AI的“一招鲜”：如何只训练一次，就让模型十项全能？

[CL] You Only Fine-tune Once：Many-Shot In-Context Fine-Tuning for Large Language Model

[Google & University of Florida]

arxiv.org