fix: handle inhomogeneous fold counts in statistical analysis

webapp/benchmark.py

@@ -352,16 +352,25 @@ def _statistical_analysis(results: dict, task: str) -> dict:
         return {}
 
     # Extract scores per fold for each model
-    # Matrix: rows = folds, cols = models
-    matrix = []
-    n_folds = 0
+    # Only include models that have a consistent number of folds
+    temp_matrix = {}
+    max_folds = 0
     for name in model_names:
         folds = results[name].get("folds", [])
-        n_folds = len(folds)
+        if not folds: continue
         scores = [f.get(primary, 0) for f in folds]
-        matrix.append(scores)
+        temp_matrix[name] = scores
+        max_folds = max(max_folds, len(scores))
     
-    matrix = np.array(matrix).T  # Now (n_folds, n_models)
+    if max_folds == 0: return {}
+
+    # Final list of models that have the full fold count
+    valid_names = [n for n, s in temp_matrix.items() if len(s) == max_folds]
+    if len(valid_names) < 2: return {}
+
+    matrix = np.array([temp_matrix[n] for n in valid_names]).T  # Shape: (n_folds, n_models)
+    n_folds = max_folds
+    model_names = valid_names  # Update model_names to match matrix columns
     
     # Calculate ranks for each fold (row)
     # Higher score = lower rank (1 is best). Using method='min' for competition ranking (ties get same best rank)