Public validation of Collapse Index (CI) on SST-2 dataset: 42.8% flip rate, AUC 0.698. Reveals model brittleness beyond 90%+ accuracy under perturbations!
-
Updated
Jan 7, 2026 - Python
Public validation of Collapse Index (CI) on SST-2 dataset: 42.8% flip rate, AUC 0.698. Reveals model brittleness beyond 90%+ accuracy under perturbations!
Public validation of SRI (Structural Retention) and CI (Collapse Index) metrics on AG News dataset: 90.4% accuracy, 9.2% flip rate, revealing model brittleness beyond standard benchmarks with AUC 0.874!
AI-powered system to detect fraudulent transactions in e-commerce using machine learning. Includes data preprocessing, feature engineering, and classification models like Random Forest and XGBoost. Achieved high accuracy with interpretable results for real-time detection.
Add a description, image, and links to the model-evaluation-and-metrics topic page so that developers can more easily learn about it.
To associate your repository with the model-evaluation-and-metrics topic, visit your repo's landing page and select "manage topics."