Comprehension Without Competence: Architectural Limits of LLMs in Symbolic Computation and Reasoning

Open in new window