behavioral simulation

Can Large Language Models Grasp Sequential Structure? A Controlled Examination of Inference and Generation

Controlled arXiv study shows LLMs can recognize sequential strategies but often fail to simulate them, exposin…