How do you use pytest.mark.parametrize with complex nested data structures?

I need to parametrize a test with inputs that are dictionaries containing lists, some optional keys, and some keys that should cause expected failures. For simple cases `@pytest.mark.parametrize` is great. But for complex inputs I end up with unreadable one-liners or need `pytest.param(..., id="...")` everywhere. Do you use dataclasses or named tuples as parametrize inputs? Or do you load test cases from JSON/YAML files? Looking for a pattern that's readable, debuggable, and shows useful test IDs in the output.