vmal/med-advisor-conversation-4B
vmal/med-advisor-conversation-4B is a 4 billion parameter Qwen3-based language model, fully fine-tuned on medical conversations. It is designed to act as a medical explainer, adjusting its responses based on the audience (patient, caregiver, science-literate adult, medical student, or healthcare worker). The model supports multi-turn conversations and features both 'thinking' and 'non-thinking' modes, excelling at providing nuanced medical explanations and safety escalations.
Loading preview...
Med Advisor Conversation 4B
Med Advisor Conversation 4B is a 4 billion parameter medical explainer model, built on Qwen3-4B and extensively fine-tuned using reinforcement learning (RL100 checkpoint) on medical conversations. Its core capability is to adapt its explanations to different audiences, including patients, caregivers, science-literate adults, medical students, and healthcare workers.
Key Capabilities
- Audience-Adjusted Explanations: Provides medical information tailored to five distinct user types, modifying depth and vocabulary accordingly.
- Multi-Turn Conversations: Capable of engaging in extended dialogues, updating explanations based on user corrections or follow-up information.
- Thinking and Non-thinking Modes: Features an
enable_thinkingswitch, allowing the model to reason internally before generating a visible answer, which significantly improves performance (e.g., 59.9% in thinking mode vs. 54.4% non-thinking on HealthBench Consensus). - Broad Medical Scope: Covers 16 medical domains and 6 types of requests, including explanations, lab interpretations, medication reasoning, self-care advice, and safety escalation.
- Safety-Oriented: Trained to prioritize emergency escalation and to decline requests for diagnosis, prescribing, or personal medical advice.
Good For
- Educational Tools: Ideal for applications requiring clear, evidence-aware medical and scientific explanations.
- Research: Serves as a valuable artifact for studying conversation-level reinforcement learning in specialized domains.
- Information Dissemination: Useful for explaining medical terms, mechanisms, evidence strength, and guiding users on when to seek clinical advice.
- Simulated Medical Scenarios: Can be used in training environments for medical students or healthcare workers to practice communication and information retrieval.