Kimi K 2.6 submission

[mubbytheback]

Team: Kimi K 2.6\nModel: GraphSAGE Inductive GNN\nType: llm

[github-actions]

🏆 Submission Scored

🎯 GNN Challenge Scoring Script
============================================================

📂 Submission: /tmp/pr_predictions.csv
   Ground truth: data/test_labels.csv

============================================================
  Submission Evaluation
============================================================

📊 Basic Metrics:
  Accuracy:           0.6126
  Precision:          0.5000
  Recall:             0.1395

🎯 Difficult Metrics (Recommended):
  Macro F1-Score:     0.4803    ← (equal weight per class)
  Balanced Accuracy:  0.5256  ← (avg per-class recalls)
  MCC:                0.0805          ← (strictest metric)

📈 Other F1 Variants:
  F1 Binary:          0.2182
  F1 Weighted:        0.5394

🔍 Clinical Metrics:
  Sensitivity (TPR):  0.1395
  Specificity (TNR):  0.9118
  Youden's Index:     0.0513
  Cohen's Kappa:      0.0591

🔗 Confusion Matrix:
  TN= 62  FP=  6
  FN= 37  TP=  6

📊 Class Distribution (y_true):
  Class 0: 68 samples
  Class 1: 43 samples
============================================================

Check the leaderboard for rankings!