AI-Assisted Requirements Engineering: An Empirical Evaluation Relative to Expert Judgment — ThinkLLM