|
|
--- |
|
|
license: mit |
|
|
tags: |
|
|
- unsloth |
|
|
- trl |
|
|
- grpo |
|
|
datasets: |
|
|
- FreedomIntelligence/Medical-R1-Distill-Data |
|
|
language: |
|
|
- en |
|
|
base_model: |
|
|
- unsloth/phi-4 |
|
|
pipeline_tag: text-generation |
|
|
library_name: transformers |
|
|
--- |
|
|
|
|
|
This model is a [Phi-4](https://huggingface.co/unsloth/phi-4) Large Language Model (LLM) that has been fine-tuned for medical reasoning tasks. It leverages the architecture of the Phi-4 model and has been specifically adapted to understand and process medical information, answer medical questions, and potentially assist in clinical decision-making support (for research purposes only). |
|
|
|
|
|
The fine-tuning process was carried out using [Unsloth AI](https://unsloth.ai/), a library designed for efficient and accessible LLM fine-tuning. Unsloth AI simplifies the process of adapting powerful models like Phi-4 for specific downstream tasks, making it easier to create specialized LLMs like this medical reasoning model. |
|
|
|
|
|
**Key Features:** |
|
|
|
|
|
* Based on the Phi-4 architecture, known for its efficiency and performance. |
|
|
* Fine-tuned on a medical reasoning dataset to enhance its capabilities in the medical domain. |
|
|
* Trained using Unsloth AI for optimized fine-tuning. |
|
|
|
|
|
## Dataset |
|
|
|
|
|
The model was fine-tuned on [FreedomIntelligence/Medical-R1-Distill-Data](https://huggingface.co/datasets/FreedomIntelligence/Medical-R1-Distill-Data) dataset. |
|
|
|
|
|
* **Dataset Description:** SFT dataset distilled from Deepseek-R1 (Full Power Version), based on medical verifiable problems from HuatuoGPT-o1. |
|
|
* **Source:** [FreedomIntelligence/Medical-R1-Distill-Data](https://huggingface.co/datasets/FreedomIntelligence/Medical-R1-Distill-Data) |
|
|
* **Dataset Size:** 22k rows. |
|
|
|
|
|
**Important Note on Dataset Bias:** It's important to acknowledge that medical datasets can contain biases that reflect real-world healthcare disparities. This model may inherit and amplify such biases. Users should be aware of this limitation and interpret the model's outputs with caution. |
|
|
|
|
|
## Intended Use |
|
|
|
|
|
This model is intended for **research purposes only**. Potential use cases include: |
|
|
|
|
|
* **Medical Education:** Assisting medical students and professionals in learning and reviewing medical concepts. |
|
|
* **Clinical Decision Support Research:** Exploring the potential of LLMs in providing support for clinical decision-making (always under the supervision of qualified professionals and not for direct clinical use). |
|
|
* **Medical Information Retrieval:** Improving the accuracy and relevance of medical information retrieval systems. |
|
|
* **Medical Question Answering:** Developing systems that can answer complex medical questions. |
|
|
|
|
|
**Disclaimer:** This model is **not intended for clinical use** and should **not be used as a substitute for professional medical advice, diagnosis, or treatment.** Always consult with a qualified healthcare provider for any health concerns or before making any decisions related to your health or treatment. |
|
|
|
