diff --git a/tests/unit/test_inference_pair_aggregate.py b/tests/unit/test_inference_pair_aggregate.py
index fb2faf1..7d07aa8 100644
--- a/tests/unit/test_inference_pair_aggregate.py
+++ b/tests/unit/test_inference_pair_aggregate.py
@@ -274,7 +274,7 @@ class TestUnitLevelAggregation:
             _make_question("s2", answer="A"),
         ]
         records = [_make_record("s1", prediction="B", answer="B")]  # 缺 s2 的 record
-        with pytest.raises(ValueError, match=r"缺 prediction.*聚合不变量被破坏"):
+        with pytest.raises(ValueError, match=r"s2.*缺 prediction.*聚合不变量被破坏"):
             _aggregate_results(records, questions, "run-broken-invariant")
 
     def test_empty_records_returns_zero(self) -> None: