Watson & Holmes: A Naturalistic Benchmark for Comparing Human and LLM Reasoning

Thatchawin Leelawat, Lewis D Griffin

Existing benchmarks for AI reasoning provide limited insight into how closely these capabilities resemble human reasoning in naturalistic contexts. We present an adaptation of the…