PromptSource

Prompts from PromptSource

• squad: answer_given_context_and_question

  text

  PromptSource template for the squad dataset.

  by BigScience Workshop

• social_i_qa: Show choices and generate index

  text

  PromptSource template for the social_i_qa dataset.

  by BigScience Workshop

• social_i_qa: I was wondering

  text

  PromptSource template for the social_i_qa dataset.

  by BigScience Workshop

• social_i_qa: Generate answer

  text

  PromptSource template for the social_i_qa dataset.

  by BigScience Workshop

• snli: based on the previous passage

  text

  Adapted from the BoolQ prompts in Schick & Schütze 2021.

  by BigScience Workshop

• snli: MNLI crowdsource

  text

  Adapted from Williams et al. 2018's instructions to crowdsourcing workers.

  by BigScience Workshop

• snli: GPT-3 style

  text

  Same as reported in Figure G7 of the GPT-3 paper, except that there is no task identifying tokens like "anli R1: ".

  by BigScience Workshop

• sms_spam: spam_label

  text

  PromptSource template for the sms_spam dataset.

  by BigScience Workshop

• sms_spam: marked as

  text

  PromptSource template for the sms_spam dataset.

  by BigScience Workshop

• sick: B entails A?

  text

  PromptSource template for the sick dataset.

  by BigScience Workshop

• sick: A entails_neutral_contradict B?

  text

  PromptSource template for the sick dataset.

  by BigScience Workshop

• sick: A entails B?

  text

  PromptSource template for the sick dataset.

  by BigScience Workshop

• sem_eval_2014_task_1: related_rate

  text

  PromptSource template for the sem_eval_2014_task_1 dataset.

  by BigScience Workshop

• sem_eval_2014_task_1: premise_imply_hypothesis

  text

  PromptSource template for the sem_eval_2014_task_1 dataset.

  by BigScience Workshop

• sem_eval_2014_task_1: premise_agree_hypothesis

  text

  PromptSource template for the sem_eval_2014_task_1 dataset.

  by BigScience Workshop

• sem_eval_2010_task_8: semantic relations without options

  text

  relationship between two nomimals; official metric: macro-averaged F1-score

  by BigScience Workshop

• sem_eval_2010_task_8: semantic relations nominials without options

  text

  mention e1,e2 after; macro-averaged F1-score official metric

  by BigScience Workshop

• scitail/tsv_format: …does the previous passage support the claim that

  text

  PromptSource template for the scitail/tsv_format dataset.

  by BigScience Workshop

• scitail/tsv_format: Suppose… Can we infer that…

  text

  PromptSource template for the scitail/tsv_format dataset.

  by BigScience Workshop

• sciq: Direct Question

  text

  PromptSource template for the sciq dataset.

  by BigScience Workshop

• scientific_papers/pubmed: generate_first_three_sentences_of_an_article_from_abstract

  text

  Given the abstract, generate first three sentences of the article.

  by BigScience Workshop

• scientific_papers/pubmed: generate_first_line_from_abstract

  text

  Given the abstract, generate the first line of the article.

  by BigScience Workshop

• scientific_papers/pubmed: generate_first_100_words_for_article

  text

  Given the abstract, generate the first 100 words of the article.

  by BigScience Workshop

• scientific_papers/arxiv: generate_first_three_sentences_of_an_article_from_abstract

  text

  Given the abstract, generate first three sentences of the article.

  by BigScience Workshop

• scientific_papers/arxiv: generate_first_line_from_abstract

  text

  Given the abstract, generate the first line of the article.

  by BigScience Workshop

• scientific_papers/arxiv: generate_first_100_words_for_article

  text

  Given the abstract, generate the first 100 words of the article.

  by BigScience Workshop

• samsum: Sum up the following dialogue

  text

  PromptSource template for the samsum dataset.

  by BigScience Workshop

• samsum: Given the above dialogue write a summary

  text

  PromptSource template for the samsum dataset.

  by BigScience Workshop

• samsum: Generate a summary for this dialogue

  text

  PromptSource template for the samsum dataset.

  by BigScience Workshop

• rotten_tomatoes: Reviewer Enjoyment

  text

  PromptSource template for the rotten_tomatoes dataset.

  by BigScience Workshop