diff --git a/mmeval/core/base_metric.py b/mmeval/core/base_metric.py
index 0b53cd88..ce9b5fbc 100644
--- a/mmeval/core/base_metric.py
+++ b/mmeval/core/base_metric.py
@@ -126,8 +126,8 @@ def compute(self, size: Optional[int] = None) -> Dict:
 
         # NOTE: Needs discussion or investigation @yancong at 9/13/2022, 4:20:48 PM  # noqa: E501
         # If the intermediate results stored in ``self._results`` are not
-        # correspond one-to-one with samples (e.g. confusion matrix), the size
-        # here may not work anymore.
+        # correspond one-to-one with samples (e.g. a total confusion matrix),
+        # the size here may not work anymore.
         if size is not None:
             collected_results = collected_results[:size]
 
@@ -145,8 +145,12 @@ def add(self, *args, **kwargs):
         """Override this method to add the intermediate results to
         ``self._results``.
 
-        For performance issue, what you add to the ``self._results`` should be
-        as simple as possible.
+        Note:
+            For performance issues, what you add to the ``self._results``
+            should be as simple as possible. But be aware that the intermediate
+            result stored in ``self._results`` should correspond one-to-one
+            with the samples, in that we need to remove the padded samples for
+            the most accurate result.
         """
 
     @abstractmethod