disi-unibo-nlp/medgenie

The First Generate-then-Read Framework for Multiple-Choice Question Answering in Medicine

17
/ 100
Experimental

Generated contexts enhance medical QA through domain-specific language models (PMC-LLaMA-13B) that create relevant passages before answer selection, eliminating dependency on retrieval. The framework offers two reader variants—Fusion-in-Decoder (FID) for supervised learning with 250M parameters and in-context learning with LLaMA-2/Zephyr-7B—operating efficiently within 24GB VRAM on MedQA-USMLE, MedMCQA, and MMLU medical subsets. Multi-view context generation combines question-focused and option-focused perspectives to improve answer accuracy.

No commits in the last 6 months.

No License Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 5 / 25
Maturity 1 / 25
Community 11 / 25

How are scores calculated?

Stars

13

Forks

2

Language

Python

License

Last pushed

May 27, 2024

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/llm-tools/disi-unibo-nlp/medgenie"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.