ningpy/redflag-detection-V2.0
ningpy/redflag-detection-V2.0 is a 7.6 billion parameter medical red-flag symptom extraction model, fine-tuned from Qwen2.5-7B-Instruct. It specializes in extracting structured medical information from user messages, including age, patient type, conditions, symptoms, and modifiers, to feed into rule engines for emergency red-flag determination. The model demonstrates strong performance with a 90.0% exact match and 0.950 Micro F1 score on a 100-case out-of-distribution test set, making it highly effective for automated medical triage systems. It supports both English and Bahasa Melayu, trained on 8355 samples covering 70 red-flag rules.
Loading preview...
Overview
ningpy/redflag-detection-V2.0 is a specialized 7.6 billion parameter language model, fine-tuned from Qwen2.5-7B-Instruct, designed for medical red-flag symptom extraction. Its primary function is to parse user messages and extract structured medical data, which can then be used by rule engines to identify potential emergency conditions.
Key Capabilities
- Structured Medical Information Extraction: Extracts detailed information such as age, patient type (self/third_party), medical conditions, symptoms (including denied symptoms), and modifiers (e.g., onset, fever temperature).
- Rule Engine Integration: Outputs data in a JSON schema, making it directly usable for automated red-flag rule evaluation.
- Performance: Achieves a 90.0% Exact Match and 0.950 Micro F1 score on a 100-case out-of-distribution test set, indicating high accuracy in identifying and extracting relevant medical flags.
- Multilingual Support: Trained on data in both English and Bahasa Melayu, covering 70 distinct red-flag rules.
Good For
- Automated Medical Triage: Ideal for systems that need to quickly identify critical medical symptoms from free-text input.
- Healthcare Applications: Useful in scenarios requiring the structured extraction of patient information for clinical decision support or emergency service routing.
- Research and Development: Provides a robust base for further fine-tuning on specific medical datasets or expanding red-flag detection capabilities.