123

app.py

@@ -64,11 +64,20 @@ def format_table(df):
 # Function to calculate top 3 models based on combined score (average of numeric columns)
 def get_top_3_models(robustness_df, context_grounding_df):
     # Combine numeric columns from both datasets
-    numeric_cols_robustness = ["Baseline", "Robustness (Δ)"]  # Updated to match the actual column name
+    numeric_cols_robustness = ["Baseline", "Robustness (Δ)"]  # Columns with numeric or string-numeric data
     numeric_cols_context = ["Irrelevant Ctx", "No Ctx", "Ctx Grounding QA", "Ctx Grounding TG", "Ctx Grounding", "Robustness", "Compliance"]  # From context grounding
     
-    # Extract numeric values for scoring
-    robustness_scores = robustness_df[numeric_cols_robustness].apply(lambda x: float(str(x).split(" (")[0]) if " (" in str(x) else float(x), axis=1)
+    # Extract numeric values for each column in robustness_df
+    robustness_scores = pd.DataFrame()
+    for col in numeric_cols_robustness:
+        if any(" (" in str(x) for x in robustness_df[col]):
+            # Handle string values with deltas (e.g., "0.95 (0.0)")
+            robustness_scores[col] = robustness_df[col].apply(lambda x: float(str(x).split(" (")[0]) if " (" in str(x) else float(x))
+        else:
+            # Handle direct float values
+            robustness_scores[col] = robustness_df[col].astype(float)
+    
+    # Extract numeric values for context_grounding_df (all are already float values)
     context_scores = context_grounding_df[numeric_cols_context].astype(float)
     
     # Combine scores by averaging