Update dabstep_benchmark/leaderboard.py

The submission validator rejects valid JSONL files with:
"Columns with non-string data type: task_id, agent_answer".

The issue seems to be in dabstep_benchmark/leaderboard.py:
submission_df = pd.read_json(submission_path, lines=True, dtype=str)
non_string_columns = [col for col in submission_df.columns if submission_df[col].dtype != 'object']

In newer pandas / infer_string mode, string columns have dtype "str" or StringDtype instead of object, so valid string columns are rejected.

Suggested fix:
from pandas.api.types import is_string_dtype
non_string_columns = [
col for col in submission_df.columns
if not is_string_dtype(submission_df[col])
]

dabstep_benchmark/leaderboard.py

@@ -11,6 +11,7 @@ from datasets import load_dataset
 from huggingface_hub import HfApi
 
 from dabstep_benchmark.utils import format_log, format_error, format_warning, is_valid_https_url, evaluate
+from pandas.api.types import is_object_dtype, is_string_dtype
 
 OWNER = "adyen"
 
@@ -80,7 +81,10 @@ def validate_submission(submission_df: pd.DataFrame):
         return format_error("Submission contains NaN values. Please ensure no missing data.")
 
     # Check if all columns are of string type
-    non_string_columns = [col for col in submission_df.columns if submission_df[col].dtype != 'object']
+    non_string_columns = [
+        col for col in submission_df.columns
+        if not (is_object_dtype(submission_df[col]) or is_string_dtype(submission_df[col]))
+    ]
     if non_string_columns:
         return format_error(f"Columns with non-string data type: {', '.join(non_string_columns)}")