Open skpig opened 2 years ago
Hi,
In the learning process of target task, we only train the cluster key and prompt key (corresponding to a source prompt), and the task query (corresponding to the target task). Through the above three parameters, we can compute the retrieved prompt ~p (Eq. 4). The mention '~p is frozen' in our paper maybe a little weird, just skip it.
Thanks for the remarkable work. I've got questions about the paper. It'll be appreciated if you could help me.