leaderboard

src/populate.py

@@ -14,9 +14,16 @@ def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchm
     all_data_json = [v.to_dict() for v in raw_data]
 
     df = pd.DataFrame.from_records(all_data_json)
-    # Sort primarily by LLM exact-match Pass@1 metric
-    sort_col = AutoEvalColumn.pass_at_1.name if hasattr(AutoEvalColumn, "pass_at_1") else AutoEvalColumn.average.name
-    df = df.sort_values(by=[sort_col], ascending=False)
+    # Sort primarily by LLM exact-match Pass@1 metric; if not present, fall back to average
+    preferred_cols = []
+    if hasattr(AutoEvalColumn, "pass_at_1"):
+        preferred_cols.append(AutoEvalColumn.pass_at_1.name)
+    preferred_cols.append(AutoEvalColumn.average.name)
+
+    for col in preferred_cols:
+        if col in df.columns:
+            df = df.sort_values(by=[col], ascending=False)
+            break
     df = df[cols].round(decimals=2)
 
     # filter out if any of the benchmarks have not been produced