Tommi09/MedicalChatBot-Qwen3-4b
Tommi09/MedicalChatBot-Qwen3-4b is a 4 billion parameter causal language model fine-tuned from the Qwen3-4b architecture. It is specifically optimized as a medical chatbot, trained on a custom dataset derived from Huatuo26M-Lite with additional adversarial data. This model is designed for medical question-answering, with a particular recommendation for use with quantized GGUF versions in LM Studio for stable performance.
Loading preview...
Overview
Tommi09/MedicalChatBot-Qwen3-4b is a 4-billion parameter model based on the Qwen3-4b architecture, specifically fine-tuned for medical question-answering. The model was trained using a self-constructed dataset of 1100 QA pairs, which includes 1000 entries from the Huatuo26M-Lite dataset and 100 adversarial data points covering out-of-domain questions, prompt injections, and model attacks.
Key Capabilities
- Medical Q&A: Designed to provide concise and professional answers to medical questions.
- Refusal of Non-Medical Queries: Programmed to decline requests outside the medical domain.
- Ethical Compliance: Adheres to safety policies and ethical guidelines in its responses.
Performance Notes
The developers note that this 4B parameter model, after fine-tuning, appears to perform better than a fine-tuned DeepSeek-7B-Base model. However, they also caution that due to the base model's inherent capabilities, it can sometimes produce "disastrous" answers. The most stable performance is observed with the quantized q4-gguf version when run in LM Studio with a carefully crafted system prompt.
Recommended Usage
For optimal and stable interaction, it is highly recommended to use the quantized qwen3_4b_model.gguf-q4.gguf file within LM Studio. Users should configure LM Studio to use the ChatML prompt template and apply a specific system prompt to guide the model's behavior, ensuring professional, medical-only responses and adherence to ethical standards.