Benchmarks

Task NameDataset NameSOTA ResultTrendResults

Glucose forecasting

AI-READI Average across groups

9.39iGlu-CE

Glucose forecasting

AI-READI (Insulin T2D)

13.04iGlu-CE

Glucose forecasting

AI-READI (Non-Insulin T2D)

10.31iGlu-CE

Glucose forecasting

AI-READI (Pre-T2D)

8.14iGlu-CE

Glucose forecasting

AI-READI (Healthy)

6.08iGlu-CE

Glucose level forecasting (PH = 3 hours)

AI-READI Avg across groups

24.15Mean Absolute Error (MAE)

Glucose level forecasting (PH = 3 hours)

AI-READI Insulin T2D group

33.32MAE

Glucose level forecasting (PH = 3 hours)

AI-READI Non-Insulin T2D group

23.62MAE

Glucose level forecasting (PH = 3 hours)

AI-READI Pre-T2D group

20.05MAE

Glucose level forecasting (PH = 3 hours)

AI-READI Healthy group

19.5MAE

Diabetes Categorization

AI-READI (five random seeds)

78.15Macro AUROC

Counterfactual Generation

AI-READI (Class 1)

98Validity

Counterfactual Generation

AI-READI Class 0

0.99Validity

Classification

AI-READI Scenario C — Dual-Class Undersampling (train)

21.47ACC

Classification

AI-READI Scenario B — Negative-Class Undersampling (train)

17.16Accuracy

Classification

AI-READI Scenario A — Positive-Class Undersampling (train)

21ACC

Membership Inference Attack

AI-READI

100AUC

Fidelity Evaluation

AI-READI (test)

1.352CWC

Classification

AI-READI Full (train)

71.8Accuracy