by OpenMed
Medical Reasoning SFT Mega is OpenMed's English dataset of about 1.79 million medical question and answer samples with chain-of-thought reasoning. It merges outputs from seven source models.
DatasetFree
Step 1: Install the datasets library.
Step 2: Load it with load_dataset("OpenMed/Medical-Reasoning-SFT-Mega").
Step 3: Read each sample's messages list and check the role of each turn.
Step 4: Use the reasoning_content field on assistant turns for chain-of-thought training, or filter on it.
Step 5: Review outputs with medical experts before relying on a trained model for anything clinical.
It has 1,789,998 samples after deduplication and about 3.78 billion tokens. The card lists 1,237,711 unique questions.
Apache 2.0.
No. The card says the data is generated by AI models, may contain inaccuracies, and is not intended for clinical decision-making without expert review.
Yes. The card lists separate datasets for each of the seven source models.