id
stringlengths 1
12
| question
stringlengths 22
566
| golden_answers
dict | data_source
stringclasses 2
values | prompt
listlengths 1
1
| ability
stringclasses 6
values | reward_model
dict | extra_info
dict | metadata
dict |
|---|---|---|---|---|---|---|---|---|
exfever_1389
|
Hotel Transylvania is Todd Durham's first animated film.
|
{
"evidence": [
"Hotel Transylvania",
"Andy Samberg",
"Celeste and Jesse Forever"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information above.\n\n<think>\n...\n</think>\n- Use for every piece of reasoning.\n- During reasoning, you must verify the claim step by step based on the given information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer>\n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\nVerify the claim:\nHotel Transylvania is Todd Durham's first animated film.\n<information>\n[[Andy Samberg]]: Andrew David Samberg (born David A. J. Samberg; August 18, 1978) is an American actor, filmmaker, musician and comedian.\n\n[[Re-Animated]]: Re-Animated is a live-action/animated television film that aired on Cartoon Network on December 8, 2006. It was the first original production on the genre for the network. The movie was released on DVD September 11, 2007, and serves as pilot for the follow-up series \"Out of Jimmy's Head\", which premiered on September 14, 2007.\n\n[[Celeste and Jesse Forever]]: Celeste and Jesse Forever is a 2012 American romantic comedy-drama film directed by Lee Toland Krieger. It stars Rashida Jones and Andy Samberg, and was written by Jones and Will McCormack, who also has a role in the film. It was released on August 3, 2012, in New York and Los Angeles.\n\n[[Ruth Paine]]: Ruth Hyde Paine (born September 3, 1932) was a friend of Marina Oswald, who was living with her at the time of the JFK assassination. According to five government investigations, Lee Harvey Oswald stored the 6.5 mm caliber Carcano rifle that he used to assassinate U.S. President John F. Kennedy in Ruth Paine's garage, unbeknownst to her and her husband, Michael Paine.\n\n[[Codex Vaticanus 2066]]: Codex Vaticanus 2066, designed by 046 (in the Gregory-Aland numbering), α 1070 (von Soden), formerly it was known also as \"Codex Basilianus\", previously it was designated by B or B. It is a Greek uncial manuscript of the New Testament written on vellum. The manuscript Paleographically has been assigned to the 10th century by the INTF, though some palaeographers proposed the 9th century. Scrivener proposed even the 8th century.\n\n[[Bakmi]]: Bakmi consists of two Hokkien Chinese words literally translated to English as \"meat noodles\" (肉麵 , ). \"Bakmi\" is a wheat based noodle which was brought to Southeast Asia by Chinese immigrants with Fujian or Hokkien origin, generally prepared and topped with minced pork seasoned in soy sauce and few sliced of \"char siu\" (叉燒) or barbecued pork, with addition of Chinese green vegetables and a bowl of broth.\n\n[[Hotel Transylvania]]: Hotel Transylvania is a 2012 American 3D computer-animated fantasy comedy film produced by Sony Pictures Animation for Columbia Pictures. It was directed by Genndy Tartakovsky and produced by Michelle Murdocca, and stars the voices of Adam Sandler, Andy Samberg, Selena Gomez, Kevin James, Steve Buscemi and Cee Lo Green. The film tells a story of Count Dracula, the owner of a hotel called Hotel Transylvania where the world's monsters can take a rest from human civilization. Dracula invites some of the most famous monsters to celebrate the 118th birthday of his daughter Mavis. When the \"human-free hotel\" is unexpectedly visited by an ordinary 21-year-old traveler named Jonathan, Dracula must protect Mavis from falling in love with him before the hotel's guests learn that there is a human in the castle, which may jeopardize the hotel's future and his career.\n\n[[2007 Speedway World Cup]]: The 2007 Speedway World Cup (SWC) was the 7th FIM Speedway World Cup season. The Final took place on July 21, 2007 in the Alfred Smoczyk Stadium in Leszno, Poland. The tournament was won by host team Poland (55 pts) and they beat defending champion Denmark (52 pts), Australia (29 pts) and Great Britain (15 pts) in the Final.\n</information>\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hotel Transylvania",
"Andy Samberg",
"Celeste and Jesse Forever"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 500,
"split": "test_off"
}
|
{
"annotator_operations": []
}
|
61613
|
The town, where a fossil of the world's largest moth was discovered in 1966, holds an annual Mary Anning Day and Lyme Regis Fossil Festival.
|
{
"evidence": [
"Lyme Regis_sentence_45",
"Lyme Regis_sentence_46"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information above.\n\n<think>\n...\n</think>\n- Use for every piece of reasoning.\n- During reasoning, you must verify the claim step by step based on the given information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer>\n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\nVerify the claim:\nThe town, where a fossil of the world's largest moth was discovered in 1966, holds an annual Mary Anning Day and Lyme Regis Fossil Festival.\n<information>\n[[Lyme Regis_sentence_3]]: The harbour wall, known as The Cobb, appears in [[Jane_Austen|Jane Austen]]'s novel [[Persuasion_(novel)|Persuasion]], the [[John_Fowles|John Fowles]] novel [[The_French_Lieutenant's_Woman|The French Lieutenant's Woman]] and the 1981 [[The_French_Lieutenant's_Woman_(film)|film of that name]], partly shot in the town.\n[[Lyme Regis_sentence_20]]: [[Lassie#Radio|Lassie]], the owner's dog, licked the face of Seaman Cowan, who was believed dead, and seemingly brought him back to life.\n[[Lyme Regis_sentence_23]]: The station was dismantled and rebuilt at [[Alresford_(Hampshire)_railway_station|Alresford]], on the [[Mid_Hants_Watercress_Railway|Mid Hants Watercress Railway]] in [[Hampshire|Hampshire]].\n[[Lyme Regis_sentence_45]]: The town holds an annual Mary Anning Day and Lyme Regis Fossil Festival.\n[[Lyme Regis_sentence_46]]: A fossil of the world's largest moth was discovered there in 1966.\n[[Lyme Regis_sentence_80]]: Well-sited for trade with France, the port's most prosperous period was from the 16th century until the end of the 18th century.\n[[Lyme Regis_sentence_122]]: Bonfire night celebrations include a torchlight procession, bonfire on the beach and firework display.\n[[Lyme Regis_sentence_135]]: It also features in [[A._S._Byatt|A.\n</information>\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Lyme Regis_sentence_45",
"Lyme Regis_sentence_46"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 501,
"split": "test_off"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.418",
"value": "?search="
},
{
"operation": "search",
"time": "18.171",
"value": "Lyme Regis Fossil Festival"
},
{
"operation": "Now on",
"time": "19.648",
"value": "Lyme Regis Fossil Festival"
},
{
"operation": "Highlighting",
"time": "25.929",
"value": "Lyme Regis Fossil Festival_sentence_0"
},
{
"operation": "Highlighting",
"time": "25.93",
"value": "Lyme Regis Fossil Festival_title"
},
{
"operation": "hyperlink",
"time": "31.282",
"value": "Lyme Regis"
},
{
"operation": "Now on",
"time": "32.946",
"value": "Lyme Regis"
},
{
"operation": "Page search",
"time": "61.977",
"value": "mary anning"
},
{
"operation": "Highlighting",
"time": "78.005",
"value": "Lyme Regis_sentence_45"
},
{
"operation": "Highlighting",
"time": "90.197",
"value": "Lyme Regis_sentence_46"
},
{
"operation": "search",
"time": "104.338",
"value": "Lyme Regis Fossil Festival"
},
{
"operation": "Now on",
"time": "105.832",
"value": "Lyme Regis Fossil Festival"
},
{
"operation": "Highlighting",
"time": "108.443",
"value": "Lyme Regis Fossil Festival_sentence_0"
},
{
"operation": "Highlighting",
"time": "108.444",
"value": "Lyme Regis Fossil Festival_title"
},
{
"operation": "search",
"time": "118.321",
"value": "Contains...Mary Anning Day"
},
{
"operation": "Now on",
"time": "120.51",
"value": "undefined"
},
{
"operation": "finish",
"time": "147.922",
"value": "finish"
}
]
}
|
exfever_3100
|
"Summertime" is an aria composed in 1935 by George Gershwin for the 1935 opera novel written by DuBose Heyward, an American author whose wife is an American playwright who co-authored the novel.
|
{
"evidence": [
"Summertime (George Gershwin song)",
"Porgy (novel)",
"Dorothy Heyward"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information above.\n\n<think>\n...\n</think>\n- Use for every piece of reasoning.\n- During reasoning, you must verify the claim step by step based on the given information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer>\n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\nVerify the claim:\n\"Summertime\" is an aria composed in 1935 by George Gershwin for the 1935 opera novel written by DuBose Heyward, an American author whose wife is an American playwright who co-authored the novel.\n<information>\n[[William Humble (disambiguation)]]: William Humble was an English cricketer.\n\n[[Mniów]]: Mniów is a village in Kielce County, Świętokrzyskie Voivodeship, in south-central Poland. It is the seat of the gmina (administrative district) called Gmina Mniów. It lies approximately 17 km north-west of the regional capital Kielce.\n\n[[Summertime (George Gershwin song)]]: \"Summertime\" is an aria composed in 1934 by George Gershwin for the 1935 opera \"Porgy and Bess\". The lyrics are by DuBose Heyward, the author of the novel \"Porgy\" on which the opera was based, although the song is also co-credited to Ira Gershwin by ASCAP.\n\n[[Tallapoosa-class cutter]]: The \"Tallapoosa\"-class cutters is a group of two Coast Guard cutters thst served with the United States Coast Guard from the 1920s to the late 1940s.\n\n[[The Meyerowitz Stories]]: The Meyerowitz Stories (New and Selected) is a 2017 American comedy-drama film directed and written by Noah Baumbach. The film stars Adam Sandler, Ben Stiller, Dustin Hoffman and Emma Thompson.\n\n[[Dorothy Heyward]]: Dorothy Heyward née Kuhns, (June 6, 1890 – November 19, 1961) was an American playwright.\n\n[[University of Oregon College of Arts and Sciences]]: The University of Oregon College of Arts and Sciences (CAS) is a public school of science and liberal arts at the University of Oregon (UO) in Eugene, Oregon, United States. The main offices of the college are located in Friendly Hall on the UO campus. The college serves an estimated 11,000 undergraduates at any given time.\n\n[[Porgy (novel)]]: Porgy is a novel written by the American author DuBose Heyward and published by the George H. Doran Company in 1925.\n</information>\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Summertime (George Gershwin song)",
"Porgy (novel)",
"Dorothy Heyward"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 502,
"split": "test_off"
}
|
{
"annotator_operations": []
}
|
exfever_7015
|
Trolls is an American computer-animated musical comedy film featuring the voice of an American singer who is a recipient of the Contemporary Icon Award by the American institution founded in 1969 by songwriter Johnny Mercer.
|
{
"evidence": [
"Trolls (film)",
"Justin Timberlake",
"Songwriters Hall of Fame"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information above.\n\n<think>\n...\n</think>\n- Use for every piece of reasoning.\n- During reasoning, you must verify the claim step by step based on the given information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer>\n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\nVerify the claim:\nTrolls is an American computer-animated musical comedy film featuring the voice of an American singer who is a recipient of the Contemporary Icon Award by the American institution founded in 1969 by songwriter Johnny Mercer.\n<information>\n[[William Boyle Barbour]]: William Boyle Barbour (22 October 1828 – 13 May 1891) was Scottish merchant and Liberal politician.\n\n[[William Msiska]]: William Msiska (born 10 October 1947) is a Malawian sprinter. He competed in the men's 400 metres at the 1972 Summer Olympics.\n\n[[Lakayym]]: Lakayym is the name of a settlement in Fujairah.\n\n[[Trolls (film)]]: Trolls is a 2016 American 3D computer-animated musical comedy film based on the Troll dolls created by Thomas Dam. The film was directed by Mike Mitchell and co-directed by Walt Dohrn, written by Jonathan Aibel and Glenn Berger and based on a story by Erica Rivinoja. The film features the voices of Anna Kendrick, Justin Timberlake, Zooey Deschanel, Russell Brand, James Corden and Gwen Stefani. The film revolves around two trolls on a quest to save their village from destruction by the Bergens, creatures who eat trolls.\n\n[[Justin Timberlake]]: Justin Randall Timberlake (born January 31, 1981) is an American singer-songwriter, actor and record producer. Born and raised in Tennessee, he appeared on the television shows \"Star Search\" and \"The All-New Mickey Mouse Club\" as a child. In the late 1990s, Timberlake rose to prominence as one of the two lead vocalists and youngest member of NSYNC, which eventually became one of the best-selling boy bands of all time. Timberlake began to adopt a more mature image as an artist with the release of his debut solo album, the R&B-focused \"Justified\" (2002), which yielded the successful singles \"Cry Me a River\" and \"Rock Your Body\", and earned his first two Grammy Awards.\n\n[[Songwriters Hall of Fame]]: The Songwriters Hall of Fame (SHOF), was founded in 1969 by songwriter Johnny Mercer and music publisher/songwriter Abe Olman and publisher/executive Howie Richmond to honor those whose work represents and maintains the heritage and legacy of a spectrum of the most beloved songs from the world's popular music songbook. It not only celebrates these established songwriters, but is also involved on the development of new songwriting talent through workshops, showcases and scholarships. There are many programs designed to teach and discover new songwriters.\n\n[[Ereta Peak]]: Ereta Peak (Bulgarian: \"връх Ерета\" , ‘Vrah Ereta’ \\'vr&h e-'re-ta\\) is the peak rising to 2300 m at the north extremity of Bastien Range in Ellsworth Mountains, Antarctica. It has steep and partly ice-free north and southwest slopes, and surmounts upper Nimitz Glacier to the southeast.\n\n[[28th PMPC Star Awards for Movies]]: The 28th Star Awards for Movies by the Philippine Movie Press Club (PMPC), honored the best Filipino films of 2011. The ceremony took place on March 14, 2012, at the Meralco Theater in Ortigas, Pasig City. The PMPC Star Awards for Movies was hosted by KC Concepcion, Derek Ramsay, Aga Muhlach and Anne Curtis.<br>\n</information>\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Trolls (film)",
"Justin Timberlake",
"Songwriters Hall of Fame"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 503,
"split": "test_off"
}
|
{
"annotator_operations": []
}
|
20593
|
The 2008 album "Nude with Boots is the 16th album of the rock band, Melvins, whose music has been described as "classical" and has been deemed "favorable" in review scores by Billboard and Crawdaddy!
|
{
"evidence": [
"Melvins_sentence_131"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information above.\n\n<think>\n...\n</think>\n- Use for every piece of reasoning.\n- During reasoning, you must verify the claim step by step based on the given information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer>\n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\nVerify the claim:\nThe 2008 album \"Nude with Boots is the 16th album of the rock band, Melvins, whose music has been described as \"classical\" and has been deemed \"favorable\" in review scores by Billboard and Crawdaddy!\n<information>\n[[Melvins_sentence_3]]: For other uses, see [[Melvin_(disambiguation)|Melvin (disambiguation)]].\n[[Melvins_sentence_54]]: 1999 saw the beginning of a partnership with [[Mike_Patton|Mike Patton]]'s [[Ipecac_Recordings|Ipecac Recordings]], which began remastering and reissuing much of the band's back catalog.\n[[Melvins_sentence_131]]: Variously described as a [[Sludge_metal|sludge metal]], [[Grunge|grunge]], [[Experimental_rock|experimental rock]], [[Alternative_metal|alternative metal]] and [[Doom_metal|doom metal]] band, Melvins explored a variety of different styles throughout its career, including [[Noise_rock|noise rock]], [[Dark_ambient|dark ambient]], [[Noise_music|noise]], jazz-rock, [[Avant-garde_music|avant-garde music]], [[Electroacoustic_music|electroacoustic music]] and [[Cowpunk|punk country]].\n[[Melvins_sentence_134]]: Buzz also listed [[Swans_(band)|Swans]] and [[Venom_(band)|Venom]] as early influences on the band.\n</information>\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Melvins_sentence_131"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 504,
"split": "test_off"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.667",
"value": "?search="
},
{
"operation": "search",
"time": "18.619",
"value": "Nude with Boots"
},
{
"operation": "Now on",
"time": "19.321",
"value": "Nude with Boots"
},
{
"operation": "Highlighting",
"time": "53.38",
"value": "Nude with Boots_sentence_0"
},
{
"operation": "Highlighting",
"time": "53.384",
"value": "Nude with Boots_title"
},
{
"operation": "hyperlink",
"time": "65.38",
"value": "Melvins"
},
{
"operation": "Now on",
"time": "66.237",
"value": "Melvins"
},
{
"operation": "Page search",
"time": "158.147",
"value": "classical"
},
{
"operation": "hyperlink",
"time": "184.829",
"value": "2 Musical style and influence Melvins_section_7"
},
{
"operation": "Highlighting",
"time": "197.704",
"value": "Melvins_sentence_131"
},
{
"operation": "finish",
"time": "214.815",
"value": "finish"
}
]
}
|
exfever_5591
|
Argo is a drama thriller film based on a memoir of a technical operations officer for a civilian foreign intelligence service of the federal government of the United Kingdom.
|
{
"evidence": [
"Argo (2012 film)",
"Tony Mendez",
"Central Intelligence Agency"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information above.\n\n<think>\n...\n</think>\n- Use for every piece of reasoning.\n- During reasoning, you must verify the claim step by step based on the given information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer>\n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\nVerify the claim:\nArgo is a drama thriller film based on a memoir of a technical operations officer for a civilian foreign intelligence service of the federal government of the United Kingdom.\n<information>\n[[Tony Mendez]]: Antonio Joseph \"Tony\" Mendez (born November 15, 1940) is an American CIA technical operations officer, now retired, who specialized in support of clandestine and covert CIA operations. He has written three memoirs about his CIA experiences.\n\n[[Ramchandrapur, West Bengal]]: Ramchandrapur is a census town in Sankrail CD Block under Sankrail police station in Sadar subdivision of Howrah district in the Indian state of West Bengal.\n\n[[Vasefjan]]: Vasefjan (Persian: واصفجان , also Romanized as Vāşefjān, Vāşfejān, and Vāşf Jān) is a village in Karasht Rural District, in the Bumehen District of Pardis County, Tehran Province, Iran. At the 2006 census, its population was 96, in 28 families.\n\n[[Edgar Elliot]]: Edgar William Elliot (9 July 1878 – 23 March 1931) was an English cricketer and rugby union player. In cricket, Elliot was a right-handed batsman who bowled right-arm medium fast, while in rugby union he played on the wing. He was born in Roker, County Durham and died in Vancouver, British Columbia, Canada on 23 March 1931.\n\n[[Central Intelligence Agency]]: The Central Intelligence Agency (CIA) is a civilian foreign intelligence service of the United States federal government, tasked with gathering, processing, and analyzing national security information from around the world, primarily through the use of human intelligence (HUMINT). As one of the principal members of the U.S. Intelligence Community (IC), the CIA reports to the Director of National Intelligence and is primarily focused on providing intelligence for the President and Cabinet.\n\n[[Central University of Gujarat]]: Central University of Gujarat in Gandhinagar, Gujarat, India was established through an Act of Parliament, Central Universities Act, 2009 by Government of India.\n\n[[Argo (2012 film)]]: Argo is a 2012 American historical drama film directed by Ben Affleck. Screenwriter Chris Terrio adapted the screenplay from the book by the U.S. Central Intelligence Agency operative Tony Mendez, \"The Master of Disguise\", and the\n</information>\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Argo (2012 film)",
"Tony Mendez",
"Central Intelligence Agency"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 505,
"split": "test_off"
}
|
{
"annotator_operations": []
}
|
93894
|
Holden (General Motors) started importing automobiles in 1972.
|
{
"evidence": [
"Holden_sentence_2",
"Holden_sentence_39",
"Holden_sentence_186"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information above.\n\n<think>\n...\n</think>\n- Use for every piece of reasoning.\n- During reasoning, you must verify the claim step by step based on the given information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer>\n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\nVerify the claim:\nHolden (General Motors) started importing automobiles in 1972.\n<information>\n[[Holden_sentence_2]]: Holden, formerly known as General Motors-Holden, was an Australian automobile marque and former [[Automobile_manufacturer|automobile manufacturer]], which manufactured cars in Australia before switching to importing cars under the Holden brand.\n[[Holden_sentence_14]]: The consolidation of final assembly at Elizabeth was completed in 1988, but some assembly operations continued at [[Dandenong|Dandenong]] until 1994.\n[[Holden_sentence_23]]: In 1879 J A Holden's eldest son [[Henry_James_Holden|Henry James (HJ) Holden]], became a partner and effectively managed the company.\n[[Holden_sentence_39]]: In 1926, General Motors (Australia) Limited was established with assembly plants at [[Newstead,_Queensland|Newstead, Queensland]]; [[Marrickville,_New_South_Wales|Marrickville, New South Wales]]; City Road, Melbourne, Victoria; [[Birkenhead,_South_Australia|Birkenhead, South Australia]]; and [[Cottesloe,_Western_Australia|Cottesloe, Western Australia]] using bodies produced by HMBB and imported [[Complete_knock_down|complete knock down]] chassis.\n[[Holden_sentence_51]]: GM's study was ultimately chosen because of its low level of government intervention.\n[[Holden_sentence_157]]: This process began in the 1970s and by 1982 Holden's commercial vehicle arm no longer offered any Bedford products.\n[[Holden_sentence_185]]: Besides manufacturing Australia's best selling car, which was exported in significant numbers, Holden continued to export many locally produced engines to power cars made elsewhere.\n[[Holden_sentence_186]]: In this decade, Holden adopted a strategy of importing cars it needed to offer a full range of competitive vehicles.\n[[Holden_sentence_193]]: For 1991, Holden updated the Statesman and Caprice with a range of improvements, including the introduction of four-wheel [[Anti-lock_braking_system|anti-lock brakes]] (ABS); although, a rear-wheel system had been standard on the Statesman Caprice from March 1976.\n[[Holden_sentence_248]]: In March 2012, Holden was given a $270 million lifeline by the Australian, South Australian and Victorian governments.\n[[Holden_sentence_259]]: Later that year, Holden also announced plans to sell the European Astra and the Korean Cruze alongside each other from 2017.\n[[Holden_sentence_275]]: Holden began to export vehicles in 1954, sending the FJ to New Zealand.\n</information>\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Holden_sentence_2",
"Holden_sentence_39",
"Holden_sentence_186"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 506,
"split": "test_off"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.733",
"value": "?search="
},
{
"operation": "search",
"time": "15.01",
"value": "Holden"
},
{
"operation": "Now on",
"time": "17.753",
"value": "Holden"
},
{
"operation": "Page search",
"time": "48.761",
"value": "import"
},
{
"operation": "Highlighting",
"time": "55.89",
"value": "Holden_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "56.846",
"value": "Holden_sentence_2"
},
{
"operation": "Page search",
"time": "169.779",
"value": "1972"
},
{
"operation": "Page search",
"time": "214.445",
"value": "import"
},
{
"operation": "Highlighting",
"time": "218.417",
"value": "Holden_sentence_2"
},
{
"operation": "Highlighting",
"time": "230.695",
"value": "Holden_sentence_39"
},
{
"operation": "Highlighting",
"time": "250.906",
"value": "Holden_sentence_186"
},
{
"operation": "finish",
"time": "263.639",
"value": "finish"
}
]
}
|
60187
|
Shelley Hirsch was born and raised in New York City in 1952, her work was commissioned by New American Radio and won the Prix Futura International Media Competition in Berlin.
|
{
"evidence": [
"Shelley Hirsch_sentence_0",
"Shelley Hirsch_sentence_3",
"Shelley Hirsch_sentence_5",
"Shelley Hirsch_sentence_6"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information above.\n\n<think>\n...\n</think>\n- Use for every piece of reasoning.\n- During reasoning, you must verify the claim step by step based on the given information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer>\n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\nVerify the claim:\nShelley Hirsch was born and raised in New York City in 1952, her work was commissioned by New American Radio and won the Prix Futura International Media Competition in Berlin.\n<information>\n[[Shelley Hirsch_sentence_0]]: Shelley Hirsch (born June 9, 1952 in [[Brooklyn|Brooklyn]], [[New_York_(state)|New York]]) is an American vocalist, performance artist, composer, improviser, and writer.\n[[Shelley Hirsch_sentence_1]]: She won a DAAD Residency Grant in Berlin 1992, a Prix Futura award in 1993, and multiple awards from Creative Capital, the Foundation for Contemporary Arts, the New York State Council for the Arts, Four from NYFA and six from Harvestworks Digital Media Arts Center.\n[[Shelley Hirsch_sentence_2]]: She was a recipient of the Guggenheim Fellowship in Music Composition in 2017.\n[[Shelley Hirsch_sentence_3]]: Born and raised in New York City, Hirsch dropped out of high school and moved to San Francisco, California, where she acted in the Experimental theater Theater of Man and explored extended vocal techniques, and began composing pieces for voice.\n[[Shelley Hirsch_sentence_5]]: Several of Hirsch's works were commissioned by New American Radio, including #39 and The Vidzer Family (1991), and O Little Town of East New York, which was originally staged at Dance Theater Workshop in New York City (1991).\n[[Shelley Hirsch_sentence_6]]: As a radio play, O Little Town won first prize at the Prix Futura International Media Competition in Berlin.\n[[Shelley Hirsch_sentence_8]]: Hirsch has appeared at festivals throughout Europe and has performed at the Kitchen, Roulette, BAM, and many other venues around and beyond New York City.\n[[Shelley Hirsch_sentence_12]]: with Simon Ho (Tzadik 2012).\n[[Shelley Hirsch_sentence_13]]: In 2008 her piece, \"In the Basement,\" was included on the compilation album Crosstalk: American Speech Music (Bridge Records) produced by [[Mendi_&_Keith_Obadike|Mendi + Keith Obadike]].\n[[Shelley Hirsch_sentence_14]]: Hirsch's many collaborators include DJ Olive, [[Mark_Dresser|Mark Dresser]], Greetje Bijma, Chantal Dumas, [[David_Moss_(musician)|David Moss]], Min Xiao Fen, Jerry Hunt, Toshio Kajiwara, [[Jin_Hi_Kim|Jin Hi Kim]], Marina Rosenfeld and Ned Rothenberg, as well as visual artists Barbara Bloom and [[Jim_Hodges_(artist)|Jim Hodges]], choreographer [[Noémie_Lafrance|Noémie Lafrance]], and filmmakers [[Nina_Danino|Nina Danino]] (Temenos soundtrack with Sainkho Namtchylak, 2001), [[Zoe_Beloff|Zoe Beloff]], Abigail Childs, and Lee Sachs.\n[[Shelley Hirsch_sentence_15]]: She recorded with September Band (Rüdiger Carl, [[Hans_Reichel|Hans Reichel]] and Paul Lovens) and with X-Communication ([[Butch_Morris|Butch Morris]], Martin Schütz and Hans Koch), and she has recorded interpretations of works by other composers, such as Cathy Berberian's, Stripsody, and Alvin Curran's, Philharmonie.\n</information>\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Shelley Hirsch_sentence_0",
"Shelley Hirsch_sentence_3",
"Shelley Hirsch_sentence_5",
"Shelley Hirsch_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 507,
"split": "test_off"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.828",
"value": "?search="
},
{
"operation": "search",
"time": "141.573",
"value": "Shelley Hirsch"
},
{
"operation": "Now on",
"time": "142.98",
"value": "Shelley Hirsch"
},
{
"operation": "Highlighting",
"time": "148.457",
"value": "Shelley Hirsch_sentence_0"
},
{
"operation": "Highlighting",
"time": "157.569",
"value": "Shelley Hirsch_sentence_3"
},
{
"operation": "Highlighting",
"time": "184.593",
"value": "Shelley Hirsch_sentence_5"
},
{
"operation": "Highlighting",
"time": "188.813",
"value": "Shelley Hirsch_sentence_6"
},
{
"operation": "finish",
"time": "201.306",
"value": "finish"
}
]
}
|
3993
|
In 2007 the Wax Simulacra (single from the album The Bedlam in Goliath by The Mars Volta) was released for download on the iTunes Store and the Zune Marketplace and in 2009, it was performed on MTV Live.
|
{
"evidence": [
"Wax Simulacra_sentence_0",
"Wax Simulacra_sentence_8",
"Wax Simulacra_sentence_6"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information above.\n\n<think>\n...\n</think>\n- Use for every piece of reasoning.\n- During reasoning, you must verify the claim step by step based on the given information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer>\n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\nVerify the claim:\nIn 2007 the Wax Simulacra (single from the album The Bedlam in Goliath by The Mars Volta) was released for download on the iTunes Store and the Zune Marketplace and in 2009, it was performed on MTV Live.\n<information>\n[[Wax Simulacra_sentence_0]]: \"Wax Simulacra\" is a single from the album [[The_Bedlam_in_Goliath|The Bedlam in Goliath]] by [[The_Mars_Volta|The Mars Volta]].\n[[Wax Simulacra_sentence_1]]: The single debuted on November 18, 2007, on [[San_Francisco|San Francisco]]'s Live 105 FM.\n[[Wax Simulacra_sentence_2]]: It was previously played live by the band under the name \"Idle Tooth\".\n[[Wax Simulacra_sentence_3]]: The B-side \"Pulled to Bits\" is a cover of a [[Siouxsie_and_the_Banshees|Siouxsie and the Banshees]] song.\n[[Wax Simulacra_sentence_6]]: On November 20, 2007, the song was released for download on the [[ITunes_Store|iTunes Store]] and the [[Zune_Marketplace|Zune Marketplace]].\n[[Wax Simulacra_sentence_7]]: On January 17, 2008, the song was performed on the [[Late_Show_with_David_Letterman|Late Show with David Letterman]].\n[[Wax Simulacra_sentence_8]]: On January 22, 2009, the song was performed on MTV Live.\n</information>\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Wax Simulacra_sentence_0",
"Wax Simulacra_sentence_8",
"Wax Simulacra_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 508,
"split": "test_off"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "22.479",
"value": "Wax Simulacra"
},
{
"operation": "Highlighting",
"time": "28.468",
"value": "Wax Simulacra_sentence_0"
},
{
"operation": "Highlighting",
"time": "40.965",
"value": "Wax Simulacra_sentence_8"
},
{
"operation": "Highlighting",
"time": "53.096",
"value": "Wax Simulacra_sentence_6"
},
{
"operation": "finish",
"time": "90.061",
"value": "finish"
}
]
}
|
exfever_3573
|
Blue Velvet blends film noir with a subgenre of horror and psychological fiction that often uses disturbed psychological states, including an instinctual process that is heavily influenced by anxiety.
|
{
"evidence": [
"Blue Velvet (film)",
"Psychological horror",
"Paranoia"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information above.\n\n<think>\n...\n</think>\n- Use for every piece of reasoning.\n- During reasoning, you must verify the claim step by step based on the given information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer>\n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\nVerify the claim:\nBlue Velvet blends film noir with a subgenre of horror and psychological fiction that often uses disturbed psychological states, including an instinctual process that is heavily influenced by anxiety.\n<information>\n[[Blue Velvet (film)]]: Blue Velvet is a 1986 American neo-noir mystery film, written and directed by David Lynch. Blending psychological horror with film noir, the film stars Kyle MacLachlan, Isabella Rossellini, Dennis Hopper and Laura Dern. The title is taken from Bobby Vinton's 1963 song of the same name.\n\n[[List of Pride FC fighters]]: This is a list of fighters who have participated in the Pride Fighting Championships.\n\n[[Paranoia]]: Paranoia is a thought process believed to be heavily influenced by anxiety or fear, often to the point of delusion and irrationality. Paranoid thinking typically includes persecutory, or beliefs of conspiracy concerning a perceived threat towards oneself (e.g. \"\"Everyone is out to get me\"\"). Paranoia is distinct from phobias, which also involve irrational fear, but usually no blame. Making false accusations and the general distrust of others also frequently accompany paranoia. For example, an incident most people would view as an accident or coincidence, a paranoid person might believe was intentional. Paranoia is a central symptom of psychosis.\n\n[[Willow River, Minnesota]]: Willow River is a city in Pine County, Minnesota, United States, at the confluence of the Kettle and Willow Rivers. The population was 415 at the 2010 census.\n\n[[Psychological horror]]: Psychological horror is a subgenre of horror and psychological fiction that relies on mental, emotional and psychological states to frighten, disturb, or unsettle readers, viewers, or players. The subgenre frequently overlaps with the related subgenre of psychological thriller, and it often uses mystery elements and characters with unstable, unreliable, or disturbed psychological states to enhance the suspense, drama, action and horror of the setting and plot and to provide an overall unpleasant, unsettling, or distressing atmosphere.\n\n[[Union Avenue Opera]]: Union Avenue Opera (previously Union Avenue Opera Theatre) is an opera company based in St. Louis, Missouri. The company was founded in 1994 by Scott Schoonover, the music director of Union Avenue Christian Church, which serves as the company's venue in St. Louis' Visitation Park neighborhood.\n</information>\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Blue Velvet (film)",
"Psychological horror",
"Paranoia"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 509,
"split": "test_off"
}
|
{
"annotator_operations": []
}
|
exfever_2225
|
Coldplay is an Australian rock band consisting of talented musicians, including an English drummer and a songwriter best known as the drummer and backing vocalist of the rock band Coldplay.
|
{
"evidence": [
"Coldplay",
"Will Champion"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information above.\n\n<think>\n...\n</think>\n- Use for every piece of reasoning.\n- During reasoning, you must verify the claim step by step based on the given information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer>\n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\nVerify the claim:\nColdplay is an Australian rock band consisting of talented musicians, including an English drummer and a songwriter best known as the drummer and backing vocalist of the rock band Coldplay.\n<information>\n[[Dabara]]: Dabara is an administrative unit, known as Union council, of Tank District in the Khyber Pakhtunkhwa province of Pakistan.\n\n[[Nathan Boyle]]: Nathan Boyle (born 14 April 1994 in Derry, Northern Ireland) is a football striker who currently plays for Derry City in the League of Ireland.\n\n[[Will Champion]]: William Champion (born 31 July 1978) is an English musician. multi-instrumentalist and singer. He is best known as the drummer and backing vocalist for the rock band Coldplay.\n\n[[The King of Masks]]: The King of Masks () is a 1996 Chinese film directed by Wu Tianming.\n\n[[Coldplay]]: Coldplay are a British rock band formed in 1996 by lead vocalist and keyboardist Chris Martin and lead guitarist Jonny Buckland at University College London (UCL). After they formed under the name Pectoralz, Guy Berryman joined the group as bassist and they changed their name to Starfish. Will Champion joined as drummer and backing vocalist, completing the lineup. Creative director Phil Harvey is often referred to as the fifth member by the band. The band renamed themselves \"Coldplay\" in 1998, before recording and releasing three EPs: \"Safety\" in 1998 and \"Brothers & Sisters\" and \"The Blue Room\" in 1999. \"The Blue Room\" was their first release on a major label, after signing to Parlophone.\n\n[[Abraham Jefferson Seay]]: Abraham Jefferson Seay (November 28, 1832 – December 12, 1915) was an American lawyer, soldier, judge, and politician. Seay attained the rank of colonel in the Union army during the American Civil War. Later, Seay would serve as an associate justice on Oklahoma Territory's supreme court and as the second Governor of the Oklahoma Territory.\n\n[[Kanadukathan]]: Kanadukathan is a Town Panchayat in the Karaikudi Taluk of the Indian state of Tamil Nadu.\n</information>\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Coldplay",
"Will Champion"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 510,
"split": "test_off"
}
|
{
"annotator_operations": []
}
|
45889
|
Akameros was the "archon of the Sclavenes of Belzetia", mentioned only once by Theophanes the Confessor, and planned to marry Empress Irene of Athens.
|
{
"evidence": [
"Akameros_sentence_0",
"Akameros_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information above.\n\n<think>\n...\n</think>\n- Use for every piece of reasoning.\n- During reasoning, you must verify the claim step by step based on the given information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer>\n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\nVerify the claim:\nAkameros was the \"archon of the Sclavenes of Belzetia\", mentioned only once by Theophanes the Confessor, and planned to marry Empress Irene of Athens.\n<information>\n[[Akameros_sentence_0]]: Akameros ([[Greek_language|Greek]]: Ἀκάμηρος, fl. 799 AD)—his original name was probably Akamir—was the \"[[Archon|archon]] of the [[Sclaviniae|Sclavenes]] of [[Belzetia|Belzetia]]\" (ό των Σκλαυινών της Βελζητίας άρχων), an autonomous [[South_Slavs|South Slavic]] community in [[Central_Greece|Central Greece]] under [[Byzantine|Byzantine]] sovereignty, in the late 8th century.\n[[Akameros_sentence_1]]: He is mentioned only once, by [[Theophanes_the_Confessor|Theophanes the Confessor]], as leading a plot involving the sons of [[Constantine_V|Constantine V]] (r. 741–775) – the former [[Caesar_(title)|Caesares]] [[Nikephoros_(Caesar)|Nikephoros]] and Christopher, and their younger brothers Niketas, Anthimos and Eudokimos – who had been deposed, [[Political_mutilation_in_Byzantine_culture|mutilated]] by their elder brother, the emperor [[Leo_IV_the_Khazar|Leo IV the Khazar]] (r. 775–780).\n[[Akameros_sentence_3]]: In March 799, Akameros, in collusion with troops from the local [[Theme_(Byzantine_district)|theme]] of [[Hellas_(theme)|Hellas]], planned to seize them and declare one of them emperor.\n[[Akameros_sentence_4]]: The plot was foiled however as Irene was informed of it, and the Empress sent a trusted kinsman to Athens: the brothers were blinded and moved to the island of Panormos in the [[Marmara_Sea|Marmara Sea]].\n[[Akameros_sentence_5]]: Nothing further is heard of Akameros.\n[[Akameros_sentence_6]]: The area of Belzetia has been identified by some scholars with the area of settlement of the [[Belegezitai|Belegezitai]] around [[Demetrias|Demetrias]] in eastern [[Thessaly|Thessaly]], but this is probably an error, and the origin of Belzetia has to be sought with another Slavic tribe, the Berzetai.\n</information>\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Akameros_sentence_0",
"Akameros_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 511,
"split": "test_off"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.697",
"value": "?search="
},
{
"operation": "search",
"time": "134.279",
"value": "Akameros was the \"archon of the Sclavenes of Belzetia"
},
{
"operation": "Now on",
"time": "135.288",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "136.741",
"value": "Akameros"
},
{
"operation": "Now on",
"time": "137.316",
"value": "Akameros"
},
{
"operation": "Highlighting",
"time": "139.65",
"value": "Akameros_sentence_0"
},
{
"operation": "Highlighting",
"time": "156.794",
"value": "Akameros_sentence_1"
},
{
"operation": "Page search",
"time": "162.402",
"value": "planned"
},
{
"operation": "Page search",
"time": "173.715",
"value": "marry"
},
{
"operation": "Page search",
"time": "183.321",
"value": "empress"
},
{
"operation": "Page search",
"time": "195.402",
"value": "irene"
},
{
"operation": "Page search",
"time": "256.678",
"value": "marry"
},
{
"operation": "Page search",
"time": "317.565",
"value": "irene"
},
{
"operation": "finish",
"time": "339.249",
"value": "finish"
}
]
}
|
exfever_3803
|
London is the largest city in England that stands on the shortest river entirely in England that flows into the North Sea near a town in the borough of Thurrock, Essex, England.
|
{
"evidence": [
"London",
"River Thames",
"Tilbury"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLondon is the largest city in England that stands on the shortest river entirely in England that flows into the North Sea near a town in the borough of Thurrock, Essex, England.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"London",
"River Thames",
"Tilbury"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 0,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
50241
|
In 2010 Iraq's Ministry of Health reported maternal mortality in Iraq at two/100,000 birth, then in 2011 CIA World Factbook and the WHO reported the maternal mortality rate at one death per 100,000 births.
|
{
"evidence": [
"List of countries by maternal mortality ratio_sentence_4",
"Maternal death_sentence_101",
"The World Factbook_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn 2010 Iraq's Ministry of Health reported maternal mortality in Iraq at two/100,000 birth, then in 2011 CIA World Factbook and the WHO reported the maternal mortality rate at one death per 100,000 births.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"List of countries by maternal mortality ratio_sentence_4",
"Maternal death_sentence_101",
"The World Factbook_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.907",
"value": "?search="
},
{
"operation": "search",
"time": "23.972",
"value": "iraq's Ministry of Health"
},
{
"operation": "Now on",
"time": "27.485",
"value": "undefined"
},
{
"operation": "search",
"time": "56.37",
"value": "Maternal mortality"
},
{
"operation": "Now on",
"time": "58.023",
"value": "Maternal death"
},
{
"operation": "Page search",
"time": "82.301",
"value": "2010"
},
{
"operation": "Page search",
"time": "88.611",
"value": "Ministry of Health"
},
{
"operation": "search",
"time": "126.651",
"value": "CIA World Fact Book"
},
{
"operation": "Now on",
"time": "127.22",
"value": "The World Factbook"
},
{
"operation": "search",
"time": "219.468",
"value": "Maternal mortality in India"
},
{
"operation": "Now on",
"time": "220.566",
"value": "Maternal mortality in India"
},
{
"operation": "Page search",
"time": "222.894",
"value": "IRAQ"
},
{
"operation": "back-button-clicked",
"time": "230.167",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "230.607",
"value": "The World Factbook"
},
{
"operation": "search",
"time": "238.557",
"value": "\"maternal mortality rate at one death per 100,000 births\""
},
{
"operation": "Now on",
"time": "240.159",
"value": "undefined"
},
{
"operation": "search",
"time": "250.412",
"value": "Maternal mortality rate"
},
{
"operation": "Now on",
"time": "251.551",
"value": "Maternal death"
},
{
"operation": "Page search",
"time": "253.574",
"value": "IRAQ"
},
{
"operation": "Page search",
"time": "257.758",
"value": "CIA"
},
{
"operation": "search",
"time": "268.931",
"value": ""
},
{
"operation": "Now on",
"time": "269.278",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "270.044",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "271.928",
"value": "Maternal death"
},
{
"operation": "Page search",
"time": "273.206",
"value": "WHO"
},
{
"operation": "Page search",
"time": "352.767",
"value": "2011"
},
{
"operation": "search",
"time": "497.665",
"value": "Ministry of Health (Iraq)"
},
{
"operation": "Now on",
"time": "498.563",
"value": "Ministry of Health (Iraq)"
},
{
"operation": "Page search",
"time": "501.424",
"value": "Maternal"
},
{
"operation": "Page search",
"time": "507.039",
"value": "Mortality"
},
{
"operation": "search",
"time": "525.912",
"value": "Maternal mortality ratio"
},
{
"operation": "Now on",
"time": "529.897",
"value": "Maternal mortality ratio"
},
{
"operation": "Page search",
"time": "534.928",
"value": "Iraq"
},
{
"operation": "hyperlink",
"time": "554.758",
"value": "List of countries by maternal mortality ratio"
},
{
"operation": "Now on",
"time": "556.624",
"value": "List of countries by maternal mortality ratio"
},
{
"operation": "Page search",
"time": "563.665",
"value": "Iraw"
},
{
"operation": "Page search",
"time": "564.728",
"value": "Iraq"
},
{
"operation": "Highlighting",
"time": "571.22",
"value": "List of countries by maternal mortality ratio_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "572.089",
"value": "List of countries by maternal mortality ratio_sentence_0"
},
{
"operation": "Highlighting",
"time": "577.922",
"value": "List of countries by maternal mortality ratio_cell_0_107_1"
},
{
"operation": "Highlighting",
"time": "578.06",
"value": "List of countries by maternal mortality ratio_header_cell_0_0_1"
},
{
"operation": "Highlighting",
"time": "593.576",
"value": "List of countries by maternal mortality ratio_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "597.965",
"value": "List of countries by maternal mortality ratio_cell_0_107_1"
},
{
"operation": "Highlighting deleted",
"time": "598.096",
"value": "List of countries by maternal mortality ratio_header_cell_0_0_1"
},
{
"operation": "hyperlink",
"time": "607.49",
"value": "maternal mortality ratio"
},
{
"operation": "Now on",
"time": "608.057",
"value": "Maternal mortality ratio"
},
{
"operation": "Page search",
"time": "634.45",
"value": "2010"
},
{
"operation": "hyperlink",
"time": "635.806",
"value": "Maternal mortality rate"
},
{
"operation": "Now on",
"time": "636.196",
"value": "Maternal mortality rate"
},
{
"operation": "Page search",
"time": "643.96",
"value": "201"
},
{
"operation": "Page search",
"time": "645.279",
"value": "2010"
},
{
"operation": "Highlighting",
"time": "703.446",
"value": "Maternal death_sentence_101"
},
{
"operation": "search",
"time": "712.188",
"value": "Maternal mortality in the United States"
},
{
"operation": "Now on",
"time": "713.505",
"value": "Maternal mortality in the United States"
},
{
"operation": "Page search",
"time": "730.678",
"value": "WHO"
},
{
"operation": "search",
"time": "826.899",
"value": "List of countries by maternal mortality ratio"
},
{
"operation": "Now on",
"time": "827.782",
"value": "List of countries by maternal mortality ratio"
},
{
"operation": "hyperlink",
"time": "833.754",
"value": "World Bank"
},
{
"operation": "Now on",
"time": "835.573",
"value": "World Bank"
},
{
"operation": "back-button-clicked",
"time": "849.667",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "849.86",
"value": "List of countries by maternal mortality ratio"
},
{
"operation": "search",
"time": "863.705",
"value": "CIA World Fact Book"
},
{
"operation": "Now on",
"time": "864.163",
"value": "CIA World Fact Book"
},
{
"operation": "Highlighting",
"time": "869.247",
"value": "The World Factbook_sentence_1"
},
{
"operation": "search",
"time": "897.646",
"value": "Maternal death"
},
{
"operation": "Now on",
"time": "898.154",
"value": "Maternal death"
},
{
"operation": "Page search",
"time": "900.759",
"value": "IRAQ"
},
{
"operation": "search",
"time": "945.19",
"value": "\"two/100,000 birth\""
},
{
"operation": "Now on",
"time": "946.373",
"value": "undefined"
},
{
"operation": "search",
"time": "948.51",
"value": "two/100,000 birth"
},
{
"operation": "Now on",
"time": "950.413",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "954.26",
"value": "Maternal death"
},
{
"operation": "Now on",
"time": "954.653",
"value": "Maternal death"
},
{
"operation": "finish",
"time": "969.793",
"value": "finish"
}
]
}
|
68547
|
Ron Roberts never coached the 2014 Southeastern Louisiana Lions football team who played Jacksonville on August 30 with 3,822 people in attendance.
|
{
"evidence": [
"2014 Southeastern Louisiana Lions football team_sentence_1"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRon Roberts never coached the 2014 Southeastern Louisiana Lions football team who played Jacksonville on August 30 with 3,822 people in attendance.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"2014 Southeastern Louisiana Lions football team_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.485",
"value": "?search="
},
{
"operation": "search",
"time": "15.318",
"value": "2014 Southeastern Louisiana Lions football team"
},
{
"operation": "Now on",
"time": "16.432",
"value": "2014 Southeastern Louisiana Lions football team"
},
{
"operation": "Highlighting",
"time": "46.278",
"value": "2014 Southeastern Louisiana Lions football team_sentence_1"
},
{
"operation": "finish",
"time": "53.93",
"value": "finish"
}
]
}
|
79231
|
Big Load received a 7/10 rating from NME, a British entertainment website and brand founded in 1952.
|
{
"evidence": [
"NME_sentence_2"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBig Load received a 7/10 rating from NME, a British entertainment website and brand founded in 1952.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"NME_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.882",
"value": "?search="
},
{
"operation": "search",
"time": "12.416",
"value": "Big Load"
},
{
"operation": "Now on",
"time": "14.569",
"value": "undefined"
},
{
"operation": "search",
"time": "46.25",
"value": "Big Load Sog"
},
{
"operation": "Now on",
"time": "47.232",
"value": "undefined"
},
{
"operation": "search",
"time": "53.807",
"value": "Big Load Song"
},
{
"operation": "Now on",
"time": "55.373",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "79.859",
"value": "next 20"
},
{
"operation": "Now on",
"time": "81.842",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "96.008",
"value": "50"
},
{
"operation": "Now on",
"time": "97.457",
"value": "undefined"
},
{
"operation": "search",
"time": "103.583",
"value": "big loud"
},
{
"operation": "Now on",
"time": "105.005",
"value": "Big Loud"
},
{
"operation": "search",
"time": "151.558",
"value": "Big Load Ratings"
},
{
"operation": "Now on",
"time": "153.095",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "170.001",
"value": "big loud rankings"
},
{
"operation": "Now on",
"time": "172.45",
"value": "undefined"
},
{
"operation": "search",
"time": "185.942",
"value": "NME"
},
{
"operation": "Now on",
"time": "194.414",
"value": "NME"
},
{
"operation": "Highlighting",
"time": "200.482",
"value": "NME_sentence_2"
},
{
"operation": "search",
"time": "219.14",
"value": "Big Load NME"
},
{
"operation": "Now on",
"time": "220.982",
"value": "undefined"
},
{
"operation": "finish",
"time": "257.837",
"value": "finish"
}
]
}
|
48202
|
Hany Shaker began his musical career at a very early age and learned the violin as a child.
|
{
"evidence": [
"Hany Shaker_sentence_2"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHany Shaker began his musical career at a very early age and learned the violin as a child.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Hany Shaker_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.766",
"value": "?search="
},
{
"operation": "search",
"time": "8.738",
"value": "Hany Shaker"
},
{
"operation": "Now on",
"time": "9.67",
"value": "Hany Shaker"
},
{
"operation": "Highlighting",
"time": "21.106",
"value": "Hany Shaker_sentence_2"
},
{
"operation": "Page search",
"time": "64.623",
"value": "violin"
},
{
"operation": "finish",
"time": "105.542",
"value": "finish"
}
]
}
|
26331
|
The Derbyshire County Cricket Club in 1920 included 39 members with only two members playing over ten matches.
|
{
"evidence": [
"Derbyshire County Cricket Club in 1920_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Derbyshire County Cricket Club in 1920 included 39 members with only two members playing over ten matches.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Derbyshire County Cricket Club in 1920_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.083",
"value": "?search="
},
{
"operation": "search",
"time": "13.932",
"value": "Derbyshire County Cricket Club"
},
{
"operation": "Now on",
"time": "15.789",
"value": "undefined"
},
{
"operation": "search",
"time": "35.381",
"value": "Derbyshire County Cricket Club in 1920"
},
{
"operation": "Now on",
"time": "36.175",
"value": "Derbyshire County Cricket Club in 1920"
},
{
"operation": "Highlighting",
"time": "71.412",
"value": "Derbyshire County Cricket Club in 1920_sentence_5"
},
{
"operation": "finish",
"time": "86.732",
"value": "finish"
}
]
}
|
86414
|
Oreta bicolor, a moth first described by William Warren, has greyish brown forewings with sprinkling towards the outer part of the wing.
|
{
"evidence": [
"Oreta bicolor_sentence_1",
"Oreta bicolor_sentence_0",
"Oreta bicolor_sentence_4"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nOreta bicolor, a moth first described by William Warren, has greyish brown forewings with sprinkling towards the outer part of the wing.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Oreta bicolor_sentence_1",
"Oreta bicolor_sentence_0",
"Oreta bicolor_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.776",
"value": "undefined"
},
{
"operation": "search",
"time": "23.57",
"value": "Oreta bicolor"
},
{
"operation": "Now on",
"time": "24.553",
"value": "Oreta bicolor"
},
{
"operation": "Highlighting",
"time": "30.281",
"value": "Oreta bicolor_sentence_1"
},
{
"operation": "Highlighting",
"time": "33.113",
"value": "Oreta bicolor_sentence_0"
},
{
"operation": "Highlighting",
"time": "36.944",
"value": "Oreta bicolor_sentence_4"
},
{
"operation": "finish",
"time": "51.897",
"value": "finish"
}
]
}
|
exfever_6042
|
Jamie Foxx is a Canadian actor who became widely known for his portrayal in the 2004 biographical film Ray, for which he won an award presented by the Critics' Choice Association and was first presented in 1995.
|
{
"evidence": [
"Jamie Foxx",
"Critics' Choice Movie Award for Best Actor"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJamie Foxx is a Canadian actor who became widely known for his portrayal in the 2004 biographical film Ray, for which he won an award presented by the Critics' Choice Association and was first presented in 1995.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jamie Foxx",
"Critics' Choice Movie Award for Best Actor"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
56975
|
Tewksbury Mills, planned to have between 1,100,000 and 2,200,000 sq ft of gross leasable area, is proposed to be an enclosed mall with two levels and an entertainment complex.
|
{
"evidence": [
"Tewksbury Mills_sentence_4",
"Tewksbury Mills_sentence_2",
"Tewksbury Mills_sentence_3"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nTewksbury Mills, planned to have between 1,100,000 and 2,200,000 sq ft of gross leasable area, is proposed to be an enclosed mall with two levels and an entertainment complex.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Tewksbury Mills_sentence_4",
"Tewksbury Mills_sentence_2",
"Tewksbury Mills_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.936",
"value": "?search="
},
{
"operation": "search",
"time": "25.494",
"value": "Tewksbury Mills"
},
{
"operation": "Now on",
"time": "26.037",
"value": "Tewksbury Mills"
},
{
"operation": "Highlighting",
"time": "45.452",
"value": "Tewksbury Mills_sentence_4"
},
{
"operation": "Highlighting",
"time": "90.13",
"value": "Tewksbury Mills_sentence_2"
},
{
"operation": "Highlighting",
"time": "91.434",
"value": "Tewksbury Mills_sentence_3"
},
{
"operation": "finish",
"time": "97.938",
"value": "finish"
}
]
}
|
12387
|
SOD3 is a medication that helps the brain, lungs, and tissues.
|
{
"evidence": [
"SOD3_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSOD3 is a medication that helps the brain, lungs, and tissues.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"SOD3_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 9,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.383",
"value": "?search="
},
{
"operation": "search",
"time": "8.834",
"value": "SOD3"
},
{
"operation": "Now on",
"time": "9.866",
"value": "SOD3"
},
{
"operation": "Highlighting",
"time": "18.244",
"value": "SOD3_sentence_3"
},
{
"operation": "Highlighting",
"time": "27.116",
"value": "SOD3_sentence_0"
},
{
"operation": "Highlighting",
"time": "34.84",
"value": "SOD3_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "36.321",
"value": "SOD3_sentence_0"
},
{
"operation": "Page search",
"time": "43.447",
"value": "medication"
},
{
"operation": "Highlighting deleted",
"time": "48.103",
"value": "SOD3_sentence_3"
},
{
"operation": "finish",
"time": "67.446",
"value": "finish"
}
]
}
|
42480
|
From June to October 1967, Dylan and The Band recorded a huge number of cover songs and original Dylan material in the basement (a collection of over 100 songs recorded by Bob Dylan and the Band), wherein they also developed their distinctive sound for the first time.
|
{
"evidence": [
"Music from Big Pink_sentence_6",
"List of Basement Tapes songs_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nFrom June to October 1967, Dylan and The Band recorded a huge number of cover songs and original Dylan material in the basement (a collection of over 100 songs recorded by Bob Dylan and the Band), wherein they also developed their distinctive sound for the first time.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Music from Big Pink_sentence_6",
"List of Basement Tapes songs_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 10,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.712",
"value": "?search="
},
{
"operation": "search",
"time": "13.999",
"value": "List of artists who have covered Bob Dylan songs"
},
{
"operation": "Now on",
"time": "14.905",
"value": "List of artists who have covered Bob Dylan songs"
},
{
"operation": "Page search",
"time": "30.573",
"value": "Dylan and The Band"
},
{
"operation": "Page search",
"time": "30.799",
"value": "Dylan and The Band"
},
{
"operation": "Page search",
"time": "31.307",
"value": "Dylan and The Band"
},
{
"operation": "Page search",
"time": "31.57",
"value": "Dylan and The Band"
},
{
"operation": "Page search",
"time": "31.756",
"value": "Dylan and The Band"
},
{
"operation": "Page search",
"time": "31.917",
"value": "Dylan and The Band"
},
{
"operation": "search",
"time": "32.966",
"value": "Dylan and The Band"
},
{
"operation": "Now on",
"time": "36.475",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "56.015",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "56.089",
"value": "List of artists who have covered Bob Dylan songs"
},
{
"operation": "Page search",
"time": "65.841",
"value": "basement"
},
{
"operation": "Page search",
"time": "66.052",
"value": "basement"
},
{
"operation": "Page search",
"time": "66.232",
"value": "basement"
},
{
"operation": "Page search",
"time": "66.487",
"value": "basement"
},
{
"operation": "Page search",
"time": "66.724",
"value": "basement"
},
{
"operation": "Page search",
"time": "66.913",
"value": "basement"
},
{
"operation": "search",
"time": "91.095",
"value": "Contains...The Band – Music from Big Pink"
},
{
"operation": "Now on",
"time": "93.474",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "94.552",
"value": "Music from Big Pink"
},
{
"operation": "Now on",
"time": "95.223",
"value": "Music from Big Pink"
},
{
"operation": "Page search",
"time": "100.387",
"value": "Dylan material in the basement"
},
{
"operation": "Highlighting",
"time": "103.476",
"value": "Music from Big Pink_sentence_6"
},
{
"operation": "Page search",
"time": "129.964",
"value": "collection"
},
{
"operation": "Page search",
"time": "130.171",
"value": "collection"
},
{
"operation": "Page search",
"time": "130.367",
"value": "collection"
},
{
"operation": "Page search",
"time": "130.749",
"value": "collection"
},
{
"operation": "Page search",
"time": "130.921",
"value": "collection"
},
{
"operation": "Page search",
"time": "131.084",
"value": "collection"
},
{
"operation": "Page search",
"time": "145.464",
"value": "collectionbasement"
},
{
"operation": "Page search",
"time": "149.357",
"value": "basement"
},
{
"operation": "Page search",
"time": "178.806",
"value": "first time"
},
{
"operation": "Page search",
"time": "179.007",
"value": "first time"
},
{
"operation": "hyperlink",
"time": "183.445",
"value": "Bob Dylan"
},
{
"operation": "Now on",
"time": "184.544",
"value": "Bob Dylan"
},
{
"operation": "Page search",
"time": "190.051",
"value": "recorded by Bob Dylan and the Band"
},
{
"operation": "Page search",
"time": "190.27",
"value": "recorded by Bob Dylan and the Band"
},
{
"operation": "Page search",
"time": "196.681",
"value": "100"
},
{
"operation": "Page search",
"time": "196.856",
"value": "100"
},
{
"operation": "search",
"time": "264.724",
"value": "List of Basement Tapes songs"
},
{
"operation": "Now on",
"time": "265.697",
"value": "List of Basement Tapes songs"
},
{
"operation": "Highlighting",
"time": "268.535",
"value": "List of Basement Tapes songs_sentence_0"
},
{
"operation": "Page search",
"time": "283.908",
"value": "distinctive"
},
{
"operation": "Page search",
"time": "284.134",
"value": "distinctive"
},
{
"operation": "Page search",
"time": "284.308",
"value": "distinctive"
},
{
"operation": "Page search",
"time": "284.908",
"value": "distinctive"
},
{
"operation": "Page search",
"time": "285.095",
"value": "distinctive"
},
{
"operation": "back-button-clicked",
"time": "291.036",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "291.245",
"value": "Bob Dylan"
},
{
"operation": "back-button-clicked",
"time": "292.806",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "292.809",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "292.897",
"value": "Music from Big Pink"
},
{
"operation": "Page search",
"time": "294.419",
"value": "distinctive"
},
{
"operation": "Page search",
"time": "294.686",
"value": "distinctive"
},
{
"operation": "Page search",
"time": "294.902",
"value": "distinctive"
},
{
"operation": "Page search",
"time": "295.448",
"value": "distinctive"
},
{
"operation": "Page search",
"time": "295.651",
"value": "distinctive"
},
{
"operation": "Page search",
"time": "308.101",
"value": "June"
},
{
"operation": "Page search",
"time": "308.304",
"value": "June"
},
{
"operation": "finish",
"time": "314.613",
"value": "finish"
}
]
}
|
73752
|
In October 1785, a peace treaty between the Spanish and Comanches was enacted, achieving an acceptable peace in the north, however, the Apaches still were a threat in the South, and the lands spread until Pecos
|
{
"evidence": [
"Comanche–Mexico Wars_sentence_11"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn October 1785, a peace treaty between the Spanish and Comanches was enacted, achieving an acceptable peace in the north, however, the Apaches still were a threat in the South, and the lands spread until Pecos\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Comanche–Mexico Wars_sentence_11"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 11,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.903",
"value": "?search="
},
{
"operation": "search",
"time": "12.417",
"value": "October 1785 Peace treaty"
},
{
"operation": "Now on",
"time": "15.038",
"value": "undefined"
},
{
"operation": "search",
"time": "48.563",
"value": "peace treaty Spanish Comanche"
},
{
"operation": "Now on",
"time": "49.919",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "52.467",
"value": "Meusebach–Comanche treaty Comanche_section_10"
},
{
"operation": "Now on",
"time": "54.048",
"value": "Comanche#Meusebach.E2.80.93Comanche treaty Comanche section 10"
},
{
"operation": "hyperlink",
"time": "129.091",
"value": "Comanche Wars"
},
{
"operation": "Now on",
"time": "130.29",
"value": "Comanche Wars"
},
{
"operation": "search",
"time": "282.79",
"value": "Spanish-Comanche Treaty"
},
{
"operation": "Now on",
"time": "284.038",
"value": "undefined"
},
{
"operation": "search",
"time": "303.974",
"value": "Comanche Treaty 1785"
},
{
"operation": "Now on",
"time": "304.764",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "310.488",
"value": "Intertribal warfare 1785-1845 Mandan_section_7"
},
{
"operation": "Now on",
"time": "311.657",
"value": "Mandan#Intertribal warfare 1785-1845 Mandan section 7"
},
{
"operation": "search",
"time": "336.768",
"value": "Comanche history"
},
{
"operation": "Now on",
"time": "338.227",
"value": "Comanche history"
},
{
"operation": "hyperlink",
"time": "401.152",
"value": "Comanche-Mexico Wars"
},
{
"operation": "Now on",
"time": "402.008",
"value": "Comanche–Mexico Wars"
},
{
"operation": "Highlighting",
"time": "424.836",
"value": "Comanche–Mexico Wars_sentence_11"
},
{
"operation": "hyperlink",
"time": "431.741",
"value": "Pedro Vial"
},
{
"operation": "Now on",
"time": "432.426",
"value": "Pedro Vial"
},
{
"operation": "back-button-clicked",
"time": "468.767",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "470.026",
"value": "Comanche–Mexico Wars"
},
{
"operation": "search",
"time": "497.791",
"value": "1785 Apache"
},
{
"operation": "Now on",
"time": "499.317",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "506.881",
"value": "Intertribal warfare 1785-1845 Mandan_section_7"
},
{
"operation": "Now on",
"time": "507.473",
"value": "Mandan#Intertribal warfare 1785-1845 Mandan section 7"
},
{
"operation": "finish",
"time": "527.727",
"value": "finish"
}
]
}
|
exfever_3598
|
"Diamonds" is a song recorded by Barbadian singer Rihanna and was available in the digital transfer of music via the Internet after it premiered on September 26, 2012.
|
{
"evidence": [
"Diamonds (Rihanna song)",
"Music download"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n\"Diamonds\" is a song recorded by Barbadian singer Rihanna and was available in the digital transfer of music via the Internet after it premiered on September 26, 2012.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Diamonds (Rihanna song)",
"Music download"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 12,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_2570
|
The Beach Boys are one of the most popular American rock bands in California.
|
{
"evidence": [
"The Beach Boys",
"Mike Love",
"Lyricist"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Beach Boys are one of the most popular American rock bands in California.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Beach Boys",
"Mike Love",
"Lyricist"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 13,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
49607
|
Ray Beck played three years at Georgia Tech, a public research university and institute of technology in Atlanta, and was named All-America as well as Most Valuable Lineman.
|
{
"evidence": [
"Ray Beck_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRay Beck played three years at Georgia Tech, a public research university and institute of technology in Atlanta, and was named All-America as well as Most Valuable Lineman.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ray Beck_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 14,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.767",
"value": "?search="
},
{
"operation": "search",
"time": "7.7",
"value": "Ray Beck"
},
{
"operation": "Now on",
"time": "8.735",
"value": "Ray Beck"
},
{
"operation": "Highlighting",
"time": "179.244",
"value": "Ray Beck_sentence_2"
},
{
"operation": "finish",
"time": "623.879",
"value": "finish"
}
]
}
|
72453
|
Henry Christie, the youngest son of Henry Christie of Craigton, was consecrated at Dundee on 28 April 1709 as a college bishop in the Scottish Episcopal Church to maintain the Episcopal succession without being committed to a particular Episcopal see.
|
{
"evidence": [
"Henry Christie_sentence_1",
"Henry Christie_sentence_7",
"Henry Christie_sentence_8"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHenry Christie, the youngest son of Henry Christie of Craigton, was consecrated at Dundee on 28 April 1709 as a college bishop in the Scottish Episcopal Church to maintain the Episcopal succession without being committed to a particular Episcopal see.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Henry Christie_sentence_1",
"Henry Christie_sentence_7",
"Henry Christie_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 15,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.044",
"value": "?search="
},
{
"operation": "search",
"time": "19.663",
"value": "Henry Christie"
},
{
"operation": "Now on",
"time": "20.728",
"value": "Henry Christie"
},
{
"operation": "Highlighting",
"time": "27.461",
"value": "Henry Christie_sentence_1"
},
{
"operation": "Highlighting",
"time": "65.555",
"value": "Henry Christie_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "68.095",
"value": "Henry Christie_sentence_0"
},
{
"operation": "Highlighting",
"time": "82.466",
"value": "Henry Christie_sentence_7"
},
{
"operation": "Highlighting",
"time": "94.697",
"value": "Henry Christie_sentence_8"
},
{
"operation": "finish",
"time": "113.197",
"value": "finish"
}
]
}
|
exfever_8671
|
Donnie Wahlberg is the vocalist of the boy band New Kids on the Block.
|
{
"evidence": [
"Donnie Wahlberg",
"Zookeeper (film)"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDonnie Wahlberg is the vocalist of the boy band New Kids on the Block.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Donnie Wahlberg",
"Zookeeper (film)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 16,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
86332
|
Debbie Blackburn is a Democratic politician which is one of the two major contemporary political parties in the United States.
|
{
"evidence": [
"Debbie Blackburn_sentence_0",
"Democratic Party (United States)_sentence_1"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDebbie Blackburn is a Democratic politician which is one of the two major contemporary political parties in the United States.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Debbie Blackburn_sentence_0",
"Democratic Party (United States)_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 17,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.111",
"value": "?search="
},
{
"operation": "search",
"time": "4.628",
"value": "Debbie Blackburn"
},
{
"operation": "Now on",
"time": "5.269",
"value": "Debbie Blackburn"
},
{
"operation": "Highlighting",
"time": "8.524",
"value": "Debbie Blackburn_sentence_1"
},
{
"operation": "Highlighting",
"time": "9.101",
"value": "Debbie Blackburn_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "10.128",
"value": "Debbie Blackburn_sentence_1"
},
{
"operation": "hyperlink",
"time": "10.942",
"value": "Democratic"
},
{
"operation": "Now on",
"time": "15.974",
"value": "Democratic Party (United States)"
},
{
"operation": "Highlighting",
"time": "17.955",
"value": "Democratic Party (United States)_sentence_1"
},
{
"operation": "finish",
"time": "26.142",
"value": "finish"
}
]
}
|
exfever_7639
|
Andrew Patrick Wood was the back up singer and lyricist of an American rock band that formed in Seattle, Washington in 1988.
|
{
"evidence": [
"Andrew Wood (singer)",
"Mother Love Bone"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAndrew Patrick Wood was the back up singer and lyricist of an American rock band that formed in Seattle, Washington in 1988.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Andrew Wood (singer)",
"Mother Love Bone"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 18,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_6199
|
John F. Kennedy served at the height of the period of geopolitical tension between the United States and the Soviet Union, together with their respective allies, including the group of socialist states.
|
{
"evidence": [
"John F. Kennedy",
"Cold War",
"Eastern Bloc"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJohn F. Kennedy served at the height of the period of geopolitical tension between the United States and the Soviet Union, together with their respective allies, including the group of socialist states.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"John F. Kennedy",
"Cold War",
"Eastern Bloc"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 19,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
36544
|
Howard Wandrei's Time Burial was re-published in 1995 by Fedogan & Bremer.
|
{
"evidence": [
"Time Burial_sentence_0",
"Time Burial_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHoward Wandrei's Time Burial was re-published in 1995 by Fedogan & Bremer.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Time Burial_sentence_0",
"Time Burial_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 20,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.919",
"value": "?search="
},
{
"operation": "search",
"time": "15.649",
"value": "Time Burial"
},
{
"operation": "Now on",
"time": "16.501",
"value": "Time Burial"
},
{
"operation": "Highlighting",
"time": "28.548",
"value": "Time Burial_sentence_0"
},
{
"operation": "Page search",
"time": "42.341",
"value": "1995"
},
{
"operation": "Highlighting",
"time": "62.688",
"value": "Time Burial_sentence_1"
},
{
"operation": "finish",
"time": "259.809",
"value": "finish"
}
]
}
|
exfever_5683
|
Ji Chang-wook starred in a 2011 South Korean fusion historical/action television series.
|
{
"evidence": [
"Ji Chang-wook",
"Warrior Baek Dong-soo"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJi Chang-wook starred in a 2011 South Korean fusion historical/action television series.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ji Chang-wook",
"Warrior Baek Dong-soo"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 21,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
51441
|
Edward Charles Stirling had no children, he introduced a formal bill for women's suffrage but the bill did not pass.
|
{
"evidence": [
"Edward Charles Stirling_sentence_17"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nEdward Charles Stirling had no children, he introduced a formal bill for women's suffrage but the bill did not pass.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Edward Charles Stirling_sentence_17"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 22,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.453",
"value": "?search="
},
{
"operation": "search",
"time": "45.031",
"value": "Edward Charles Sterling"
},
{
"operation": "Now on",
"time": "47.285",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "56.989",
"value": "edward charles stirling"
},
{
"operation": "Now on",
"time": "57.943",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "63.532",
"value": "Edward Charles Stirling"
},
{
"operation": "Now on",
"time": "64.494",
"value": "Edward Charles Stirling"
},
{
"operation": "Highlighting",
"time": "123.593",
"value": "Edward Charles Stirling_sentence_14"
},
{
"operation": "Highlighting",
"time": "147.361",
"value": "Edward Charles Stirling_sentence_17"
},
{
"operation": "Highlighting deleted",
"time": "156.291",
"value": "Edward Charles Stirling_sentence_14"
},
{
"operation": "finish",
"time": "191.368",
"value": "finish"
}
]
}
|
exfever_3419
|
Fox Broadcasting Company is a major broadcasting company that is now owned by Disney.
|
{
"evidence": [
"New Girl",
"Jake Johnson",
"Critics' Choice Television Award for Best Actor in a Comedy Series"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nFox Broadcasting Company is a major broadcasting company that is now owned by Disney.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"New Girl",
"Jake Johnson",
"Critics' Choice Television Award for Best Actor in a Comedy Series"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 23,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_6308
|
Anderson Cooper is a broadcast journalist, a political commentator, and an actor.
|
{
"evidence": [
"Anderson Cooper",
"Channel One News"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAnderson Cooper is a broadcast journalist, a political commentator, and an actor.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Anderson Cooper",
"Channel One News"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 24,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_4362
|
The Nile is a major north-flowing river whose drainage basin covers ten countries, including a landlocked country in East Africa.
|
{
"evidence": [
"Nile",
"Uganda"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Nile is a major north-flowing river whose drainage basin covers ten countries, including a landlocked country in East Africa.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Nile",
"Uganda"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 25,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_2684
|
Arkham Asylum appears commonly in stories featuring Batman, a character that possesses abilities beyond those of ordinary people, dedicating themselves to fighting an unlawful act punishable by the state.
|
{
"evidence": [
"Arkham Asylum",
"Superhero",
"Crime"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nArkham Asylum appears commonly in stories featuring Batman, a character that possesses abilities beyond those of ordinary people, dedicating themselves to fighting an unlawful act punishable by the state.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Arkham Asylum",
"Superhero",
"Crime"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 26,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_5980
|
The G1 Climax is the biggest professional wrestling tournament in Japan.
|
{
"evidence": [
"G1 Climax",
"Round-robin tournament"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe G1 Climax is the biggest professional wrestling tournament in Japan.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"G1 Climax",
"Round-robin tournament"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 27,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_8563
|
Illumination Entertainment produced the film Minions, a 2015 American computer-animated film that is in a category of film that emphasizes humor, which is derived from the classical genre in a collaborative form of performing art that uses live performers.
|
{
"evidence": [
"Minions (film)",
"Comedy film",
"Theatre"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIllumination Entertainment produced the film Minions, a 2015 American computer-animated film that is in a category of film that emphasizes humor, which is derived from the classical genre in a collaborative form of performing art that uses live performers.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Minions (film)",
"Comedy film",
"Theatre"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 28,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
52157
|
The Super Bowl was held at the Miami Orange Bowl three times in the 1970s.
|
{
"evidence": [
"Super Bowl_sentence_28"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Super Bowl was held at the Miami Orange Bowl three times in the 1970s.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Super Bowl_sentence_28"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 29,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.868",
"value": "?search="
},
{
"operation": "search",
"time": "16.636",
"value": "The Super Bowl"
},
{
"operation": "Now on",
"time": "19.068",
"value": "Super Bowl"
},
{
"operation": "Highlighting",
"time": "75.076",
"value": "Super Bowl_sentence_28"
},
{
"operation": "hyperlink",
"time": "81.54",
"value": "Orange Bowl"
},
{
"operation": "Now on",
"time": "88.326",
"value": "Orange Bowl"
},
{
"operation": "finish",
"time": "109.209",
"value": "finish"
}
]
}
|
57980
|
Connecticut Route 94, an east–west state highway in Connecticut, passes by the Macclain Earth Products factory as it traverses the rural part of Glastonbury and crosses into the town of Hebron about 3.3 miles (5.3 km) east of Roaring Brook becoming known as Gilead Street, where it ends 0.8 miles (1.3 km) east of the town line at an intersection with Route 85.
|
{
"evidence": [
"Connecticut Route 94_sentence_0",
"Connecticut Route 94_sentence_5",
"Connecticut Route 94_sentence_6"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nConnecticut Route 94, an east–west state highway in Connecticut, passes by the Macclain Earth Products factory as it traverses the rural part of Glastonbury and crosses into the town of Hebron about 3.3 miles (5.3 km) east of Roaring Brook becoming known as Gilead Street, where it ends 0.8 miles (1.3 km) east of the town line at an intersection with Route 85.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Connecticut Route 94_sentence_0",
"Connecticut Route 94_sentence_5",
"Connecticut Route 94_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 30,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.983",
"value": "?search="
},
{
"operation": "search",
"time": "25.752",
"value": "connecticut route 94"
},
{
"operation": "Now on",
"time": "26.476",
"value": "Connecticut Route 94"
},
{
"operation": "Highlighting",
"time": "33.043",
"value": "Connecticut Route 94_sentence_0"
},
{
"operation": "Highlighting",
"time": "40.489",
"value": "Connecticut Route 94_sentence_5"
},
{
"operation": "Highlighting",
"time": "67.058",
"value": "Connecticut Route 94_sentence_6"
},
{
"operation": "finish",
"time": "90.937",
"value": "finish"
}
]
}
|
68763
|
The Three Faces of Eve, a 1957 mystery drama, features a plot in which a mother learns she has multiple personalities.
|
{
"evidence": [
"The Three Faces of Eve_sentence_0",
"The Three Faces of Eve_sentence_6",
"The Three Faces of Eve_sentence_7"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Three Faces of Eve, a 1957 mystery drama, features a plot in which a mother learns she has multiple personalities.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"The Three Faces of Eve_sentence_0",
"The Three Faces of Eve_sentence_6",
"The Three Faces of Eve_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 31,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.792",
"value": "?search="
},
{
"operation": "search",
"time": "14.204",
"value": "Three Faces of Eve"
},
{
"operation": "Now on",
"time": "16.472",
"value": "The Three Faces of Eve"
},
{
"operation": "Highlighting",
"time": "21.827",
"value": "The Three Faces of Eve_sentence_0"
},
{
"operation": "Highlighting",
"time": "133.625",
"value": "The Three Faces of Eve_sentence_6"
},
{
"operation": "Highlighting",
"time": "141.328",
"value": "The Three Faces of Eve_sentence_7"
},
{
"operation": "finish",
"time": "279.174",
"value": "finish"
}
]
}
|
exfever_5456
|
A Thousand Suns is the fourth studio album by American rock band Linkin Park which was recorded at a recording facility located in North Hollywood, California, that was created by producer and mixer Jay Baumgardner in 1992.
|
{
"evidence": [
"A Thousand Suns",
"NRG Recording Studios"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nA Thousand Suns is the fourth studio album by American rock band Linkin Park which was recorded at a recording facility located in North Hollywood, California, that was created by producer and mixer Jay Baumgardner in 1992.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"A Thousand Suns",
"NRG Recording Studios"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 32,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
92435
|
American rock band Good Charlotte faced "a lot of criticism from all sides" and felt out of place in the current pop scene and the pop punk genre they had come from, the members internalised this while they made the album.
|
{
"evidence": [
"Good Charlotte (album)_sentence_1",
"The Young and the Hopeless_sentence_10",
"Good Charlotte_sentence_9"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAmerican rock band Good Charlotte faced \"a lot of criticism from all sides\" and felt out of place in the current pop scene and the pop punk genre they had come from, the members internalised this while they made the album.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Good Charlotte (album)_sentence_1",
"The Young and the Hopeless_sentence_10",
"Good Charlotte_sentence_9"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 33,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.333",
"value": "undefined"
},
{
"operation": "search",
"time": "28.083",
"value": "Good Charlotte"
},
{
"operation": "Now on",
"time": "29.974",
"value": "Good Charlotte"
},
{
"operation": "Page search",
"time": "65.622",
"value": "criti"
},
{
"operation": "Page search",
"time": "65.802",
"value": "criti"
},
{
"operation": "Page search",
"time": "66.015",
"value": "criti"
},
{
"operation": "Page search",
"time": "66.186",
"value": "criti"
},
{
"operation": "Page search",
"time": "66.37",
"value": "criti"
},
{
"operation": "Page search",
"time": "66.562",
"value": "criti"
},
{
"operation": "Page search",
"time": "66.73",
"value": "criti"
},
{
"operation": "Page search",
"time": "167.485",
"value": "pop"
},
{
"operation": "Page search",
"time": "167.918",
"value": "pop"
},
{
"operation": "Page search",
"time": "168.171",
"value": "pop"
},
{
"operation": "Page search",
"time": "168.419",
"value": "pop"
},
{
"operation": "Page search",
"time": "168.683",
"value": "pop"
},
{
"operation": "Page search",
"time": "168.964",
"value": "pop"
},
{
"operation": "Page search",
"time": "169.219",
"value": "pop"
},
{
"operation": "hyperlink",
"time": "383.4",
"value": "Good Charlotte"
},
{
"operation": "Now on",
"time": "384.703",
"value": "Good Charlotte (album)"
},
{
"operation": "Highlighting",
"time": "392.568",
"value": "Good Charlotte (album)_sentence_1"
},
{
"operation": "back-button-clicked",
"time": "402.066",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "402.849",
"value": "Good Charlotte"
},
{
"operation": "hyperlink",
"time": "441.433",
"value": "The Young and the Hopeless"
},
{
"operation": "Now on",
"time": "442.585",
"value": "The Young and the Hopeless"
},
{
"operation": "Highlighting",
"time": "463.204",
"value": "The Young and the Hopeless_sentence_10"
},
{
"operation": "back-button-clicked",
"time": "483.542",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "484.402",
"value": "Good Charlotte"
},
{
"operation": "Highlighting",
"time": "504.196",
"value": "Good Charlotte_sentence_9"
},
{
"operation": "Page search",
"time": "621.032",
"value": "pop punk"
},
{
"operation": "Page search",
"time": "621.298",
"value": "pop punk"
},
{
"operation": "Page search",
"time": "621.537",
"value": "pop punk"
},
{
"operation": "Page search",
"time": "621.735",
"value": "pop punk"
},
{
"operation": "Page search",
"time": "621.94",
"value": "pop punk"
},
{
"operation": "Page search",
"time": "622.122",
"value": "pop punk"
},
{
"operation": "Page search",
"time": "622.276",
"value": "pop punk"
},
{
"operation": "Page search",
"time": "716.796",
"value": "punk"
},
{
"operation": "Page search",
"time": "716.994",
"value": "punk"
},
{
"operation": "Page search",
"time": "717.217",
"value": "punk"
},
{
"operation": "Page search",
"time": "717.394",
"value": "punk"
},
{
"operation": "Page search",
"time": "717.545",
"value": "punk"
},
{
"operation": "finish",
"time": "755.613",
"value": "finish"
}
]
}
|
77123
|
The Medway River in England is 70 miles and the Medway Navigation is 19 miles with ten active locks and one demolished lock.
|
{
"evidence": [
"River Medway_sentence_1",
"River Medway_sentence_24",
"River Medway_sentence_2",
"River Medway_sentence_18"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Medway River in England is 70 miles and the Medway Navigation is 19 miles with ten active locks and one demolished lock.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"River Medway_sentence_1",
"River Medway_sentence_24",
"River Medway_sentence_2",
"River Medway_sentence_18"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 34,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.033",
"value": "undefined"
},
{
"operation": "search",
"time": "17.949",
"value": "Medway River"
},
{
"operation": "Now on",
"time": "19.27",
"value": "Medway River"
},
{
"operation": "hyperlink",
"time": "34.53",
"value": "River Medway"
},
{
"operation": "Now on",
"time": "36.104",
"value": "River Medway"
},
{
"operation": "Highlighting",
"time": "44.77",
"value": "River Medway_sentence_0"
},
{
"operation": "hyperlink",
"time": "48.811",
"value": "Medway River (disambiguation)"
},
{
"operation": "Now on",
"time": "49.653",
"value": "Medway River (disambiguation)"
},
{
"operation": "hyperlink",
"time": "81.539",
"value": "Midway River"
},
{
"operation": "Now on",
"time": "82.375",
"value": "Midway River"
},
{
"operation": "back-button-clicked",
"time": "88.716",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "89.183",
"value": "Medway River"
},
{
"operation": "Highlighting",
"time": "151.836",
"value": "Medway River_item_7_5"
},
{
"operation": "Highlighting",
"time": "151.841",
"value": "Medway River_title"
},
{
"operation": "Highlighting deleted",
"time": "154.362",
"value": "Medway River_item_7_5"
},
{
"operation": "Highlighting deleted",
"time": "154.369",
"value": "Medway River_title"
},
{
"operation": "hyperlink",
"time": "155.306",
"value": "River Medway"
},
{
"operation": "Now on",
"time": "156.423",
"value": "River Medway"
},
{
"operation": "Highlighting",
"time": "164.135",
"value": "River Medway_sentence_1"
},
{
"operation": "Page search",
"time": "177.759",
"value": "navigation"
},
{
"operation": "Page search",
"time": "177.929",
"value": "navigation"
},
{
"operation": "Highlighting",
"time": "188.854",
"value": "River Medway_sentence_24"
},
{
"operation": "Page search",
"time": "205.654",
"value": "lock"
},
{
"operation": "Page search",
"time": "206.018",
"value": "lock"
},
{
"operation": "Page search",
"time": "206.244",
"value": "lock"
},
{
"operation": "Highlighting",
"time": "252.611",
"value": "River Medway_sentence_2"
},
{
"operation": "Highlighting",
"time": "269.899",
"value": "River Medway_sentence_18"
},
{
"operation": "finish",
"time": "322.175",
"value": "finish"
}
]
}
|
exfever_6546
|
Joseph Barbera was an Angolan artist who created multiple frames that gave the illusion of movement.
|
{
"evidence": [
"Joseph Barbera",
"Animator"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJoseph Barbera was an Angolan artist who created multiple frames that gave the illusion of movement.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Joseph Barbera",
"Animator"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 35,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
21592
|
In 1829, Sir James Murray used a "condensed solution of fluid magnesia" preparation of his own design to treat the Lord Lieutenant of Ireland of stomach pain; this was so successful that he was appointed resident physician to Anglesey and two subsequent Lords Lieutenant, and knighted.
|
{
"evidence": [
"Magnesium hydroxide_sentence_22",
"Magnesium hydroxide_sentence_23"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn 1829, Sir James Murray used a \"condensed solution of fluid magnesia\" preparation of his own design to treat the Lord Lieutenant of Ireland of stomach pain; this was so successful that he was appointed resident physician to Anglesey and two subsequent Lords Lieutenant, and knighted.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Magnesium hydroxide_sentence_22",
"Magnesium hydroxide_sentence_23"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 36,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.283",
"value": "?search="
},
{
"operation": "search",
"time": "18.137",
"value": "James Murray"
},
{
"operation": "Now on",
"time": "19.668",
"value": "James Murray"
},
{
"operation": "search",
"time": "51.622",
"value": "Sir James Murray"
},
{
"operation": "Now on",
"time": "53.139",
"value": "James Murray"
},
{
"operation": "search",
"time": "83.505",
"value": "condensed solution of fluid magnesia"
},
{
"operation": "Now on",
"time": "85.636",
"value": "undefined"
},
{
"operation": "search",
"time": "108.908",
"value": "treat the Lord Lieutenant of Ireland of stomach pain"
},
{
"operation": "Now on",
"time": "111.01",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "117.001",
"value": "History of milk of magnesia Magnesium hydroxide_section_5"
},
{
"operation": "Now on",
"time": "118.563",
"value": "Magnesium hydroxide#History of milk of magnesia Magnesium hydroxide section 5"
},
{
"operation": "Highlighting",
"time": "128.875",
"value": "Magnesium hydroxide_sentence_22"
},
{
"operation": "Highlighting",
"time": "139.427",
"value": "Magnesium hydroxide_sentence_23"
},
{
"operation": "finish",
"time": "154.621",
"value": "finish"
}
]
}
|
64679
|
PPP2R2A consist of a common heteromeric core enzyme, and it is implicated in the negative control of cell growth.
|
{
"evidence": [
"PPP2R2A_sentence_3",
"PPP2R2A_sentence_2"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPPP2R2A consist of a common heteromeric core enzyme, and it is implicated in the negative control of cell growth.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"PPP2R2A_sentence_3",
"PPP2R2A_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 37,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "3.939",
"value": "?search="
},
{
"operation": "search",
"time": "25.108",
"value": "PPP2R2A"
},
{
"operation": "Now on",
"time": "25.792",
"value": "PPP2R2A"
},
{
"operation": "Highlighting",
"time": "60.35",
"value": "PPP2R2A_sentence_3"
},
{
"operation": "Highlighting",
"time": "75.799",
"value": "PPP2R2A_sentence_2"
},
{
"operation": "finish",
"time": "83.524",
"value": "finish"
}
]
}
|
43382
|
Edward Austin Kent, an architect who died on the Titanic, helped found the Buffalo Society of Architects.
|
{
"evidence": [
"Edward Austin Kent_sentence_0",
"Edward Austin Kent_sentence_1",
"Edward Austin Kent_sentence_8"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nEdward Austin Kent, an architect who died on the Titanic, helped found the Buffalo Society of Architects.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Edward Austin Kent_sentence_0",
"Edward Austin Kent_sentence_1",
"Edward Austin Kent_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 38,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.821",
"value": "?search="
},
{
"operation": "search",
"time": "6.091",
"value": "Edward Austin Kent"
},
{
"operation": "Now on",
"time": "6.949",
"value": "Edward Austin Kent"
},
{
"operation": "Highlighting",
"time": "10.797",
"value": "Edward Austin Kent_sentence_0"
},
{
"operation": "Highlighting",
"time": "15.384",
"value": "Edward Austin Kent_sentence_1"
},
{
"operation": "Highlighting",
"time": "35.011",
"value": "Edward Austin Kent_sentence_8"
},
{
"operation": "finish",
"time": "48.442",
"value": "finish"
}
]
}
|
78043
|
Győző Leblanc’s surname means “white” in French, which is a very rare surname in France.
|
{
"evidence": [
"Le Blanc (disambiguation)_sentence_1",
"Le Blanc (surname)_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nGyőző Leblanc’s surname means “white” in French, which is a very rare surname in France.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Le Blanc (disambiguation)_sentence_1",
"Le Blanc (surname)_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 39,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.005",
"value": "?search="
},
{
"operation": "search",
"time": "13.042",
"value": "Győző Leblanc"
},
{
"operation": "Now on",
"time": "13.564",
"value": "Győző Leblanc"
},
{
"operation": "Page search",
"time": "118.882",
"value": "rare"
},
{
"operation": "search",
"time": "129.674",
"value": "Leblanc"
},
{
"operation": "Now on",
"time": "130.373",
"value": "Le Blanc (disambiguation)"
},
{
"operation": "Highlighting",
"time": "135.873",
"value": "Le Blanc (disambiguation)_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "151.617",
"value": "Le Blanc (disambiguation)_sentence_1"
},
{
"operation": "hyperlink",
"time": "153.217",
"value": "Le Blanc (surname)"
},
{
"operation": "Now on",
"time": "153.818",
"value": "Le Blanc (surname)"
},
{
"operation": "Highlighting",
"time": "157.954",
"value": "Le Blanc (surname)_sentence_0"
},
{
"operation": "Page search",
"time": "195.266",
"value": "french"
},
{
"operation": "search",
"time": "217.346",
"value": "french surnames"
},
{
"operation": "Now on",
"time": "218.675",
"value": "undefined"
},
{
"operation": "search",
"time": "236.747",
"value": "Leblanc french surname"
},
{
"operation": "Now on",
"time": "237.728",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "253.379",
"value": "A Le Blanc (surname)_section_0"
},
{
"operation": "Now on",
"time": "253.655",
"value": "Le Blanc (surname)#A Le Blanc .28surname.29 section 0"
},
{
"operation": "Highlighting deleted",
"time": "258.787",
"value": "Le Blanc (surname)_sentence_0"
},
{
"operation": "search",
"time": "265.587",
"value": "leblanc"
},
{
"operation": "Now on",
"time": "266.021",
"value": "Leblanc"
},
{
"operation": "Highlighting",
"time": "268.321",
"value": "Le Blanc (disambiguation)_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "273.12",
"value": "Le Blanc (disambiguation)_sentence_1"
},
{
"operation": "search",
"time": "337.12",
"value": "french language surnames"
},
{
"operation": "Now on",
"time": "338.436",
"value": "undefined"
},
{
"operation": "search",
"time": "362.171",
"value": "french-language surname"
},
{
"operation": "Now on",
"time": "362.641",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "367.546",
"value": "Chappuis"
},
{
"operation": "Now on",
"time": "367.953",
"value": "Chappuis"
},
{
"operation": "search",
"time": "384.122",
"value": "Leblanc"
},
{
"operation": "Now on",
"time": "384.649",
"value": "Leblanc"
},
{
"operation": "Highlighting",
"time": "388.69",
"value": "Le Blanc (disambiguation)_sentence_1"
},
{
"operation": "hyperlink",
"time": "397.368",
"value": "Le Blanc (surname)"
},
{
"operation": "Now on",
"time": "397.714",
"value": "Le Blanc (surname)"
},
{
"operation": "Highlighting",
"time": "398.937",
"value": "Le Blanc (surname)_sentence_0"
},
{
"operation": "finish",
"time": "407.88",
"value": "finish"
}
]
}
|
exfever_8881
|
See You on the Other Side is the seventh studio album by American nu metal band from Bakersfield, California, formed in 1993.
|
{
"evidence": [
"See You on the Other Side (Korn album)",
"Korn"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSee You on the Other Side is the seventh studio album by American nu metal band from Bakersfield, California, formed in 1993.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"See You on the Other Side (Korn album)",
"Korn"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 40,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
62827
|
The 1973 Tasman Series, was named after the sea between two countries, had a total of eight rounds held in New Zealand and Australia.
|
{
"evidence": [
"Tasman Series_sentence_2",
"1973 Tasman Series_sentence_3",
"1973 Tasman Series_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe 1973 Tasman Series, was named after the sea between two countries, had a total of eight rounds held in New Zealand and Australia.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Tasman Series_sentence_2",
"1973 Tasman Series_sentence_3",
"1973 Tasman Series_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 41,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.14",
"value": "?search="
},
{
"operation": "search",
"time": "32.836",
"value": "Contains...1973 Tasman Series"
},
{
"operation": "Now on",
"time": "34.067",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "38.77",
"value": "1973 Tasman Series"
},
{
"operation": "Now on",
"time": "39.479",
"value": "1973 Tasman Series"
},
{
"operation": "Page search",
"time": "56.895",
"value": "named"
},
{
"operation": "Page search",
"time": "61.919",
"value": "sea"
},
{
"operation": "hyperlink",
"time": "121.7",
"value": "Tasman Formula"
},
{
"operation": "Now on",
"time": "122.752",
"value": "Tasman Formula"
},
{
"operation": "Highlighting",
"time": "132.196",
"value": "Tasman Series_sentence_2"
},
{
"operation": "back-button-clicked",
"time": "136.068",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "136.145",
"value": "1973 Tasman Series"
},
{
"operation": "Page search",
"time": "138.952",
"value": "eight"
},
{
"operation": "Highlighting",
"time": "147.078",
"value": "1973 Tasman Series_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "157.094",
"value": "1973 Tasman Series_sentence_1"
},
{
"operation": "Page search",
"time": "159.098",
"value": "eight"
},
{
"operation": "Highlighting",
"time": "161.057",
"value": "1973 Tasman Series_sentence_3"
},
{
"operation": "Highlighting",
"time": "167.09",
"value": "1973 Tasman Series_sentence_0"
},
{
"operation": "finish",
"time": "174.934",
"value": "finish"
}
]
}
|
exfever_3850
|
A 2020 survey by the World Values Survey found that 96.6% of Iranians believe in Islam.
|
{
"evidence": [
"Allen Ginsberg",
"Eastern religions",
"Traditional African religions"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nA 2020 survey by the World Values Survey found that 96.6% of Iranians believe in Islam.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Allen Ginsberg",
"Eastern religions",
"Traditional African religions"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 42,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
49161
|
History of Ferencvárosi TC drew with Fluminense and won against Brazil and Palestra Itália (founded on January 7, 1921).
|
{
"evidence": [
"History of Ferencvárosi TC_sentence_54",
"History of Ferencvárosi TC_sentence_55"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHistory of Ferencvárosi TC drew with Fluminense and won against Brazil and Palestra Itália (founded on January 7, 1921).\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"History of Ferencvárosi TC_sentence_54",
"History of Ferencvárosi TC_sentence_55"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 43,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.745",
"value": "undefined"
},
{
"operation": "Now on",
"time": "26.112",
"value": "undefined"
},
{
"operation": "search",
"time": "65.287",
"value": "Ferencvárosi TC"
},
{
"operation": "Now on",
"time": "66.581",
"value": "Ferencvárosi TC"
},
{
"operation": "Page search",
"time": "80.029",
"value": "flumineense"
},
{
"operation": "Page search",
"time": "81.501",
"value": "flumineense"
},
{
"operation": "Page search",
"time": "127.572",
"value": "fluminense"
},
{
"operation": "Page search",
"time": "153.547",
"value": "palestra"
},
{
"operation": "hyperlink",
"time": "169.688",
"value": "History of Ferencvárosi TC"
},
{
"operation": "Now on",
"time": "170.873",
"value": "History of Ferencvárosi TC"
},
{
"operation": "Page search",
"time": "183.853",
"value": "Fluminense"
},
{
"operation": "Highlighting",
"time": "194.301",
"value": "History of Ferencvárosi TC_sentence_53"
},
{
"operation": "Page search",
"time": "222.044",
"value": "Brazil and Palestra Itália"
},
{
"operation": "Page search",
"time": "242.803",
"value": "Palestra Itália"
},
{
"operation": "Page search",
"time": "245.346",
"value": "Palestra Itália"
},
{
"operation": "Highlighting",
"time": "260.818",
"value": "History of Ferencvárosi TC_sentence_54"
},
{
"operation": "Highlighting deleted",
"time": "262.062",
"value": "History of Ferencvárosi TC_sentence_54"
},
{
"operation": "Highlighting",
"time": "262.68",
"value": "History of Ferencvárosi TC_sentence_54"
},
{
"operation": "Highlighting",
"time": "263.94",
"value": "History of Ferencvárosi TC_sentence_55"
},
{
"operation": "hyperlink",
"time": "279.78",
"value": "Palestra Itália"
},
{
"operation": "Now on",
"time": "280.619",
"value": "Palestra Itália Futebol Clube"
},
{
"operation": "back-button-clicked",
"time": "283.459",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "283.637",
"value": "History of Ferencvárosi TC"
},
{
"operation": "Highlighting deleted",
"time": "286.338",
"value": "History of Ferencvárosi TC_sentence_53"
},
{
"operation": "finish",
"time": "295.661",
"value": "finish"
}
]
}
|
89219
|
In the Trinity By-Election 2005 in Chelmsford City, Conservative candidate Martin Bracken beating Paul Hutchinson of the Liberal Democrats party (which formed a UK coalition government with the Conservative Party from 2010 to 2015).
|
{
"evidence": [
"Chelmsford (UK Parliament constituency)_sentence_0",
"Chelmsford (UK Parliament constituency)_sentence_4"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn the Trinity By-Election 2005 in Chelmsford City, Conservative candidate Martin Bracken beating Paul Hutchinson of the Liberal Democrats party (which formed a UK coalition government with the Conservative Party from 2010 to 2015).\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Chelmsford (UK Parliament constituency)_sentence_0",
"Chelmsford (UK Parliament constituency)_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 44,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.1",
"value": "undefined"
},
{
"operation": "search",
"time": "57.133",
"value": "Contains...Martin Bracken"
},
{
"operation": "Now on",
"time": "58.792",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "64.181",
"value": "West Chelmsford (UK Parliament constituency)"
},
{
"operation": "Now on",
"time": "65.463",
"value": "West Chelmsford (UK Parliament constituency)"
},
{
"operation": "search",
"time": "136.63",
"value": "Trinity"
},
{
"operation": "Now on",
"time": "138.09",
"value": "Trinity"
},
{
"operation": "hyperlink",
"time": "154.771",
"value": "Trinity (disambiguation)"
},
{
"operation": "Now on",
"time": "155.651",
"value": "Trinity (disambiguation)"
},
{
"operation": "search",
"time": "216.191",
"value": "Chelmsford City"
},
{
"operation": "Now on",
"time": "217.393",
"value": "Chelmsford City F.C."
},
{
"operation": "back-button-clicked",
"time": "244.366",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "244.482",
"value": "Trinity (disambiguation)"
},
{
"operation": "hyperlink",
"time": "257.156",
"value": "8.2 United Kingdom Trinity (disambiguation)_section_12"
},
{
"operation": "Highlighting",
"time": "267.867",
"value": "Trinity (disambiguation)_item_10_54"
},
{
"operation": "Highlighting",
"time": "267.868",
"value": "Trinity (disambiguation)_title"
},
{
"operation": "Highlighting",
"time": "267.868",
"value": "Trinity (disambiguation)_section_10"
},
{
"operation": "Highlighting",
"time": "267.868",
"value": "Trinity (disambiguation)_section_12"
},
{
"operation": "Highlighting deleted",
"time": "269.926",
"value": "Trinity (disambiguation)_item_10_54"
},
{
"operation": "Highlighting deleted",
"time": "269.927",
"value": "Trinity (disambiguation)_title"
},
{
"operation": "Highlighting deleted",
"time": "269.928",
"value": "Trinity (disambiguation)_section_10"
},
{
"operation": "Highlighting deleted",
"time": "269.928",
"value": "Trinity (disambiguation)_section_12"
},
{
"operation": "Highlighting",
"time": "272.051",
"value": "Trinity (disambiguation)_item_10_54"
},
{
"operation": "Highlighting",
"time": "272.051",
"value": "Trinity (disambiguation)_title"
},
{
"operation": "Highlighting",
"time": "272.051",
"value": "Trinity (disambiguation)_section_10"
},
{
"operation": "Highlighting",
"time": "272.051",
"value": "Trinity (disambiguation)_section_12"
},
{
"operation": "Highlighting deleted",
"time": "290.875",
"value": "Trinity (disambiguation)_item_10_54"
},
{
"operation": "Highlighting deleted",
"time": "290.876",
"value": "Trinity (disambiguation)_title"
},
{
"operation": "Highlighting deleted",
"time": "290.876",
"value": "Trinity (disambiguation)_section_10"
},
{
"operation": "Highlighting deleted",
"time": "290.877",
"value": "Trinity (disambiguation)_section_12"
},
{
"operation": "search",
"time": "317.167",
"value": "Contains...By-Election 2005 in Chelmsford City"
},
{
"operation": "Now on",
"time": "319.652",
"value": "undefined"
},
{
"operation": "search",
"time": "334.241",
"value": "Chelmsford City"
},
{
"operation": "Now on",
"time": "335.189",
"value": "Chelmsford City F.C."
},
{
"operation": "search",
"time": "464.833",
"value": "Contains...Trinity By-Election 2005 in Chelmsford City"
},
{
"operation": "Now on",
"time": "465.749",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "479.476",
"value": "Chelmsford (UK Parliament constituency)"
},
{
"operation": "Now on",
"time": "481.581",
"value": "Chelmsford (UK Parliament constituency)"
},
{
"operation": "Highlighting",
"time": "486.33",
"value": "Chelmsford (UK Parliament constituency)_sentence_0"
},
{
"operation": "Highlighting",
"time": "592.092",
"value": "Chelmsford (UK Parliament constituency)_sentence_4"
},
{
"operation": "search",
"time": "609.583",
"value": "Contains...Trinity By-Election 2005 in Chelmsford"
},
{
"operation": "Now on",
"time": "610.638",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "617.003",
"value": "Vicky Ford"
},
{
"operation": "Now on",
"time": "617.693",
"value": "Vicky Ford"
},
{
"operation": "back-button-clicked",
"time": "630.641",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "631.497",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "646.255",
"value": "List of electoral wards in Essex"
},
{
"operation": "Now on",
"time": "648.412",
"value": "List of electoral wards in Essex"
},
{
"operation": "hyperlink",
"time": "659.065",
"value": "4.5 Chelmsford List of electoral wards in Essex_section_23"
},
{
"operation": "finish",
"time": "761.633",
"value": "finish"
}
]
}
|
exfever_7823
|
Damages is an American legal thriller that premiered on FX and aired for three seasons before moving to an American multichannel video programming distributor based in El Segundo, Mexico.
|
{
"evidence": [
"Damages (TV series)",
"DirecTV"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDamages is an American legal thriller that premiered on FX and aired for three seasons before moving to an American multichannel video programming distributor based in El Segundo, Mexico.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Damages (TV series)",
"DirecTV"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 45,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
69844
|
James Joseph Patterson was the son of Joseph Medill Patterson, the founder of the New York Daily News which was the first U.S. daily newspaper printed in tabloid format, and was the eleventh-highest circulated newspaper in the United States as of 2019.
|
{
"evidence": [
"James Joseph Patterson_sentence_3",
"New York Daily News_sentence_3",
"New York Daily News_sentence_4"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJames Joseph Patterson was the son of Joseph Medill Patterson, the founder of the New York Daily News which was the first U.S. daily newspaper printed in tabloid format, and was the eleventh-highest circulated newspaper in the United States as of 2019.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"James Joseph Patterson_sentence_3",
"New York Daily News_sentence_3",
"New York Daily News_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 46,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.064",
"value": "?search="
},
{
"operation": "search",
"time": "114.92",
"value": "Joseph Medill Patterson"
},
{
"operation": "Now on",
"time": "115.946",
"value": "Joseph Medill Patterson"
},
{
"operation": "hyperlink",
"time": "132.11",
"value": "James Joseph Patterson"
},
{
"operation": "Now on",
"time": "133.42",
"value": "James Joseph Patterson"
},
{
"operation": "Highlighting",
"time": "161.824",
"value": "James Joseph Patterson_sentence_3"
},
{
"operation": "hyperlink",
"time": "189.118",
"value": "New York Daily News"
},
{
"operation": "Now on",
"time": "189.833",
"value": "New York Daily News"
},
{
"operation": "Highlighting",
"time": "243.597",
"value": "New York Daily News_sentence_3"
},
{
"operation": "Highlighting",
"time": "262.621",
"value": "New York Daily News_sentence_4"
},
{
"operation": "finish",
"time": "311.462",
"value": "finish"
}
]
}
|
exfever_5383
|
A counterculture is a culture whose values and norms of behavior differ substantially from those of mainstream society with a specific political movement during the English Civil War as an example.
|
{
"evidence": [
"Counterculture",
"Levellers"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nA counterculture is a culture whose values and norms of behavior differ substantially from those of mainstream society with a specific political movement during the English Civil War as an example.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Counterculture",
"Levellers"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 47,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_2408
|
Role Models has film stars, including an American actor and comedian who also appeared in a 2005 American action-comedy road film directed by Jay Chandrasekhar.
|
{
"evidence": [
"Role Models",
"Seann William Scott",
"The Dukes of Hazzard (film)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRole Models has film stars, including an American actor and comedian who also appeared in a 2005 American action-comedy road film directed by Jay Chandrasekhar.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Role Models",
"Seann William Scott",
"The Dukes of Hazzard (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 48,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
51431
|
The Samfälligheten för Nordisk Sed which was formally founded in 1997 recognizes other religions and collaborates with other modern Pagan organisations like the Forn Sed Assembly and the Nordic Asa-Community.
|
{
"evidence": [
"Samfälligheten för Nordisk Sed_sentence_4",
"Samfälligheten för Nordisk Sed_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Samfälligheten för Nordisk Sed which was formally founded in 1997 recognizes other religions and collaborates with other modern Pagan organisations like the Forn Sed Assembly and the Nordic Asa-Community.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Samfälligheten för Nordisk Sed_sentence_4",
"Samfälligheten för Nordisk Sed_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 49,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.612",
"value": "?search="
},
{
"operation": "search",
"time": "31.883",
"value": "Samfälligheten för Nordisk Sed"
},
{
"operation": "Now on",
"time": "32.899",
"value": "Samfälligheten för Nordisk Sed"
},
{
"operation": "Highlighting",
"time": "49.097",
"value": "Samfälligheten för Nordisk Sed_sentence_2"
},
{
"operation": "Highlighting",
"time": "53.653",
"value": "Samfälligheten för Nordisk Sed_sentence_4"
},
{
"operation": "Highlighting",
"time": "72.497",
"value": "Samfälligheten för Nordisk Sed_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "73.901",
"value": "Samfälligheten för Nordisk Sed_sentence_2"
},
{
"operation": "finish",
"time": "79.886",
"value": "finish"
}
]
}
|
exfever_5832
|
Vincent Cassel performed as a troubled French Jewish youth in a film by an Italian director and founder of MNP Entreprise.
|
{
"evidence": [
"Vincent Cassel",
"Mathieu Kassovitz"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nVincent Cassel performed as a troubled French Jewish youth in a film by an Italian director and founder of MNP Entreprise.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Vincent Cassel",
"Mathieu Kassovitz"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 50,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_6259
|
Drax the Destroyer is Jim Starlin's first created?fictional character for Marvel Comics.
|
{
"evidence": [
"Drax the Destroyer",
"American comic book",
"Television"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDrax the Destroyer is Jim Starlin's first created?fictional character for Marvel Comics.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Drax the Destroyer",
"American comic book",
"Television"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 51,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_5082
|
Bosnia and Herzegovina is the third country in Europe that is mostly mountainous.
|
{
"evidence": [
"Bosnia and Herzegovina",
"Sarajevo"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBosnia and Herzegovina is the third country in Europe that is mostly mountainous.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Bosnia and Herzegovina",
"Sarajevo"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 52,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
34985
|
Charles Herbert Ruffing, one of five children of John and Frances Ruffing, was born on May 3, 1905, in Granville, a village in Putnam County, Illinois, United States.
|
{
"evidence": [
"Red Ruffing_sentence_16",
"Red Ruffing_sentence_15",
"Granville, Illinois_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nCharles Herbert Ruffing, one of five children of John and Frances Ruffing, was born on May 3, 1905, in Granville, a village in Putnam County, Illinois, United States.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Red Ruffing_sentence_16",
"Red Ruffing_sentence_15",
"Granville, Illinois_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 53,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "4.004",
"value": "?search="
},
{
"operation": "search",
"time": "16.355",
"value": "Charles Herbert Ruffing"
},
{
"operation": "Now on",
"time": "17.904",
"value": "Red Ruffing"
},
{
"operation": "Highlighting",
"time": "41.302",
"value": "Red Ruffing_sentence_16"
},
{
"operation": "Highlighting",
"time": "54.989",
"value": "Red Ruffing_sentence_0"
},
{
"operation": "Page search",
"time": "76.542",
"value": "Grainville"
},
{
"operation": "Page search",
"time": "76.804",
"value": "Grainville"
},
{
"operation": "Page search",
"time": "77.404",
"value": "Grainville"
},
{
"operation": "Page search",
"time": "77.66",
"value": "Grainville"
},
{
"operation": "Page search",
"time": "77.899",
"value": "Grainville"
},
{
"operation": "Page search",
"time": "78.132",
"value": "Grainville"
},
{
"operation": "Page search",
"time": "78.284",
"value": "Grainville"
},
{
"operation": "Highlighting",
"time": "108.907",
"value": "Red Ruffing_sentence_15"
},
{
"operation": "Highlighting deleted",
"time": "119.314",
"value": "Red Ruffing_sentence_0"
},
{
"operation": "hyperlink",
"time": "134.663",
"value": "Granville, Illinois"
},
{
"operation": "Now on",
"time": "135.394",
"value": "Granville, Illinois"
},
{
"operation": "Highlighting",
"time": "137.248",
"value": "Granville, Illinois_sentence_0"
},
{
"operation": "finish",
"time": "147.289",
"value": "finish"
}
]
}
|
exfever_7953
|
Andre Agassi is the first of two men to achieve the Grand Slam, which consists of winning all four major championships in one discipline in the same calendar year in the racket sport, which can be played individually against a single opponent or between two teams of two players each.
|
{
"evidence": [
"Andre Agassi",
"Grand Slam (tennis)",
"Tennis"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAndre Agassi is the first of two men to achieve the Grand Slam, which consists of winning all four major championships in one discipline in the same calendar year in the racket sport, which can be played individually against a single opponent or between two teams of two players each.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Andre Agassi",
"Grand Slam (tennis)",
"Tennis"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 54,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_7442
|
Interscope Records was founded by multi-platinum producers and songwriters such as Benny Blanco.
|
{
"evidence": [
"Lady Gaga",
"KonLive Distribution"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nInterscope Records was founded by multi-platinum producers and songwriters such as Benny Blanco.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lady Gaga",
"KonLive Distribution"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 55,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
19496
|
In 2005, Jan Arild Ellingsen, who is currently the Progress Party's spokesperson on matters relating to justice and crime, accepted an Iraqi death sentence over a Norwegian Iraqi, drawing a sharp rebuke from fellow Progress Party MP André Kvakkestad.
|
{
"evidence": [
"Jan Arild Ellingsen_sentence_2",
"Jan Arild Ellingsen_sentence_3"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn 2005, Jan Arild Ellingsen, who is currently the Progress Party's spokesperson on matters relating to justice and crime, accepted an Iraqi death sentence over a Norwegian Iraqi, drawing a sharp rebuke from fellow Progress Party MP André Kvakkestad.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Jan Arild Ellingsen_sentence_2",
"Jan Arild Ellingsen_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 56,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.226",
"value": "?search="
},
{
"operation": "search",
"time": "34.346",
"value": "Jan Arild Ellingsen"
},
{
"operation": "Now on",
"time": "35.213",
"value": "Jan Arild Ellingsen"
},
{
"operation": "Highlighting",
"time": "90.695",
"value": "Jan Arild Ellingsen_sentence_2"
},
{
"operation": "Highlighting",
"time": "126.078",
"value": "Jan Arild Ellingsen_sentence_3"
},
{
"operation": "finish",
"time": "139.178",
"value": "finish"
}
]
}
|
21670
|
Kewal Singh (an Indian diplomat, Foreign Treasurer and India's ambassador to the USSR, Pakistan and USA) took office on October 21, 1954 then left on November 16, 1956; before M.K. Kripalani took over on November 17, 1956.
|
{
"evidence": [
"Kewal Singh_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nKewal Singh (an Indian diplomat, Foreign Treasurer and India's ambassador to the USSR, Pakistan and USA) took office on October 21, 1954 then left on November 16, 1956; before M.K. Kripalani took over on November 17, 1956.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Kewal Singh_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 57,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.948",
"value": "?search="
},
{
"operation": "search",
"time": "13.687",
"value": "Contains...Kawal Singh"
},
{
"operation": "Now on",
"time": "14.655",
"value": "undefined"
},
{
"operation": "search",
"time": "37.598",
"value": "Kewal Singh"
},
{
"operation": "Now on",
"time": "38.576",
"value": "Kewal Singh"
},
{
"operation": "Highlighting",
"time": "49.682",
"value": "Kewal Singh_sentence_0"
},
{
"operation": "search",
"time": "120.722",
"value": "Contains...M.K. Kripalani"
},
{
"operation": "Now on",
"time": "122.387",
"value": "undefined"
},
{
"operation": "search",
"time": "133.335",
"value": "Contains...kripalani"
},
{
"operation": "Now on",
"time": "133.945",
"value": "undefined"
},
{
"operation": "search",
"time": "154.813",
"value": "Indian diplomatic history"
},
{
"operation": "Now on",
"time": "158.789",
"value": "Foreign relations of India"
},
{
"operation": "Page search",
"time": "163.476",
"value": "kewal singh"
},
{
"operation": "search",
"time": "246.65",
"value": "Contains...J.B. kripalani"
},
{
"operation": "Now on",
"time": "247.514",
"value": "undefined"
},
{
"operation": "search",
"time": "254.979",
"value": "Contains...kripalani"
},
{
"operation": "Now on",
"time": "255.669",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "258.458",
"value": "Janata Party"
},
{
"operation": "Now on",
"time": "259.618",
"value": "Janata Party"
},
{
"operation": "Page search",
"time": "270.161",
"value": "kripalani"
},
{
"operation": "Highlighting",
"time": "278.931",
"value": "Janata Party_sentence_9Janata Party_sentence_10"
},
{
"operation": "Highlighting",
"time": "278.932",
"value": "Janata Party_title"
},
{
"operation": "Highlighting deleted",
"time": "279.933",
"value": "Janata Party_sentence_9Janata Party_sentence_10"
},
{
"operation": "Highlighting deleted",
"time": "279.935",
"value": "Janata Party_title"
},
{
"operation": "search",
"time": "289.229",
"value": "Contains...Jivatram kripalani"
},
{
"operation": "Now on",
"time": "289.827",
"value": "undefined"
},
{
"operation": "search",
"time": "334.497",
"value": "Contains...\"October 21, 1954\" \"Kewal\""
},
{
"operation": "Now on",
"time": "335.373",
"value": "undefined"
},
{
"operation": "search",
"time": "354.379",
"value": "Contains...\"October 21\" \"Kewal\""
},
{
"operation": "Now on",
"time": "354.842",
"value": "undefined"
},
{
"operation": "finish",
"time": "384.957",
"value": "finish"
}
]
}
|
47285
|
Sam Concepcion appeared in less than 20 television shows from 2011 to 2020.
|
{
"evidence": [
"Sam Concepcion_sentence_3"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSam Concepcion appeared in less than 20 television shows from 2011 to 2020.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sam Concepcion_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 58,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.152",
"value": "?search="
},
{
"operation": "search",
"time": "27.294",
"value": "Sam Concepcion"
},
{
"operation": "Now on",
"time": "28.589",
"value": "Sam Concepcion"
},
{
"operation": "Highlighting",
"time": "45.588",
"value": "Sam Concepcion_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "187.136",
"value": "Sam Concepcion_sentence_3"
},
{
"operation": "search",
"time": "196.989",
"value": "Sam Concepcion filmography"
},
{
"operation": "Now on",
"time": "198.636",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "209.366",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "209.473",
"value": "Sam Concepcion"
},
{
"operation": "Highlighting",
"time": "233.244",
"value": "Sam Concepcion_sentence_3"
},
{
"operation": "finish",
"time": "286.872",
"value": "finish"
}
]
}
|
13214
|
Christian Maicon Hening was born 25 August 1980 in Blumenau, Brazil and played with the Coritiba team.
|
{
"evidence": [
"Christian Maicon Hening_sentence_0"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nChristian Maicon Hening was born 25 August 1980 in Blumenau, Brazil and played with the Coritiba team.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Christian Maicon Hening_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 59,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.843",
"value": "?search="
},
{
"operation": "search",
"time": "14.921",
"value": "Christian Maicon Hening"
},
{
"operation": "Now on",
"time": "15.934",
"value": "Christian Maicon Hening"
},
{
"operation": "Highlighting",
"time": "36.976",
"value": "Christian Maicon Hening_sentence_0"
},
{
"operation": "finish",
"time": "51.34",
"value": "finish"
}
]
}
|
73955
|
Bangladeshi /Dhallywood films (based in Dhaka, Bangladesh) evolved in drama, action, war, and romance in 2011.
|
{
"evidence": [
"Cinema of Bangladesh_sentence_2"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBangladeshi /Dhallywood films (based in Dhaka, Bangladesh) evolved in drama, action, war, and romance in 2011.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Cinema of Bangladesh_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 60,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.728",
"value": "?search="
},
{
"operation": "search",
"time": "10.888",
"value": "Contains...Bangladeshi /Dhallywood films"
},
{
"operation": "Now on",
"time": "13.353",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "17.24",
"value": "bangladesh dhallywood films"
},
{
"operation": "Now on",
"time": "18.818",
"value": "undefined"
},
{
"operation": "search",
"time": "34.448",
"value": "Contains...Bangladeshi Dhallywood films"
},
{
"operation": "Now on",
"time": "35.483",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "41.818",
"value": "International recognition of Bangladeshi cinema Cinema of Bangladesh_section_14"
},
{
"operation": "Now on",
"time": "43.916",
"value": "Cinema of Bangladesh#International recognition of Bangladeshi cinema Cinema of Bangladesh section 14"
},
{
"operation": "Page search",
"time": "52.953",
"value": "Bangladeshi /Dhallywood films"
},
{
"operation": "search",
"time": "109.818",
"value": "\"evolved in drama, action, war, and romance \""
},
{
"operation": "Now on",
"time": "111.304",
"value": "undefined"
},
{
"operation": "search",
"time": "122.69",
"value": "Contains...\"based in Dhaka, Bangladesh\""
},
{
"operation": "Now on",
"time": "125.18",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "132.602",
"value": "Cinema of Bangladesh"
},
{
"operation": "Now on",
"time": "133.433",
"value": "Cinema of Bangladesh"
},
{
"operation": "Highlighting",
"time": "135.594",
"value": "Cinema of Bangladesh_sentence_2"
},
{
"operation": "Page search",
"time": "157.595",
"value": "evolved"
},
{
"operation": "Page search",
"time": "218.9",
"value": "2011"
},
{
"operation": "Page search",
"time": "220.079",
"value": "2011"
},
{
"operation": "Page search",
"time": "282.402",
"value": "drama, action"
},
{
"operation": "finish",
"time": "301.686",
"value": "finish"
}
]
}
|
50534
|
Westminster Schools of Augusta fields 16 varsity interscholastic sport teams which has never won state and regional championships.
|
{
"evidence": [
"Westminster Schools of Augusta_sentence_8",
"Westminster Schools of Augusta_sentence_9"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWestminster Schools of Augusta fields 16 varsity interscholastic sport teams which has never won state and regional championships.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Westminster Schools of Augusta_sentence_8",
"Westminster Schools of Augusta_sentence_9"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 61,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.179",
"value": "?search="
},
{
"operation": "search",
"time": "9.416",
"value": "Westminster Schools of Augusta"
},
{
"operation": "Now on",
"time": "10.365",
"value": "Westminster Schools of Augusta"
},
{
"operation": "Highlighting",
"time": "31.824",
"value": "Westminster Schools of Augusta_sentence_10"
},
{
"operation": "Highlighting",
"time": "36.264",
"value": "Westminster Schools of Augusta_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "37.472",
"value": "Westminster Schools of Augusta_sentence_10"
},
{
"operation": "Highlighting",
"time": "38.352",
"value": "Westminster Schools of Augusta_sentence_9"
},
{
"operation": "finish",
"time": "55.535",
"value": "finish"
}
]
}
|
exfever_6406
|
Millie Bobby Brown was the most famous actress in Stranger Things.
|
{
"evidence": [
"Millie Bobby Brown",
"Netflix",
"Latin America"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMillie Bobby Brown was the most famous actress in Stranger Things.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Millie Bobby Brown",
"Netflix",
"Latin America"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 62,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_6339
|
Outlander is a historical drama television series based on the ongoing novel series of the same name by an American author whose books merge multiple genres, including elements of?a fiction genre where the nature of an event remains mysterious until the end of the story.
|
{
"evidence": [
"Outlander (TV series)",
"Diana Gabaldon",
"Mystery fiction"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nOutlander is a historical drama television series based on the ongoing novel series of the same name by an American author whose books merge multiple genres, including elements of?a fiction genre where the nature of an event remains mysterious until the end of the story.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Outlander (TV series)",
"Diana Gabaldon",
"Mystery fiction"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 63,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_2963
|
Jimmy Wales first founded Bomis, a web portal primarily known for featuring adult content.
|
{
"evidence": [
"Jimmy Wales",
"Huntsville, Alabama"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJimmy Wales first founded Bomis, a web portal primarily known for featuring adult content.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jimmy Wales",
"Huntsville, Alabama"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 64,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_8457
|
The human condition is all of the characteristics that compose the essentials of human existence, including the time during which one or more offspring develops inside a woman's mammary gland, after which delivery is the termination of a pregnancy in which one or more babies leave the uterus via vaginal delivery or Caesarean section.?
|
{
"evidence": [
"Human condition",
"Childbirth",
"Pregnancy"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe human condition is all of the characteristics that compose the essentials of human existence, including the time during which one or more offspring develops inside a woman's mammary gland, after which delivery is the termination of a pregnancy in which one or more babies leave the uterus via vaginal delivery or Caesarean section.?\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Human condition",
"Childbirth",
"Pregnancy"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 65,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_3001
|
Wyatt Berry Stapp Earp was a Korean lawman and the brother of deputy U.S. Marshal and Tombstone, Arizona City Marshal, who led his younger brothers and Wyatt Earp's American gambler friend.
|
{
"evidence": [
"Wyatt Earp",
"Virgil Earp",
"Doc Holliday"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWyatt Berry Stapp Earp was a Korean lawman and the brother of deputy U.S. Marshal and Tombstone, Arizona City Marshal, who led his younger brothers and Wyatt Earp's American gambler friend.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Wyatt Earp",
"Virgil Earp",
"Doc Holliday"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 66,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
37760
|
PLD2 is activated by substrate presentation and the nutrients is palmitoylated, which drives PLD2 to lipid rafts.
|
{
"evidence": [
"PLD2_sentence_5",
"PLD2_sentence_6"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPLD2 is activated by substrate presentation and the nutrients is palmitoylated, which drives PLD2 to lipid rafts.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"PLD2_sentence_5",
"PLD2_sentence_6"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 67,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.917",
"value": "undefined"
},
{
"operation": "search",
"time": "16.112",
"value": "PLD2"
},
{
"operation": "Now on",
"time": "17.698",
"value": "PLD2"
},
{
"operation": "Page search",
"time": "36.902",
"value": "activa"
},
{
"operation": "Highlighting",
"time": "54.911",
"value": "PLD2_sentence_0"
},
{
"operation": "Highlighting",
"time": "57.334",
"value": "PLD2_sentence_1"
},
{
"operation": "Highlighting",
"time": "57.339",
"value": "PLD2_section_0"
},
{
"operation": "Highlighting",
"time": "77.935",
"value": "PLD2_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "78.371",
"value": "PLD2_sentence_1"
},
{
"operation": "Highlighting",
"time": "102.05",
"value": "PLD2_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "103.834",
"value": "PLD2_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "103.844",
"value": "PLD2_section_0"
},
{
"operation": "Highlighting deleted",
"time": "105.183",
"value": "PLD2_sentence_0"
},
{
"operation": "Highlighting",
"time": "119.686",
"value": "PLD2_sentence_6"
},
{
"operation": "Page search",
"time": "132.653",
"value": "nutrients"
},
{
"operation": "Page search",
"time": "133.604",
"value": "nutrients"
},
{
"operation": "Page search",
"time": "133.907",
"value": "nutrients"
},
{
"operation": "Page search",
"time": "134.179",
"value": "nutrients"
},
{
"operation": "finish",
"time": "157.675",
"value": "finish"
}
]
}
|
exfever_7571
|
Keanu Reeves' first film is Johnny Mnemonic.
|
{
"evidence": [
"Johnny Mnemonic (film)",
"Robert Longo"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nKeanu Reeves' first film is Johnny Mnemonic.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Johnny Mnemonic (film)",
"Robert Longo"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 68,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
65079
|
MSX1 have roles in limb-pattern formation, there is also strong evidence from sequencing studies and mutation in this gene associated with diseases which was first described in 1965.
|
{
"evidence": [
"MSX1_sentence_6",
"MSX1_sentence_7",
"MSX1_sentence_8"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMSX1 have roles in limb-pattern formation, there is also strong evidence from sequencing studies and mutation in this gene associated with diseases which was first described in 1965.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"MSX1_sentence_6",
"MSX1_sentence_7",
"MSX1_sentence_8"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 69,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.404",
"value": "?search="
},
{
"operation": "search",
"time": "12.51",
"value": "MSX1"
},
{
"operation": "Now on",
"time": "16.083",
"value": "MSX1"
},
{
"operation": "Highlighting",
"time": "67.709",
"value": "MSX1_sentence_6"
},
{
"operation": "Highlighting",
"time": "78.029",
"value": "MSX1_sentence_7"
},
{
"operation": "hyperlink",
"time": "166.812",
"value": "cleft lip and palate"
},
{
"operation": "Now on",
"time": "182.244",
"value": "Cleft lip and cleft palate"
},
{
"operation": "search",
"time": "191.023",
"value": "MSX1"
},
{
"operation": "Now on",
"time": "191.561",
"value": "MSX1"
},
{
"operation": "Highlighting",
"time": "254.835",
"value": "MSX1_sentence_8"
},
{
"operation": "hyperlink",
"time": "283.867",
"value": "homeobox"
},
{
"operation": "Now on",
"time": "297.839",
"value": "Homeobox"
},
{
"operation": "search",
"time": "303.398",
"value": "MSX1"
},
{
"operation": "Now on",
"time": "304.777",
"value": "MSX1"
},
{
"operation": "hyperlink",
"time": "319.668",
"value": "Witkop syndrome"
},
{
"operation": "Now on",
"time": "320.073",
"value": "Tooth and nail syndrome"
},
{
"operation": "search",
"time": "334.093",
"value": "MSX1"
},
{
"operation": "Now on",
"time": "335.19",
"value": "MSX1"
},
{
"operation": "hyperlink",
"time": "358.292",
"value": "hypodontia"
},
{
"operation": "Now on",
"time": "359.17",
"value": "Hypodontia"
},
{
"operation": "search",
"time": "375.702",
"value": "MSX1"
},
{
"operation": "Now on",
"time": "376.263",
"value": "MSX1"
},
{
"operation": "hyperlink",
"time": "416.892",
"value": "Msh homeobox 2"
},
{
"operation": "Now on",
"time": "418.621",
"value": "Msh homeobox 2"
},
{
"operation": "finish",
"time": "426.443",
"value": "finish"
}
]
}
|
exfever_3559
|
Mark Zuckerberg founded Facebook in 2004, along with fellow Harvard College students and roommates, one of whom invested in early-stage startups such as Qwiki, and an online mobile payments and identity verification company.
|
{
"evidence": [
"Facebook",
"Eduardo Saverin",
"Jumio"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMark Zuckerberg founded Facebook in 2004, along with fellow Harvard College students and roommates, one of whom invested in early-stage startups such as Qwiki, and an online mobile payments and identity verification company.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Facebook",
"Eduardo Saverin",
"Jumio"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 70,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_8411
|
American singer Rihanna is featured in Chris Brown's Graffiti album.
|
{
"evidence": [
"Graffiti (Chris Brown album)",
"Jive Records"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAmerican singer Rihanna is featured in Chris Brown's Graffiti album.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Graffiti (Chris Brown album)",
"Jive Records"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 71,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_4302
|
Tunisia is the nearest Arab country that trades with Africa.
|
{
"evidence": [
"Tunisia",
"Libya"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nTunisia is the nearest Arab country that trades with Africa.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tunisia",
"Libya"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 72,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
65248
|
Andre Adams debuted at the highest level for New Zealand (nicknamed the Black Caps) in March 2005, and later signed with Essex where he stayed for the 2005 and 2006 seasons.
|
{
"evidence": [
"Andre Adams_sentence_1"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAndre Adams debuted at the highest level for New Zealand (nicknamed the Black Caps) in March 2005, and later signed with Essex where he stayed for the 2005 and 2006 seasons.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Andre Adams_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 73,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.582",
"value": "?search="
},
{
"operation": "search",
"time": "16.614",
"value": "Andre Adams"
},
{
"operation": "Now on",
"time": "18.059",
"value": "Andre Adams"
},
{
"operation": "Highlighting",
"time": "26.673",
"value": "Andre Adams_sentence_1"
},
{
"operation": "hyperlink",
"time": "40.32",
"value": "New Zealand"
},
{
"operation": "Now on",
"time": "54.924",
"value": "Andre Adams"
},
{
"operation": "hyperlink",
"time": "56.899",
"value": "New Zealand"
},
{
"operation": "Now on",
"time": "59.538",
"value": "New Zealand national cricket team"
},
{
"operation": "Highlighting",
"time": "61.059",
"value": "New Zealand national cricket team_sentence_1"
},
{
"operation": "Highlighting",
"time": "61.07",
"value": "New Zealand national cricket team_title"
},
{
"operation": "back-button-clicked",
"time": "63.578",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "63.859",
"value": "Andre Adams"
},
{
"operation": "finish",
"time": "81.338",
"value": "finish"
}
]
}
|
exfever_1403
|
A Thousand Suns are the fourth studio album by British rock band Linkin Park that took place at NRG Recording Studios in a state in the Western United States that borders to the east the Western region of the United States known as the 7th-most extensive, the 32nd-most populous, and the 9th-least densely populated of the U.S. states.
|
{
"evidence": [
"A Thousand Suns",
"California",
"Nevada"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nA Thousand Suns are the fourth studio album by British rock band Linkin Park that took place at NRG Recording Studios in a state in the Western United States that borders to the east the Western region of the United States known as the 7th-most extensive, the 32nd-most populous, and the 9th-least densely populated of the U.S. states.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"A Thousand Suns",
"California",
"Nevada"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 74,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
82324
|
Lanjera is a small village and in 2011 was home to 207 houses of 527 males, 498 females and 138 teenagers, according to the Census India
|
{
"evidence": [
"Lanjera_sentence_3",
"Lanjera_sentence_5"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLanjera is a small village and in 2011 was home to 207 houses of 527 males, 498 females and 138 teenagers, according to the Census India\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Lanjera_sentence_3",
"Lanjera_sentence_5"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 75,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.83",
"value": "?search="
},
{
"operation": "search",
"time": "5.935",
"value": "Lanjera"
},
{
"operation": "Now on",
"time": "6.878",
"value": "Lanjera"
},
{
"operation": "Page search",
"time": "10.166",
"value": "small"
},
{
"operation": "Highlighting",
"time": "22.885",
"value": "Lanjera_sentence_3"
},
{
"operation": "Highlighting",
"time": "55.14",
"value": "Lanjera_sentence_5"
},
{
"operation": "finish",
"time": "357.41",
"value": "finish"
}
]
}
|
exfever_2113
|
Warner Bros. Records U.S. The company owns and operates some of the largest and most successful labels in the world.
|
{
"evidence": [
"One More Light",
"Linkin Park"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWarner Bros. Records U.S. The company owns and operates some of the largest and most successful labels in the world.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"One More Light",
"Linkin Park"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 76,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_360
|
Himmler's first name, Heinrich, was that of his godfather.
|
{
"evidence": [
"Heinrich Himmler",
"Nazi Party"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHimmler's first name, Heinrich, was that of his godfather.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Heinrich Himmler",
"Nazi Party"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 77,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_2382
|
Michelle Williams was born in Kalispell, Montana, United States.
|
{
"evidence": [
"Michelle Williams (actress)",
"Dawson's Creek"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMichelle Williams was born in Kalispell, Montana, United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Michelle Williams (actress)",
"Dawson's Creek"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 78,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_8788
|
Allure is an American women's magazine focused on beauty, published monthly by Condé Nast in the most populous city in the United States.
|
{
"evidence": [
"Allure (magazine)",
"New York City"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAllure is an American women's magazine focused on beauty, published monthly by Condé Nast in the most populous city in the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Allure (magazine)",
"New York City"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 79,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
42465
|
The Burlingame Treaty amended the Treaty of Tientsin which ended the first phase of the Second Opium War.
|
{
"evidence": [
"Burlingame Treaty_sentence_0",
"Second Opium War_sentence_76"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Burlingame Treaty amended the Treaty of Tientsin which ended the first phase of the Second Opium War.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Burlingame Treaty_sentence_0",
"Second Opium War_sentence_76"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 80,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.206",
"value": "?search="
},
{
"operation": "search",
"time": "32.196",
"value": "The Burlingame Treaty"
},
{
"operation": "Now on",
"time": "33.717",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "37.558",
"value": "Burlingame Treaty"
},
{
"operation": "Now on",
"time": "38.448",
"value": "Burlingame Treaty"
},
{
"operation": "Highlighting",
"time": "45.074",
"value": "Burlingame Treaty_sentence_0"
},
{
"operation": "Page search",
"time": "73.993",
"value": "Second Opium War"
},
{
"operation": "search",
"time": "90.537",
"value": "Second Opium War"
},
{
"operation": "Now on",
"time": "91.962",
"value": "Second Opium War"
},
{
"operation": "Page search",
"time": "112.276",
"value": "Burlingame Treaty"
},
{
"operation": "Highlighting",
"time": "150.327",
"value": "Second Opium War_sentence_76"
},
{
"operation": "finish",
"time": "203.306",
"value": "finish"
}
]
}
|
exfever_4734
|
Cinderella is a film suitable for all ages.
|
{
"evidence": [
"Cinderella (2015 Disney film)",
"Walt Disney Pictures",
"Film studio"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nCinderella is a film suitable for all ages.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Cinderella (2015 Disney film)",
"Walt Disney Pictures",
"Film studio"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 81,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
20615
|
In 1860, twenty-six years before the Haymarket Riot, Chicago's German population was 22,230.
|
{
"evidence": [
"Haymarket affair_sentence_4"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn 1860, twenty-six years before the Haymarket Riot, Chicago's German population was 22,230.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Haymarket affair_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 82,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.718",
"value": "?search="
},
{
"operation": "search",
"time": "24.829",
"value": "chicago population 1860"
},
{
"operation": "Now on",
"time": "26.434",
"value": "undefined"
},
{
"operation": "search",
"time": "53.898",
"value": "Haymarket Riot"
},
{
"operation": "Now on",
"time": "55.062",
"value": "Haymarket affair"
},
{
"operation": "Page search",
"time": "79.609",
"value": "german"
},
{
"operation": "search",
"time": "174.699",
"value": "1860 german population chicago"
},
{
"operation": "Now on",
"time": "175.604",
"value": "undefined"
},
{
"operation": "search",
"time": "244.601",
"value": "chicago population"
},
{
"operation": "Now on",
"time": "245.375",
"value": "undefined"
},
{
"operation": "search",
"time": "258.974",
"value": "german population chicago"
},
{
"operation": "Now on",
"time": "259.892",
"value": "undefined"
},
{
"operation": "search",
"time": "285.483",
"value": "1860 Chicago mayoral election"
},
{
"operation": "Now on",
"time": "286.342",
"value": "1860 Chicago mayoral election"
},
{
"operation": "search",
"time": "311.243",
"value": "haymarket riot"
},
{
"operation": "Now on",
"time": "312.418",
"value": "Haymarket affair"
},
{
"operation": "Highlighting",
"time": "327.165",
"value": "Haymarket affair_sentence_4"
},
{
"operation": "Page search",
"time": "331.779",
"value": "1860"
},
{
"operation": "finish",
"time": "359.31",
"value": "finish"
}
]
}
|
76565
|
Pirmin Strasser was promoted to SK Rapid Wien in 2011, after the departure of Kevin Fend.
|
{
"evidence": [
"Pirmin Strasser_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPirmin Strasser was promoted to SK Rapid Wien in 2011, after the departure of Kevin Fend.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Pirmin Strasser_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 83,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.198",
"value": "?search="
},
{
"operation": "search",
"time": "13.352",
"value": "Pirmin Strasser"
},
{
"operation": "Now on",
"time": "14.217",
"value": "Pirmin Strasser"
},
{
"operation": "Highlighting",
"time": "20.592",
"value": "Pirmin Strasser_sentence_5"
},
{
"operation": "finish",
"time": "37.759",
"value": "finish"
}
]
}
|
exfever_3868
|
The Walt Disney Company is an English multinational entertainment known for its major entertainment company or motion picture company that has its own privately owned studio facility or facilities that are used to make films.
|
{
"evidence": [
"The Walt Disney Company",
"Film studio"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Walt Disney Company is an English multinational entertainment known for its major entertainment company or motion picture company that has its own privately owned studio facility or facilities that are used to make films.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Walt Disney Company",
"Film studio"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 84,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
50887
|
Joanne Lunn's recordings of Bach cantatas with John Eliot Gardiner and the Monteverdi Choir include Herr, wie du willt, so schicks mit mir, and BWV 73 and she also recorded Bach cantatas with the Bach Collegium Japan (composed of an orchestra and a chorus specializing in Baroque music, playing on period instruments).
|
{
"evidence": [
"Joanne Lunn_sentence_6",
"Joanne Lunn_sentence_7",
"Bach Collegium Japan_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJoanne Lunn's recordings of Bach cantatas with John Eliot Gardiner and the Monteverdi Choir include Herr, wie du willt, so schicks mit mir, and BWV 73 and she also recorded Bach cantatas with the Bach Collegium Japan (composed of an orchestra and a chorus specializing in Baroque music, playing on period instruments).\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Joanne Lunn_sentence_6",
"Joanne Lunn_sentence_7",
"Bach Collegium Japan_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 85,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.178",
"value": "?search="
},
{
"operation": "search",
"time": "22.306",
"value": "Joanne Lunn"
},
{
"operation": "Now on",
"time": "23.012",
"value": "Joanne Lunn"
},
{
"operation": "Page search",
"time": "45.175",
"value": "John Eliot Gardiner"
},
{
"operation": "Highlighting",
"time": "53.354",
"value": "Joanne Lunn_sentence_6"
},
{
"operation": "Highlighting",
"time": "62.962",
"value": "Joanne Lunn_sentence_7"
},
{
"operation": "hyperlink",
"time": "71.858",
"value": "Bach Collegium Japan"
},
{
"operation": "Now on",
"time": "72.521",
"value": "Bach Collegium Japan"
},
{
"operation": "Highlighting",
"time": "75.163",
"value": "Bach Collegium Japan_sentence_0"
},
{
"operation": "finish",
"time": "90.875",
"value": "finish"
}
]
}
|
exfever_8004
|
One World Trade Center is the main building of the rebuilt World Trade Center complex in a state located in the most populous city in the United States.
|
{
"evidence": [
"One World Trade Center",
"New York City"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nOne World Trade Center is the main building of the rebuilt World Trade Center complex in a state located in the most populous city in the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"One World Trade Center",
"New York City"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 86,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
42478
|
TAFII25 binds to TAFII259 and inhibits its acetyltransferase activity and the exact role of TAFII25 is currently uknown but studies have shown that it interacts with the C-jun pathway.
|
{
"evidence": [
"Histone acetyltransferase_sentence_12"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nTAFII25 binds to TAFII259 and inhibits its acetyltransferase activity and the exact role of TAFII25 is currently uknown but studies have shown that it interacts with the C-jun pathway.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Histone acetyltransferase_sentence_12"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 87,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.362",
"value": "?search="
},
{
"operation": "search",
"time": "16.631",
"value": "TAFII25"
},
{
"operation": "Now on",
"time": "17.718",
"value": "undefined"
},
{
"operation": "search",
"time": "30.237",
"value": "TAFII25"
},
{
"operation": "Now on",
"time": "30.999",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "37.766",
"value": "Histone acetyltransferase"
},
{
"operation": "Now on",
"time": "39.357",
"value": "Histone acetyltransferase"
},
{
"operation": "Page search",
"time": "50.691",
"value": "TAFII25"
},
{
"operation": "Page search",
"time": "51.02",
"value": "TAFII25"
},
{
"operation": "Page search",
"time": "51.244",
"value": "TAFII25"
},
{
"operation": "Page search",
"time": "51.412",
"value": "TAFII25"
},
{
"operation": "Highlighting",
"time": "60.2",
"value": "Histone acetyltransferase_sentence_12"
},
{
"operation": "Highlighting deleted",
"time": "68.116",
"value": "Histone acetyltransferase_sentence_12"
},
{
"operation": "search",
"time": "73.788",
"value": "TAFII259"
},
{
"operation": "Now on",
"time": "74.523",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "79.298",
"value": "O-GlcNAc"
},
{
"operation": "Now on",
"time": "80.84",
"value": "O-GlcNAc"
},
{
"operation": "Page search",
"time": "87.717",
"value": "TAFII259"
},
{
"operation": "Page search",
"time": "87.881",
"value": "TAFII259"
},
{
"operation": "back-button-clicked",
"time": "118.54",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "119.42",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "121.739",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "121.773",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "122.393",
"value": "Histone acetyltransferase"
},
{
"operation": "Page search",
"time": "129.681",
"value": "TAFII25"
},
{
"operation": "Page search",
"time": "130.278",
"value": "TAFII25"
},
{
"operation": "Page search",
"time": "130.47",
"value": "TAFII25"
},
{
"operation": "Page search",
"time": "130.653",
"value": "TAFII25"
},
{
"operation": "Highlighting",
"time": "166.449",
"value": "Histone acetyltransferase_sentence_12"
},
{
"operation": "finish",
"time": "172.035",
"value": "finish"
}
]
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.