diff --git a/olmo/eval/downstream.py b/olmo/eval/downstream.py
index a53c875c1..fddd2c3f8 100644
--- a/olmo/eval/downstream.py
+++ b/olmo/eval/downstream.py
@@ -1866,6 +1866,7 @@ def doc_to_label(self, doc) -> int:
 
 # This standardizes the metrics we should eval for the ladder.
 # Train and test sets are added when applicable.
+# No subsampling happens in these sets.
 label_to_task_map_new = {
     "arc_challenge_train_rc_5shot": (
         OEEvalTask,