Evaluation Results

205-sample benchmark across 15+ categories -- real metrics computed from methodD statistical analysis

ROC Curve

Receiver Operating Characteristic -- AUC = 0.781

Confusion Matrix

205 total predictions

Predicted

Actual

AIHuman
AI
61TP
44FN
Human
15FP
85TN

Overall accuracy: 71.2%

Summary Statistics

Key performance metrics

MetricValue
Accuracy71.2%
Precision80.3%
Recall58.1%
F1 Score67.4%
AUC-ROC0.781
Cohen's Kappa0.428
Cohen's d1.184
Samples205

Per-Domain Accuracy

Classification accuracy across content categories

Ablation Study

Full signal vs. individual sub-signal F1 scores

Evaluation performed on a curated 205-sample benchmark spanning 15 content domains. All metrics computed from real methodD statistical analysis with threshold optimized via Youden's J statistic. AUC computed via trapezoidal integration.