diff --git a/xiaobai-datahub/tests/test_revision_review.py b/xiaobai-datahub/tests/test_revision_review.py
index 02eae2c..7547673 100644
--- a/xiaobai-datahub/tests/test_revision_review.py
+++ b/xiaobai-datahub/tests/test_revision_review.py
@@ -151,7 +151,19 @@ class RevisionReviewTests(unittest.TestCase):
         return None if row is None else row["turnover_rate"]
 
     def _batch_ids(self, db: HubDB, day: str) -> set[str]:
-        return {str(row["batch_id"]) for row in db.fetchall("SELECT batch_id FROM batches WHERE trade_date=?", (day,))}
+        # HEL-478/HEL-463 integration: extended soft datasets (limit_events,
+        # dragon_tiger, sector_daily, popularity) publish on their own schedule
+        # inside the same tick and are outside the valuation review boundary.
+        # Scope assertions to the A-group + stocks boundary the review republishes.
+        boundary = ("daily", "valuation", "moneyflow", "auction", "stocks")
+        placeholders = ",".join("?" for _ in boundary)
+        return {
+            str(row["batch_id"])
+            for row in db.fetchall(
+                f"SELECT batch_id FROM batches WHERE trade_date=? AND dataset IN ({placeholders})",
+                (day, *boundary),
+            )
+        }
 
     def test_no_change_does_not_create_a_new_batch(self) -> None:
         tmp, db, transport, pipe, sched = make_revision_env()