Hugging Face Overview
Fleming-R1-7B: Medical Reasoning Model
Fleming-R1-7B is a specialized 7 billion parameter model developed by UbiquantAI, based on the Qwen2.5-7B architecture, engineered for advanced medical reasoning. It is designed to analyze complex medical problems step-by-step and provide reliable answers.
Key Capabilities
- Medical Reasoning: Excels in performing detailed analysis for medical scenarios, achieving state-of-the-art results on various medical benchmarks for its size class.
- Advanced Training: Utilizes a "chain-of-thought cold start" approach, distilling high-quality reasoning traces from teacher models, combined with two-stage reinforcement learning that includes adaptive hard-negative mining to strengthen problem-solving.
- Data Strategy: Incorporates public medical datasets with knowledge graphs to enhance coverage of rare diseases, medications, and multi-hop reasoning chains.
Good For
- Medical Research: Analyzing complex medical cases and generating reasoning traces for non-clinical reference.
- Educational Tools: Developing applications that require step-by-step medical problem-solving and explanation.
- Benchmarking: Evaluating medical reasoning capabilities against other models, particularly in Chinese medical tasks where its larger counterpart (Fleming-R1-32B) shows strong performance.