Submitted preprint of IOLBENCH, a benchmark to test LLM performance on linguistic reasoning. arXiv:2501.04249