NEU-HAI/mental-alpaca
NEU-HAI/mental-alpaca is a 7 billion parameter Alpaca-based large language model developed by the Northeastern University Human-Centered AI Lab. Fine-tuned on 4 high-quality datasets (Dreaddit, DepSeverity, SDCNL, CCRS-Suicide), this model specializes in mental health prediction using online text data. It is intended for research purposes in English, leveraging a 4096-token context length for text generation tasks.
Loading preview...
Overview
NEU-HAI/mental-alpaca is a 7 billion parameter language model developed by the Northeastern University Human-Centered AI Lab. It is fine-tuned from an Alpaca base model, which itself is based on Llama-2-7b, for the specific task of mental health prediction from online text data. The model leverages a 4096-token context length and is designed for research applications.
Key Capabilities
- Mental Health Prediction: Specialized in analyzing online text for mental health indicators.
- Fine-tuned Datasets: Trained on Dreaddit, DepSeverity, SDCNL, and CCRS-Suicide datasets to enhance its predictive accuracy.
- English Language Support: Primarily developed and intended for use with English text.
Intended Use Cases
- Research Purposes: Designed for academic and research exploration in mental health prediction.
- Text Analysis: Suitable for analyzing online textual data to identify patterns related to mental health.
Limitations and Considerations
- Research Use Only: Not intended for clinical diagnosis or direct patient care.
- Compliance: Use must adhere to the restrictions and licenses of the original Stanford Alpaca and Llama-2-7b projects.
- Bias and Risks: Inherits biases and limitations from its base models, as detailed in the respective project documentations. Further details on the fine-tuning process and prompts are available in the associated research paper.