Study meta-learning stabilized prompting with PLMs for few-shot learning on dialogue
History (list) - list of past utterances, including current utterance - each utterance is a [speaker + text] Target (dictionary) - keys include - domain - slot - value - global_id