Многоуровневая настройка на основе подсказок

AbstractMultitask prompted finetuning (MTF) has been shown to help large language models generalize to new tasks in a zero-shot setting, but so far explorations of MTF have focused on English …...