Search is not available for this dataset
conversations
listlengths
2
2
[ { "from": "human", "value": "In this task, you're shown a three-part story, with a beginning, middle, and ending. Your job is to slightly modify the middle part, so that the whole story becomes unlikely, improbable, or inconsistent. Generated sentences must minimally alter the given middle, with at most 4 new words added/existing words removed. Your sentence should be grammatically and syntactically correct. Also stick to the context of the given story. For example, if the story talks about \"doctors\", you can talk about \"health\" or \"diagnosis\", but don't mention \"aliens\". Avoid introducing any extra/irrelevant information, and use names instead of pronouns (e.g. he / she) wherever possible. Avoid simply negating the original middle, for example: \"She likes cakes.\" is changed to \"She doesn't like cakes.\"\nQ: Beginning: I have the greatest wife in the world! Middle: My wife gave me a fantastic gift on my birthday. Ending: I'll have to get her a gift as a thank you!\nA: " }, { "from": "gpt", "value": "My wife gave me a mediocre gift on my birthday." } ]
[ { "from": "human", "value": "Q: In this task, you are given inputs i and A, where i is an integer and A is a list. You need to list all the elements of A preceding the last i elements. i will always have a value less than the length of A.\n2, ['d', 'f', '1767', 'p', 'K', 'V', 'A']\nA:" }, { "from": "gpt", "value": "d, f, 1767, p, K" } ]
[ { "from": "human", "value": "Detailed Instructions: In this task, you're given a pair of sentences, sentence 1 and sentence 2. Your job is to classify the genre of the sentences from the 10 options displayed. Indicate your answer by entering the number of the choice as 1-10.\nSee one example below:\nProblem: Sentence 1: Next to the MGM Grand you will find M and M World. Sentence 2: The candy has many fans who love its attractions. Choices: 1. FACE-TO-FACE, 2. GOVERNMENT, 3. LETTERS, 4. 9/11, 5. SLATE, 6. TELEPHONE, 7. TRAVEL, 8. VERBATIM, 9. OUP, 10. FICTION.\nSolution: 1\nExplanation: The places are specifically pointed out and referred to as attractions.\n\nProblem: Sentence 1: He'd had a series of victories in various types of powered vehicles, including a victory on land at Daytona Beach in Florida, before coming to Coniston in 1967 to attempt to break the world record in speed boating. Sentence 2: He attempted to break the speed boating world record in 1967 but failed. Choices: 1. FACE-TO-FACE, 2. GOVERNMENT, 3. LETTERS, 4. 9/11, 5. SLATE, 6. TELEPHONE, 7. TRAVEL, 8. VERBATIM, 9. OUP, 10. FICTION.\nSolution:" }, { "from": "gpt", "value": "6" } ]
[ { "from": "human", "value": "You will be given a definition of a task first, then some input of the task.\nIn this task, you are given news articles and you need to generate a highlight, i.e., a short summary, with a maximum length of 10 lines.\n\n(CNN) -- On Monday, news broke that about 200 girls had been kidnapped from their school in Chibok, in the northeastern state of Borno -- a region at the center of Nigeria's five-year terrorist insurgency. The very next day, the Nigerian military announced that all but nine of the girls had been rescued. This turned out to be untrue. The school's principal and the girls' parents complained that the girls were still missing. In August last year, a military spokesman announced the death of Abubakar Shekau, the leader of Islamist extremist group Boko Haram, at the hands of the military. But like the news of the release of the schoolgirls, it proved to be fiction. 'Misleading sources' Incidents like this have come to shape the way the military is perceived in the wildly unpredictable battle against Boko Haram. Posts on the military's social media accounts regularly boast of \"smoking out\" or ambushing terrorists or recovering weapons -- often written in way that brings to mind Iraq's infamous former information minister under Saddam Hussein, dubbed \"Comical Ali,\" who claimed coalition forces were in retreat even as American tanks rolled almost unchallenged across the country in 2003. Like him they've also ended up pinning the blame for inaccurate reporting on \"misleading\" sources. Even when there is truth to its narratives -- and there have been major successes, including a crackdown that started in early 2013 that killed several top Boko Harem commanders and driven others across the border into Chad, Niger and Cameroon -- the triumphalism seems odd when juxtaposed with the harsh reality of events like the Chibok abduction, or the one at another government school in neighboring Yobe State in February, in which more than fifty schoolboys were murdered in their dormitories, with nary a soldier in sight until several hours later. Perhaps the military resorts to this impulsively buoyant tone because it believes it cannot afford to sound anything otherwise. Or perhaps it's simply because it can get away with it -- because the Nigerian authorities have a long and remarkable history of getting away with anything they say. That tenuous relationship with fact makes it relatively easy for senior government officials to publicly dispute government finances to the tune of billions of dollars; and for an assortment of newspaper headlines to display wildly differing casualty figures the morning after a bombing incident. Death 'cheap and plentiful' And then there's the scale of Nigeria's tragedies. For a country that is not at war, death is cheap and plentiful. So cheap and so sweeping in its audacity that Nigerians readily make jokes about it. That might help explain the trademark blunted edge of Nigerian outrage. If it happens often enough, the mind is soon inured, and eagerly accepting of the sense of resignation that might offer the best protection against the emotional impact of the next cycle of negative breaking news. All of the above combine to create the context in which the Nigerian military -- wielding political power for 29 of Nigeria's first 39 years after independence -- has learned to operate. Not since the civil war, almost 50 years ago, has it been tested this much. The closest it got were the lengthy tours of duty in Liberia and Sierra Leone, in which it played a prominent role at the head of the West African ECOMOG Force, intended more as a \"peacekeeping\" unit than a combat force. Indeed what we are seeing may be evidence of its struggle to adapt to new rules of engagement, fighting an enemy driven by convictions much deeper than those displayed by the pro-democracy activists and diamond-obsessed rebels it contended with in the 1990s; possessing access to sophisticated weapons, and operating in terrain far better suited to insurgents than conventional armies. Abuse claims . The increasing militarization of the troubled zones has since spurred accusations of human rights abuses, from local and international observers. An International Crisis Group report from April 2014 has called for an end to the use of \"heavy-handed military and police methods that risk pushing yet more restless, jobless and frustrated youths into violence and extremism.\" Communication strategies also require overhaul. In the age of social media, the military needs to realize that propaganda is now a lot more likely to be found out and discredited. News reports suggest that the military, long hampered by aging hardware, is now acquiring new weapons and equipment. That's heartwarming. In a country where institutional graft is the rule and not the exception, it is crucial to ensure that the military budgets are spent to boost the military's capability, and troops' morale -- and not pocketed by bigwigs. International cooperation also needs to be stepped up; and it does seem that the government is now more willing than ever to work with Europe and America. Nigerians have long been wary of allowing the American military the sort of foothold it has in countries like Yemen and Pakistan, but there's certainly room for more intense cooperation that does not involve abdicating total control. Finally the military will need to prepare to adapt itself to the reality of the government's planned shift to a \"soft\" counter-terrorism strategy, embodied in a document unveiled by National Security Adviser, Sambo Dasuki, a retired Army Colonel, in March. Amid the backlash it has faced recently, the beleaguered military can count on the support of a growing number of Nigerians, who think that it is being under appreciated for the work it is doing. Just this week a \"Support The Nigerian Military\" page launched on Facebook, in honor of \"our military men and women on the field who risk their lives daily to keep us safe.\"\nOutput:" }, { "from": "gpt", "value": "Ogunlesi: Nigeria's military has issued incorrect statements in fight against extremists .\nOften blame their inaccurate reporting on \"misleading sources,\" he says .\nOgunlesi says the military may be trying to paint a constantly positive picture .\nOgunlesi says it has had some successes in its five-year war with Islamist extremists ." } ]
[ { "from": "human", "value": "In this task, you need to replace a letter in the sentence with another given letter.\n\nInput: Consider Input: Sentence: 'children trying little hamburgers from a tray at school'. Replace the letter 'y' with 'l' in the sentence.\n\nOutput: children trling little hamburgers from a tral at school\n\n\nInput: Consider Input: Sentence: 'a couch, footrest, tv and other furniture are arranged in a living room'. Replace the letter 'o' with 'g' in the sentence.\n\nOutput: a cguch, fggtrest, tv and gther furniture are arranged in a living rggm\n\n\nInput: Consider Input: Sentence: 'graffiti on a train compartment and a metal container placed nearby'. Replace the letter 'i' with 'f' in the sentence.\n\n" }, { "from": "gpt", "value": "Output: graffftf on a trafn compartment and a metal contafner placed nearby\n\n\n" } ]
[ { "from": "human", "value": "Definition: In this task, you are given a question and an answer. Answer \"Yes\" if the given answer correctly answers the question, otherwise answer \"No\".\nInput: how big did girl scout cookie boxes used to be, Answer: A mound of Girl Scout cookies.\nOutput:" }, { "from": "gpt", "value": "No" } ]
[ { "from": "human", "value": "Definition: In this task, you will be shown an extract from a movie plot. You need to read the extract and create questions that can be answered from the extract. Once you create a question you also need to provide the correct answer for it. Write your answer like \"Question: your question, Answer: your answer\". The questions should be well-formed and grammatically correct. The questions should be completely answerable from the given passage and should not require any external knowledge. Subjective questions are not allowed. Short answers containing words that are present in the passage are preferred. Create questions that result in factoid answers. A simple rule of thumb to decide whether a question is a factoid or not is to see if two different people with average reading/comprehension skills would come up with the same answer after reading the passage.\nInput: Mural featuring Frederick Douglass in Belfast, Northern Ireland.Frederick Douglass and the White Negro is a documentary telling the story of ex-slave, abolitionist, writer and politician Frederick Douglass and his escape to Ireland from America in the 1840s. The film follows Douglass' life from slavery as a young man through to his time in Ireland where he befriended Daniel O'Connell, toured the country spreading the message of abolition and was treated as a human being for the first time by white people. His arrival in Ireland coincided with the Great Famine and he witnessed white people in what he considered to be a worse state than his fellow African Americans back in the US. The film follows Douglass back to America where he buys his freedom with funds raised in Ireland and Britain. Fellow passengers on his return journey include the Irish escaping the famine who arrive in their millions and would go on to play a major role in the New York Draft Riot of 1863 which Douglass could only despair over. The film examines (with contributions from the author of How The Irish Became White Noel Ignatiev amongst others) the turbulent relationship between African Americans and Irish Americans during the American Civil War, what drew them together and what drove them apart and how this would shape the America of the twentieth century and beyond.\nOutput:" }, { "from": "gpt", "value": "Question: How did Douglas buy freedom in America?, Answer: With funds raised in Ireland." } ]
[ { "from": "human", "value": "instruction:\nIn this task, you are given a text from a social media post. Your task is to classify the given post into two categories: 1) yes if the given post is sexually offensive, 2) no, otherwise. Emphasis on sexually offensive or any lewd reference. Generate label 'no' for offensive posts that do not reference sexually explicit content. Warning: the examples and instances may contain offensive language.\nquestion:\nSo my sister's boyfriend wants to fight me He walked in on me fucking his girl\nanswer:\nYes\n\n\nquestion:\nHello new followers. I am anime loving nigger who is too poor to afford a laptop, resulting in being a mw3 theater player for Darth.\nanswer:\nNo\n\n\nquestion:\nNo half-breed spic has any right to say anything to any Anglo ever, and should be killed for doing so.\nanswer:\n" }, { "from": "gpt", "value": "No\n\n\n" } ]
[ { "from": "human", "value": "Instructions: You are given a short text as a title. Your task is to generate a poem as output that is related to the given title and should feel like written by kids. The output should be a run-on sentence (two or more complete sentences connected without any punctuation). The poem should not be too long or too complex, because it should feel like it is written by younger person without high level of literature education.\nInput: Idiot\nOutput:" }, { "from": "gpt", "value": "there be an idiot name kyle who think dump girl be worthwile until he get slap and everyone clap now heartbreaking be n't his style" } ]
[ { "from": "human", "value": "In this task, you will be given two sentences sentence1 and sentence2. You should decide whether the second sentence is entailed(agreed) by the first sentence. If it does entail, answer \"yes\", else answer \"no\".\nOne example is below.\nQ: sentence1:region is nation in location\nsentence2:region is country in location\nA: yes\nRationale: A nation is a community of people of similar characteristics/descent with a common government. A country is a region sharing a common government. Hence, a nation can be defined as a country.\nQ: sentence1:award_winner is trailing employer sentence1:award_winner is leading employer \nA:" }, { "from": "gpt", "value": "no" } ]
[ { "from": "human", "value": "In this task, you're given a review from Amazon and rating for the product on a scale of 1-5 based on the review. The rating means 1: extremely poor, 2: poor, 3: neutral, 4: good, 5: extremely good. Your task is to generate whether the rating matches the review or not. Answer with \"True\" if the sentence belongs to that section, otherwise answer with \"False\"\n\nReview: I will never this life changing book. The experience was almost transcendental, this author has a permanent place on my bookshelf. \n Rating: 1\nFalse\n\nReview: Doesn't fit my Note 8 or my wife's Galaxy S6. The neck needs to be at least an 1.25 inches longer to accommodate either phone without a case one \n Rating: 2\nTrue\n\nReview: Lasted for about 2 months, now most of them are not working or will work sparingly. Also the charging port connector was easily bender and damaged . \n Rating: 1\n" }, { "from": "gpt", "value": "True\n\n" } ]
[ { "from": "human", "value": "Q: Given a story, answer the question about the story. The question is the last sentence in the input. The story has one of the three following scenarios: the first is when the individual's belief matches reality, the second is when the individual's belief does not match reality, and the third is when an individual has a false belief about another individual's beliefs. The question will ask about the location of an object in the story with respect to either none or one of the three scenarios. Note that there are distractor sentences in each story that are unrelated to the question and are designed to confuse the reader.\nEvelyn entered the staircase. Mia entered the staircase. The lime is in the red_bucket. Evelyn moved the lime to the blue_pantry. Where does Evelyn think that Mia searches for the lime?\nA:" }, { "from": "gpt", "value": "blue_pantry" } ]
[ { "from": "human", "value": "You are given a background paragraph that describes one or more causal or qualitative relationships such as a relationship in economics or a scientific law and a story that makes use of the concepts or the relationship described in the provided paragraph. You are also given a question about the story that requires an understanding of the relationship described in the background paragraph and the story. You need to come up with an answer to the given question; the answer will be a span from either the question or the story. In order to correctly answer the given question, you need to understand the relationship mentioned in the background paragraph and should be able to use it to understand that in the story. Your answer can not consist of any word that is not mentioned in any of these: the background paragraph, the story, or the question. You can directly copy and paste a span from the story or the question while answering the given question.\n\nExample Input: Background Paragraph: Plucking, also referred to as quarrying, is a glacial phenomenon that is responsible for the erosion and transportation of individual pieces of bedrock, especially large \"joint blocks\". This occurs in a type of glacier called a \"valley glacier\". As a glacier moves down a valley, friction causes the basal ice of the glacier to melt and infiltrate joints (cracks) in the bedrock. The freezing and thawing action of the ice enlarges, widens, or causes further cracks in the bedrock as it changes volume across the ice/water phase transition (a form of hydraulic wedging), gradually loosening the rock between the joints. This produces large pieces of rock called joint blocks. Eventually these joint blocks come loose and become trapped in the glacier. \nStory: Rob likes to visit high mountainous locations. Last year he visited the the Alps mountains in Austria. There he saw an unique type of glacier called valley glacier. This year he went further east to visit the Ural mountains in Russia. There he didn't see any valley glacier. \nQuestion: Which mountain would not see erosion of bedrocks?\nExample Output: Ural.\n\nExample Input: Background Paragraph: Mantle melting can come about three ways: (1) when the temperature rises, (2) if the pressure lowers (which lowers the melting point), and (3) if water is added, which lowers the melting point. Two of these (1 and 3) might explain why there are volcanoes at convergent plate boundaries the temperature of the subducting plate increases as it sinks into the hot mantle (1). Sediments lying on top of the subducting plate contain water. As the sediments subduct, the water rises into the overlying mantle material. This lowers the melting temperature of the mantle (3). When the mantle above the subducting plate melts, volcanoes form above it. This leads to volcanoes in an island arc or continental arc. \nStory: Two brothers Syd and Dan went to college. Syd studied plate tectonics and mantle melting. His brother Dan got to study types of soil and pedology. \nQuestion: Which brother studied more about the influence sediments have on mantle melting?\nExample Output: Syd.\n\nExample Input: Background Paragraph: Current Situation : Currently the heating effect of extra greenhouse gases (since the start of the industrial revolution) is equal to about . Thus the recent period has recorded parallel increases in concentration of carbon dioxide and average global temperature. As more greenhouse gases are put into the atmosphere the temperature will increase further. There are certain effects of a warmer Earth (discussed more in optional section), which could accelerate the process, even if no more greenhouse gases are put into the atmosphere (an unlikely prospect for the foreseeable future). \nStory: Due to an increase in the number of factories releasing carbon dioxide the concentration of it has increased. Recent reports show a correlation between increasing levels of carbon dioxide and the average global temperature increasing. \nQuestion: Given the average temperature increases even more, will the concentration of carbon dioxide have increased or decreased?\nExample Output:" }, { "from": "gpt", "value": "increased.\n\n" } ]
[ { "from": "human", "value": "In this task, you will be given two sentences separated by \", so\". You should decide whether the first sentence can be the cause of the second sentence. If you can see a possible causation, answer with \"plausible\", otherwise answer with \"not plausible\".\n\nExample Input: The boy murmured his reply to his mother, so he was able to talk\nExample Output: not plausible\n\nExample Input: The speaker disarmed her debate opponent, so of course she agreed to speak\nExample Output: not plausible\n\nExample Input: The offender violated parole, so the case was opened\nExample Output:" }, { "from": "gpt", "value": "plausible\n\n" } ]
[ { "from": "human", "value": "Teacher:You are given a paragraph (Passage), a question (Question) and two answer options (Option1 and Option2). Your task is to find the correct answer (and return the string of the correct option, not option1/2) for the given question from the given options and based on the given passage. Answer of the question can be found directly from the passage.\nTeacher: Now, understand the problem? Solve this instance: Passage: Growing vegetables is a great way to get fresh vegetables we all like . Before you start the planting process you need to decided what vegetables you want to grow and whether you need seeds or starter plants and layout your garden and/or containers with them , then purchase your seeds or starter plants from a store that sells them . To grow vegetables the ground and/or container must be prepared by rototilling the ground or filling the container with potting soil . Then place seeds/plants in the prepared rows or container . When all seeds/starter plants are in the rows or holes made in the container dirt , cover them dirt as instructed on the package , fertilize and water the area around the plant and seeds and water . Depending on the weather you will need to water your garden every 2 or 3 days and fertilize about every 10 days . Most of your seeds will sprout in 7 to 10 days and be ready to harvest in the days stated on the package . Question: What needs to be spaced out? Option1: the planting days Option2: the seeds\nStudent:" }, { "from": "gpt", "value": "the seeds" } ]
[ { "from": "human", "value": "Q: In this task you will be given a list of integers. You should find the minimum absolute difference between 2 integers in the list. The absolute difference is the absolute value of one integer subtracted by another. The output should be a single integer which is the smallest possible absolute distance.\n[-54, -28, 83, -2, 12]\nA:" }, { "from": "gpt", "value": "14" } ]
[ { "from": "human", "value": "Instructions: In this task, you are given an input i,A where i is an integer and A is an array. You need to find every ith element of A starting with the 1st element.\nInput: 3, ['4977', '8695', 'u', '3265', '3979', 'i', 'Z', '5111', 'r', 'X']\nOutput:" }, { "from": "gpt", "value": "4977, 3265, Z, X" } ]
[ { "from": "human", "value": "Detailed Instructions: Given a scientific question, generate an incorrect answer to the given question. The incorrect answer should be a plausible alternative to the correct answer. The answer should be some other item belonging to the same category as the correct answer but should be incorrect.\nProblem:The properties of the alkali metals are similar to each other as expected for elements in the same what?\nSolution:" }, { "from": "gpt", "value": "class" } ]
[ { "from": "human", "value": "In this task you are given a sentence pair from wikipedia that have high lexical overlap. If the sentences have the same meaning and are just paraphrases of each other label them as \"Paraphrase\" , if not label them as \"Not paraphrase\". The two sentences are seperated by a new line.\nExample: In January 2011 , the Deputy Secretary General of FIBA Asia , Hagop Khajirian , inspected the venue together with SBP - President Manuel V. Pangilinan .\nIn January 2011 , FIBA Asia deputy secretary general Hagop Khajirian along with SBP president Manuel V. Pangilinan inspected the venue .\nExample solution: Paraphrase\nExample explanation: The second sentence has the same meaning as the first sentence and is just a paraphrased version of it.\n\nProblem: Hunt returned to New York until he moved to Venice in 1976 , California .\nHunt returned to Venice , California until 1976 when he moved to New York .\n" }, { "from": "gpt", "value": "Solution: Not paraphrase" } ]
[ { "from": "human", "value": "In this task you will be given a list, of lists, of integers. For every inner list contained in the input list, you should multiply every even number in that list. The output should be a list of integers with the same length as the number of lists in the input list. If there are no even numbers in an inner list you should output 0 for that list.\n\n[[-25, 20, -31], [-5, 48, -22], [37, -46, -35, -30], [-22, -35], [-21, 19, 44, -6], [46, -22, 21], [15, 25, 42, -44], [-27, 44, 28], [-31, 19, -49, 30, -13], [-2, -1, 6], [-28, 31, -41, -12, -47], [42, 11, -1, 39], [5, 43], [-24, -10]]" }, { "from": "gpt", "value": "[20, -1056, 1380, -22, -264, -1012, -1848, 1232, 30, -12, 336, 42, 0, 240]" } ]
[ { "from": "human", "value": "Definition: In this task, you will be given sentences in which your task is to recognize the name of the drug or medicine. Drugs are substances that change a person's mental or physical state. They can affect how your brain works, how you feel and behave, your understanding, and your senses. Although there might be several correct answers, you need to write one of them.\nInput: Only after three subsequent episodes of severe, symptomatic thrombocytopenia over the next four weeks did he say, upon repeat questioning, that he had continued to take quinine for night leg cramps.\nOutput:" }, { "from": "gpt", "value": "quinine" } ]
[ { "from": "human", "value": "Teacher:You need to answer a given question containing a blank (_). Your answer must be one of the two objects mentioned in the question, for example \"trophy\" and \"suitcase\". Your answer must not contain a word that is not present in the question. Please don't use articles (e.g., the, a) before the answer.\nTeacher: Now, understand the problem? Solve this instance: Jacob wanted his outside garden table to be plastic rather than wood, as the _ would warp with exposure to weather.\nStudent:" }, { "from": "gpt", "value": "wood" } ]
[ { "from": "human", "value": "Given news headlines, an edited word and funniness labels. The original sentence has word with the given format {word}. Create new headlines by replacing {word} in the original sentence with edit word. Your task is to generate answer \"Yes\" if an edited sentence matches funniness label, otherwise generate answer \"No\". \n\nNews Headline: Diesel cars are 10 times more toxic than trucks and {buses} , data shows \n Edit: girlfriends \n Label: Not Funny" }, { "from": "gpt", "value": "No" } ]
[ { "from": "human", "value": "Given the task definition and input, reply with output. This task involves creating questions from a given passage that involve some kind of complex reasoning (including numerical reasoning). \nThe generated questions must require looking at more than one part of the passage to answer. Try to use a variety of reasoning types in your questions (some of the sample reasoning types are illustrated in the 'positive examples' field) and also have a variety of answer types (spans, numbers, dates). A span is a continuous phrase taken directly from the passage or question. In such questions, the answer spans are recommended to be five words. Questions with answer type \"number\", are expected to start with \"How many\". For instance, \"How long did an event last?\" can be phrased as \"How many years did an event last?\". To increase diversity, try to create more questions for which the answer will be multiple spans. Please create questions that require AT LEAST 2 arithmetic operations as illustrated in the 'positive examples' field. You are encouraged to try DIFFERENT COMBINATIONS of operations and not just limit to the combinations shown in those examples. The more challenging questions you create, the better it is.\n\nPassage: On 19 November 1921, about 100 Moplah prisoners were sent by train from Malabar to Coimbatore. They were packed in goods wagons which, of course, had no ventilation. So, when the doors were opened after a five hours' journey, the prisoners were found to be in a state of collapse, with horrible wounds inflicted by bites and bows in one another by the struggling men. In all, 82 men died. In the aftermath of this ethnic cleansing, the Suddhi Movement was created by the Arya Samaj. They converted over 2000 Hindus who had been forcibly converted to Islam by the Moplahs. However their leader, Swami Shraddhananda, had to pay with his life, due to this effort. The aged Swami was stabbed on 23 December 1926 by a person named Abdul Rashid at his Ashram.\n" }, { "from": "gpt", "value": "How many Moplah prisoners survived the train from Malabar to Coimbatore?" } ]
[ { "from": "human", "value": "Detailed Instructions: Read the given query and classify it as a 'Good' or 'Bad' query depending on how well the query is formed, 'Bad' being the expected output for a not so well formed query and 'Good' being the expected output for a well formed query. A query may be wrong based on common sense or general facts, but if it is well formed, you should answer with Good.\nQ: Columbus have kids ?\nA:" }, { "from": "gpt", "value": "Bad" } ]
[ { "from": "human", "value": "Instructions: Categorize the comment on the basis of obscenity. If the comment is obscene output Yes, otherwise output No.\nInput: Nope. Just insatiable eclectic curiosity.\nOutput:" }, { "from": "gpt", "value": "No" } ]
[ { "from": "human", "value": "In this task, you are given an ambiguous question/query (which can be answered in more than one way) and a clarification statement to understand the query more precisely. Your task to classify that if the given clarification accurately clarifies the given query or not and based on that provide 'Yes' or 'No'.\n\nInput: Consider Input: Query: I'm interested in dinosaurs \n Clarification: are you looking for a list of movies ron howard directed\n\nOutput: No\n\n\nInput: Consider Input: Query: What is martha stewart and imclone? \n Clarification: are you looking for homes or demographic information\n\nOutput: No\n\n\nInput: Consider Input: Query: Tell me about kansas city mo \n Clarification: are you looking for netflix\n\n" }, { "from": "gpt", "value": "Output: No\n\n\n" } ]
[ { "from": "human", "value": "Instructions: In this task, you are given an input list. A list contains several comma-separated items written within brackets. You need to return the position of all the alphabetical elements in the given list in order. Assume the position of the 1st element to be 1. Return -1 if no alphabetical element is in the list.\nInput: ['483', 'E', 'V', 'Z', '1113', '9861', '4035', '9577', 'v', '1195', 't', '9687', 'i', '1469', 't', '9169', '8801', '4081', 'b', '3569', '4723', '9517', 'k', '6701', 'y', 'm', '6901']\nOutput:" }, { "from": "gpt", "value": "2, 3, 4, 9, 11, 13, 15, 19, 23, 25, 26" } ]
[ { "from": "human", "value": "You will be given a definition of a task first, then an example. Follow the example to solve a new instance of the task.\nIn this task, based on the given context word, you are asked to create a pair of sentences each containing a blank (_) and their corresponding answer. The sentence pair should look similar, and should be about two related but different objects; for example \"trophy\" and \"suitcase\". Additionally, the two sentences must be different in terms of trigger words (e.g., \"small\" and \"big\") which express contrasting properties about the two objects. The answer must not be associated with the trigger word; instead, it should depend on the context present in twin sentences. Also, the answers should not be ambiguous. For each sentence, there should be an agreed upon answer to fill in the blank. Your generations should NOT contain potentially explicit, offensive, or adult content. Do not use animals and proper nouns (e.g., New York, Macbook, Jeff Bezos, McDonald's, ...) as your objects. Avoid repeating the same style or phrase in twin sentences. E.g., a contrasting sentence can always be created using simple negation i.e. by adding not, never, etc. Instead, try to increase diversity. Both twin sentences must contain at least 15 and at most 30 words. Twin sentences must have at least 70% overlapping words. You must utilize the given context word while writing the twin sentences. Each of the twin sentences must contain only one blank. Make sure that ObjectX and Y have the same number e.g. when ObjectX is singular, ObjectY must be singular, too. The two objects (ObjectX & ObjectY) should be used ONCE in each sentence. Here is a list of contrastive words that may be used as trigger words. You should create more such trigger words and use them in your twin sentences.\n| Attribute | triggerword | contrastive triggerword | \n| age | old | new | \n| altitude | low | high |\n| area | small | vast | \n| brightness | dark | light | \n| clarity | obscure | clear | \n| cleanness | dirty | clean | \n| complexity | simple | complex | \n| cost | cheap | expensive |\n| density | sparse | dense |\n| depth | shallow | deep |\n| distance | near | far | \n| electric conductivity | low | high |\n| flexibility | rigid | flexible |\n| granularity | fine | coarse | \n| hardness | soft | hard | \n| length | short | long | \n| magnitude | small | large | \n| mass | small | large | \n| odor | weak | strong | \n| pressure | low | high | \n| resistance | low | high | \n| shape | round | sharp |\n| shape | flat | spiky |\n| size | small | large | \n| sound | quiet | loud | \n| sound pitch | low | high | \n| speed | slow | fast | \n| stability | unstable | stable | \n| strength | weak | strong | \n| temperature | low | high | \n| texture | smooth | rough | \n| thermal conductivity | low | high |\n| thickness | thin | thick | \n| volume | small | large |\n| weight | light | heavy | \n| width | narrow | wide | \n| location | in | out |\n| location | up | down |\n| location | above | below |\n| location | on | off |\n| location | to | from |\n\n\nContext word: fit.\nSolution: Sentence 1: The trophy doesn't fit into the brown suitcase because _ is too large. \nAnswer1: trophy. \nSentence 2: The trophy doesn't fit into the brown suitcase because _ is too small. \nAnswer2: suitcase.\nWhy? Here, the two objects are \"trophy\" and \"suitcase\". The trigger words are \"large\" and \"small\". Everything else (apart from the trigger words) are the same in the two sentences. The two sentences have different answers, which are the two objects (\"trophy\" and \"suitcase\").\n\nNew input: Context Word: cement.\nSolution:" }, { "from": "gpt", "value": "Sentence 1: The cement could not be used to plaster the whole wall because the _ is small. \nAnswer1: cement. \nSentence 2: The cement could not be used to plaster the whole wall because the _ is wide. \nAnswer2: wall." } ]
[ { "from": "human", "value": "In this task, you will be presented with a question and you have to answer the question based on your knowledge. Your answers should be as short as possible.\n\nExample Input: What was the first year the country hosted the Games that also hosted the Games when a competition was held at the Bucheon Gymnasium ?\nExample Output: 1986\n\nExample Input: What is the release date of the game that is a shooter game for the SNES ?\nExample Output: August 12 , 2008\n\nExample Input: The event involving the player who topped the money list in 2002 moved to where in 1973 ?\nExample Output:" }, { "from": "gpt", "value": "Tokyo Yomiuri Country Club\n\n" } ]
[ { "from": "human", "value": "instruction:\nThis task is about writing a correct answer for the reading comprehension task. Based on the information provided in a given passage, you should identify the shortest continuous text span from the passage that serves as an answer to the given question. Avoid answers that are incorrect or provides incomplete justification for the question.\nquestion:\nPassage: Pascal Boyer argues that while there is a wide array of supernatural concepts found around the world, in general, supernatural beings tend to behave much like people. The construction of gods and spirits like persons is one of the best known traits of religion. He cites examples from Greek mythology, which is, in his opinion, more like a modern soap opera than other religious systems. Bertrand du Castel and Timothy Jurgensen demonstrate through formalization that Boyer's explanatory model matches physics' epistemology in positing not directly observable entities as intermediaries. Anthropologist Stewart Guthrie contends that people project human features onto non-human aspects of the world because it makes those aspects more familiar. Sigmund Freud also suggested that god concepts are projections of one's father. Question: Who argues that there are lots of supernatural things in the world?\nanswer:\nPascal Boyer\n\n\nquestion:\nPassage: Kenneth Gergen formulated additional classifications, which include the strategic manipulator, the pastiche personality, and the relational self. The strategic manipulator is a person who begins to regard all senses of identity merely as role-playing exercises, and who gradually becomes alienated from his or her social \"self\". The pastiche personality abandons all aspirations toward a true or \"essential\" identity, instead viewing social interactions as opportunities to play out, and hence become, the roles they play. Finally, the relational self is a perspective by which persons abandon all sense of exclusive self, and view all sense of identity in terms of social engagement with others. For Gergen, these strategies follow one another in phases, and they are linked to the increase in popularity of postmodern culture and the rise of telecommunications technology. Question: In what classification is a person who regards all experiences are role play and becomes alienated from his or her social self?\nanswer:\nThe strategic manipulator\n\n\nquestion:\nPassage: The system is currently undergoing a period of refurbishment and modernization, entitled 'Metro: All Change.' The programme has replaced all ticket machines and introduced ticket gates at the busiest stations - part of the transition to smart ticketing. All Metro trains are being completely refurbished and most stations are undergoing improvement works (or in some cases complete reconstruction, for example North Shields). In addition; tracks, signalling and overhead wires are also being overhauled. Longer term plans include the procurement of an entirely new fleet of trains and further extensions to the system. Proposed routes include to Newcastle's west end, to the Cobalt business park in North Tyneside, to the Metrocentre in Gateshead and to additional locations in Gateshead, South Tyneside and Sunderland. Several of the proposed routes would require trams as opposed to the current light rail trains. Question: What does the replacement of the ticket machines and introduction of ticket gates herald the transition to?\nanswer:\n" }, { "from": "gpt", "value": "smart ticketing.\n\n\n" } ]
[ { "from": "human", "value": "In this task, you're given a pair of sentences, sentence 1 and sentence 2, that contradict each other. Your job is to alter sentence 2 so that the pair agree with each other. Generated sentences must be short, with less than 15 words. New information can be introduced. Avoid using pronouns to confuse the subject of the sentence.\n\nInput: Consider Input: Sentence 1: Two people are standing outside of a wine shop. Sentence 2: Two people sitting in a restaurant.\n\nOutput: Two women are near a store.\n\n\nInput: Consider Input: Sentence 1: A man in a yellow shirt and blue jeans is passed out, sleeping, dead or resting his eyes on a sidewalk. Sentence 2: A man is sleeping on a car.\n\nOutput: A man is laying on the sidewalk.\n\n\nInput: Consider Input: Sentence 1: A man in yellow and green shorts jumping in a living room. Sentence 2: The horse swats flies with his tail.\n\n" }, { "from": "gpt", "value": "Output: A man is in a house.\n\n\n" } ]
[ { "from": "human", "value": "Instructions: In this task, you're given a question, a context passage, and four options which are terms from the passage. After reading a passage, you will get a brief understanding of the terms. Your job is to determine by searching and reading further information of which term you can answer the question. Indicate your choice as 'a', 'b', 'c', or 'd'. If you think more than one option is plausible, choose the more probable option to help you answer the question.\nInput: Question: How many years had the Nexstar Broadcasting Group been in business at the time of their purchase of KSEE assets? Passage:On February 6, 2013, Granite sold KSEE's non-license assets to the Nexstar Broadcasting Group, with Nexstar also intending to purchase KSEE's license following Federal Communications Commission approval; in the interim, Nexstar operated the station via a time brokerage agreement. The deal made KSEE a sister station to CBS affiliate KGPE, which Nexstar had just acquired from Newport Television. Normally, duopolies between two \"Big Four\" affiliates, let alone \"Big Three,\" would not be allowed because those stations are usually the four highest-rated stations in a market, which FCC regulations do not allow any common ownership of. However, according to Nielsen, in 2013 KGPE was ranked as the fourth highest-rated station in the market and KSEE fifth, after KFSN (ABC), KFTV (Univision), and KMPH-TV (Fox), allowing a duopoly to be formed between the stations. This marked the second instance (after the Gannett Company purchased ABC affiliate WJXX in Jacksonville, Florida, creating a duopoly with that market's NBC affiliate WTLV, in 2000) in which a single company owns a duopoly involving two stations that are affiliated by a Big Three television network; and is also Nexstar's first true Big Three duopoly (Nexstar's other Big Three duopolies are virtually formed, in which the other station is owned by Mission Broadcasting). The merger was approved on April 17, and completed by May 31.\n Links: a. Big Three television networks b. Local marketing agreement c. Nexstar Media Group d. Mission Broadcasting\nOutput:" }, { "from": "gpt", "value": "c" } ]
[ { "from": "human", "value": "Part 1. Definition\nGiven a scientific question, generate a correct answer to it.\nPart 2. Example\nWho proposed the theory of evolution by natural selection?\nAnswer: darwin\nExplanation: This is a direct fact that Charles Darwin proposed the theory of evolution.\nPart 3. Exercise\nTranslation is the second part of the central dogma of what?\nAnswer:" }, { "from": "gpt", "value": "molecular biology" } ]
[ { "from": "human", "value": "In this task you are given a statement and an explanation giving you further knowledge about an entity in the statement. You must judge whether the statement is true or false based on the explanation. Label an instance as \"True\" if the explanation confirms the statement or doesn't disprove it. Label an instance as \"False\" if the explanation disproves the statement. The statement and explanation are separated by a newline character.\n\nInput: Consider Input: A person must purchase home Insurance for their house, or they could serve jail time.\nIt is not illegal to have you house be uninsured. This can be confirmed from an online search.\n\nOutput: False\n\n\nInput: Consider Input: Errol Flynn sold home owner's insurance for a living.\nErrol Flynn was an actor (although he may have sold insurance prior to becoming famous).\n\nOutput: False\n\n\nInput: Consider Input: Farmers harvest Pecan for its piquant juice.\nPecans are a dry nut.\n\n" }, { "from": "gpt", "value": "Output: False\n\n\n" } ]
[ { "from": "human", "value": "You are given a short poem which is written by a kid. Your task is to predict if the kid is in elementary school or high school. Anyone who is from grade 1st to 6th-grade will be considered as in elementary school, and 7th to 12th-grade kids will be considered as in high school. There are only two possible outputs, i.e., elementary and high. All inputs have at least one output and it cannot have both categories at a time.\n\nExample input: he look into your eye that same look just take your breathe away you feel at that moment that he be definitely the one for you but just when you realize that just as fast a he take your breathe away he take your heart and break it into tiny little piece and watch it burn\nExample output: high\nExample explanation: The poem is long and linguistically well structured, and the prediction is correct.\nQ: there once be a fellow his name be trouble he 's always run on the double he run and jump all over the house just once i wish he be quiet a a mouse he sniff and snort make strange sound then late at night he ca n't be find i know you think i be talk about a frog you silly thing it 's just my sister 's dog\nA:" }, { "from": "gpt", "value": "high" } ]
[ { "from": "human", "value": "Given the task definition and input, reply with output. In this task, you are given a piece of an article. Your task is to generate a short summary of the text. Try to give the summary in just one sentence.\n\nKarachi police in their continued drive against criminals, arrested 210 accused from various parts of the city during the past three days.\n" }, { "from": "gpt", "value": "Karachi police, arrested 210 accused during the past three days." } ]
[ { "from": "human", "value": "Given the task definition and input, reply with output. In this task you are given a sentence. You must judge whether there exist two consecutive words within the sentence with flipped orders, that is, whether the sentence will make sense and be correct if the order of two consecutive words changes. Label the instances as \"Inversion\" or \"Original\" based on your judgment.\n\nPeople were awoken in shifts.\n" }, { "from": "gpt", "value": "Original" } ]
[ { "from": "human", "value": "In this task, you will be presented with a question, a word, and a POS tag. You have to determine whether the part-of-speech tag of the given word in the question is equal to the given POS tag or not. Give your answer with True or False. Here is the Alphabetical list of part-of-speech tags used in this task: CC: Coordinating conjunction, CD: Cardinal number, DT: Determiner, EX: Existential there, FW: Foreign word, IN: Preposition or subordinating conjunction, JJ: Adjective, JJR: Adjective, comparative, JJS: Adjective, superlative, LS: List item marker, MD: Modal, NN: Noun, singular or mass, NNS: Noun, plural, NNP: Proper noun, singular, NNPS: Proper noun, plural, PDT: Predeterminer, POS: Possessive ending, PRP: Personal pronoun, PRP$: Possessive pronoun, RB: Adverb, RBR: Adverb, comparative, RBS: Adverb, superlative, RP: Particle, SYM: Symbol, TO: to, UH: Interjection, VB: Verb, base form, VBD: Verb, past tense, VBG: Verb, gerund or present participle, VBN: Verb, past participle, VBP: Verb, non-3rd person singular present, VBZ: Verb, 3rd person singular present, WDT: Wh-determiner, WP: Wh-pronoun, WP$: Possessive wh-pronoun, WRB: Wh-adverb\nOne example is below.\nQ: Who were the builders of the mosque in Herat with fire temples ? \n, Word: Who \n, POS tag: IN\nA: False\nRationale: This is a good example. POS tag of who should be WP(Wh-pronoun), and not IN(Preposition).\nQ: In what state is a company incorporated that is noted by an Eagle Scout who served as the 69th United States Secretary of State ? \n, Word: by \n, POS tag: IN\nA:" }, { "from": "gpt", "value": "True" } ]
[ { "from": "human", "value": "Teacher: In this task, you need to write an incorrect answer to the given question. Along with the question, you are also given the correct answer(s) and the associated paragraph. An incorrect answer should be closely related to the content of the paragraph and/or the question but will not truthfully answer the question. Your incorrect answers should be similar to the given correct answer such that it will pose a challenge for students to distinguish the two. A good incorrect answer should prevent those who skip the paragraph from answering the question. The incorrect answer should not be trivial, in the sense that they should not be easily eliminated without reading the paragraph.\nTeacher: Now, understand the problem? If you are still confused, see the following example:\nParagraph- Sent 1: It was hot that day.\nSent 2: The temperature on the wall of the backyard was showing something well over 100 F.\nSent 3: Meanwhile Tom, at home, was trying finish the remainder of carrots from last night, and packing for his trip to Chicago tomorrow.\nSent 4: As employees of the Art Museum, Tom and his older cousin often had to travel to Chicago. Question: What was the temperature outside, when Tom was eating carrots? Correct Answer: well over 100 F.\nSolution: Not very hot.\nReason: This is a good incorrect answer, as most of the the words of this answer appear in the the paragraph. They are also related to the content of the question (temperature), yet do not answer the question correctly. Consequently, one has to carefully read the paragraph understand that \"not very hot\" does not answer the question.\n\nNow, solve this instance: Paragraph- Sent 1: Iwasaki's careful respect for the Long Beach program and its lawyers earned him the political capital he needed to complete his takeover in a matter of weeks.\nSent 2: \"The Long Beach program had strong support in the community so in a situation like that, one has to recognize that it's not like a takeover where I have all the answers and I know best,\" Iwasaki said.\nSent 3: \"The people in the community who are working there have the contacts and the knowledge that will allow service to continue.\"Sent 4: Things have gone less smoothly across town.\nSent 5: There, Dudovitz, a longtime poverty lawyer and executive director of the San Fernando Valley's 36-year-old legal aid program, continues to struggle with his hostile takeover of the neighboring San Gabriel-Pomona Valleys service area one year after it was accomplished.\nSent 6: On the bright side, Dudovitz has extended his respected program to clients in the San Gabriel-Pomona Valley, and he now operates on a much larger budget, $6.5 million last year.\nSent 7: However, his clash with the old San Gabriel program resulted in litigation, bitter feelings and a mission that some say is not clearly focused on serving poor people.\nSent 8: \"It was a difficult situation that was probably mishandled by everyone,\" a longtime observer of the public interest community said of the San Fernando Valley-San Gabriel-Pomona Valley merger.\nSent 9: \"There are very few people who come out as the heroes.\nSent 10: Personalities got involved when they shouldn't have.\nSent 11: Things were said that caused bad feelings and couldn't be unsaid.\"Sent 12: Iwasaki's merger with the smaller, 48-year-old Long Beach program was friendly and fast, and no one - not even Long Beach board members - lost a job.\nSent 13: When it was over, Iwasaki had $1 million more in federal dollars and two new offices.\nSent 14: Long Beach clients regained services they had lost years ago when federal budget cuts and dwindling grants reduced the staff of 15 lawyers to five and cut immigration and consumer law programs.\nSent 15: Iwasaki said, \"[I judged the transition] better than I could have hoped for.\". \nQuestion: What is the budget Dudovitz now operates on?.\nStudent:" }, { "from": "gpt", "value": "$7.5 million." } ]
[ { "from": "human", "value": "In this task you will be given a list, of lists, of numbers. Each list is presented with brackets and comma-separated values. For example, a list of numbers is [1,2,3] and a list of two lists of numbers is [[1,2],[3,4,6]]. You should sum up the values at every position of every inner list and put it as an element of your answer, that is, you should sum the first members of each inner list together and put it as the first element of the list, then sum the second members of each inner list together and put it as the second element of your list, etc. The output should be a list of numbers with the same length as the longest inner list.\nOne example is below.\nQ: [[6, 3, 8], [7, -5]]\nA: [13, -2, 8]\nRationale: We start by summing up the first elements of all the lists: 6 + 7 = 13, so the first element of the output is 13. For the second element of the output, we sum up the second elements of all the lists: 3 + (-5) = -2. One of the lists doesn't have a third element, so the third element of the answer is: 8 + 0 = 8.\nQ: [[-7, -79, -31, 18], [-53, 92, 48, 87], [-54, -71, -39, 21, -29, 13, 98], [81, -32, 30, -4, 22, 96, -88]]\nA:" }, { "from": "gpt", "value": "[-33, -90, 8, 122, -7, 109, 10]" } ]
[ { "from": "human", "value": "Detailed Instructions: Given an trivia question precisely answer the question with a word/phrase/name. External resources such as Wikipedia could be used to obtain the facts.\nQ: Formed in the 1980s and having the leader Hassan Nasrallah, which radical Shi-ite Muslim organisation in Lebanon has, for years, been involved in guerilla warfare against Israel?\nA:" }, { "from": "gpt", "value": "hezbollah" } ]
[ { "from": "human", "value": "Definition: In this task you are given a small conversation between two persons and 4 options on how the conversation should continue. Your job is to choose the most reasonable option. The conversation and the options are separated by a newline character. Each dialogue in the conversation are separated by a comma. F and M indicate female and male, respectively.\nInput: M: I've been busy lately, I've been spending a lot of time talking to people in chat rooms. ,F: So have you met someone yet? ,M: No, yesterday over 100 people wanted to talk to me. One even said that I was his dream girlfriend. ,F: What? Wait a minute, don't you mean dream boyfriend? ,M: Oh, I forgot to tell you I pretended to be a 16-year-old girl. I had to, nobody wants to talk to me as a guy.\n(A) F: So you've been busy looking for chat rooms because more than 200 people want to talk to you? (B) F: You mean you pretend to be a girl who is busy chatting with others online? (C) F: I am afraid I did not quite catch what you were saying. Please repeat it. (D) F: You mean you recently pretended to be an 18-year-old girl and were busy visiting an online friend?\nOutput:" }, { "from": "gpt", "value": "B" } ]
[ { "from": "human", "value": "You will be given a definition of a task first, then some input of the task.\nIn this task, you are given inputs i,j, and A, where i and j are integers and A is a list. You need to return the sum of all the numerical elements in the list A between the positions i and j (including positions i and j). Return 0 if no numerical element is present in the list between the given ranges. i and j will be non-negative, and will always have a value less than the length of A. i will always be less than j.\n\n2, 3, ['7575', 'r', '5341', 'O', '8323', '69']\nOutput:" }, { "from": "gpt", "value": "5341" } ]
[ { "from": "human", "value": "Part 1. Definition\nGiven a story, answer the question about the story. The question is the last sentence in the input. These stories can be difficult due to their length and how each story has at least one of the three following scenarios: the first is when the individual's belief matches reality, the second is when the individual's belief does not match reality, and the third is when an individual has a false belief about another individual's beliefs. The question will ask about the location of an object in the story with respect to either none or one of the three scenarios.\nPart 2. Example\nJacob entered the dining_room. William entered the dining_room. The tomato is in the green_drawer. William exited the dining_room. Jacob moved the tomato to the blue_cupboard. Jacob is in the dining_room. Olivia entered the dining_room. The cucumber is in the blue_cupboard. Olivia exited the dining_room. Jacob moved the cucumber to the green_drawer. William entered the pantry. Jacob entered the pantry. The asparagus is in the red_cupboard. Jacob exited the pantry. William moved the asparagus to the green_pantry. Abigail entered the hall. William entered the hall. The persimmon is in the blue_pantry. William exited the hall. Abigail moved the persimmon to the blue_envelope. Where does Abigail think that William searches for the persimmon?\nAnswer: blue_pantry\nExplanation: The persimmon was last in the blue_pantry before William exited the hall. After William exited the hall, Abigail moved the persimmon to the blue_envelope, so she knows where William will look for it.\nPart 3. Exercise\nAbigail entered the attic. Jackson entered the attic. The strawberry is in the green_bottle. Jackson exited the attic. Abigail moved the strawberry to the green_drawer. Where will Jackson look for the strawberry? Evelyn entered the crawlspace. Abigail entered the crawlspace. The tangerine is in the blue_cupboard. Abigail exited the crawlspace. Evelyn moved the tangerine to the green_crate. Evelyn exited the crawlspace. Abigail entered the crawlspace. Where is the tangerine really?\nAnswer:" }, { "from": "gpt", "value": "green_crate" } ]
[ { "from": "human", "value": "In this task, you're given a pair of sentences, sentence 1 and sentence 2. Sentence 2 contradicts sentence 1. Your job is to alter sentence 2 so that the pair neither agree nor contradict each other. Generated sentences must be short, with less than 15 words. New information can be introduced. Avoid using pronouns to confuse the subject of the sentence.\n\nSentence 1: Two white and brown dogs barking on the shoreline while waves roll in behind them. Sentence 2: 2 cats are in the park" }, { "from": "gpt", "value": "Some animals are on the beach" } ]
[ { "from": "human", "value": "In mathematics, the absolute value of a number is the non-negative value of that number, without regarding its sign. For example, the absolute value of -2 is 2, and the absolute value of 5 is 5. In this task you will be given a list of numbers and you need to return the element with highest absolute value. If a negative and positive element have the same absolute value you should return the positive element. The absolute value for negative numbers can be found by multiplying them by -1. After finding the element with the maximum absolute value you should return the value of that element before you applied the absolute value.\n\n[ 2.834 10.134 -50.613 -50.056]" }, { "from": "gpt", "value": "-50.613" } ]
[ { "from": "human", "value": "Instructions: In this task, you are given a question and answer for that. The answer will fill in the blank or will complete the unfinished question. Your task is to provide an explanation based on the given question or fill-in-the-blank statement, and answer.\nInput: Question: Dan is watching the ship leave the dock. While the ship is near him, it is massive. However, as the ship floats away, it begins to look \n Answer: smaller\nOutput:" }, { "from": "gpt", "value": "If you are standing on the shore and a ship is going out to sea, the ship gets smaller as it moves further away from you." } ]
[ { "from": "human", "value": "In this task, you are given an input list. A list contains several comma-separated items written within brackets. You need to return the position of all the numerical elements in the given list in order. Assume the position of the 1st element to be 1. Return -1 if no numerical element is in the list.\n\n['S', '5461', 'H', '3523', 'q', 'W', '9669', '1159', '9773', 'U', 'A', 'X', '2115', 'B', 'i', 'X', 'p', 'o', '3053', 'T', '3285', '2491', 'Y', 'd', 'o', '777', '7659', 'p', 'f', 'p', '4693', '1847', '5941', 'F', '6907']" }, { "from": "gpt", "value": "2, 4, 7, 8, 9, 13, 19, 21, 22, 26, 27, 31, 32, 33, 35" } ]
[ { "from": "human", "value": "Given the task definition, example input & output, solve the new input case.\nIn this task, you are given inputs i,j, and A, where i and j are integers and A is a list. You need to concatenate all elements of A from the ith element to the jth element, and print the resultant string. i and j will be non-negative, and will always have a value less than the length of A. i will always be less than j.\nExample: 6, 14, ['9545', '5185', '1667', 'w', 's', '9559', 'T', 'u', 'U', '6843', 'L', 'l', '6365', '7103', 'A', '2381', '5037', 'X', '4487', 'd']\nOutput: 9559TuU6843Ll63657103\nHere, concatenating all the elements between the 6th element and 14th element gives '9559TuU6843Ll63657103'.\n\nNew input case for you: 2, 4, ['759', '3797', 'P', '6119', '7147']\nOutput:" }, { "from": "gpt", "value": "3797P6119" } ]
[ { "from": "human", "value": "Given the task definition and input, reply with output. In this task, you are given inputs i and A, where i is an integer and A is a list. You need to list all the elements of A after the first i elements. i will always have a value less than the length of A.\n\n2, ['W', 'y', '5663', '8229', '5563', '2953']\n" }, { "from": "gpt", "value": "5663, 8229, 5563, 2953" } ]
[ { "from": "human", "value": "Definition: Given a statement about date and time, state whether the statement is true or false. The number of date/time operands in the statement ranges between 2 and 3. Let's say the values are denoted by t1, t2 and t3. The statements follow one of the following ten templates: 't1 occurs before t2, t1 doesn't occur before t2, t1 occurs after t2, t1 doesn't occur after t2, t1 occurs between t2 and t3, t1 doesn't occur between t2 and t3, t1 occured before t2 but after t3, t1 occured after t2 but before t3, t1 didn't occur before t2 but after t3, t1 didn't occur after t2 but before t3'. The output should be either 'True' or 'False'.\nInput: 22:57:09 occured before 16:07:22 but after 6:17:45 AM\nOutput:" }, { "from": "gpt", "value": "True" } ]
[ { "from": "human", "value": "Q: In this task, you're given a question, a context passage, and four options which are terms from the passage. After reading a passage, you will get a brief understanding of the terms. Your job is to determine by searching and reading further information of which term you can answer the question. Indicate your choice as 'a', 'b', 'c', or 'd'. If you think more than one option is plausible, choose the more probable option to help you answer the question.\nQuestion: Who won the American Civil War? Passage:The Habeas Corpus Suspension, (1863), entitled An Act relating to Habeas Corpus, and regulating Judicial Proceedings in Certain Cases, was an Act of Congress that authorized the president of the United States to suspend the privilege of the writ of habeas corpus in response to the American Civil War and provided for the release of political prisoners. It began in the House of Representatives as an indemnity bill, introduced on December 5, 1862, releasing the president and his subordinates from any liability for having suspended habeas corpus without congressional approval. The Senate amended the House's bill, and the compromise reported out of the conference committee altered it to qualify the indemnity and to suspend habeas corpus on Congress's own authority. Abraham Lincoln signed the bill into law on March 3, 1863, and suspended habeas corpus under the authority it granted him six months later. The suspension was partially lifted with the issuance of Proclamation 148 by Andrew Johnson, and the Act became inoperative with the end of the Civil War. The exceptions to his Proclamation 148 were the States of Virginia, Kentucky, Tennessee, North Carolina, South Carolina, Georgia, Florida, Alabama, Mississippi, Louisiana, Arkansas, and Texas, the District of Columbia, and the Territories of New Mexico and Arizona.\n Links: a. American Civil War b. Act of Congress c. Legal liability d. President of the United States\nA:" }, { "from": "gpt", "value": "a" } ]
[ { "from": "human", "value": "Teacher:Given the sentence, generate \"yes, and\" response. \"Yes, and\" is a rule-of-thumb in improvisational comedy that suggests that a participant in a dialogue should accept what another participant has stated (\"Yes\") and then expand on that line of thought or context (\"and...\"). 1 In short, a \"Yes, and\" is a dialogue exchange in which a speaker responds by adding new information on top of the information/setting that was constructed by another speaker. Note that a \"Yes, and\" does not require someone explicitly saying 'yes, and...' as part of a dialogue exchange, although it could be the case if it agrees with the description above. There are many ways in which a response could implicitly/explicitly agree to the prompt without specifically saying 'yes, and...'. \nTeacher: Now, understand the problem? Solve this instance: We got married right away. It was a busy day at Tommy Bahama. You wore your best cream shirt, and you wore your best tan.\nStudent:" }, { "from": "gpt", "value": "We were almost like camouflage, you couldn't tell either of us apart." } ]
[ { "from": "human", "value": "Detailed Instructions: Given a math problem with context and a question and 5 answer choices, the task is to provide the correct answer choice based on the problem. You must choose one of the given answer choices by letter: a, b, c, d, or e; anything else is invalid.\nSee one example below:\nProblem: Problem: a multiple choice test consists of 4 questions, and each question has 5 answer choices. in how many r ways can the test be completed if every question is unanswered ?\nOptions: a. 24, b. 120, c. 625, d. 720, e. 1024\nSolution: c\nExplanation: 5 choices for each of the 4 questions, thus total r of 5 * 5 * 5 * 5 = 5 ^ 4 = 625 ways to answer all of them. Hence, the answer is c.\n\nProblem: Problem: what is 15 % of 2 / 3 of 0.6 ?\nOptions: a. 0.06, b. 0.9, c. 9, d. 90, e. none of the above\nSolution:" }, { "from": "gpt", "value": "a" } ]
[ { "from": "human", "value": "In this task, you are given commands (in terms of logical operations) to select relevant rows from the given table. Your job is to generate a natural language interpretation of the given command. \n Here are the definitions of logical operators: \n 1. count: returns the number of rows in the view \n 2. only: returns whether there is exactly one row in the view \n 3. hop: returns the value under the header column of the row \n 4. and: returns the boolean operation result of two arguments \n 5. max/min/avg/sum: returns the max/min/average/sum of the values under the header column \n 6. nth max/nth min: returns the n-th max/n-th min of the values under the header column \n 7. argmax/argmin: returns the row with the max/min value in header column \n 8. nth argmax/nth argmin: returns the row with the n-th max/min value in header column \n 9. eq/not eq: returns if the two arguments are equal \n 10. round eq: returns if the two arguments are roughly equal under certain tolerance \n 11. greater/less: returns if argument 1 is greater/less than argument 2 \n 12. diff: returns the difference between two arguments \n 13. filter eq/not eq: returns the subview whose values under the header column is equal/not equal to argument 3 \n 14. filter greater/less: returns the subview whose values under the header column is greater/less than argument 3 \n 15. filter greater eq /less eq: returns the subview whose values under the header column is greater/less or equal than argument 3 \n 16. filter all: returns the view itself for the case of describing the whole table \n 17. all eq/not eq: returns whether all the values under the header column are equal/not equal to argument 3 \n 18. all greater/less: returns whether all the values under the header column are greater/less than argument 3 \n 19. all greater eq/less eq: returns whether all the values under the header column are greater/less or equal to argument 3 \n 20. most eq/not eq: returns whether most of the values under the header column are equal/not equal to argument 3 \n 21. most greater/less: returns whether most of the values under the header column are greater/less than argument 3 \n 22. most greater eq/less eq: returns whether most of the values under the header column are greater/less or equal to argument 3\n\nonly { filter_eq { all_rows ; month ; april } } = true" }, { "from": "gpt", "value": "select the rows whose month record fuzzily matches to april . there is only one such row in the table ." } ]
[ { "from": "human", "value": "You're given a sentence and your task is to classify whether the sentence is acceptable or not. Any sentence which is grammatically correct, has a naturalistic text, is written by a native speaker and which minimizes superfluous content is acceptable, otherwise unacceptable. If the sentence is acceptable then write \"acceptable\", otherwise \"unacceptable\".\n\nYuko rubbed the pizza with a garlic clove." }, { "from": "gpt", "value": "acceptable" } ]
[ { "from": "human", "value": "In this task you are given a sentence with one coordinating conjunction (for, and, nor, but, or, yet, and so). You must judge whether the order of two coordinated clausal conjoints have been inverted or not. Inversion would result in the sentence sounding unnatural. Label the instances as \"Original\" or \"Inversion\" based on your judgment.\n\nEx Input:\nShe then released her hand and I fell to the floor .\n\nEx Output:\nOriginal\n\n\nEx Input:\nWe fog up the windows and there 's a hand pressed against the glass just like in that one movie .\n\nEx Output:\nOriginal\n\n\nEx Input:\nThough the kiss was brief , he breathed in the scent of her , and she stiffened as his mouth touched her skin .\n\nEx Output:\n" }, { "from": "gpt", "value": "Inversion\n\n\n" } ]
[ { "from": "human", "value": "You will be given a definition of a task first, then an example. Follow the example to solve a new instance of the task.\nIn this task, you are given a dialogue from a conversation between an agent and a customer. Your task is to determine the speaker of the dialogue. Answer with \"agent\" or \"customer\".\n\nHello. How may I aid you today?\nSolution: agent\nWhy? This is a pretty straightforward dialogue, and the speaker wants to help the other. So, the agent is the speaker.\n\nNew input: Sure, Susan Allen here.\nSolution:" }, { "from": "gpt", "value": "customer" } ]
[ { "from": "human", "value": "Detailed Instructions: In this task, you are given two lists A,B. Find the longest common subsequence in the lists A and B.\nProblem:[1043, 4021, 'b', 'e', '5055', 'N', 7081], ['o', 'S', 'b', 'e', '5055', 'q', 8711, 9911]\nSolution:" }, { "from": "gpt", "value": "b, e, 5055" } ]
[ { "from": "human", "value": "Definition: Given a prompt and two completions from a language model, select the completion that has the most similar style to the prompt. Style is defined as the tone, word choice, grammar, and sentence structure throughout the prompt-completion pair. If a prompt is colloquial, then the completion should also be colloquial, as opposed to a completion that is encyclopedic or overly formal. Generate either Completion A or Completion B.\nInput: Prompt: Yes, I agree with that, but for the winter I can deal with \nCompletion A: the ice cold, so I may spend any of all night on the beach. But, it's quite a time to keep the weather clear of bad \nCompletion B: them. \"So it's the same, but the winter will not be like the summer. I think it would be better to not us\nOutput:" }, { "from": "gpt", "value": "Completion A" } ]
[ { "from": "human", "value": "Definition: You will be given two sentences. One of them is created by paraphrasing the original one, with changes on an aspect, or using synonyms. Your task is to decide what is the difference between two sentences. Types of change are explained below:\n Tense: The verbs in the sentence are changed in tense.\nNumber: Plural nouns, verbs and pronouns are changed into single ones or the other way around.\nVoice: If the verbs are in active voice, they're changed to passive or the other way around.\nAdverb: The paraphrase has one adverb or more than the original sentence.\nGender: The paraphrase differs from the original sentence in the gender of the names and pronouns.\nSynonym: Some words or phrases of the original sentence are replaced with synonym words or phrases. Changes in the names of people are also considered a synonym change. Classify your answers into Tense, Number, Voice, Adverb, Gender, and Synonym.\nInput: original sentence: Steve follows Fred's example in everything . He admires him hugely . paraphrase: Steve closely follows Fred's example in everything . He admires him hugely .\nOutput:" }, { "from": "gpt", "value": "Adverb" } ]
[ { "from": "human", "value": "You will be given a definition of a task first, then some input of the task.\nIn this task, you are given an input list A. You need to convert all the alphabets in the list with a number representing their position in the English alphabet. E.g., replace A by 1, B by 2, a by 1, b by 2, and so on.\n\n['2519', '3909', 'a', 'D', '4429', 'N', '2511', '9803', 'O', '8793', '4221', 'I', 'i', 'J', 'O', '6517', 'V', 'M', 'w', '1197', '2169', '5097', 'U', 'Q']\nOutput:" }, { "from": "gpt", "value": "2519, 3909, 1, 4, 4429, 14, 2511, 9803, 15, 8793, 4221, 9, 9, 10, 15, 6517, 22, 13, 23, 1197, 2169, 5097, 21, 17" } ]
[ { "from": "human", "value": "In this task you will be given a list of numbers and you should remove all duplicates in the list. If every number is repeated in the list an empty list should be returned. Your list should be numbers inside brackets, just like the given list.\n\nExample input: [0,1,0,2,5,1]\nExample output: [2,5]\nExample explanation: The only elements that are not duplicated is 2 and 5. This is a good example.\nQ: [3, 1, 1, 7, 3, 4, 5, 5, 0, 7]\nA:" }, { "from": "gpt", "value": "[4, 0]" } ]
[ { "from": "human", "value": "Part 1. Definition\nYou are given an elementary science question along with a few answer options. List all the words from the question that are essential and sufficient for choosing the correct answer option. Keep in mind that essential words are necessary to understand the meaning of the given question. However, non-essential words provide little information needed to select the correct answer. You can think about it like this: If the question was in another language, knowing the meaning of which words were necessary and enough for you to understand the question and answer it?. If your answer consists of multiple spans of words from the question, separate those using a comma (\", \"). Do not generate any word that is not present in the question. Make sure that your response is contained in the given question and the words are in the same order as the question.\n Grammatical correctness is not important.\nPart 2. Example\nQuestion: Which type of energy does a person use to pedal a bicycle? (A) light (B) sound (C) mechanical (D) electrical?\nAnswer: energy, pedal, bicycle\nExplanation: Clearly the words \"energy\", \"pedal\" and \"bicycle\" are necessary and enough to understand the category of the energy associated with this process, i.e. \"mechanical\". Other words are not necessary to inferring the correct answer.\nPart 3. Exercise\nQuestion: Which is a process that most likely formed the Appalachian Mountains found in the eastern part of North America? (A) earthquake (B) folded rock (C) plate tectonics (D) volcanic activity?\nAnswer:" }, { "from": "gpt", "value": "process, formed, Appalachian, Mountains" } ]
[ { "from": "human", "value": "Detailed Instructions: Given an trivia question precisely answer the question with a word/phrase/name. External resources such as Wikipedia could be used to obtain the facts.\nQ: Which popular Welsh entertainer wrote the First World War song Keep the Home Fires Burning?\nA:" }, { "from": "gpt", "value": "ivor novello" } ]
[ { "from": "human", "value": "You are given a password and you need to generate the number of steps required to convert the given password to a strong password. A password is considered strong if (a) it has at least 6 characters and at most 20 characters; (b) it contains at least one lowercase letter and one uppercase letter, and at least one digit; (c) it does not contain three repeating characters in a row. In one step you can: (1) Insert one character to password, (2) delete one character from password, or (3) replace one character of password with another character.\n\npassword = 68QcMTwQrh!2aHS\n0\n\npassword = zRmtAsXySpsjj.n\n1\n\npassword = 1.r4OkqGYlOkVIaDUNaYe.4RlGF2hwXajb9nBsJYp3s\n" }, { "from": "gpt", "value": "23\n\n" } ]
[ { "from": "human", "value": "In this task, you are given a premise, a hypothesis, and an update. The premise sentence describes a real-world situation and is always assumed to be true. The hypothesis sentence describes an assumption or inference that you might make about that situation having read the premise. The update provides additional information about the situation that might weaken or strengthen the hypothesis. A weakener is a statement that weakens the hypothesis, that is, it makes you much less likely to believe the hypothesis is true. A strengthener is a statement that strengthens the hypothesis, that is, it makes you much more likely to believe the hypothesis is true. Your task is to answer with 'strengthener' or 'weakener' if the update strengthens or weakens the hypothesis, respectively.\n\nPremise: An expensive car is parked in what looks like a car-wash with a man standing at the open passenger door side.\nHypothesis: The man is on his cell phone.\nUpdate: The man is holding his cell phone to his ear." }, { "from": "gpt", "value": "strengthener" } ]
[ { "from": "human", "value": "Given a command in a limited form of natural language, provide the correct sequence of actions that executes the command to thus navigate an agent in its environment. A command can be broken down into many different actions. Actions are uppercase and are individual steps that serve as the building blocks for a command. For commands, 'left' and 'right' are used to denote the direction of an action. The word 'opposite' turns the agent backward in the specified direction. The word 'around' makes the agent execute an action while turning around in the specified direction. The word 'and' means to execute the next scope of the command following the previous scope of the command. The word 'after' signifies to execute the previous scope of the command following the next scope of the command. The words 'twice' and 'thrice' trigger repetition of a command that they scope over two times or three times, respectively. There are only six actions: 'I_LOOK', 'I_WALK', 'I_RUN', 'I_JUMP', 'I_TURN_LEFT', and 'I_TURN_RIGHT'. These actions respectively align with the commands 'look', 'walk', 'run', 'jump', 'turn left', and 'turn right'. Actions and commands do not have quotations in the input and output.\nQ: walk around right thrice and jump opposite right twice\nA: " }, { "from": "gpt", "value": "I_TURN_RIGHT I_WALK I_TURN_RIGHT I_WALK I_TURN_RIGHT I_WALK I_TURN_RIGHT I_WALK I_TURN_RIGHT I_WALK I_TURN_RIGHT I_WALK I_TURN_RIGHT I_WALK I_TURN_RIGHT I_WALK I_TURN_RIGHT I_WALK I_TURN_RIGHT I_WALK I_TURN_RIGHT I_WALK I_TURN_RIGHT I_WALK I_TURN_RIGHT I_TURN_RIGHT I_JUMP I_TURN_RIGHT I_TURN_RIGHT I_JUMP" } ]
[ { "from": "human", "value": "Given the task definition and input, reply with output. Given a question and a context passage, generate the answer having the word or phrase from the context passage. Here, the answer should be a shortest continous span from the passage.\n\nContext: Ca2+ is a ubiquitous intracellular messenger responsible for controlling numerous cellular processes including fertilization, mitosis, neuronal transmission, contraction and relaxation of muscles, gene transcription, and cell death. At rest, the cytoplasmic Ca2+ concentration [Ca2+]i is approximately 100 nM, but this level rises to 500-1,000 nM upon activation. In osteoblasts, the elevation of [Ca2+]i is a result of an increase in the release of Ca2+ from endoplasmic reticulum and/or extracellular Ca2+ influx through voltage gated Ca2+ channels. Many of the cellular effects of Ca2+ are mediated by the Ca2+ binding protein, calmodulin (CaM). Upon binding up to four calcium ions, CaM undergoes a conformational change, which enables it to bind to specific proteins eliciting a specific response. Calmodulin kinase II (CaMKII) is a major target of the Ca(2+)/CaM second messenger system. Once bound to Ca(2+)/CaM, the multimeric CaMKII is released from its autoinhibitory status and maximally activated, which then leads to an intraholoenzyme autophosphorylation reaction. Calcineurin (Cn) is another major target protein that is activated by Ca(2+)/CaM. Cn is a serine-threonine phosphatase that consists of a heterodimeric protein complex composed of a catalytic subunit (CnA) and a regulatory subunit (CnB). Upon activation, Cn directly binds to, and dephosphorylates nuclear factor of activated T cells (NFAT) transcription factors within the cytoplasm allowing them to translocate to the nucleus and participate in the regulation of gene expression. This review will examine the potential mechanisms by which calcium, CaM, CaMKII, and Cn/NFAT control osteoblast proliferation and differentiation. \n Question: Which calcium/calmodulin dependent protein phosphatase is involved in the activation of the family of NFAT transcription factors (Nuclear Factors of Activated T cells)?\n" }, { "from": "gpt", "value": "can" } ]
[ { "from": "human", "value": "You need to answer the question 'Are the given steps in order?', given a set of steps describing a process. Your answer must be either Yes or No. If the answer is No, that means the steps are out of order and do not make sense in the order they are in. If the answer is Yes, that means the steps are in order and make sense in the order that they are in. A set of steps are not in order if the steps reference information that is introduced in a later step\n\n ['The rock is broken down into smaller pieces', 'Erosion begins.', 'The roots of the growing plants start to break up the rock', 'The plant acids dissolve the rock', 'Plants start growing on or near the rocks'] " }, { "from": "gpt", "value": " No " } ]
[ { "from": "human", "value": "Given the task definition and input, reply with output. In this task, you are given a sentence in the Bulgarian language and corresponding English translation of this sentence. Here, your job is to generate label \"yes\" if translation is right, otherwise generate label \"no\".\n\nBulgarian: 3., English: Amendment of regulation on a standard financial regulation for executive agencies (Rule 131) (vote)\n" }, { "from": "gpt", "value": "no" } ]
[ { "from": "human", "value": "In this task, you need to provide the parts-of-speech tag of a word present in a sentence specified within curly braces ( '{{ ... }}' ). The parts-of-speech tags are coarse labels that represent a category of words with similar grammatical properties. The list of part-of-speech tags i.e. tagset of this corpus is 'ADJ': Adjectives are words that typically modify nouns and specify their properties or attributes, 'ADP': Adposition is a cover term for prepositions and postpositions, 'ADV': Adverbs are words that typically modify verbs for such categories as time, place, direction or manner, 'AUX': An auxiliary is a function word that accompanies the lexical verb of a verb phrase and expresses grammatical distinctions not carried by the lexical verb, such as person, number, tense, mood, aspect, voice or evidentiality, 'CCONJ': A coordinating conjunction is a word that links words or larger constituents without syntactically subordinating one to the other and expresses a semantic relationship between them, 'DET': Determiners are words that modify nouns or noun phrases and express the reference of the noun phrase in context, 'INTJ': An interjection is a word that is used most often as an exclamation or part of an exclamation, 'NOUN': Nouns are a part of speech typically denoting a person, place, thing, animal or idea, 'NUM': A numeral is a word, functioning most typically as a determiner, adjective or pronoun, that expresses a number and a relation to the number, such as quantity, sequence, frequency or fraction, 'PART': Particles are function words that must be associated with another word or phrase to impart meaning and that do not satisfy definitions of other universal parts of speech, 'PRON': Pronouns are words that substitute for nouns or noun phrases, whose meaning is recoverable from the linguistic or extralinguistic context, 'PROPN': A proper noun is a noun (or nominal content word) that is the name (or part of the name) of a specific individual, place, or object, 'PUNCT': Punctuation marks are non-alphabetical characters and character groups used in many languages to delimit linguistic units in printed text, 'SCONJ': A subordinating conjunction is a conjunction that links constructions by making one of them a constituent of the other. The subordinating conjunction typically marks the incorporated constituent which has the status of a (subordinate) clause, 'SYM': A symbol is a word-like entity that differs from ordinary words by form, function or both, 'VERB': A verb is a member of the syntactic class of words that typically signal events and actions, can constitute a minimal predicate in a clause, and govern the number and types of other constituents which may occur in the clause, 'X': The tag X is used for words that for some reason cannot be assigned a real part-of-speech category.\n\nSentence: Then , of course , there is the evidence the jury did not hear about in {{ the }} Robinson case ... \nWord: the" }, { "from": "gpt", "value": "DET" } ]
[ { "from": "human", "value": "In this task, you need to provide the correct option for a given problem from the provided options.\nOne example: Problem: jo ' s collection contains us , indian and british stamps . if the ratio of us to indian stamps is 6 to 2 and the ratio of indian to british stamps is 5 to 1 , what is the ratio of us to british stamps ? \nOptions: a ) 5 : 1 , b ) 30 : 2 , c ) 15 : 2 , d ) 20 : 2 , e ) 25 : 2\nSolution is here: b\nExplanation: \"u / i = 6 / 2 i / b = 5 / 1 since i is multiple of both 2 ( as per first ratio ) and 5 ( as per second ratio ) so let ' s assume that i = 10 i . e . multiplying teh first ratio by 5 and second ration by 2 in each numerator and denominator then , u : i : b = 30 : 18 : 2 i . e . u : b = 30 : 2 answer : option b\"\n\nNow, solve this: Problem: mrs . evans gave a test to her freshmen economics class , which has 29 students enrolled and 24 of them answered question 1 correctly . if 22 answered question 2 correctly and 5 did not take the test then how many answered both questions correctly ? \nOptions: a ) 18 , b ) 19 , c ) 22 , d ) 20 , e ) 19\nSolution:" }, { "from": "gpt", "value": "c" } ]
[ { "from": "human", "value": "Given the task definition and input, reply with output. You are given a science question (easy-level) and four answer options (associated with \"A\", \"B\", \"C\", \"D\"). Your task is to find the correct answer based on scientific facts, knowledge, and reasoning. Do not generate anything else apart from one of the following characters: 'A', 'B, 'C', 'D'. There is only one correct answer for each question.\n\nHow does ice change the shape of rocks? \n (A) It dissolves the rocks by pooling on surfaces. (B) It breaks the rocks by expanding in openings. (C) It smooths the rocks by colliding with them. (D) It moves the rocks by pressing on them.\n" }, { "from": "gpt", "value": "B" } ]
[ { "from": "human", "value": "Definition: In this task, given 2 input sentences, you must classify the relation between them. If the second sentence has a similar meaning to that of the first sentence then the output is ' B_entails_A', if the second sentence has the opposite meaning to the first sentence then it is classified as ' B_contradicts_A'. If you cannot clearly ascertain agreement/disagreement between the two sentences, the label is ' B_neutral_A'.\nInput: sentence_A: There is no guy leaping into the air in a wooded area. sentence_B: A guy is leaping into the air in a wooded area\nOutput:" }, { "from": "gpt", "value": "B_contradicts_A" } ]
[ { "from": "human", "value": "TASK DEFINITION: In this task you will be given a list of integers. You should remove any integer that is not prime. A prime integer is an integer that is only divisible by '1' and itself. The output should be the list of prime numbers in the input list. If there are no primes in the input list an empty list (\"[]\") should be returned.\nPROBLEM: [929, 96, 461, 143, 117, 271, 285, 257]\n\nSOLUTION: [929, 461, 271, 257]\n\nPROBLEM: [181, 822, 433, 71, 26, 827, 767, 877, 271, 671, 157, 200, 659]\n\nSOLUTION: [181, 433, 71, 827, 877, 271, 157, 659]\n\nPROBLEM: [618, 863, 724, 855, 173, 59, 617, 575, 451, 131, 766, 823, 71]\n\nSOLUTION:" }, { "from": "gpt", "value": "[863, 173, 59, 617, 131, 823, 71]\n\n" } ]
[ { "from": "human", "value": "Instructions: In this task, you will be presented with a passage, and you need to write an **implausible** answer to to fill in the place of \"_\". Your answer should be incorrect, but should not be out of context. Try using words that are related to the context of the passage, but are not the correct answer. Even though there exist multiple wrong answers, we only need a single wrong answer.\nInput: (CNN) Judges in two separate cases have ruled in favor of pharma giant Johnson & Johnson, overturning large financial judgments awarded to plaintiffs who believe the company's Baby Powder and Shower to Shower powder products caused their ovarian cancer. One award was to a California woman for $417 million, and another was to an Alabama woman for $72 million. In the California case, which was decided by a jury in August, Eva Echeverria testified that she had been using talcum powder as a regular part of her feminine hygiene routine for more than 50 years, since she was 11. She developed ovarian cancer and stopped using the powder in 2016, after she read a news story about another woman who used it and had ovarian cancer.Judge overturns a $417 million verdict against Johnson & Johnson in talcum powder caseA woman claimed that regular use of baby powder caused her ovarian cancerThousands of similar cases are pending in courts across the country \n Question:That case involved another lifelong talc user who had lived in _ and died four months before the trial.\nOutput:" }, { "from": "gpt", "value": "Johnson & Johnson" } ]
[ { "from": "human", "value": "Part 1. Definition\nGiven a question and a context passage, generate the answer having the word or phrase from the context passage. Here, the answer should be a shortest continous span from the passage.\nPart 2. Example\nContext: Architecturally, the school has a Catholic character. Atop the Main Building's gold dome is a golden statue of the Virgin Mary. Immediately in front of the Main Building and facing it, is a copper statue of Christ with arms upraised with the legend 'Venite Ad Me Omnes'. Next to the Main Building is the Basilica of the Sacred Heart. Immediately behind the basilica is the Grotto, a Marian place of prayer and reflection. It is a replica of the grotto at Lourdes, France where the Virgin Mary reputedly appeared to Saint Bernadette Soubirous in 1858. At the end of the main drive (and in a direct line that connects through 3 statues and the Gold Dome), is a simple, modern stone statue of Mary. \n Question: To whom did the Virgin Mary allegedly appear in 1858 in Lourdes France?\nAnswer: Saint Bernadette Soubirous\nExplanation: This is the correct answer. Because the question is asking for the name of the person and following sentence from the passage answers it: It is a replica of the grotto at Lourdes, France where the Virgin Mary reputedly appeared to Saint Bernadette Soubirous in 1858.\nPart 3. Exercise\nContext: Ultrasound is sound that has a wave frequency higher than the human ear can detect. It includes all sound with wave frequencies higher than 20,000 waves per second, or 20,000 hertz (Hz). Although we cant hear ultrasound, it is very useful to humans and some other animals. Uses of ultrasound include echolocation, sonar, and ultrasonography. Animals such as bats and dolphins send out ultrasound waves and use their echoes, or reflected waves, to identify the locations of objects they cannot see. This is called echolocation. Animals use echolocation to find prey and avoid running into objects in the dark. You can see in the Figure 1.1 how a bat uses echolocation to find insect prey. Sonar uses ultrasound in a way that is similar to echolocation. Sonar stands for sound navigation and ranging. It is used to locate underwater objects such as submarines. Thats how the ship pictured in the Figure 1.2 is using it. A sonar device is both a sender and a receiver. It sends out ultrasound waves and detects the waves after they reflect from underwater objects. The distance to underwater objects can be calculated from the known speed of sound in water and the time it takes for the sound waves to travel to the object. The equation for distance traveled when speed and time are known is: Distance = Speed Time Consider the ship and submarine pictured in the Figure 1.2. If an ultrasound wave travels from the ship to the submarine and back again in 2 seconds, what is the distance from the ship to the submarine? The sound wave travels from the ship to the submarine in just 1 second, or half the time it takes to make the round trip. The speed of sound waves through ocean water is 1437 m/s. Therefore, the distance from the ship to the submarine is: Q: Now assume that the sonar device on the ship sends an ultrasound wave to the bottom of the water. If the sound wave is reflected back to the device in 4 seconds, how deep is the water? A: The time it takes the wave to reach the bottom is 2 seconds. So the distance from the ship to the bottom of the water is: Distance = 1437 m/s 2 s = 2874 m Another use of ultrasound is to see inside the human body. This use of ultrasound is called ultrasonography. Harmless ultrasound waves are sent inside the body, and the reflected waves are used to create an image on a screen. This technology is used to examine internal organs and unborn babies without risk to the patient. You can see a doctor using ultrasound in the Figure 1.3. \n Question: the speed of sound waves through ocean water is 1437 m/s. assume that ultrasound waves take 2 seconds to travel from a sonar device on a ship to the bottom of the water and back to the ship again. how deep is the water?\nAnswer:" }, { "from": "gpt", "value": "1437 m" } ]
[ { "from": "human", "value": "Teacher: In this task, You are given an amazon review and its polarity (positive or negative). Your task is to answer \"True\" if the specified sentence and its polarity match; otherwise, answer \"False\"\nTeacher: Now, understand the problem? If you are still confused, see the following example:\nReview: These are junk! Both bulbs have burned out within a two month period!! I cannot believe that this company can be in business with such poor quality. I have used infrared lights for my ball python for many years now and I get a varied range of months from a bulb, but I have never gone through two bulbs in a matter of two months! I am very disappointed. \n Polarity: negative\nSolution: True\nReason: The user seems very disappointed in the product because the bulb they ordered has been burned out earlier than usual. So, it's a negative review, and True is the correct answer.\n\nNow, solve this instance: Review: Speechless!!! i cannot believe i haven't heard of this before, Best Quick portable charger with the convenience to charge my device on the go! the design is so slick and the device fits right in effortlessly , It charges up extremely fast and has a great feel to it as well, Must have for anyone with a Juul Device \n Polarity: positive\nStudent:" }, { "from": "gpt", "value": "True" } ]
[ { "from": "human", "value": "In this task, you will use your knowledge about language (and common sense) to determine what element the marked number refers to. The numbers are marked with two underlines around them, like: _ number _. There are several possible answers, you'll need to choose the proper one. Carefully read the given text, pay special attention to the marked number, think about what (unwritten) information the marked number holds inside, choose the most adequate word(s) from the optional answers. If none of them seems right to you, there's also an option for other. If your answer is \"REFERENCE\", also write the reference entity, otherwise write the implicit option name. Options to choose from are:\nREFERENCE: Some object which is being mentioned in the text before or after the target number. The reference answer has a higher priority than any other. If both Reference and another answer are possible, prioritize the Reference.\nYEAR: Describing a calendric year\nAGE: Describing someone's age\nCURRENCY: Reference to some monetary value e.g dollar, euro etc.\nPEOPLE: Describing a single/plural persons\nTIME: Describing a time of the day. Usually you can add the word o'clock after those numbers.\nOTHER: Some other option, which isn't listed here.\n\nExample Input: Broton: Strength ?\nZygon: Diastellic reading seven oh _ three _ .\nBroton: Increase the sonic core tone by three remars .\nZygon: Increased tone , three remars . Contact firm .\nBroton: Check directional pulse .\nZygon: Pulse correct to within one Earth mile . Closing .\nExample Output: OTHER\n\nExample Input: Kat: You must be Mr. Big .\nAndy: And who are you ? Mrs. Big Mouth ?\nKat: Got it in _ one _ .\nExample Output: OTHER\n\nExample Input: Bulma: Hey , do n't touch my Pearl !\nSon Goku: Is that Pearl yours ?\nBulma: Of course it 's mine , my father gave it to me !\nSon Goku: I 've got one at home too !\nBulma: Huh ? Do you really ? Mine has five stars , how many stars does yours have ?\nSon Goku: It 's got _ four _ !\nBulma: WOW ! I FOUND IT ! I FOUND IT ! YAY , I FOUND IT !\nExample Output:" }, { "from": "gpt", "value": "REFERENCE stars\n\n" } ]
[ { "from": "human", "value": "Given a premise, an initial context, an original ending, and a new ending, the task is to generate the counterfactual context that is aligned with the new ending. Each instance consists of a five-sentence story. The premise is the first sentence of a story, and the second sentence, which is the initial context, provides more information about the story's context. The original ending is the last three sentences of the story. Also, you are given a new ending that can provide the new story with the same premise. You should write a counterfactual context to explain the small differences between the original and new endings. More specifically, a story context contains the general plot of the story. And a counterfactual context is a slight modification to the initial context. To sum up, you should write the second sentence of a story based on the premise(first sentence) and the new ending(last three sentences) of the story.\n\nExample Input: Premise: Steve loved to play his guitar on the sidewalk. \n Initial Context: He went out every day and people would give him tips. \n Original Ending: One day a music producer heard him performing. The producer asked Steve to meet with him to discuss business. Steve got to sign to a record label and make his first album. \n New ending: One day a movie producer heard him performing. The producer asked Steve to meet with him to discuss business. Steve got to sign to a contract and record the music for a movie. \nExample Output: People would never give him any money.\n\nExample Input: Premise: Sally wanted a child very badly, but she had no husband! \n Initial Context: She was growing older and becoming desperate. \n Original Ending: One sad night, Sally drank too much and had a one night stand. That night she conceived a child! She raised the baby all alone, finally happy to be a mommy! \n New ending: One happy day, a doctor told her she was pregnant. One night she conceived a child! She raised the baby all alone, finally happy to be a mommy! \nExample Output: She was growing older so decided to visit a sperm bank for donors.\n\nExample Input: Premise: I used to sell things on eBay. \n Initial Context: I would check the recycle bins outside our dumpster. \n Original Ending: One day a person left a hurling stick. It had some chips but was still solid. I sold it on eBay for $80. \n New ending: A person had left the hurling stick near the trash. It had some chips but was still solid. I tried to sell it on eBay for $80. \nExample Output:" }, { "from": "gpt", "value": "It's harder to sell stuff on ebay now because of all the stores there so I couldn't sell the hurling stick.\n\n" } ]
[ { "from": "human", "value": "In this task, you are presented with a term, a description of the term, and an expected answer ('yes' or 'no'). You should write a yes-no question about the given term such that the answer is the one provided to you (i.e., If the answer is \"No\", you should ask a question that its answer would be \"No\", and if the answer is \"Yes\", you should ask a question that its answer is \"Yes\". ). The question should have a definitive answer (as opposed to ambiguous or subjective questions, e.g., Is Batman a good movie?). Create a question such that its answer can not be found easily on a single web page (e.g., mentioned in a Wikipedia page). This can be accomplished if answering the question requires more than one fact (facts = statements that can be found on a Wikipedia page or Google). For example, answering the question 'did Aristotle use a laptop?', one needs the know about the invention of the laptop and the death of Aristotle. Avoid questions should not just compare the properties of objects (e.g., Is a door bigger than an elephant?) or those that refer to details in the given description.\nExample: Term: Cooking oil, Description: Oil consumed by humans, from vegetable or animal origin., Answer:No\nExample solution: Can all types of cooking oil be poured?\nExample explanation: This is a good question. For answering this question, you need to know different cooking oils and whether they can be poured.\n\nProblem: Term: History of Europe, Description: History of Europe from the beginnings of recorded history, Answer:No\n" }, { "from": "gpt", "value": "Solution: Does the history of Europe include the age of dinosaurs?" } ]
[ { "from": "human", "value": "Detailed Instructions: In this task, you're given a context passage. Your job is to generate relevant questions that can be answered by directly referring to the passage.\nQ: Remy said the picture looked like it was going down hill .\nA:" }, { "from": "gpt", "value": "How would Remy feel afterwards?" } ]
[ { "from": "human", "value": "Given a set of four words, generate the category that the words belong to. Words are separated by commas. The possible categories are social gathering, accomodation, physical property, measurement unit, corporate, nutritional value, boats, police punishment, location (proximity), card games, outdoor sport activity, military rank, baby animals, weather, consumer product, animals, boxing terminology, food, chocolate food, bird, type of sandwich, date status, body part, ocean, employment activity, moral characteristic, elements, poker, computers, construction, guitar part, shady activities, physical activity, kitchenware, temperature, type of rigidity, emotional status, season, mode of transportation, window material, activity, emotional display, geographical property, fried food, store status, widespread, aesthetic characteristic, alcoholic drinks, secretary duty, direction, personal characteristic, and animal.\n\n[Q]: lettuce, toast, burger, club\n[A]: type of sandwich\n\n\n[Q]: basic, typical, regular, general\n[A]: widespread\n\n\n[Q]: vitamin c, salt, fat, big\n[A]:" }, { "from": "gpt", "value": "nutritional value\n\n\n" } ]
[ { "from": "human", "value": "Given the task definition, example input & output, solve the new input case.\nIn this task, you are given commands (in terms of logical operations) to select relevant rows from the given table. Your job is to generate a natural language interpretation of the given command. \n Here are the definitions of logical operators: \n 1. count: returns the number of rows in the view \n 2. only: returns whether there is exactly one row in the view \n 3. hop: returns the value under the header column of the row \n 4. and: returns the boolean operation result of two arguments \n 5. max/min/avg/sum: returns the max/min/average/sum of the values under the header column \n 6. nth max/nth min: returns the n-th max/n-th min of the values under the header column \n 7. argmax/argmin: returns the row with the max/min value in header column \n 8. nth argmax/nth argmin: returns the row with the n-th max/min value in header column \n 9. eq/not eq: returns if the two arguments are equal \n 10. round eq: returns if the two arguments are roughly equal under certain tolerance \n 11. greater/less: returns if argument 1 is greater/less than argument 2 \n 12. diff: returns the difference between two arguments \n 13. filter eq/not eq: returns the subview whose values under the header column is equal/not equal to argument 3 \n 14. filter greater/less: returns the subview whose values under the header column is greater/less than argument 3 \n 15. filter greater eq /less eq: returns the subview whose values under the header column is greater/less or equal than argument 3 \n 16. filter all: returns the view itself for the case of describing the whole table \n 17. all eq/not eq: returns whether all the values under the header column are equal/not equal to argument 3 \n 18. all greater/less: returns whether all the values under the header column are greater/less than argument 3 \n 19. all greater eq/less eq: returns whether all the values under the header column are greater/less or equal to argument 3 \n 20. most eq/not eq: returns whether most of the values under the header column are equal/not equal to argument 3 \n 21. most greater/less: returns whether most of the values under the header column are greater/less than argument 3 \n 22. most greater eq/less eq: returns whether most of the values under the header column are greater/less or equal to argument 3\nExample: eq { hop { nth_argmax { all_rows ; attendance ; 3 } ; competition } ; danish superliga 2005 - 06 } = true\nOutput: select the row whose attendance record of all rows is 3rd maximum. the competition record of this row is danish superliga 2005-06.\nHere, the generated sentence interprets given logical form correctly that 3rd maximum should be selected from given table rows.\n\nNew input case for you: eq { hop { nth_argmin { all_rows ; discontinued ; 2 } ; model } ; lisa 2 } = true\nOutput:" }, { "from": "gpt", "value": "select the row whose discontinued record of all rows is 2nd minimum . the model record of this row is lisa 2 ." } ]
[ { "from": "human", "value": "Q: In this task you will be given a list of integers. A list contains numbers separated by a comma. You need to round every integer to the closest power of 2. A power of 2 is a number in the form '2^n', it is a number that is the result of multiplying by 2 n times. The following are all powers of 2, '2, 4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096'. If an integer is exactly in equally far from two different powers of 2 then you should output the larger power of 2. The output should be a list of integers that is the result of rounding each integer int the input list to the closest power of 2. The output should include a '[' to denote the start of the output list and ']' to denote the end of the output list.\n[132, 956, 3949, 1357, 16, 35]\nA:" }, { "from": "gpt", "value": "[128, 1024, 4096, 1024, 16, 32]" } ]
[ { "from": "human", "value": "You are given two sentences(Sentence1 and Sentence2). Answer \"Yes\" if these sentences are a paraphrase of one another, otherwise answer \"No\".\n\nInput: Consider Input: Sentence1: The group ambitiously expects to issue guidelines by the end of the year and release products by late 2004 ., Sentence2: The group expects to roll out the first products by the second half of 2004 .\n\nOutput: No\n\n\nInput: Consider Input: Sentence1: Country-music station KKCS has suspended two disc jockeys for playing songs by the Dixie Chicks in violation of a ban imposed after one group member criticized President George Bush ., Sentence2: A radio station has suspended two disc jockeys for locking themselves in the studio and continuously playing Dixie Chicks songs , violating the station 's two-month-old ban on the group 's music .\n\nOutput: No\n\n\nInput: Consider Input: Sentence1: \" It ' ' s absurd , \" Funny Cide 's trainer Barclay Tagg said ., Sentence2: Meanwhile , Funny Cide 's trainer , Barclay Tagg , called the allegations \" ridiculous . \"\n\n" }, { "from": "gpt", "value": "Output: Yes\n\n\n" } ]
[ { "from": "human", "value": "TASK DEFINITION: In this task, you're given a pair of sentences, sentence 1 and sentence 2. Your job is to choose whether the two sentences clearly agree (entailment)/disagree (contradiction) with each other, or if this cannot be determined (neutral). Your answer must be in the form of the letters E, C, and N respectively.\nPROBLEM: Sentence 1: Child playing on a playground, hanging from bars. Sentence 2: A dog is eating some food.\n\nSOLUTION: C\n\nPROBLEM: Sentence 1: Two girls in colourful dresses blowing bubbles at a park. Sentence 2: Two girls blow bubbles.\n\nSOLUTION: C\n\nPROBLEM: Sentence 1: The dog has red straps in its back. Sentence 2: The dog with red stripes can fly.\n\nSOLUTION:" }, { "from": "gpt", "value": "N\n\n" } ]
[ { "from": "human", "value": "In this task, you are given a passage and a question regarding that passage. You must determine whether or not the question is answerable from the given passage. If a question is answerable, output should be 'True', otherwise 'False'. You must not consider any other information that is not provided in the passage while labelling True or False.\n\nEx Input:\nPassage: The weight of boilers and condensers generally makes the power-to-weight ratio of a steam plant lower than for internal combustion engines. For mobile applications steam has been largely superseded by internal combustion engines or electric motors. However, most electric power is generated using steam turbine plant, so that indirectly the world's industry is still dependent on steam power. Recent concerns about fuel sources and pollution have incited a renewed interest in steam both as a component of cogeneration processes and as a prime mover. This is becoming known as the Advanced Steam movement.[citation needed] Question: What plants create most electric power?\n\nEx Output:\nTrue\n\n\nEx Input:\nPassage: One of the more notable applications of printing technology was the chao, the paper money of the Yuan. Chao were made from the bark of mulberry trees. The Yuan government used woodblocks to print paper money, but switched to bronze plates in 1275. The Mongols experimented with establishing the Chinese-style paper monetary system in Mongol-controlled territories outside of China. The Yuan minister Bolad was sent to Iran, where he explained Yuan paper money to the Il-khanate court of Gaykhatu. The Il-khanate government issued paper money in 1294, but public distrust of the exotic new currency doomed the experiment. Question: When did the Il-khanate stop experimenting with paper money?\n\nEx Output:\nFalse\n\n\nEx Input:\nPassage: Connection-oriented transmission requires a setup phase in each involved node before any packet is transferred to establish the parameters of communication. The packets include a connection identifier rather than address information and are negotiated between endpoints so that they are delivered in order and with error checking. Address information is only transferred to each node during the connection set-up phase, when the route to the destination is discovered and an entry is added to the switching table in each network node through which the connection passes. The signaling protocols used allow the application to specify its requirements and discover link parameters. Acceptable values for service parameters may be negotiated. Routing a packet requires the node to look up the connection id in a table. The packet header can be small, as it only needs to contain this code and any information, such as length, timestamp, or sequence number, which is different for different packets. Question: How is error checking involved in delivery? \n\nEx Output:\n" }, { "from": "gpt", "value": "False\n\n\n" } ]
[ { "from": "human", "value": "In this task, you need to reverse all words of a given length in the sentence. The number of letters in a word determine its length. For example, the length of the word \"apple\" is 5.\nQ: Sentence: 'a giraffe standing in the middle of a forest'. Reverse all words of length '2' in the given sentence.\nA: " }, { "from": "gpt", "value": "a giraffe standing ni the middle fo a forest" } ]
[ { "from": "human", "value": "This task involves annotating the answer type to a given question that involve some kind of complex reasoning (including numerical reasoning). Note that the questions require looking at more than one part of the passage to answer. There are 3 possible answer types (i) spans, (ii) numbers and (iii) dates. If the answer can be found in the passage, label it as \"span\". If the answer is a number, label as \"number\". Similarly, label \"date\" if you think the answer to the given question is a date.\n\n[Q]: Passage: During the 1970s the oil shocks (1973 oil crisis and 1979 energy crisis) hit Iceland hard. Inflation rose to 43% in 1974 and 59% in 1980, falling to 15% in 1987 but rising to 30% in 1988. Iceland experienced moderately strong GDP growth (3% on average) from 1995 to 2004. Growth slowed between 2000 and 2002, but the economy expanded by 4.3% in 2003 and grew by 6.2% in 2004. Growth in 2005 exceeded 6%. Inflation averaged merely 1.5% from 1993 to 1994, and only 1.7% from 1994 to 1995. Inflation over 2006 topped at 8.6%, with a rate of 6.9% as of January 2007. Standard & Poors reduced their rating for Iceland to AA- from A+ (long term) in December 2006, following a loosening of fiscal policy by the Icelandic government ahead of the 2007 elections. Foreign debt rose to more than five times the value of Icelands GDP, and Icelands Central Bank raised short-term interest rates to nearly 15% in 2007. Due to the plunging currency against the euro and dollar, in 2008 inflation was speculated to be at 20-25%.\nQuestion: Which year was the inflation average higher, from 1993 to 1994 or 1994 to 1995?\n[A]: span\n\n\n[Q]: Passage: On 28 March 1193 Richard was brought to Speyer and handed over to Henry VI, Holy Roman Emperor, who imprisoned him in Trifels Castle. Henry VI was aggrieved by the support the Plantagenets had given to the family of Henry the Lion and by Richard's recognition of Tancred in Sicily. Henry VI needed money to raise an army and assert his rights over southern Italy and continued to hold Richard for ransom. In response, Pope Celestine III excommunicated Henry VI, as he had Duke Leopold, for the continued wrongful imprisonment of Richard. Richard famously refused to show deference to the emperor and declared to him, \"I am born of a rank which recognises no superior but God\". Despite his complaints, the conditions of his captivity were not severe. The emperor demanded that 150,000 marks be delivered to him before he would release the king, the same amount raised by the Saladin tithe only a few years earlier, and 2-3 times the annual income for the English Crown under Richard. Richard's mother, Eleanor of Aquitaine, worked to raise the ransom. Both clergy and laymen were taxed for a quarter of the value of their property, the gold and silver treasures of the churches were confiscated, and money was raised from the scutage and the carucage taxes. At the same time, John, Richard's brother, and King Philip of France offered 80,000 marks for the Emperor to hold Richard prisoner until Michaelmas 1194. The emperor turned down the offer. The money to rescue the King was transferred to Germany by the emperor's ambassadors, but \"at the king's peril\" , and finally, on 4 February 1194 Richard was released. Philip sent a message to John: \"Look to yourself; the devil is loose\".\nQuestion: How many different people with the name Henry are referenced?\n[A]: number\n\n\n[Q]: Passage: Salin turned out to be the last battle of the war. After the successive battlefield defeats, the bickering between the Confederation's two main powers broke out in full force. Sawlon II of Mohnyin, who had only grudgingly agreed to Hkonmaing's takeover of the Ava throne, blamed Hkonmaing for the defeats, and now planned to put Sithu Kyawhtin I of Salin on the Ava throne. In April/May 1545, Sawlon II sent Sithu Kyawhtin with an army , which went on to occupy Sagaing, the city directly across the Irrawaddy from Ava. During the rainy season, c. September 1545, Hkonmaing died, and was succeeded by his son Narapati III. The new king promptly sent a mission to Pegu to secure friendly relations in exchange for his recognition of the new de facto border between the two kingdoms. Tabinshwehti accepted the offer. Toungoo's decisive victory gave the upstart kingdom control of all of central Burma, and cemented its emergence as the largest polity in Burma since the fall of Pagan Empire in 1287. Indeed, \"there was once more a king in Burma\". In the following years, Ava and Mohnyin-backed Sagaing would be locked in a war until 1551, while an emboldened Toungoo would turn its attention to conquering Arakan in 1545-47, and Siam in 1547-49.\nQuestion: In what year did the Toungoo's conquering of Arakan and Siam overlap?\n[A]:" }, { "from": "gpt", "value": "date\n\n\n" } ]
[ { "from": "human", "value": "Instructions: In this task, a passage will be given and the goal is to generate a question about temporal relations based on that passage. A temporal relation describes the relation between two things with respect to time e.g., something happens/starts/terminates/... immediately/shortly/usually/... before/after/during/... something else.\n Your question should include an event, or ask about an event, and should not be a yes/no question.\nInput: Passage: The mission includes 13 Italians, despite political uncertainty in Rome over whether Italy's leadership of the force will get the required backing of parliament in a vote which was to be held later Wednesday. The advance team arrived in Tirana late Tuesday.\nOutput:" }, { "from": "gpt", "value": "What event has already finished?" } ]
[ { "from": "human", "value": "In this task, you're given a review from Amazon and rating for the product on a scale of 1-5 based on the review. The rating means 1: extremely poor, 2: poor, 3: neutral, 4: good, 5: extremely good. Your task is to generate whether the rating matches the review or not. Answer with \"True\" if the sentence belongs to that section, otherwise answer with \"False\"\nQ: Review: Have owned these slippers before and they were sooooooooo comfy. This pair were very stiff and uncomfortable. Not nearly as soft as my original pair. Very difficult to walk in. \n Rating: 2\nA: " }, { "from": "gpt", "value": "True" } ]
[ { "from": "human", "value": "In this task, you are given commands (in terms of logical operations) to select relevant rows from the given table. Your job is to generate a natural language interpretation of the given command. \n Here are the definitions of logical operators: \n 1. count: returns the number of rows in the view \n 2. only: returns whether there is exactly one row in the view \n 3. hop: returns the value under the header column of the row \n 4. and: returns the boolean operation result of two arguments \n 5. max/min/avg/sum: returns the max/min/average/sum of the values under the header column \n 6. nth max/nth min: returns the n-th max/n-th min of the values under the header column \n 7. argmax/argmin: returns the row with the max/min value in header column \n 8. nth argmax/nth argmin: returns the row with the n-th max/min value in header column \n 9. eq/not eq: returns if the two arguments are equal \n 10. round eq: returns if the two arguments are roughly equal under certain tolerance \n 11. greater/less: returns if argument 1 is greater/less than argument 2 \n 12. diff: returns the difference between two arguments \n 13. filter eq/not eq: returns the subview whose values under the header column is equal/not equal to argument 3 \n 14. filter greater/less: returns the subview whose values under the header column is greater/less than argument 3 \n 15. filter greater eq /less eq: returns the subview whose values under the header column is greater/less or equal than argument 3 \n 16. filter all: returns the view itself for the case of describing the whole table \n 17. all eq/not eq: returns whether all the values under the header column are equal/not equal to argument 3 \n 18. all greater/less: returns whether all the values under the header column are greater/less than argument 3 \n 19. all greater eq/less eq: returns whether all the values under the header column are greater/less or equal to argument 3 \n 20. most eq/not eq: returns whether most of the values under the header column are equal/not equal to argument 3 \n 21. most greater/less: returns whether most of the values under the header column are greater/less than argument 3 \n 22. most greater eq/less eq: returns whether most of the values under the header column are greater/less or equal to argument 3\nQ: and { only { filter_eq { all_rows ; competition ; 2014 fifa world cup qualification } } ; eq { hop { filter_eq { all_rows ; competition ; 2014 fifa world cup qualification } ; date } ; 23 march 2013 } } = true\nA: " }, { "from": "gpt", "value": "select the rows whose competition record fuzzily matches to 2014 fifa world cup qualification . there is only one such row in the table . the date record of this unqiue row is 23 march 2013 ." } ]
[ { "from": "human", "value": "Given an trivia question precisely answer the question with a word/phrase/name. External resources such as Wikipedia could be used to obtain the facts.\n\nAshtanga Vinyasa, Bikram, and Hatha Vidya are styles of what?" }, { "from": "gpt", "value": "yoga" } ]
[ { "from": "human", "value": "In this task you will be given a list of numbers. You should remove any number that is not an integer (whole number). If every number is not an whole number then an empty list (\"[]\") should be returned. Otherwise, answer with the list of whole numbers separated by comma inside brackets.\nExample: [73.059, 7, 10.659, 18.459, 11]\nExample solution: [7, 11]\nExample explanation: 7 and 11 are the only whole numbers in the list.\n\nProblem: [94, 0.634, 1.092, 50, 15.447, -18, 70]\n" }, { "from": "gpt", "value": "Solution: [94, 50, -18, 70]" } ]
[ { "from": "human", "value": "Given a piece of financial news and its polarity, classify it into 'true' if the polarity is correct and classify into 'false' if the polarity is incorrect. Output must be 'true' or 'false'. \n\nnews:As an overall effect of the above-mentioned measures Glaston will recognize non-recurring expenses of EUR 12 million in the fourth quarter of 2008 .\npolarity:negative" }, { "from": "gpt", "value": "false" } ]
[ { "from": "human", "value": "Instructions: In this task, you're given a fill-in-the-blank question that contains two object names. Additionally, you're given one answer which is one of the objects present in the question. In this task, you need to minimally change the given question so that the answer flips to another object in the question. Remember the question has to be about both objects which are related but different; for example, \"trophy\" and \"suitcase\". This task typically involves replacing one word i.e., the 'trigger word' with its antonym (e.g., changing from \"small\" to \"big\"). You should not change any content in the given question beyond a word or two, i.e., the trigger word/phrase. The expected answer to your question must not be associated with just the trigger word; instead, it should depend on the context present in the question. The expected answer should not be ambiguous. For your question, there should be an agreed upon answer to fill in the blank. Your generations should NOT contain potentially explicit, offensive, or adult content. In your generated question, retain the object names from the input question. Avoid repeating the same style or phrase in generating your modified question e.g. this task can be always solved using a simple negation i.e. by adding not, never, etc. Instead, try to increase the word diversity. Your question must contain at least 15 and at most 30 words. Your question must have at least 70% overlapping words with the given question. You must utilize the given context word while writing the question. Your question must contain only one blank. The two objects should be used ONCE in your question. Here is a list of attributes and associated contrastive words that may help write contrastive trigger words. Note that contrastive trigger words help in flipping the label.\n| Attribute | triggerword | contrastive triggerword | \n| age | old | new | \n| altitude | low | high |\n| area | small | vast | \n| brightness | dark | light | \n| clarity | obscure | clear | \n| cleanness | dirty | clean | \n| complexity | simple | complex | \n| cost | cheap | expensive |\n| density | sparse | dense |\n| depth | shallow | deep |\n| distance | near | far | \n| electric conductivity | low | high |\n| flexibility | rigid | flexible |\n| granularity | fine | coarse | \n| hardness | soft | hard | \n| length | short | long | \n| magnitude | small | large | \n| mass | small | large | \n| odor | weak | strong | \n| pressure | low | high | \n| resistance | low | high | \n| shape | round | sharp |\n| shape | flat | spiky |\n| size | small | large | \n| sound | quiet | loud | \n| sound pitch | low | high | \n| speed | slow | fast | \n| stability | unstable | stable | \n| strength | weak | strong | \n| temperature | low | high | \n| texture | smooth | rough | \n| thermal conductivity | low | high |\n| thickness | thin | thick | \n| volume | small | large |\n| weight | light | heavy | \n| width | narrow | wide | \n| location | in | out |\n| location | up | down |\n| location | above | below |\n| location | on | off |\n| location | to | from |\n\nInput: Context Word: race. \nQuestion: He had to pull his car out of the race after a tire blew becuase the _ is useless. \nAnswer: car.\nOutput:" }, { "from": "gpt", "value": "He had to pull his car out of the race after a tire blew becuase the _ is damaged." } ]
[ { "from": "human", "value": "You will be given a definition of a task first, then some input of the task.\nIn this task, you're given a short article. Your job is to classify the article based on its category. Use the following classification labels, 0. World, 1. Sports, 2. Business, 3. Science or Technical. Label the text \"0\" if it contains information related to world. Label the text \"1\" if it contains information related to sports. Label the text \"2\" if it contains information related business. Label the text \"3\" if it contains science or technical related information. \n\nLondon shares up as Wall St climbs, WPP slips after Grey deal (AFP) AFP - Blue chips closed comfortably higher, lifted in late afternoon trade by a positive early performance by Wall Street, while a firm media sector and recovering oil stocks also boosted the FTSE 100, dealers said.\nOutput:" }, { "from": "gpt", "value": "0" } ]