unicornftk/Doctor-R1
Doctor-R1 by unicornftk is an 8B parameter AI doctor agent, built on Qwen3-8B, designed for strategic, multi-turn patient inquiries and diagnostic decision-making. It is fine-tuned using Experiential Agentic Reinforcement Learning to unify communication and medical decision-making skills. This model excels in dynamic clinical consultation benchmarks like HealthBench and MAQuE, outperforming larger open-source models.
Loading preview...
Doctor-R1: AI for Clinical Inquiry
Doctor-R1 is an 8B parameter AI agent, based on Qwen3-8B, specifically engineered to simulate the complete, dynamic consultation process of a human physician. It integrates both strategic patient inquiry and accurate medical decision-making within a single framework, a significant departure from traditional static medical QA models.
Key Capabilities & Innovations
- Unified Clinical Skills: Holistically combines conversational quality (soft skills) and diagnostic accuracy (hard skills) through a dual-competency reward system.
- Experiential Reinforcement Learning: Utilizes a novel closed-loop framework where the agent continuously learns and improves from its own high-quality experiences.
- State-of-the-Art Performance: Achieves leading performance among open-source models on dynamic benchmarks such as HealthBench (36.29 Avg. Score) and MAQuE (60.00% Accuracy), even surpassing some proprietary models at a significantly smaller scale.
Ideal Use Cases
- Medical Consultation Simulation: For training and research in AI-driven patient interaction.
- Diagnostic Support Systems: As an intelligent agent to assist in preliminary patient assessment and inquiry.
- Healthcare AI Development: For developers seeking a highly capable and efficient model for complex medical dialogue and decision-making tasks.