Dataset · lingshu-medical-mllm
Medical reasoning with CoT
lingshu-medical-mllm/ReasonMed
370k question-answer pairs with step-by-step CoT rationales and concise summaries for medical tasks.
What's inside
370k high-quality question-answer examples with multi-step chain-of-thought rationales and concise summaries. Examples are distilled from 1.75 million initial reasoning paths generated by three strong LLMs, including Qwen-2.5-72B. Claimed to be the largest open medical reasoning dataset; the language is English.
What you can do with it
- Medical LLM SFT for CoT-based answers
- LoRA training for medical question-answering
- Multi-step medical reasoning benchmark
- Tasks
- Question answering, Text generation
- Languages
- English
- Size
- 1M–10M rows
- Downloads this month
- 642
- Likes
- 96
Models trained on it
- mradermacher/gpt-oss-20b-MedReason-i1-GGUF 4.1K
- mradermacher/Sanad-1.0-i1-GGUF 2K
- mradermacher/gpt-oss-20b-MedReason-GGUF 1.2K
- mradermacher/LlamaMed-3.1-8B-Reasoner-i1-GGUF 680
- Rumiii/gpt-oss-20b-MedReasonText generation 614
- mradermacher/Sanad-1.0-GGUF 421
- mradermacher/LlamaMed-3.1-8B-Reasoner-GGUF 227
- Rumiii/LlamaMed-3.1-8B-ReasonerText generation 30
How to load it
from datasets import load_dataset
ds = load_dataset("lingshu-medical-mllm/ReasonMed")