAG) systems.
The model generates domain-specific hypothetical documents from user queries to improve dense retrieval performance and reduce hallucinations.
Training Data
- Dataset: MIRIAD medical QA dataset
- Specialties:
- Cardiology
- Cardiothoracic Surgery
- Vascular Surgery
- ~15,000 question–answer pairs
Training Objective
Sequence-to-sequence generation:
