LLMs Exhibit Cognitive Limitations – BIS
When faced with logical puzzles that require reasoning about others' knowledge and counterfactual scenarios, large language models (LLMs) demonstrate a "distinctive and revealing pattern of...