From 5854e1b6e13ca6fe79a672296d1dfb4398be5857 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=86=AF=E5=9F=BA=E9=AD=81?= <1412414664@qq.com>
Date: Mon, 22 Jun 2026 23:36:24 +0800
Subject: [PATCH] fix: load eval jsonl files with utf-8 bom

---
 .../promptflow/evals/evaluate/_evaluate.py    | 12 +++++------
 .../tests/evals/unittests/test_evaluate.py    | 20 ++++++++++++++++++-
 2 files changed, 25 insertions(+), 7 deletions(-)

diff --git a/src/promptflow-evals/promptflow/evals/evaluate/_evaluate.py b/src/promptflow-evals/promptflow/evals/evaluate/_evaluate.py
index 578236c527f..f50c7b0cd15 100644
--- a/src/promptflow-evals/promptflow/evals/evaluate/_evaluate.py
+++ b/src/promptflow-evals/promptflow/evals/evaluate/_evaluate.py
@@ -191,12 +191,12 @@ def _validate_and_load_data(target, data, evaluators, output_path, azure_ai_proj
         if not isinstance(evaluation_name, str):
             raise ValueError("evaluation_name must be a string.")
 
-    try:
-        initial_data_df = pd.read_json(data, lines=True)
-    except Exception as e:
-        raise ValueError(
-            f"Failed to load data from {data}. Please validate it is a valid jsonl data. Error: {str(e)}."
-        ) from e
+    try:
+        initial_data_df = pd.read_json(data, lines=True, encoding="utf-8-sig")
+    except Exception as e:
+        raise ValueError(
+            f"Failed to load data from {data}. Please validate it is a valid jsonl data. Error: {str(e)}."
+        ) from e
 
     return initial_data_df
 
diff --git a/src/promptflow-evals/tests/evals/unittests/test_evaluate.py b/src/promptflow-evals/tests/evals/unittests/test_evaluate.py
index 8f5d38db1d3..9283eec34c7 100644
--- a/src/promptflow-evals/tests/evals/unittests/test_evaluate.py
+++ b/src/promptflow-evals/tests/evals/unittests/test_evaluate.py
@@ -11,7 +11,12 @@
 from promptflow.client import PFClient
 from promptflow.evals._constants import DEFAULT_EVALUATION_RESULTS_FILE_NAME
 from promptflow.evals.evaluate import evaluate
-from promptflow.evals.evaluate._evaluate import _aggregate_metrics, _apply_target_to_data, _rename_columns_conditionally
+from promptflow.evals.evaluate._evaluate import (
+    _aggregate_metrics,
+    _apply_target_to_data,
+    _rename_columns_conditionally,
+    _validate_and_load_data,
+)
 from promptflow.evals.evaluate._utils import _apply_column_mapping, _trace_destination_from_project_scope
 from promptflow.evals.evaluators import (
     ContentSafetyEvaluator,
@@ -123,6 +128,19 @@ def test_evaluate_invalid_jsonl_data(self, mock_model_config, invalid_jsonl_file
         assert "Failed to load data from " in exc_info.value.args[0]
         assert "Please validate it is a valid jsonl data" in exc_info.value.args[0]
 
+    def test_validate_and_load_data_accepts_utf8_sig_jsonl(self, tmp_path):
+        data_file = tmp_path / "evaluate_test_data_utf8_sig.jsonl"
+        data_file.write_text(
+            '{"question": "How are you?", "answer": "Fine", "ground_truth": "Fine"}\n',
+            encoding="utf-8-sig",
+        )
+
+        data = _validate_and_load_data(None, str(data_file), {"g": F1ScoreEvaluator()}, None, None, None)
+
+        assert data.to_dict(orient="records") == [
+            {"question": "How are you?", "answer": "Fine", "ground_truth": "Fine"}
+        ]
+
     def test_evaluate_missing_required_inputs(self, missing_columns_jsonl_file):
         with pytest.raises(ValueError) as exc_info:
             evaluate(data=missing_columns_jsonl_file, evaluators={"g": F1ScoreEvaluator()})