Large language models (LLMs) often make discrete decisions, such as flagging phishing or escalating to a human. System-1 models, such as Jev and the OpenAI D…
机构:UT Austin
来源:arXiv 2610.21244 | AI4Papers 论文推荐平台