BUG: torch: expand_dims axis is keyword or positional argument

[ev-br]

Also add several skips to make the CI green:

• dask searchsorted scalars
• dask expand_dims tuple axes
• numpy < 2 isin scalars

---

Dask does not allow scalars as arguments to searchsorted:

$ ARRAY_API_TESTS_MODULE=array_api_compat.dask.array pytest array_api_tests/test_searching_functions.py::test_searchsorted_with_scalars --max-examples 500

...

    @given(data=st.data())
>   def test_searchsorted_with_scalars(data):
                   ^^^
...

        # call np.searchsorted for each pair of blocks in a and v
>       meta = np.searchsorted(a._meta, v._meta)
                                        ^^^^^^^
E       AttributeError: 'int' object has no attribute '_meta'
E
E       ========== FAILING CODE SNIPPET:
E       xp.searchsorted(dask.array<sort, shape=(1,), dtype=uint8, chunksize=(1,), chunktype=numpy.ndarray>, 0, sorter=None, **kw) with kw = {}
E       ====================

Just add an xfail for now.

cross-ref data-apis/array-api-tests#394

[Copilot]

Pull request overview

This PR adds an xfail entry for the test_searchsorted_with_scalars test in dask, which fails because dask does not support scalar arguments to searchsorted. The test will be skipped until proper support is implemented or the issue is resolved.

• Adds xfail for test_searchsorted_with_scalars under a new "2025.12 support" section

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.