diff --git a/tests/conftest.py b/tests/conftest.py
index af1f5e529d..0b7dec95bd 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -11,7 +11,13 @@
 
 import narwhals as nw
 from narwhals._utils import Implementation, generate_temporary_column_name
-from tests.utils import ID_PANDAS_LIKE, PANDAS_VERSION, pyspark_session, sqlframe_session
+from tests.utils import (
+    ID_PANDAS_LIKE,
+    PANDAS_VERSION,
+    any_integer_like_floats,
+    pyspark_session,
+    sqlframe_session,
+)
 
 if TYPE_CHECKING:
     from collections.abc import Callable, Sequence
@@ -102,14 +108,29 @@ def pandas_constructor(obj: Data) -> pd.DataFrame:
 def pandas_nullable_constructor(obj: Data) -> pd.DataFrame:
     import pandas as pd
 
-    return pd.DataFrame(obj).convert_dtypes(dtype_backend="numpy_nullable")
+    return pd.DataFrame(
+        {
+            k: pd.Series(v).convert_dtypes(
+                convert_integer=not any_integer_like_floats(v),
+                dtype_backend="numpy_nullable",
+            )
+            for k, v in obj.items()
+        }
+    )
 
 
 def pandas_pyarrow_constructor(obj: Data) -> pd.DataFrame:
     pytest.importorskip("pyarrow")
     import pandas as pd
 
-    return pd.DataFrame(obj).convert_dtypes(dtype_backend="pyarrow")
+    return pd.DataFrame(
+        {
+            k: pd.Series(v).convert_dtypes(
+                convert_integer=not any_integer_like_floats(v), dtype_backend="pyarrow"
+            )
+            for k, v in obj.items()
+        }
+    )
 
 
 def modin_constructor(obj: Data) -> IntoDataFrame:  # pragma: no cover
@@ -124,7 +145,14 @@ def modin_pyarrow_constructor(obj: Data) -> IntoDataFrame:  # pragma: no cover
     import modin.pandas as mpd
     import pandas as pd
 
-    df = mpd.DataFrame(pd.DataFrame(obj)).convert_dtypes(dtype_backend="pyarrow")
+    df = mpd.DataFrame(
+        {
+            k: pd.Series(v).convert_dtypes(
+                convert_integer=not any_integer_like_floats(v), dtype_backend="pyarrow"
+            )
+            for k, v in obj.items()
+        }
+    )
     return cast("IntoDataFrame", df)
 
 
diff --git a/tests/frame/to_pandas_test.py b/tests/frame/to_pandas_test.py
index 473b685c19..53329b118d 100644
--- a/tests/frame/to_pandas_test.py
+++ b/tests/frame/to_pandas_test.py
@@ -25,7 +25,9 @@ def test_convert_pandas(constructor_eager: ConstructorEager) -> None:
     if constructor_eager.__name__.startswith("pandas"):
         expected = cast("pd.DataFrame", constructor_eager(data))
     elif "modin_pyarrow" in str(constructor_eager):
-        expected = pd.DataFrame(data).convert_dtypes(dtype_backend="pyarrow")
+        from modin.pandas.io import to_pandas
+
+        expected = to_pandas(constructor_eager(data))
     else:
         expected = pd.DataFrame(data)
 
diff --git a/tests/utils.py b/tests/utils.py
index 4e4b4e133f..4f10b7944e 100644
--- a/tests/utils.py
+++ b/tests/utils.py
@@ -276,3 +276,8 @@ def xfail_if_pyspark_connect(  # pragma: no cover
 ) -> None:
     if is_pyspark_connect(constructor):
         request.applymarker(pytest.mark.xfail(reason=reason))
+
+
+def any_integer_like_floats(values: list[Any]) -> bool:
+    """Return True if any value is a float that represents an integer (e.g. 1.0, 2.0)."""
+    return any(isinstance(v, float) and v.is_integer() for v in values)
diff --git a/tests/utils_test.py b/tests/utils_test.py
index e18e3041ab..2654207f73 100644
--- a/tests/utils_test.py
+++ b/tests/utils_test.py
@@ -3,6 +3,7 @@
 import re
 import string
 from dataclasses import dataclass
+from datetime import date, datetime
 from itertools import chain
 from typing import TYPE_CHECKING, Any, ClassVar, Protocol, cast
 
@@ -19,7 +20,7 @@
     parse_version,
     requires,
 )
-from tests.utils import get_module_version_as_tuple
+from tests.utils import any_integer_like_floats, get_module_version_as_tuple
 
 pytest.importorskip("pandas")
 import pandas as pd
@@ -626,3 +627,34 @@ def fn() -> Iterator[str]:
     assert next(iter(deferred_1)) == "h"
     assert list(deferred_1) == list("hello")
     assert "".join(chain(deferred_1, deferred_2)) == "helloHELLO"
+
+
+@pytest.mark.parametrize(
+    ("values", "expected"),
+    [
+        pytest.param([1, 2, 3], False, id="all_ints"),
+        pytest.param([1.5, 2.2, 3.1], False, id="no_integer_like_floats"),
+        pytest.param([1.0, 2.0, 3.5], True, id="some_integer_like_floats"),
+        pytest.param([1.0, 2.0, 3.0], True, id="all_integer_like_floats"),
+        pytest.param([1, 2.0, 3], True, id="mixed_int_and_float"),
+        pytest.param([], False, id="empty_list"),
+        pytest.param([-1.0, -2.0], True, id="negative_integer_like_floats"),
+        pytest.param([1.0, 2.5, 3.0, 4.1], True, id="mixed_floats"),
+        # mixed non-numeric types
+        pytest.param(["a", "b", "c"], False, id="all_strings"),
+        pytest.param([1.0, "a", 2.0], True, id="floats_and_strings"),
+        pytest.param([1, "a", 2], False, id="ints_and_strings_no_floats"),
+        # datetime / date objects
+        pytest.param([date(2024, 1, 1), date(2024, 1, 2)], False, id="all_dates"),
+        pytest.param(
+            [datetime(2024, 1, 1), 1.0, datetime(2024, 1, 2)],
+            True,
+            id="floats_and_datetimes",
+        ),
+        # None / null handling
+        pytest.param([None, None], False, id="all_none"),
+        pytest.param([1.0, None, 2.0], True, id="floats_and_none"),
+    ],
+)
+def test_any_integer_like_floats(values: list[Any], expected: bool) -> None:  # noqa: FBT001
+    assert any_integer_like_floats(values) is expected