medical-qa

Multimodal Question Answering in the Medical Domain: A summary of Existing Datasets and Systems

A large-scale (194k), Multiple-Choice Question Answering (MCQA) dataset designed to address realworld medical entrance exam questions.

[EMNLP'24] EHRAgent: Code Empowers Large Language Models for Complex Tabular Reasoning on Electronic Health Records

KGARevion: AI Agent for Knowledge-Intensive Biomedical QA

🧠💊 MediMind: Medical Question Answering System powered by BioBERT + Flan-T5. Intelligent routing for 8 question types, confidence-based abstention, 200k+ medical Q&A pairs. React frontend + FastAPI backend.

Supervised fine-tuning (SFT) of Qwen3 for structured medical reasoning QA — teaching models to "think before answering" (<think>...</think>). Supports full fine-tuning & LoRA, with an end-to-end data→train→eval→compare pipeline and quantitative benchmarks (PPL, format compliance, semantic similarity, latency/throughput).

Medical-QA hallucination benchmark for 3 live LLMs via Groq (Llama 3.1 8B, Llama 3.3 70B, GPT-OSS 120B) — dual keyword + LLM-as-judge scoring, Wilson confidence intervals, phrasing-ambiguity axis. CoT reached 87.5% vs. 85.0% zero-shot accuracy (n=40, overlapping CIs).

This project presents a medical question–answering language model built by fine-tuning Google Gemma-2-2B-IT using LoRA (Low-Rank Adaptation) 🧠⚕️. The primary objective is to adapt a general-purpose large language model to the healthcare domain in a parameter-efficient, reproducible, and resource-aware manner.

Reliable medical QA with Mistral-7B, QLoRA, selective prediction, and learned abstention via warm-start SFT + DPO.

Explainable Healthcare QA Chatbot with LLM + RAG + XAI — grounded medical answers with source citations, built on FastAPI, LangChain/LangGraph, and a Next.js frontend.

Fine-tuned LLaMA 3 medical Q&A chatbot with prescription OCR analysis, served locally via Ollama + Streamlit

Paninian rule-augmented RAG for USMLE-style medical QA on MedQA. FAISS textbook retrieval + explainable clinical rule traces.

Medical QA agent mini-repro: four-way recall (semantic/keyword/KG multi-hop/rule) with explainable weighted fusion (+21pp audit accuracy) — 四路缺一不可，KG 反向两跳

Fine-tuning Qwen2.5-1.5B on MedQuAD using QLoRA for medical question answering. Two-phase training: conversational and structured output.

Fine-tuning Llama 3.3 70B (QLoRA) + FAISS RAG on USMLE exams — 80.99% on MedQA-US, within 3 points of GPT-4, using only open-source models.

Fine-tuning BERT for medical question answering (CPGQA/SQuAD) — Swinburne undergraduate research.

Instruction-tuned LLaMA 2 chatbot fine-tuned with LoRA on real medical Q&A data. Built for conversational health-related queries using Transformers and PEFT.

This project involves fine-tuning the T5 transformer model for medical question-answering tasks. The model is trained on a domain-specific dataset, enabling it to generate accurate and contextually relevant medical responses.

MedRAFT: A Chinese medical question answering system using Retrieval-Augmented Generation (RAG) and Retrieval-Augmented Fine-Tuning (RAFT) with knowledge distillation for efficient deployment.

Auditable medical on-policy distillation and capability restoration for Qwen 4B