diff --git a/doc/whats_new/v1.3.rst b/doc/whats_new/v1.3.rst
index 76138d3bea483..2ba3a3494bc6a 100644
--- a/doc/whats_new/v1.3.rst
+++ b/doc/whats_new/v1.3.rst
@@ -274,6 +274,11 @@ Changelog
 - |Fix| :func:`datasets.fetch_openml` returns improved data types when
   `as_frame=True` and `parser="liac-arff"`. :pr:`26386` by `Thomas Fan`_.
 
+- |Fix| Following the ARFF specs, only the marker `"?"` is now considered as a missing
+  values when opening ARFF files fetched using :func:`datasets.fetch_openml` when using
+  the pandas parser. The parameter `read_csv_kwargs` allows to overwrite this behaviour.
+  :pr:`26551` by :user:`Guillaume Lemaitre <glemaitre>`.
+
 - |Enhancement| Allows to overwrite the parameters used to open the ARFF file using
   the parameter `read_csv_kwargs` in :func:`datasets.fetch_openml` when using the
   pandas parser.
diff --git a/sklearn/datasets/_arff_parser.py b/sklearn/datasets/_arff_parser.py
index 26a394b95f790..7b2faa4b67f4d 100644
--- a/sklearn/datasets/_arff_parser.py
+++ b/sklearn/datasets/_arff_parser.py
@@ -387,6 +387,7 @@ def _pandas_arff_parser(
         "header": None,
         "index_col": False,  # always force pandas to not use the first column as index
         "na_values": ["?"],  # missing values are represented by `?`
+        "keep_default_na": False,  # only `?` is a missing value given the ARFF specs
         "comment": "%",  # skip line starting by `%` since they are comments
         "quotechar": '"',  # delimiter to use for quoted strings
         "skipinitialspace": True,  # skip spaces after delimiter to follow ARFF specs