Update cardinality score and C2ST naming

src/about.py

@@ -15,7 +15,7 @@ class Tasks(Enum):
     # task0 = Task("anli_r1", "acc", "ANLI")
     # task1 = Task("logiqa", "acc_norm", "LogiQA")
     # task_0 = Task("multi-table", "AggregationDetection-LogisticRegression", "AggregationDetection-LogisticRegression ⬇️")
-    task_1 = Task("multi-table", "AggregationDetection-XGBClassifier", "AggregationDetection-XGBClassifier ⬇️")
+    task_1 = Task("multi-table", "AggregationDetection-XGBClassifier", "C2ST Agg-XGBClassifier ⬇️")
     task_2 = Task("multi-table", "CardinalityShapeSimilarity", "CardinalityShapeSimilarity ⬆️")
 
 class SingleTableTasks(Enum):

src/populate.py

@@ -42,14 +42,23 @@ def get_leaderboard_df(results_path: str, cols: list, benchmark_cols: list) -> p
                 data = json.load(fp)
                 all_data_json.append(data)
 
-    multi_table_metrics = [task.value.col_name for task in Tasks]
-    single_table_metrics = [task.value.col_name for task in SingleTableTasks]
-    single_column_metrics = [task.value.col_name for task in SingleColumnTasks]
+    multi_table_metrics = [task.value.metric for task in Tasks]
+    single_table_metrics = [task.value.metric for task in SingleTableTasks]
+    single_column_metrics = [task.value.metric for task in SingleColumnTasks]
 
-    # create empty dataframe with the columns multi_table_metrics
-    multitable_df = pd.DataFrame(columns=["Dataset", "Model"] + multi_table_metrics)
-    singletable_df = pd.DataFrame(columns=["Dataset", "Model"] + single_table_metrics)
-    singlecolumn_df = pd.DataFrame(columns=["Dataset", "Table", "Model"] + single_column_metrics)
+    multi_table_metric_names = [task.value.col_name for task in Tasks]
+    single_table_metric_names = [task.value.col_name for task in SingleTableTasks]
+    single_column_metric_names = [task.value.col_name for task in SingleColumnTasks]
+
+    # Create mapping between metrics and their display names
+    multi_table_metric_mapping = dict(zip(multi_table_metrics, multi_table_metric_names))
+    single_table_metric_mapping = dict(zip(single_table_metrics, single_table_metric_names))
+    single_column_metric_mapping = dict(zip(single_column_metrics, single_column_metric_names))
+
+    # create empty dataframe with the display column names
+    multitable_df = pd.DataFrame(columns=["Dataset", "Model"] + multi_table_metric_names)
+    singletable_df = pd.DataFrame(columns=["Dataset", "Model"] + single_table_metric_names)
+    singlecolumn_df = pd.DataFrame(columns=["Dataset", "Table", "Model"] + single_column_metric_names)
 
     # iterate through all json files and add the data to the dataframe
     for data in all_data_json:
@@ -58,6 +67,16 @@ def get_leaderboard_df(results_path: str, cols: list, benchmark_cols: list) -> p
         row = {"Dataset": dataset, "Model": model}
         for metric in multi_table_metrics:
             stripped_metric = strip_emoji(metric)
+            display_name = multi_table_metric_mapping[metric]  # Get the display name for this metric
+            
+            # Special case for CardinalityShapeSimilarity which is stored under "Trends"
+            if "CardinalityShapeSimilarity" in metric:
+                if "Trends" in data["multi_table_metrics"] and "cardinality" in data["multi_table_metrics"]["Trends"]:
+                    row[display_name] = data["multi_table_metrics"]["Trends"]["cardinality"]
+                else:
+                    row[display_name] = np.nan
+                continue
+                
             if stripped_metric in data["multi_table_metrics"]:
                 metric_values = []
                 for table in data["multi_table_metrics"][stripped_metric].keys():
@@ -66,14 +85,15 @@ def get_leaderboard_df(results_path: str, cols: list, benchmark_cols: list) -> p
                     if "statistic" in data["multi_table_metrics"][stripped_metric][table]:
                         metric_values.append(data["multi_table_metrics"][stripped_metric][table]["statistic"])
 
-                row[metric] = np.mean(metric_values).round(decimals=2)
+                row[display_name] = np.mean(metric_values).round(decimals=2)  # Use display name as column
             else:
-                row[metric] = np.nan
+                row[display_name] = np.nan  # Use display name as column
         multitable_df = pd.concat([multitable_df, pd.DataFrame([row])], ignore_index=True)
 
         singletable_row = {"Dataset": dataset, "Model": model}
         for metric in single_table_metrics:
             stripped_metric = strip_emoji(metric)
+            display_name = single_table_metric_mapping[metric]  # Get the display name for this metric
             if stripped_metric in data["single_table_metrics"]:
                 metric_values = []
                 for table in data["single_table_metrics"][stripped_metric].keys():
@@ -82,15 +102,16 @@ def get_leaderboard_df(results_path: str, cols: list, benchmark_cols: list) -> p
                     if "value" in data["single_table_metrics"][stripped_metric][table]:
                         metric_values.append(data["single_table_metrics"][stripped_metric][table]["value"])
 
-                singletable_row[metric] = np.mean(metric_values).round(decimals=2)
+                singletable_row[display_name] = np.mean(metric_values).round(decimals=2)  # Use display name as column
             else:
-                singletable_row[metric] = np.nan
+                singletable_row[display_name] = np.nan  # Use display name as column
         singletable_df = pd.concat([singletable_df, pd.DataFrame([singletable_row])], ignore_index=True)
 
         singlecolumn_row = {"Dataset": dataset, "Model": model, "Table": ""}
         # insert row
         for metric in single_column_metrics:
             stripped_metric = strip_emoji(metric)
+            display_name = single_column_metric_mapping[metric]  # Get the display name for this metric
             if stripped_metric in data["single_column_metrics"]:
                 for table in data["single_column_metrics"][stripped_metric].keys():
                     # check if row where dataset = dataset, model = model, table = table exists
@@ -115,7 +136,7 @@ def get_leaderboard_df(results_path: str, cols: list, benchmark_cols: list) -> p
                     singlecolumn_df.loc[
                         (singlecolumn_df["Dataset"] == dataset) & 
                         (singlecolumn_df["Model"] == model) & 
-                        (singlecolumn_df["Table"] == table), metric] = np.mean(metric_values).round(decimals=2)
+                        (singlecolumn_df["Table"] == table), display_name] = np.mean(metric_values).round(decimals=2)  # Use display name as column
             
 
     return singlecolumn_df, singletable_df, multitable_df