This dataset contains 66 carefully selected text-only samples from the IgakuQA medical exam dataset, curated using advanced difficulty assessment and model evaluation techniques.

Dataset Description

This is a high-quality subset of the IgakuQA Japanese medical exam dataset, selected based on:

Difficulty Score: Measures how challenging the sample is for AI models Consistency Score: Evaluates response consistency across different models… See the full description on the dataset page: https://huggingface.co/datasets/japan-ai-official/igakuqa-subset-curated.