A System Phd Would Not Dream of a Girl from the Theory Department

In the first autumn after the release of GPT-4, Cheng Xubai began to study how to make LLM run faster and cheaper; Xu Qingyao has been asking: If the information itself is selected, who are we evaluating? A PhD student in systems and a girl majoring in theory, starting from disagreements in the seminar, gradually shared the same document, the same journey home, and more and more time that could not be explained by just "discussing work". Papers will be rejected, experiments will fail, and internships will come with time. But in those years when everything demanded answers as quickly as possible, they learned to ask the right questions first.