This is not a medical Q&A tool. It searches a fixed sample of 2,000 PubMed abstracts (7,627 passages), shows the closest passages, and does not generate answers or give medical advice. Type a biomedical question to see how two models rank the passages:
BGE-base (no fine-tuning) and BGE-base fine-tuned on 1,300 synthetic PubMed query–abstract pairs.
Everything runs in your browser. Passages found by both models are highlighted.
Research and code: GitHub.
Example questions the sample is known to cover:
Loading…
BGE-base (no fine-tuning)
rafiazarin/bge-base-en-v1.5-onnx (8-bit export of BAAI/bge-base-en-v1.5)