diff --git a/array_api_compat/torch/_aliases.py b/array_api_compat/torch/_aliases.py
index f2ec7b1..7af3f2a 100644
--- a/array_api_compat/torch/_aliases.py
+++ b/array_api_compat/torch/_aliases.py
@@ -30,6 +30,12 @@
     torch.int32,
     torch.int64,
 }
+try:
+    # torch >=2.3
+    _int_dtypes |= {torch.uint16, torch.uint32, torch.uint64}
+except AttributeError:
+    pass
+
 
 _array_api_dtypes = {
     torch.bool,
diff --git a/torch-xfails.txt b/torch-xfails.txt
index 44bef5a..5a7703c 100644
--- a/torch-xfails.txt
+++ b/torch-xfails.txt
@@ -8,24 +8,12 @@ array_api_tests/test_array_object.py::test_getitem
 array_api_tests/test_array_object.py::test_setitem
 # Masking doesn't suport 0 dimensions in the mask
 array_api_tests/test_array_object.py::test_getitem_masking
-# torch doesn't have uint dtypes other than uint8
-array_api_tests/test_array_object.py::test_scalar_casting[__int__(uint16)]
-array_api_tests/test_array_object.py::test_scalar_casting[__int__(uint32)]
-array_api_tests/test_array_object.py::test_scalar_casting[__int__(uint64)]
-array_api_tests/test_array_object.py::test_scalar_casting[__index__(uint16)]
-array_api_tests/test_array_object.py::test_scalar_casting[__index__(uint32)]
-array_api_tests/test_array_object.py::test_scalar_casting[__index__(uint64)]
 
 # Overflow error from large inputs
 array_api_tests/test_creation_functions.py::test_arange
 # pytorch linspace bug (should be fixed in torch 2.0)
 array_api_tests/test_creation_functions.py::test_linspace
 
-# torch doesn't have higher uint dtypes
-array_api_tests/test_data_type_functions.py::test_iinfo[uint16]
-array_api_tests/test_data_type_functions.py::test_iinfo[uint32]
-array_api_tests/test_data_type_functions.py::test_iinfo[uint64]
-
 # We cannot wrap the tensor object
 array_api_tests/test_has_names.py::test_has_names[array_method-__array_namespace__]
 array_api_tests/test_has_names.py::test_has_names[array_method-to_device]