Overview
vikash06/doctorLLM10k is a 7 billion parameter language model, fine-tuned over 10 epochs on 10,000 data points, specifically for generating creative medical responses. It is designed to handle questions and instructions that require general medical knowledge, focusing on tasks that do not necessitate external information retrieval. The model's training emphasizes following specific instructions, constraints, and guidelines within medical contexts.
Key Capabilities
- Creative Medical Response Generation: Excels at producing imaginative yet medically sound answers to doctor-related prompts.
- Instruction Following: Highly capable of adhering to detailed instructions, constraints, and guidelines provided in prompts.
- General Medical Knowledge: Leverages a broad understanding of medical concepts to formulate responses.
Performance Metrics
Evaluated on the Open LLM Leaderboard, vikash06/doctorLLM10k achieved an average score of 50.70. Notable scores include 79.94 on HellaSwag (10-Shot) and 70.01 on Winogrande (5-shot). Its performance on MMLU (5-Shot) was 44.40, and on AI2 Reasoning Challenge (25-Shot) it scored 54.95. For mathematical reasoning, it scored 10.16 on GSM8k (5-shot).
Good For
- Medical Content Generation: Ideal for applications requiring the creation of medically relevant text based on specific prompts.
- Educational Tools: Can be used in scenarios where a model needs to provide creative yet accurate medical explanations or scenarios.
- Simulated Medical Interactions: Suitable for developing tools that simulate doctor-patient interactions or medical consultations where creative problem-solving is needed.