thebnbrkr/hippocratically_small

Hippocratically Small: Medical Chatbot Model

Model: thebnbrkr/hippocratically_small
Architecture: [TinyLlama-1.1B-Chat-v1.0 or your base model]
Fine-tuned for: Medical Q&A and clinical conversation

Model Description

This model is a small language model fine-tuned for medical chat and clinical question-answering. It is designed to assist with general medical queries, symptom triage, and patient education. The model was trained on a curated dataset of patient-doctor dialogues and medical Q&A.

Intended Use

• Medical question answering for educational and research purposes
• Simulated patient-doctor conversations
• Clinical triage and symptom explanation

Limitations

• Not a substitute for professional medical advice. Always consult a healthcare provider for real medical concerns.
• May not reflect the latest clinical guidelines or local protocols.
• Answers may be generic and lack nuance for complex cases.
• May occasionally provide outdated or partially incorrect information.

Example Prompt:

prompt: | A 28-year-old woman presents with sudden onset right lower abdominal pain, mild fever, and nausea. She reports her last menstrual period was 6 weeks ago. What are the possible diagnoses, and what initial tests should be performed?

Example response:

The symptoms described could be due to several causes, including ectopic pregnancy, ovarian cyst rupture, or acute appendicitis. Given the missed period, ectopic pregnancy should be considered first. Initial tests should include a urine or serum pregnancy test and a pelvic ultrasound to assess for ectopic pregnancy or ovarian pathology. Blood tests and a physical examination are also important to evaluate for infection or other causes

How the model works:

• This model is a LoRA fine-tuned version of TinyLlama-1.1B-Chat-v1.0, trained on patient-doctor dialogues and medical Q&A datasets.
• Fine-tuning was performed using the PEFT (Parameter-Efficient Fine-Tuning) library and Hugging Face Transformers, adapting the model to generate helpful, conversational medical responses.

Why the model gave this answer:

• The model generates answers by recognizing patterns in medical data it was trained on. For acute abdominal pain, it recalled common textbook causes (like appendicitis and peptic ulcer) and described their symptoms and diagnostic steps. However, it may not always prioritize the most urgent or likely diagnosis (such as ectopic pregnancy in a woman with missed periods and lower abdominal pain), especially if such patterns were underrepresented in the training data.

Limitations:

• May provide verbose or generic answers.
• May miss critical diagnoses or not prioritize emergencies.
• Not a substitute for professional clinical judgment or guidelines.

Model Details

• Base model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
• Fine-tuning method: LoRA (Low-Rank Adaptation) using the PEFT library and Hugging Face Transformers.
• Training data: ruslanmv/ai-medical-chatbot (patient-doctor dialogues)
• Intended use: Medical Q&A, clinical triage, patient education (for research/education only)
• How it works: The model uses instruction-tuning and chat formatting. Given a user prompt, it generates a response based on patterns learned from the fine-tuning data.