diff --git a/example/evaluate_generated.py b/example/evaluate_generated.py
index 7ed87e6..5ae1a7d 100644
--- a/example/evaluate_generated.py
+++ b/example/evaluate_generated.py
@@ -123,7 +123,10 @@ def main():
     args.vocab = str((base_dir / args.vocab).resolve()) if not Path(args.vocab).is_absolute() else args.vocab
     args.out = str((base_dir / args.out).resolve()) if not Path(args.out).is_absolute() else args.out
     if args.reference and not Path(args.reference).is_absolute():
-        args.reference = str((base_dir / args.reference).resolve())
+        if any(ch in args.reference for ch in ["*", "?", "["]):
+            args.reference = str(base_dir / args.reference)
+        else:
+            args.reference = str((base_dir / args.reference).resolve())
     ref_path = resolve_reference_path(args.reference)
     split = load_json(args.split)
     time_col = split.get("time_column", "time")