luizaaca/qwen3-0.6b-clinical-screening
The luizaaca/qwen3-0.6b-clinical-screening model is a 0.8 billion parameter Qwen3-based language model fine-tuned by luizaaca for clinical screening applications. It specializes in identifying diseases from reported symptoms, utilizing a 32768 token context length. This model is optimized for lightweight, plain-text symptom-to-disease identification, distinguishing it from models that process structured JSON schemas. It was trained using QLoRA on specific medical datasets to provide auxiliary support for healthcare professionals.
Loading preview...
Qwen3-0.6B Clinical Screening: A Specialized Medical Assistant
This model, developed by luizaaca, is a compact, 0.8 billion parameter Qwen3-based language model specifically fine-tuned for clinical screening and symptom-to-disease identification. It leverages QLoRA via Unsloth on a 4-bit loaded base model, with a rank of 16 and alpha of 32, targeting key projection modules.
Key Capabilities
- Disease Identification: Processes natural-language symptom lists to suggest potential diseases.
- Specialized Training: Fine-tuned on
dhivyeshrk/diseases-and-symptoms-datasetandniyarrbarman/symptom2diseasefor medical relevance. - Output Format: Provides answers in a standardized disclaimer format: "Based on the reported symptoms, the clinical indication points to: ."
- Lightweight Deployment: Available in merged
transformersformat, PEFT LoRA adapters, and quantized GGUF forllama.cpp, Ollama, and LM Studio. - Evaluation: Validated using accuracy, macro-F1, Cohen's Kappa, confusion matrices, and BERTScore against held-out datasets.
Intended Use Cases
- Research Experiments: Ideal for exploring lightweight clinical-screening assistants.
- Prototyping: Suitable for teaching and developing symptom-to-disease prompting systems.
- Local Inference: Supports local deployment with Transformers, PEFT, GGUF-compatible runtimes, or Ollama.
Important Considerations
This model is designed as an auxiliary tool for healthcare professionals and is not intended for autonomous diagnosis, treatment decisions, emergency triage without oversight, prescribing, or as a substitute for professional medical judgment. It focuses on plain-text disease identification, unlike models designed for structured JSON outputs.