add factuality, faithfulness scores

src/display/utils.py

@@ -74,6 +74,9 @@ auto_eval_column_dict.append(["model", ColumnContent, ColumnContent("Model", "ma
 #Scores
 # auto_eval_column_dict.append(["average", ColumnContent, ColumnContent("Avg", "number", True)])
 
+auto_eval_column_dict.append(["Faithfulness", ColumnContent, ColumnContent("Faithfulness", "number", True)])
+auto_eval_column_dict.append(["Factuality", ColumnContent, ColumnContent("Factuality", "number", True)])
+
 for task in Tasks:
     auto_eval_column_dict.append([task.name, ColumnContent, ColumnContent(task.value.col_name, "number", True)])
 

src/populate.py

@@ -60,11 +60,20 @@ def get_leaderboard_df(results_path: str,
     # if AutoEvalColumn.average.name in df:
     #     df = df.sort_values(by=[AutoEvalColumn.average.name], ascending=False)
 
-    df = df[cols].round(decimals=2)
+    cols_mod = copy.deepcopy(cols)
+    cols_mod.remove('Faithfulness')
+    cols_mod.remove('Factuality')
+    df = df[cols_mod]#.round(decimals=2)
 
     # filter out if any of the benchmarks have not been produced
     df = df[has_no_nan_values(df, benchmark_cols)]
 
+    Factuality_score = df[factuality_tasks].mean(axis=1)
+    Faithfulness_score = df[faithfulness_tasks].mean(axis=1)
+    df.insert(2, 'Factuality', Factuality_score)
+    df.insert(2, 'Faithfulness', Faithfulness_score)
+    df = df.round(decimals=2)
+
     return raw_data, df