Measuring and mitigating overreliance to build human-compatible AI
arXiv:2509.08010v2 Announce Type: replace-cross Abstract: Large language models (LLMs) distinguish themselves from previous technologies by functioning as collaborative ``thought partners,'' capable o