diff --git a/ludwig/trainers/trainer_llm.py b/ludwig/trainers/trainer_llm.py
index 9540663288b..8d1c28a6b1c 100644
--- a/ludwig/trainers/trainer_llm.py
+++ b/ludwig/trainers/trainer_llm.py
@@ -205,6 +205,9 @@ def tune_batch_size(
         snapshot_weights: bool = True,
         on_best_batch_size_updated: Optional[Callable[[int, float, int], None]] = None,
     ) -> int:
+        # TODO: Implement batch size tuning for LLM, currently just returns the default batch size
+        # Compared to ECD, this just requires forward passes till we OOM.
+        # https://github.com/ludwig-ai/ludwig/issues/3525
         return MINIMUM_BATCH_SIZE
 
     @property