Chinese Medical Deepseek R1 Distill 20K

A Chinese-language dataset of 20,000 synthetic medical consultation dialogues generated by the DeepSeek-R1 model. The dataset, created by JackGao and last updated on April 15, 2025, includes patient questions, AI-synthesized responses, and department labels such as neurology and cardiology.

Use Cases

• Training department classification models based on the 'label' field for patient queries.
• Fine-tuning medical question-answering models using the 'question' and 'synthesized_text' fields.
• Analyzing the characteristics of AI-generated medical text versus real doctor replies ('medical_replay').
• Benchmarking model performance on Chinese medical dialogue tasks across different specialties.

Strengths

• Contains 20,000 entries, providing a substantial volume of synthetic data.
• Includes department labels for specialties like neurology and cardiology, enabling supervised classification tasks.
• Provides both patient questions and AI-generated responses, creating a full dialogue structure.

Limitations

• The AI-generated responses are from a DeepSeek-R1 FP8 model and have not undergone strict factual verification.
• Medical queries and department classifications lack fine-grained governance and may contain inaccuracies.
• The description notes that some doctor replies were obtained but not validated for authenticity.

Provenance

Source

huggingface

Collection Method

Synthesized by the DeepSeek-R1 model, with some doctor replies obtained but not validated.

Time Range

null

Freshness

Last updated 2025-04 15 10:49:38.

Geography

null