PromptSource

Prompts from PromptSource

• app_reviews: convert_to_star_rating

  text

  Given the review, generate a star rating.

  by BigScience Workshop

• app_reviews: convert_to_rating

  text

  Convert review to rating

  by BigScience Workshop

• anli: based on the previous passage

  text

  Adapted from the BoolQ prompts in Schick & Schütze 2021.

  by BigScience Workshop

• anli: MNLI crowdsource

  text

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

  by BigScience Workshop

• anli: 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

• ambig_qa/light: answer_prediction_all_answers_interrogative

  text

  PromptSource template for the ambig_qa/light dataset.

  by BigScience Workshop

• ambig_qa/light: answer_prediction_all_answers_affirmative

  text

  PromptSource template for the ambig_qa/light dataset.

  by BigScience Workshop

• amazon_us_reviews/Wireless_v1_00: Generate review headline based on review body

  text

  Generate review headline based on review body

  by BigScience Workshop

• amazon_us_reviews/Wireless_v1_00: Generate review headline based on rating

  text

  Generate review headline based on rating.

  by BigScience Workshop

• amazon_us_reviews/Wireless_v1_00: Generate review based on rating and category

  text

  Generate review based on rating and category

  by BigScience Workshop

• amazon_reviews_multi/en: prompt_review_to_star

  text

  Rating based on review body

  by BigScience Workshop

• amazon_reviews_multi/en: prompt_body_title_to_star

  text

  Rating based on review body,title

  by BigScience Workshop

• amazon_reviews_multi/en: prompt_body_title_category_to_star

  text

  Rating based on review body, title, category

  by BigScience Workshop

• amazon_polarity: Is_this_review_negative

  text

  PromptSource template for the amazon_polarity dataset.

  by BigScience Workshop

• amazon_polarity: Is_this_review

  text

  PromptSource template for the amazon_polarity dataset.

  by BigScience Workshop

• amazon_polarity: Is_this_product_review_positive

  text

  PromptSource template for the amazon_polarity dataset.

  by BigScience Workshop

• ag_news: which_section

  text

  PromptSource template for the ag_news dataset.

  by BigScience Workshop

• ag_news: classify_question_first

  text

  PromptSource template for the ag_news dataset.

  by BigScience Workshop

• ag_news: classify

  text

  PromptSource template for the ag_news dataset.

  by BigScience Workshop

• aeslc: the_text_below

  text

  PromptSource template for the aeslc dataset.

  by BigScience Workshop

• aeslc: generate_subject_line

  text

  Instruct to generate

  by BigScience Workshop

• aeslc: generate_subject

  text

  Instruct to generate

  by BigScience Workshop

• adversarial_qa/droberta: generate_question

  text

  Input: Context, Output: Question (generate a question)

  by BigScience Workshop

• adversarial_qa/dbidaf: generate_question

  text

  Input: Context, Output: Question (generate a question)

  by BigScience Workshop

• adversarial_qa/dbert: generate_question

  text

  Input: Context, Output: Question (generate a question)

  by BigScience Workshop

• adversarial_qa/adversarialQA: generate_question

  text

  Input: Context, Output: Question (generate a question)

  by BigScience Workshop

• ade_corpus_v2/Ade_corpus_v2_drug_dosage_relation: find-drug-and-dosage

  text

  PromptSource template for the ade_corpus_v2/Ade_corpus_v2_drug_dosage_relation dataset.

  by BigScience Workshop

• ade_corpus_v2/Ade_corpus_v2_drug_dosage_relation: find-drug

  text

  PromptSource template for the ade_corpus_v2/Ade_corpus_v2_drug_dosage_relation dataset.

  by BigScience Workshop

• ade_corpus_v2/Ade_corpus_v2_drug_dosage_relation: find-dosage

  text

  PromptSource template for the ade_corpus_v2/Ade_corpus_v2_drug_dosage_relation dataset.

  by BigScience Workshop

• ade_corpus_v2/Ade_corpus_v2_drug_ade_relation: find-drug-and-effect-two-questions

  text

  PromptSource template for the ade_corpus_v2/Ade_corpus_v2_drug_ade_relation dataset.

  by BigScience Workshop