Multiple-Choice question like CSQA dataset require select one choice from candidate choice. The previous approach was to input questions and train on the answers as labels. In the testing phase, the loss is calculated by taking the question as input and the individual options as temporary labels. The one with the smallest loss is selected as the predicted answer.
I find that there is no introduction to using openprompt on multiple choice tasks, so is there a pipeline of that type?
Thank you!
Multiple-Choice question like CSQA dataset require select one choice from candidate choice. The previous approach was to input questions and train on the answers as labels. In the testing phase, the loss is calculated by taking the question as input and the individual options as temporary labels. The one with the smallest loss is selected as the predicted answer. I find that there is no introduction to using openprompt on multiple choice tasks, so is there a pipeline of that type? Thank you!