BioGPT — Biomedical Text Generation & Mining

BioGPT — Biomedical Text Generation & Mining

BioGPT is Microsoft Research's generative pre-trained transformer trained exclusively on 15 million PubMed abstracts. It achieves state-of-the-art results on biomedical relation extraction, question answering (BioASQ, PubMedQA), and biomedical entity recognition — and can generate factually grounded biomedical text for report summarisation and literature synthesis.

Key Features

• Pre-trained on 15M PubMed abstracts (domain-specific vocabulary)
• State-of-the-art on PubMedQA, BC5CDR (NER), DDI (drug-drug interaction extraction)
• Fine-tuning recipes included for downstream biomedical tasks
• HuggingFace Transformers-compatible — drop-in for pipeline()
• Supports text generation, masked prediction, and feature extraction

Quick Start

from transformers import pipeline

bio_gen = pipeline("text-generation", model="microsoft/biogpt")
result = bio_gen("COVID-19 is caused by", max_new_tokens=50, do_sample=False)
print(result[0]["generated_text"])

npx ai-supply add biogpt-biomedical-generative-model

Curated mirror of the open-source BioGPT (MIT). Get it from the source.