MelloGPT: A Specialized Mental Health LLM
MelloGPT, developed by steve-cse, is a 7 billion parameter language model built upon the Mistral-7B-Instruct-v0.1 architecture. Its primary distinction lies in its fine-tuning on the counsel-chat dataset, which comprises mental health counseling conversations. This specialization aims to equip the model with the ability to understand and respond empathetically to mental health concerns.
Key Capabilities
- Mental Health Conversation Focus: Designed to engage in dialogues related to mental well-being.
- Empathetic Responses: Fine-tuned to provide supportive and understanding interactions.
- Ethical Considerations: Training incorporates ethical and privacy considerations relevant to sensitive mental health topics.
Performance Overview
Evaluated on the Open LLM Leaderboard, MelloGPT achieved an average score of 57.59. Notable scores include 76.12 on HellaSwag (10-Shot) and 73.88 on Winogrande (5-shot), indicating its general language understanding capabilities. However, it is crucial to note that this model is not a substitute for professional mental health assistance and should be used with this understanding.
Use Cases
- Supportive Conversational AI: Ideal for applications requiring empathetic and context-aware responses in mental health-related discussions.
- Research and Development: Can serve as a base for further research into AI applications in mental health support, with appropriate safeguards.