ThaiLLM/ThaiLLM-8B-MedApp
ThaiLLM/ThaiLLM-8B-MedApp is an 8 billion parameter language model, fine-tuned by ThaiLLM using reinforcement learning from ThaiLLM/ThaiLLM-8B-IQ. This model is specifically designed for medical applications, excelling at routing user requests to appropriate medical tool calls and information queries. It demonstrates high accuracy in medical tool use, particularly for tasks like searching medical facts and managing appointments, with a context length of 32768 tokens.
Loading preview...
Overview
ThaiLLM-8B-MedApp is an 8 billion parameter language model developed by ThaiLLM, specifically fine-tuned for medical applications. It is built upon the ThaiLLM/ThaiLLM-8B-IQ base model and utilizes reinforcement learning for its specialized capabilities.
Key Capabilities
- Medical Tool Routing: The model is optimized to accurately route user requests to the correct medical tools and information queries.
- High Accuracy in Medical Tasks: Achieves strong performance in tasks such as
search_medical_facts(99.91 F1),get_health_emergency_contact(99.54 F1), andcreate_appointment(98.83 F1). - Reinforcement Learning Fine-tuning: Trained using Prime-Intellect's
prime-rlframework, enhancing its ability to handle complex medical interactions.
Performance Highlights
Compared to other models, ThaiLLM-8B-MedApp shows competitive accuracy in tool use, with a Macro F1 score of 89.64 and a Trigger F1 of 100.00. While its performance on some specific tools like create_reminder and list_reminder is lower than ThaiLLM-8B-ToolUse, it maintains high accuracy for critical medical information retrieval and appointment management.
Good For
- Applications requiring precise routing of medical queries to specific tools.
- Building chatbots or virtual assistants for healthcare settings.
- Tasks involving searching medical facts, managing appointments, and identifying emergency contacts.