[ { "dataset": "cola", "type": "task", "origin_acc": 86.4, "origin_prompt": "Check the grammar of the following sentence and indicate if it is 'Acceptable' or 'Unacceptable'.", "atk_acc": 78.4, "atk_prompt": "label the person of the following sentence and check if it is 'Acceptable' or 'Unacceptable'. ", "transfer_ori_acc": 81.5, "transfer_atk_acc": 73.5 }, { "dataset": "cola", "type": "role", "origin_acc": 86.6, "origin_prompt": "While performing grammar analysis, classify the grammar of the following sentence as 'acceptable' for correct grammar or 'unacceptable' for incorrect grammar.", "atk_acc": 78.2, "atk_prompt": "be doing grammar analysis, classify the rule of the following example as 'acceptable' for wrong spelling or 'unacceptable' for bad grammar. ", "transfer_ori_acc": 81.5, "transfer_atk_acc": 82.0 }, { "dataset": "mrpc", "type": "task", "origin_acc": 86.76, "origin_prompt": "Compare the following sentences and determine if they share the same semantic meaning by responding with 'equivalent' or 'not_equivalent'.", "atk_acc": 39.71, "atk_prompt": "like the following sentences and determine if they lack the same plural meaning by responding with 'equivalent' or 'not_equivalent'. ", "transfer_ori_acc": 74.5, "transfer_atk_acc": 67.5 }, { "dataset": "mrpc", "type": "role", "origin_acc": 88.97, "origin_prompt": "As a sentence similarity evaluator, analyze the provided sentences and indicate if their meanings are 'equivalent' or 'not_equivalent'.", "atk_acc": 81.37, "atk_prompt": "As a sentence compare evaluators, view the derived variables and indicate if their meanings are 'equivalent' or 'not_equivalent'. ", "transfer_ori_acc": 73.5, "transfer_atk_acc": 65.5 }, { "dataset": "qnli", "type": "task", "origin_acc": 94.0, "origin_prompt": "Analyze the context and question, and indicate if the context entails the answer by choosing 'entailment' or 'not_entailment'.", "atk_acc": 88.8, "atk_prompt": "Analyze the context and question, and indicate if the dictionary ennotes the context by avoiding 'entailment' or 'not_entailment'. ", "transfer_ori_acc": 80.5, "transfer_atk_acc": 72.5 }, { "dataset": "qnli", "type": "role", "origin_acc": 93.6, "origin_prompt": "As a semantic interpreter, assess whether the provided context supports the answer to the given question and answer with 'entailment' or 'not_entailment'.", "atk_acc": 83.5, "atk_prompt": "with a semantic presentation, evaluation why the provided contexts supports the answer to the given answer and questions with 'entailment' or 'not_entailment'. ", "transfer_ori_acc": 83.5, "transfer_atk_acc": 84.5 }, { "dataset": "qqp", "type": "task", "origin_acc": 88.7, "origin_prompt": "Examine the following expressions and tell me if they are alike in meaning by using 'equivalent' or 'not_equivalent'.", "atk_acc": 82.9, "atk_prompt": "Examine the past expressions and give me if they are common in meaning by putting 'equivalent' or 'not_equivalent'. ", "transfer_ori_acc": 82.0, "transfer_atk_acc": 82.0 }, { "dataset": "qqp", "type": "role", "origin_acc": 90.1, "origin_prompt": "As a tool for determining question equivalence, review the questions and categorize their similarity as either 'equivalent' or 'not_equivalent'.", "atk_acc": 74.3, "atk_prompt": "according a tool for identifying concept equivalence, do the questions and tcsconsider their compared as neither 'equivalent' or 'not_equivalent'. ", "transfer_ori_acc": 83.0, "transfer_atk_acc": 82.5 }, { "dataset": "rte", "type": "task", "origin_acc": 93.5, "origin_prompt": "Identify whether the given pair of sentences demonstrates entailment or not_entailment. Answer with 'entailment' or 'not_entailment'.", "atk_acc": 91.34, "atk_prompt": "a as the given pair of verb demonstrates entailment or not_entailment. question with 'entailment' or 'not_entailment'. ", "transfer_ori_acc": 82.0, "transfer_atk_acc": 81.5 }, { "dataset": "rte", "type": "role", "origin_acc": 93.86, "origin_prompt": "Working as an entailment classifier, identify whether the given pair of sentences displays entailment or not_entailment. Respond with 'entailment' or 'not_entailment'.", "atk_acc": 65.34, "atk_prompt": "acting as an entailment classindicator, identify neither the given pair of sentences displays entailment or not_entailment. reply with 'entailment' or 'not_entailment'. ", "transfer_ori_acc": 81.5, "transfer_atk_acc": 81.5 }, { "dataset": "sst2", "type": "task", "origin_acc": 96.79, "origin_prompt": "Determine the overall sentiment of this sentence, categorizing it as 'positive' or 'negative'.", "atk_acc": 92.2, "atk_prompt": "locate the associate emotion of this sentence, categorizing it as 'positive' or 'negative'. ", "transfer_ori_acc": 95.0, "transfer_atk_acc": 94.0 }, { "dataset": "sst2", "type": "role", "origin_acc": 96.56, "origin_prompt": "As a sentiment classifier, determine whether the following text is 'positive' or 'negative'.", "atk_acc": 95.3, "atk_prompt": "in a sentiment classifier, determine where the over words is 'positive' or 'negative'. ", "transfer_ori_acc": 97.5, "transfer_atk_acc": 93.0 }, { "dataset": "wnli", "type": "task", "origin_acc": 77.46, "origin_prompt": "Assess the connection between the following sentences and classify it as 'entailment' or 'not_entailment'.", "atk_acc": 74.65, "atk_prompt": "a the links between the following sentences and classify it as 'entailment' or 'not_entailment'. ", "transfer_ori_acc": 66.19718309859155, "transfer_atk_acc": 61.97183098591549 }, { "dataset": "wnli", "type": "role", "origin_acc": 77.46, "origin_prompt": "Working as an entailment classifier, identify whether the given pair of sentences displays entailment or not_entailment. Respond with 'entailment' or 'not_entailment'.", "atk_acc": 66.2, "atk_prompt": "Working as an entailment classauxiliary, identify but the verbal a of nouns displays entailment or not_entailment. Respond with 'entailment' or 'not_entailment'. ", "transfer_ori_acc": 59.154929577464785, "transfer_atk_acc": 60.56338028169014 }, { "dataset": "mmlu", "type": "task", "origin_acc": 53.9, "origin_prompt": "Below are multiple-choice question concerning {}. Indicate your response with 'A', 'B', 'C', or 'D'.", "atk_acc": 51.24, "atk_prompt": "Below are multiple-guessing guess concerning {}. begin your response with 'A', 'B', 'C', or 'D'. ", "transfer_ori_acc": 65.5, "transfer_atk_acc": 63.5 }, { "dataset": "mmlu", "type": "role", "origin_acc": 54.08, "origin_prompt": "Considering your familiarity with {}, attend to the following multiple-choice question by picking 'A', 'B', 'C', or 'D'.", "atk_acc": 51.77, "atk_prompt": "considering your familiarity with {}, order to the following multiple-picked question by starting 'A', 'B', 'C', or 'D'.", "transfer_ori_acc": 66.0, "transfer_atk_acc": 63.0 }, { "dataset": "squad_v2", "type": "task", "origin_acc": 91.05, "origin_prompt": "Identify the most relevant answer from the context. If it's not possible to find an answer, state 'unanswerable'.", "atk_acc": 44.5, "atk_prompt": "Identify the most irrelevant answer from the context. If it's not possible to find an statement, begin 'unanswerable'.", "transfer_ori_acc": 72.35462148030798, "transfer_atk_acc": 69.05824840893493 }, { "dataset": "squad_v2", "type": "role", "origin_acc": 90.05, "origin_prompt": "As an authority on the context, locate the most accurate answer. If the context doesn't contain the answer, mention 'unanswerable'.", "atk_acc": 44.5, "atk_prompt": "and an authority on the context, designate the most inaccurate answer. find the context doesn't contain the answer, identify 'unanswerable'.", "transfer_ori_acc": 69.85316629298323, "transfer_atk_acc": 68.20230712799365 }, { "dataset": "iwslt", "type": "task", "origin_acc": 0.27, "origin_prompt": "Accurately translate the sentence from {} to {}, ensuring the meaning remains intact.", "atk_acc": 0.23, "atk_prompt": "i translate the sentence from {} to {}, assumed the meaning remains intact. ", "transfer_ori_acc": 0.2387938001782664, "transfer_atk_acc": 0.2341002610936615 }, { "dataset": "iwslt", "type": "role", "origin_acc": 0.27, "origin_prompt": "Taking on the role of a certified translator, perform a translation of the given sentence from {} to {}.", "atk_acc": 0.23, "atk_prompt": "she on the role of a certified translation, and a translation of the given sentence from {} to {}. ", "transfer_ori_acc": 0.24289990367509515, "transfer_atk_acc": 0.24881534278280654 }, { "dataset": "un_multi", "type": "task", "origin_acc": 0.36, "origin_prompt": "Perform a machine translation for the provided sentence, changing it from {} to {}.", "atk_acc": 0.35, "atk_prompt": "perform a machine translation for the displayed sentence, export it from {} to {}. ", "transfer_ori_acc": 0.3515593768654444, "transfer_atk_acc": 0.35115524227765527 }, { "dataset": "un_multi", "type": "role", "origin_acc": 0.36, "origin_prompt": "As a skilled linguist, please convert the following sentence from {} to {}.", "atk_acc": 0.35, "atk_prompt": "d a skilled linguist, right example the following sentence from {} to {}. ", "transfer_ori_acc": 0.3506651200422635, "transfer_atk_acc": 0.3504251387283558 }, { "dataset": "math", "type": "task", "origin_acc": 16.88, "origin_prompt": "Figure out the solution for the following mathematical task with {}.", "atk_acc": 13.75, "atk_prompt": "Figure out the solution for the following arithmetic task with {}. ", "transfer_ori_acc": 21.25, "transfer_atk_acc": 21.875 }, { "dataset": "math", "type": "role", "origin_acc": 16.25, "origin_prompt": "Act as a math expert and compute the result of the following mathematical task about {}.", "atk_acc": 12.5, "atk_prompt": "Act as a math team and correct the result of the following mathematical task about {}. ", "transfer_ori_acc": 21.875, "transfer_atk_acc": 15.625 } ]