Vincenzo2K04/NINA-Qwen3-4B
Vincenzo2K04/NINA-Qwen3-4B is a 4 billion parameter Qwen3-based language model fine-tuned for nursing documentation tasks. This model is specifically designed with strict, non-overridable guardrails to prevent clinical diagnoses, treatment recommendations, or physician impersonation. It excels as a specialized documentation tool for qualified registered nurses, ensuring safe and compliant AI assistance in healthcare settings.
Loading preview...
NINA-Qwen3-4B: Nursing Intelligent Network Assistant
NINA-Qwen3-4B is a specialized 4 billion parameter language model, fine-tuned from the Qwen/Qwen3-4B architecture. Its core purpose is to assist with nursing documentation tasks, providing a focused and safe AI tool for healthcare professionals.
Key Capabilities & Differentiators
- Strict Guardrails: NINA is engineered with absolute, non-overridable limits to ensure patient safety and regulatory compliance. It will never provide clinical diagnoses, recommend treatments or medications, alter physician orders, or impersonate medical personnel.
- Adversarial Training: The model has been rigorously trained against 22 adversarial techniques, with P1 guardrail testing before every push, enhancing its robustness against misuse.
- Specialized Fine-tuning: Trained using SFT with LoRA (r=16 RSLoRA) on approximately 6,400 balanced samples, optimizing its performance for nursing-specific language and workflows.
Ideal Use Cases
- Nursing Documentation: Primarily designed as a documentation aid for qualified registered nurses.
- Safe AI Integration: Suitable for environments requiring highly constrained and safety-critical AI applications in healthcare.
Important Disclaimer: NINA is strictly a documentation tool and is not intended for diagnostic or clinical decision-making purposes.