Code and data for crosstalk text generation tasks, exploring whether large models and pre-trained language models can understand humor.
