ningpy/redflag-detection-V4.2
ningpy/redflag-detection-V4.2 is a 7.6 billion parameter model fine-tuned from Qwen2.5-7B-Instruct, specifically designed for medical red-flag extraction. It strictly complies with the 2026-08 Medical Advisory "Red Flad_New" specification, featuring 59 rules and an 89-item closed symptom set. This model excels at identifying critical medical indicators with high precision and recall, making it suitable for clinical decision support systems.
Loading preview...
Overview
ningpy/redflag-detection-V4.2 is a specialized 7.6 billion parameter language model, fine-tuned from Qwen2.5-7B-Instruct, for the precise extraction of medical red flags. This version (V4.2) is engineered for strict compliance with the 2026-08 Medical Advisory "Red Flad_New" specification, adhering to a defined set of 59 rules and 89 symptoms.
Key Enhancements in V4.2
- Strict Spec Compliance: Removed 6 extension rules and 7 symptoms present in V4.1 to align precisely with the official 59-rule Medical Advisory scope.
- Refined Symptom Set: Features a closed symptom set of 89 items, down from 96 in V4.1, by dropping orphaned entries.
- Improved Training Data: Training data was sanitized by removing 81 samples of orphan-only cases and 160 orphan symptom mentions from labels, with an addition of 31 negative reinforcement samples.
- Performance: Achieves an Exact Match of 89.6%, Precision of 0.921, and F1 score of 0.844 on the strict validation set. For clinical-fair evaluation (where "SUSPECTED" is a True Positive), it shows a Recall of 0.865.
Use Cases
This model is ideal for applications requiring highly accurate and compliant medical red-flag detection, particularly within clinical decision support systems that must adhere to the specified 59-rule standard. It is designed to be paired with a compatible 59-rule engine, such as redflag_scripts/v20/rule_engine_v20.py.