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 |
|---|---|---|---|---|---|---|---|---|
74886
|
Le Triptyque des Monts et Châteaux is a Belgian Road race first held in 1997.
|
{
"evidence": [
"Le Triptyque des Monts et Châteaux_sentence_0",
"Le Triptyque des Monts et Châteaux_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:\nLe Triptyque des Monts et Châteaux is a Belgian Road race first held in 1997.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Le Triptyque des Monts et Châteaux_sentence_0",
"Le Triptyque des Monts et Châteaux_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4584,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.658",
"value": "?search="
},
{
"operation": "search",
"time": "13.733",
"value": "Le Triptyque des Monts et Châteaux"
},
{
"operation": "Now on",
"time": "15.121",
"value": "Le Triptyque des Monts et Châteaux"
},
{
"operation": "Highlighting",
"time": "33.404",
"value": "Le Triptyque des Monts et Châteaux_sentence_0"
},
{
"operation": "Highlighting",
"time": "45.138",
"value": "Le Triptyque des Monts et Châteaux_sentence_1"
},
{
"operation": "finish",
"time": "62.682",
"value": "finish"
}
]
}
|
exfever_43
|
The Girl on the Train is Emily Blunt's first psychological thriller film.
|
{
"evidence": [
"The Girl on the Train (2016 film)",
"Rebecca Ferguson"
],
"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 Girl on the Train is Emily Blunt's first psychological thriller film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Girl on the Train (2016 film)",
"Rebecca Ferguson"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4585,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
8647
|
Walter Restrepo played three seasons for the Chicago Cubs (an American professional soccer team that was founded in 2006), and scored ten goals his entire career.
|
{
"evidence": [
"Chicago Cubs_sentence_35"
],
"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:\nWalter Restrepo played three seasons for the Chicago Cubs (an American professional soccer team that was founded in 2006), and scored ten goals his entire career.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Chicago Cubs_sentence_35"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4586,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.934",
"value": "?search="
},
{
"operation": "search",
"time": "14.777",
"value": "Wálter Restrepo"
},
{
"operation": "Now on",
"time": "15.632",
"value": "Wálter Restrepo"
},
{
"operation": "search",
"time": "134.164",
"value": "Chicago Cubs"
},
{
"operation": "Now on",
"time": "135.708",
"value": "Chicago Cubs"
},
{
"operation": "Highlighting",
"time": "233.37",
"value": "Chicago Cubs_sentence_35"
},
{
"operation": "finish",
"time": "249.636",
"value": "finish"
}
]
}
|
26520
|
Bernie Sanders, former Mayor of Burlington, Vermont lost to Dolores Sandoval as the Democratic bet for the 1990 US House of Representatives election in Vermont.
|
{
"evidence": [
"Bernie Sanders_sentence_19"
],
"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:\nBernie Sanders, former Mayor of Burlington, Vermont lost to Dolores Sandoval as the Democratic bet for the 1990 US House of Representatives election in Vermont.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bernie Sanders_sentence_19"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4587,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.174",
"value": "?search="
},
{
"operation": "search",
"time": "13.755",
"value": "Bernie Sanders"
},
{
"operation": "Now on",
"time": "15.704",
"value": "Bernie Sanders"
},
{
"operation": "Highlighting",
"time": "42.527",
"value": "Bernie Sanders_sentence_19"
},
{
"operation": "finish",
"time": "181.441",
"value": "finish"
}
]
}
|
92295
|
30th New York State Legislature ( the Southern District is the stronghold of the Democratic Party for 17 years while the Eastern District was the stronghold of the Republican Party for 25 years.
|
{
"evidence": [
"30th New York State Legislature_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:\n30th New York State Legislature ( the Southern District is the stronghold of the Democratic Party for 17 years while the Eastern District was the stronghold of the Republican Party for 25 years.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"30th New York State Legislature_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4588,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.097",
"value": "?search="
},
{
"operation": "search",
"time": "11.585",
"value": "30th New York State Legislature"
},
{
"operation": "Now on",
"time": "14.058",
"value": "30th New York State Legislature"
},
{
"operation": "Highlighting",
"time": "72.626",
"value": "30th New York State Legislature_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "74.619",
"value": "30th New York State Legislature_sentence_4"
},
{
"operation": "Page search",
"time": "85.311",
"value": "Southern District"
},
{
"operation": "search",
"time": "196.51",
"value": "southern District is the stronghold of the Democratic Party for 17 years"
},
{
"operation": "Now on",
"time": "202.44",
"value": "undefined"
},
{
"operation": "search",
"time": "245.702",
"value": "Southern District is the stronghold of the Democratic"
},
{
"operation": "Now on",
"time": "247.11",
"value": "undefined"
},
{
"operation": "search",
"time": "477.685",
"value": "30th New York State Legislature"
},
{
"operation": "Now on",
"time": "478.41",
"value": "30th New York State Legislature"
},
{
"operation": "search",
"time": "592.115",
"value": "Southern District of New York"
},
{
"operation": "Now on",
"time": "597.062",
"value": "United States District Court for the Southern District of New York"
},
{
"operation": "hyperlink",
"time": "610.918",
"value": "2 History United States District Court for the Southern District of New York_section_1"
},
{
"operation": "Page search",
"time": "665.586",
"value": "stronghold"
},
{
"operation": "hyperlink",
"time": "685.958",
"value": "Northern"
},
{
"operation": "Now on",
"time": "687.976",
"value": "United States District Court for the Northern District of New York"
},
{
"operation": "search",
"time": "708.978",
"value": "southern district"
},
{
"operation": "Now on",
"time": "709.805",
"value": "Southern District"
},
{
"operation": "search",
"time": "753.914",
"value": "30th New York State Legislature southern district"
},
{
"operation": "Now on",
"time": "757.048",
"value": "undefined"
},
{
"operation": "search",
"time": "767.472",
"value": "30th New York State Legislature"
},
{
"operation": "Now on",
"time": "768.282",
"value": "30th New York State Legislature"
},
{
"operation": "hyperlink",
"time": "798.453",
"value": "4.1 Districts 30th New York State Legislature_section_4"
},
{
"operation": "Highlighting",
"time": "800.613",
"value": "30th New York State Legislature_item_0_0"
},
{
"operation": "Highlighting",
"time": "800.615",
"value": "30th New York State Legislature_section_3"
},
{
"operation": "Highlighting",
"time": "800.615",
"value": "30th New York State Legislature_section_4"
},
{
"operation": "Highlighting",
"time": "860.486",
"value": "30th New York State Legislature_sentence_2"
},
{
"operation": "hyperlink",
"time": "873.037",
"value": "Democratic-Republicans"
},
{
"operation": "Now on",
"time": "874.865",
"value": "Democratic-Republican Party"
},
{
"operation": "Page search",
"time": "931.308",
"value": "district"
},
{
"operation": "finish",
"time": "965.22",
"value": "finish"
}
]
}
|
exfever_7225
|
Soundgarden was an American rock band that helped to popularize grunge music, a style of alternative rock that developed in the geographic region in western North America bounded by its coastal waters of the Pacific Ocean to the west and, loosely, by the Rocky Mountains to the east.
|
{
"evidence": [
"Soundgarden",
"Pacific Northwest"
],
"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:\nSoundgarden was an American rock band that helped to popularize grunge music, a style of alternative rock that developed in the geographic region in western North America bounded by its coastal waters of the Pacific Ocean to the west and, loosely, by the Rocky Mountains to the east.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Soundgarden",
"Pacific Northwest"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4589,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1166
|
Aileen Wuornos is the first female serial killer to have a biographical film that is shown worldwide.
|
{
"evidence": [
"Patty Jenkins",
"Monster (2003 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:\nAileen Wuornos is the first female serial killer to have a biographical film that is shown worldwide.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Patty Jenkins",
"Monster (2003 film)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4590,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
49064
|
Steve Jobs is the co-founder and managing director of Belvedere Management Limited, a financial services group based in Mauritius that claims to have $16 billion of assets under administration, and also a director of Fulhold Pharma plc subsidiary Ful Hold Ltd.
|
{
"evidence": [
"Kenneth Maillard_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:\nSteve Jobs is the co-founder and managing director of Belvedere Management Limited, a financial services group based in Mauritius that claims to have $16 billion of assets under administration, and also a director of Fulhold Pharma plc subsidiary Ful Hold Ltd.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Kenneth Maillard_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4591,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.779",
"value": "?search="
},
{
"operation": "search",
"time": "13.464",
"value": "Steve Jobs"
},
{
"operation": "Now on",
"time": "18.987",
"value": "Steve Jobs"
},
{
"operation": "Page search",
"time": "28.417",
"value": "belvedere"
},
{
"operation": "search",
"time": "54.371",
"value": "Steve Jobs (disambiguation)"
},
{
"operation": "Now on",
"time": "55.12",
"value": "Steve Jobs (disambiguation)"
},
{
"operation": "back-button-clicked",
"time": "92.609",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "92.923",
"value": "Steve Jobs"
},
{
"operation": "Page search",
"time": "107.336",
"value": "belvedere"
},
{
"operation": "search",
"time": "148.714",
"value": "Belvedere Management Limited"
},
{
"operation": "Now on",
"time": "149.587",
"value": "Belvedere Management Limited"
},
{
"operation": "Page search",
"time": "160.946",
"value": "found"
},
{
"operation": "Page search",
"time": "186.398",
"value": "director"
},
{
"operation": "Highlighting",
"time": "201.922",
"value": "Belvedere Management_sentence_5"
},
{
"operation": "Highlighting",
"time": "201.923",
"value": "Belvedere Management_title"
},
{
"operation": "Highlighting",
"time": "201.923",
"value": "Belvedere Management_section_0"
},
{
"operation": "hyperlink",
"time": "203.454",
"value": "David Cosgrove"
},
{
"operation": "Now on",
"time": "203.951",
"value": "David Dawson Cosgrove"
},
{
"operation": "Page search",
"time": "214.701",
"value": "manag"
},
{
"operation": "Page search",
"time": "224.874",
"value": "director"
},
{
"operation": "search",
"time": "259.762",
"value": "managing director of Belvedere Management Limited,"
},
{
"operation": "Now on",
"time": "261.644",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "266.358",
"value": "Kenneth Maillard"
},
{
"operation": "Now on",
"time": "267.186",
"value": "Kenneth Maillard"
},
{
"operation": "Highlighting",
"time": "279.306",
"value": "Kenneth Maillard_sentence_2"
},
{
"operation": "finish",
"time": "293.266",
"value": "finish"
}
]
}
|
exfever_3175
|
Tom Hardy has since been excluded for an award presented annually by the Academy of Motion Picture Arts and Sciences.
|
{
"evidence": [
"Tom Hardy",
"Academy Award for Best Supporting 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:\nTom Hardy has since been excluded for an award presented annually by the Academy of Motion Picture Arts and Sciences.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tom Hardy",
"Academy Award for Best Supporting Actor"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4592,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
18021
|
The peroxisome-associated thiolase SCPx protein and the intracellular lipid transfer protein SCP2 are encoded by the SCP2 gene, said to play a role in Zellweger syndrome.
|
{
"evidence": [
"SCP2_sentence_3",
"SCP2_sentence_0",
"SCP2_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:\nThe peroxisome-associated thiolase SCPx protein and the intracellular lipid transfer protein SCP2 are encoded by the SCP2 gene, said to play a role in Zellweger syndrome.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"SCP2_sentence_3",
"SCP2_sentence_0",
"SCP2_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4593,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.113",
"value": "?search="
},
{
"operation": "search",
"time": "22.422",
"value": "Contains...peroxisome-associated thiolase SCPx protein"
},
{
"operation": "Now on",
"time": "23.211",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "26.632",
"value": "SCP2"
},
{
"operation": "Now on",
"time": "27.313",
"value": "SCP2"
},
{
"operation": "Page search",
"time": "55.326",
"value": "peroxisome"
},
{
"operation": "Highlighting",
"time": "59.986",
"value": "SCP2_sentence_3"
},
{
"operation": "Page search",
"time": "80.985",
"value": "lipid"
},
{
"operation": "Highlighting",
"time": "83.779",
"value": "SCP2_sentence_0"
},
{
"operation": "Highlighting",
"time": "100.058",
"value": "SCP2_sentence_1"
},
{
"operation": "Highlighting",
"time": "117.226",
"value": "SCP2_sentence_6"
},
{
"operation": "Page search",
"time": "151.095",
"value": "intracellular"
},
{
"operation": "search",
"time": "280.377",
"value": "Contains...intracellular lipid transfer protein SCP2"
},
{
"operation": "Now on",
"time": "281.737",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "292.441",
"value": "SCP2"
},
{
"operation": "Now on",
"time": "293.419",
"value": "SCP2"
},
{
"operation": "Page search",
"time": "341.883",
"value": "intracellular"
},
{
"operation": "finish",
"time": "356.506",
"value": "finish"
}
]
}
|
exfever_2821
|
Gil Sharone is an English drummer and member of an American experimental alternative rock/post-rock group.
|
{
"evidence": [
"Gil Sharone",
"Team Sleep"
],
"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:\nGil Sharone is an English drummer and member of an American experimental alternative rock/post-rock group.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Gil Sharone",
"Team Sleep"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4594,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
66697
|
"Fall in Line" is centered around lyrical themes of female empowerment, the song received positive reviews from music critics and a lyric video was uploaded on Instagram which was founded in 2005.
|
{
"evidence": [
"Fall in Line_sentence_6"
],
"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:\n\"Fall in Line\" is centered around lyrical themes of female empowerment, the song received positive reviews from music critics and a lyric video was uploaded on Instagram which was founded in 2005.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Fall in Line_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4595,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.691",
"value": "?search="
},
{
"operation": "search",
"time": "11.159",
"value": "Fall in Line"
},
{
"operation": "Now on",
"time": "12.335",
"value": "Fall in Line"
},
{
"operation": "Highlighting",
"time": "21.415",
"value": "Fall in Line_sentence_3"
},
{
"operation": "Highlighting",
"time": "24.988",
"value": "Fall in Line_sentence_4"
},
{
"operation": "Highlighting",
"time": "37.769",
"value": "Fall in Line_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "42.951",
"value": "Fall in Line_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "43.914",
"value": "Fall in Line_sentence_3"
},
{
"operation": "hyperlink",
"time": "50.263",
"value": "YouTube"
},
{
"operation": "Now on",
"time": "53.599",
"value": "YouTube"
},
{
"operation": "back-button-clicked",
"time": "54.059",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "54.201",
"value": "Fall in Line"
},
{
"operation": "finish",
"time": "63.251",
"value": "finish"
}
]
}
|
exfever_1536
|
William Harrison Clanton was a hero Cowboy in Cochise County in a location created from the western half of the New Mexico Territory during the conflict between the Union and the Confederacy.
|
{
"evidence": [
"Billy Clanton",
"Arizona Territory",
"American Civil War"
],
"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:\nWilliam Harrison Clanton was a hero Cowboy in Cochise County in a location created from the western half of the New Mexico Territory during the conflict between the Union and the Confederacy.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Billy Clanton",
"Arizona Territory",
"American Civil War"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4596,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4715
|
Daz Dllinger is a British rapper who worked on Dr. Dre's debut solo album that was released on December 15, 1992.
|
{
"evidence": [
"Daz Dillinger",
"The Chronic"
],
"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:\nDaz Dllinger is a British rapper who worked on Dr. Dre's debut solo album that was released on December 15, 1992.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Daz Dillinger",
"The Chronic"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4597,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
5429
|
Faugli Mawa has 65 children under the age of 6 years which is 15.12% of total population of the village, while most of the people are from Schedule Caste which constitutes 33.49% of total population in the village, but does not have any Schedule Tribe so far, in addition, per census 2011, 117 people were engaged in work activities out of the total population of the village which includes 109 males and 8 females.
|
{
"evidence": [
"Faugli Mawa_sentence_5",
"Faugli Mawa_sentence_6",
"Faugli Mawa_sentence_7",
"Faugli Mawa_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:\nFaugli Mawa has 65 children under the age of 6 years which is 15.12% of total population of the village, while most of the people are from Schedule Caste which constitutes 33.49% of total population in the village, but does not have any Schedule Tribe so far, in addition, per census 2011, 117 people were engaged in work activities out of the total population of the village which includes 109 males and 8 females.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Faugli Mawa_sentence_5",
"Faugli Mawa_sentence_6",
"Faugli Mawa_sentence_7",
"Faugli Mawa_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4598,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.014",
"value": "?search="
},
{
"operation": "search",
"time": "26.843",
"value": "Faugli Mawa"
},
{
"operation": "Now on",
"time": "27.788",
"value": "Faugli Mawa"
},
{
"operation": "Highlighting",
"time": "37.521",
"value": "Faugli Mawa_sentence_5"
},
{
"operation": "Highlighting",
"time": "42.024",
"value": "Faugli Mawa_sentence_6"
},
{
"operation": "Highlighting",
"time": "47.426",
"value": "Faugli Mawa_sentence_7"
},
{
"operation": "Highlighting",
"time": "51.193",
"value": "Faugli Mawa_sentence_8"
},
{
"operation": "finish",
"time": "88.981",
"value": "finish"
}
]
}
|
11711
|
Starting in 1907, Julia Evangeline Brooks was one of nine women who wrote a constitution, motto, chose colors and created a name for the first African-American sorority that was ncorporated in 1913.
|
{
"evidence": [
"Julia Evangeline Brooks_sentence_0",
"Julia Evangeline Brooks_sentence_19",
"Alpha Kappa Alpha_sentence_0",
"Alpha Kappa Alpha_sentence_16",
"Alpha Kappa Alpha_sentence_17",
"Alpha Kappa Alpha_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:\nStarting in 1907, Julia Evangeline Brooks was one of nine women who wrote a constitution, motto, chose colors and created a name for the first African-American sorority that was ncorporated in 1913.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Julia Evangeline Brooks_sentence_0",
"Julia Evangeline Brooks_sentence_19",
"Alpha Kappa Alpha_sentence_0",
"Alpha Kappa Alpha_sentence_16",
"Alpha Kappa Alpha_sentence_17",
"Alpha Kappa Alpha_sentence_12"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4599,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "3.315",
"value": "?search="
},
{
"operation": "search",
"time": "13.363",
"value": "Julia Evangeline Brooks"
},
{
"operation": "Now on",
"time": "17.906",
"value": "Julia Evangeline Brooks"
},
{
"operation": "Highlighting",
"time": "60.563",
"value": "Julia Evangeline Brooks_sentence_0"
},
{
"operation": "Highlighting",
"time": "91.178",
"value": "Julia Evangeline Brooks_sentence_19"
},
{
"operation": "Highlighting",
"time": "103.178",
"value": "Julia Evangeline Brooks_sentence_18"
},
{
"operation": "Highlighting deleted",
"time": "110.538",
"value": "Julia Evangeline Brooks_sentence_18"
},
{
"operation": "hyperlink",
"time": "118.659",
"value": "Alpha Kappa Alpha"
},
{
"operation": "Now on",
"time": "128.459",
"value": "Alpha Kappa Alpha"
},
{
"operation": "Highlighting",
"time": "130.491",
"value": "Alpha Kappa Alpha_sentence_0"
},
{
"operation": "Highlighting",
"time": "192.723",
"value": "Alpha Kappa Alpha_sentence_16"
},
{
"operation": "Highlighting",
"time": "196.402",
"value": "Alpha Kappa Alpha_sentence_17"
},
{
"operation": "Highlighting",
"time": "227.235",
"value": "Alpha Kappa Alpha_sentence_12"
},
{
"operation": "finish",
"time": "311.978",
"value": "finish"
}
]
}
|
exfever_6982
|
WGBH-TV, virtual channel 2 is the secondary Public Broadcasting Service (PBS) member television station licensed to the 24th-most populous city in the USA.
|
{
"evidence": [
"WGBH-TV",
"Boston"
],
"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:\nWGBH-TV, virtual channel 2 is the secondary Public Broadcasting Service (PBS) member television station licensed to the 24th-most populous city in the USA.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"WGBH-TV",
"Boston"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4600,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
54265
|
Renaissance Pleasure Faire of Southern California is an annual Renaissance faire, a commercial reenactment of a 1580s market faire at Port Deptford, that opened in the spring of 1963, in Irwindale, California.
|
{
"evidence": [
"Renaissance Pleasure Faire of Southern California_sentence_0",
"Renaissance Pleasure Faire of Southern California_sentence_1",
"Renaissance Pleasure Faire of Southern California_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:\nRenaissance Pleasure Faire of Southern California is an annual Renaissance faire, a commercial reenactment of a 1580s market faire at Port Deptford, that opened in the spring of 1963, in Irwindale, California.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Renaissance Pleasure Faire of Southern California_sentence_0",
"Renaissance Pleasure Faire of Southern California_sentence_1",
"Renaissance Pleasure Faire of Southern California_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4601,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.417",
"value": "?search="
},
{
"operation": "search",
"time": "15.629",
"value": "Renaissance Pleasure Faire of Southern California"
},
{
"operation": "Now on",
"time": "19.484",
"value": "Renaissance Pleasure Faire of Southern California"
},
{
"operation": "Highlighting",
"time": "21.545",
"value": "Renaissance Pleasure Faire of Southern California_sentence_0"
},
{
"operation": "Highlighting",
"time": "29.174",
"value": "Renaissance Pleasure Faire of Southern California_sentence_1"
},
{
"operation": "hyperlink",
"time": "34.886",
"value": "Renaissance faire"
},
{
"operation": "Now on",
"time": "39.941",
"value": "Renaissance fair"
},
{
"operation": "Highlighting",
"time": "47.124",
"value": "Renaissance fair_sentence_1"
},
{
"operation": "Highlighting",
"time": "47.127",
"value": "Renaissance fair_title"
},
{
"operation": "back-button-clicked",
"time": "58.174",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "58.318",
"value": "Renaissance Pleasure Faire of Southern California"
},
{
"operation": "Highlighting",
"time": "65.386",
"value": "Renaissance Pleasure Faire of Southern California_sentence_2"
},
{
"operation": "finish",
"time": "88.388",
"value": "finish"
}
]
}
|
2
|
Luis Salgado appeared in various musicals in 2007.
|
{
"evidence": [
"Luis Salgado_sentence_13"
],
"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:\nLuis Salgado appeared in various musicals in 2007.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Luis Salgado_sentence_13"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4602,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "55.666",
"value": "Luis Salgado"
},
{
"operation": "Highlighting",
"time": "71.567",
"value": "Luis Salgado_sentence_8"
},
{
"operation": "hyperlink",
"time": "94.804",
"value": "Fame"
},
{
"operation": "Highlighting",
"time": "96.998",
"value": "Fame (musical)_sentence_0"
},
{
"operation": "Highlighting",
"time": "97.013",
"value": "Fame (musical)_title"
},
{
"operation": "back-button-clicked",
"time": "99.946",
"value": "back-button-clicked"
},
{
"operation": "Highlighting",
"time": "965.343",
"value": "Luis Salgado_sentence_10"
},
{
"operation": "Highlighting",
"time": "989.781",
"value": "Luis Salgado_sentence_13"
},
{
"operation": "Highlighting deleted",
"time": "1002.844",
"value": "Luis Salgado_sentence_13"
},
{
"operation": "Highlighting",
"time": "1018.17",
"value": "Luis Salgado_sentence_13"
},
{
"operation": "Highlighting",
"time": "1045.323",
"value": "Luis Salgado_sentence_13"
},
{
"operation": "hyperlink",
"time": "1068.557",
"value": "Off-Broadway"
},
{
"operation": "back-button-clicked",
"time": "1077.035",
"value": "back-button-clicked"
},
{
"operation": "finish",
"time": "1447.64",
"value": "finish"
}
]
}
|
93502
|
In a think tank at Aberystwyth University the director worked at the institute as a research assistant and led the DDMI from 2011 to 2013 was also known as David Davies Memorial Institute
|
{
"evidence": [
"David Davies Memorial Institute_sentence_0",
"David Davies Memorial Institute_sentence_32",
"David Davies Memorial Institute_sentence_33"
],
"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 a think tank at Aberystwyth University the director worked at the institute as a research assistant and led the DDMI from 2011 to 2013 was also known as David Davies Memorial Institute\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"David Davies Memorial Institute_sentence_0",
"David Davies Memorial Institute_sentence_32",
"David Davies Memorial Institute_sentence_33"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4603,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.099",
"value": "?search="
},
{
"operation": "search",
"time": "11.082",
"value": "David Davies Memorial Institute"
},
{
"operation": "Now on",
"time": "12.461",
"value": "David Davies Memorial Institute"
},
{
"operation": "Highlighting",
"time": "20.102",
"value": "David Davies Memorial Institute_sentence_0"
},
{
"operation": "Page search",
"time": "48.29",
"value": "director"
},
{
"operation": "Highlighting",
"time": "55.845",
"value": "David Davies Memorial Institute_sentence_32"
},
{
"operation": "Highlighting",
"time": "70.105",
"value": "David Davies Memorial Institute_sentence_33"
},
{
"operation": "finish",
"time": "94.716",
"value": "finish"
}
]
}
|
exfever_2831
|
Robert F. Kennedy is the first U. S. Senator from New York.
|
{
"evidence": [
"Robert F. Kennedy",
"Assassination of Robert F. Kennedy",
"Pacific Time Zone"
],
"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:\nRobert F. Kennedy is the first U. S. Senator from New York.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Robert F. Kennedy",
"Assassination of Robert F. Kennedy",
"Pacific Time Zone"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4604,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
20933
|
Pueblo del Arroyo is an Ancestral Puebloan great house with three hundred rooms.
|
{
"evidence": [
"Pueblo del Arroyo_sentence_0",
"Pueblo del Arroyo_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:\nPueblo del Arroyo is an Ancestral Puebloan great house with three hundred rooms.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Pueblo del Arroyo_sentence_0",
"Pueblo del Arroyo_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4605,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.643",
"value": "?search="
},
{
"operation": "search",
"time": "6.891",
"value": "Pueblo del Arroyo"
},
{
"operation": "Now on",
"time": "7.848",
"value": "Pueblo del Arroyo"
},
{
"operation": "Highlighting",
"time": "28.941",
"value": "Pueblo del Arroyo_sentence_0"
},
{
"operation": "Highlighting",
"time": "31.337",
"value": "Pueblo del Arroyo_sentence_2"
},
{
"operation": "finish",
"time": "48.34",
"value": "finish"
}
]
}
|
exfever_4836
|
Tanya Roberts was an American actress, producer, and model known as Midge Pinciotti on the American television period sitcom that aired on Fox from August 23, 1998, to May 18, 2006.
|
{
"evidence": [
"Tanya Roberts",
"That '70s Show"
],
"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:\nTanya Roberts was an American actress, producer, and model known as Midge Pinciotti on the American television period sitcom that aired on Fox from August 23, 1998, to May 18, 2006.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tanya Roberts",
"That '70s Show"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4606,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1467
|
Karl Anthony Malone spent his first 18 seasons in the NBA?with the Utah Jazz and formed a formidable duo with his teammate,?regarded as one of the greatest point guards and passers of all time.
|
{
"evidence": [
"Karl Malone",
"John Stockton"
],
"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:\nKarl Anthony Malone spent his first 18 seasons in the NBA?with the Utah Jazz and formed a formidable duo with his teammate,?regarded as one of the greatest point guards and passers of all time.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Karl Malone",
"John Stockton"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4607,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4583
|
Lung carcinoma is a malignant lung tumor in which a group of diseases involving abnormal cell growth with the potential to invade or spread to other parts of the body.
|
{
"evidence": [
"Lung cancer",
"Cancer"
],
"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:\nLung carcinoma is a malignant lung tumor in which a group of diseases involving abnormal cell growth with the potential to invade or spread to other parts of the body.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lung cancer",
"Cancer"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4608,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
44401
|
The Langham Place (All Souls) serves a population of 8,028; while St John's Wood (St John) serves a population of 8,533.
|
{
"evidence": [
"All Souls Church, Langham Place_sentence_3",
"St John's Wood Church_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:\nThe Langham Place (All Souls) serves a population of 8,028; while St John's Wood (St John) serves a population of 8,533.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"All Souls Church, Langham Place_sentence_3",
"St John's Wood Church_sentence_11"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4609,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.772",
"value": "?search="
},
{
"operation": "search",
"time": "9.003",
"value": "Contains...The Langham Place"
},
{
"operation": "Now on",
"time": "10.582",
"value": "undefined"
},
{
"operation": "search",
"time": "27.005",
"value": "All Souls"
},
{
"operation": "Now on",
"time": "27.979",
"value": "All Souls"
},
{
"operation": "search",
"time": "37.545",
"value": "St John's Wood"
},
{
"operation": "Now on",
"time": "40.822",
"value": "St John's Wood"
},
{
"operation": "hyperlink",
"time": "57.46",
"value": "St John's Wood, Queensland"
},
{
"operation": "Now on",
"time": "59.549",
"value": "St Johns Wood, Queensland"
},
{
"operation": "back-button-clicked",
"time": "65.71",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "65.849",
"value": "St John's Wood"
},
{
"operation": "Page search",
"time": "70.593",
"value": "population"
},
{
"operation": "search",
"time": "85.621",
"value": "Langham Place"
},
{
"operation": "Now on",
"time": "86.316",
"value": "Langham Place"
},
{
"operation": "hyperlink",
"time": "91.706",
"value": "Langham Place, London"
},
{
"operation": "Now on",
"time": "92.679",
"value": "Langham Place, London"
},
{
"operation": "hyperlink",
"time": "98.404",
"value": "1.3 All Souls Church Langham Place, London_section_3"
},
{
"operation": "hyperlink",
"time": "101.259",
"value": "All Souls Church, Langham Place"
},
{
"operation": "Now on",
"time": "102.149",
"value": "All Souls Church, Langham Place"
},
{
"operation": "Page search",
"time": "122.801",
"value": "serves"
},
{
"operation": "Page search",
"time": "136.254",
"value": "population"
},
{
"operation": "hyperlink",
"time": "143.114",
"value": "Langham Place"
},
{
"operation": "Now on",
"time": "143.621",
"value": "Langham Place, London"
},
{
"operation": "hyperlink",
"time": "150.921",
"value": "All Souls Church"
},
{
"operation": "Now on",
"time": "151.451",
"value": "All Souls Church, Langham Place"
},
{
"operation": "Highlighting",
"time": "155.21",
"value": "All Souls Church, Langham Place_sentence_3"
},
{
"operation": "search",
"time": "195.798",
"value": "St. John's Wood Church Grounds"
},
{
"operation": "Now on",
"time": "196.82",
"value": "St. John's Wood Church Grounds"
},
{
"operation": "back-button-clicked",
"time": "201.559",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "201.716",
"value": "All Souls Church, Langham Place"
},
{
"operation": "search",
"time": "209.718",
"value": "St John's Wood Church"
},
{
"operation": "Now on",
"time": "210.936",
"value": "St John's Wood Church"
},
{
"operation": "Highlighting",
"time": "226.693",
"value": "St John's Wood Church_sentence_11"
},
{
"operation": "hyperlink",
"time": "233.666",
"value": "List of churches in London"
},
{
"operation": "Now on",
"time": "243.724",
"value": "List of churches in London"
},
{
"operation": "Page search",
"time": "273.713",
"value": "langham place"
},
{
"operation": "Page search",
"time": "278.117",
"value": "all souls"
},
{
"operation": "Highlighting",
"time": "318.124",
"value": "List of churches in London_cell_3_19_0"
},
{
"operation": "Highlighting",
"time": "318.127",
"value": "List of churches in London_title"
},
{
"operation": "Highlighting",
"time": "318.129",
"value": "List of churches in London_section_3"
},
{
"operation": "Highlighting",
"time": "318.13",
"value": "List of churches in London_section_9"
},
{
"operation": "Highlighting",
"time": "318.139",
"value": "List of churches in London_header_cell_3_0_0"
},
{
"operation": "Highlighting",
"time": "319.807",
"value": "List of churches in London_cell_3_19_2"
},
{
"operation": "Highlighting",
"time": "319.818",
"value": "List of churches in London_header_cell_3_0_2"
},
{
"operation": "hyperlink",
"time": "323.281",
"value": "All Souls"
},
{
"operation": "Now on",
"time": "324.662",
"value": "All Souls' Day"
},
{
"operation": "back-button-clicked",
"time": "327.552",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "329.951",
"value": "List of churches in London"
},
{
"operation": "back-button-clicked",
"time": "353.02",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "353.027",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "353.232",
"value": "St John's Wood Church"
},
{
"operation": "search",
"time": "364.905",
"value": "St John's Wood"
},
{
"operation": "Now on",
"time": "366.187",
"value": "St John's Wood"
},
{
"operation": "Page search",
"time": "379.237",
"value": "8,533"
},
{
"operation": "search",
"time": "412.418",
"value": "St John's Wood Church"
},
{
"operation": "Now on",
"time": "413.392",
"value": "St John's Wood Church"
},
{
"operation": "Highlighting",
"time": "421.398",
"value": "St John's Wood Church_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "423.102",
"value": "St John's Wood Church_sentence_0"
},
{
"operation": "finish",
"time": "426.458",
"value": "finish"
}
]
}
|
58018
|
Mitchell Schet made his first team debut on August 23 2008 in the Dutch Super Cup against PSB Eindhoven and his league debut on September 13 that same year against Volendam.
|
{
"evidence": [
"Mitchell Schet_sentence_3"
],
"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:\nMitchell Schet made his first team debut on August 23 2008 in the Dutch Super Cup against PSB Eindhoven and his league debut on September 13 that same year against Volendam.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Mitchell Schet_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4610,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.852",
"value": "?search="
},
{
"operation": "search",
"time": "12.483",
"value": "Mitchell Schet"
},
{
"operation": "Now on",
"time": "13.55",
"value": "Mitchell Schet"
},
{
"operation": "Highlighting",
"time": "18.442",
"value": "Mitchell Schet_sentence_3"
},
{
"operation": "hyperlink",
"time": "25.333",
"value": "PSV Eindhoven"
},
{
"operation": "Now on",
"time": "27.468",
"value": "PSV Eindhoven"
},
{
"operation": "back-button-clicked",
"time": "31.5",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "31.573",
"value": "Mitchell Schet"
},
{
"operation": "finish",
"time": "47.291",
"value": "finish"
}
]
}
|
exfever_4040
|
Jean Grey, also known as Marvel Girl, is a member of the mutants who first appeared in the American comic book series published by DC Comics.
|
{
"evidence": [
"Jean Grey",
"Uncanny X-Men"
],
"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:\nJean Grey, also known as Marvel Girl, is a member of the mutants who first appeared in the American comic book series published by DC Comics.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jean Grey",
"Uncanny X-Men"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4611,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
14321
|
In 1834, Henry Dundas Campbell became Lieutenant-Governor of Sierra Lione, and he later become Governor; one of his major tenets was opposing female education.
|
{
"evidence": [
"Henry Dundas Campbell_sentence_8"
],
"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:\nIn 1834, Henry Dundas Campbell became Lieutenant-Governor of Sierra Lione, and he later become Governor; one of his major tenets was opposing female education.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Henry Dundas Campbell_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4612,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.579",
"value": "?search="
},
{
"operation": "search",
"time": "8.376",
"value": "Henry Dundas Campbell"
},
{
"operation": "Now on",
"time": "9.347",
"value": "Henry Dundas Campbell"
},
{
"operation": "Highlighting",
"time": "17.29",
"value": "Henry Dundas Campbell_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "21.515",
"value": "Henry Dundas Campbell_sentence_3"
},
{
"operation": "Highlighting",
"time": "22.386",
"value": "Henry Dundas Campbell_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "22.699",
"value": "Henry Dundas Campbell_sentence_7"
},
{
"operation": "Highlighting",
"time": "24.2",
"value": "Henry Dundas Campbell_sentence_6"
},
{
"operation": "Highlighting",
"time": "27.683",
"value": "Henry Dundas Campbell_sentence_7"
},
{
"operation": "Highlighting",
"time": "32.407",
"value": "Henry Dundas Campbell_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "33.027",
"value": "Henry Dundas Campbell_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "33.539",
"value": "Henry Dundas Campbell_sentence_6"
},
{
"operation": "finish",
"time": "38.197",
"value": "finish"
}
]
}
|
exfever_5875
|
Chicago is the most populous city in a state in the Midwestern region of a country primarily located in South America.
|
{
"evidence": [
"Chicago",
"Illinois",
"United States"
],
"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:\nChicago is the most populous city in a state in the Midwestern region of a country primarily located in South America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Chicago",
"Illinois",
"United States"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4613,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
55098
|
After overcoming his alcoholic addiction, Eric Clapton realized that he had wasted five years of his life, and later on he liked the recordings.
|
{
"evidence": [
"Eric Clapton_sentence_18",
"Eric Clapton_sentence_177",
"Eric Clapton_sentence_170",
"Eric Clapton_sentence_171"
],
"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:\nAfter overcoming his alcoholic addiction, Eric Clapton realized that he had wasted five years of his life, and later on he liked the recordings.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Eric Clapton_sentence_18",
"Eric Clapton_sentence_177",
"Eric Clapton_sentence_170",
"Eric Clapton_sentence_171"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4614,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.252",
"value": "undefined"
},
{
"operation": "search",
"time": "22.403",
"value": "Eric Clapton"
},
{
"operation": "Now on",
"time": "23.264",
"value": "Eric Clapton"
},
{
"operation": "Page search",
"time": "74.73",
"value": "wasted 5 years of his life"
},
{
"operation": "Page search",
"time": "106.137",
"value": "wasted"
},
{
"operation": "Page search",
"time": "137.736",
"value": "recordings"
},
{
"operation": "search",
"time": "359.678",
"value": "later on he liked the recordings"
},
{
"operation": "Now on",
"time": "362.464",
"value": "undefined"
},
{
"operation": "search",
"time": "383.838",
"value": "realized he had wasted 5 years of his life"
},
{
"operation": "Now on",
"time": "387.062",
"value": "undefined"
},
{
"operation": "search",
"time": "412.012",
"value": "eric clapton"
},
{
"operation": "Now on",
"time": "413.504",
"value": "Eric Clapton"
},
{
"operation": "search",
"time": "425.868",
"value": "overcoming"
},
{
"operation": "Now on",
"time": "426.485",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "428.932",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "429.102",
"value": "Eric Clapton"
},
{
"operation": "Page search",
"time": "432.422",
"value": "overcoming"
},
{
"operation": "Highlighting",
"time": "446.324",
"value": "Eric Clapton_sentence_18"
},
{
"operation": "hyperlink",
"time": "462.291",
"value": "Eric Clapton albums discography"
},
{
"operation": "Now on",
"time": "464.711",
"value": "Eric Clapton albums discography"
},
{
"operation": "back-button-clicked",
"time": "487.362",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "487.529",
"value": "Eric Clapton"
},
{
"operation": "hyperlink",
"time": "488.809",
"value": "Eric Clapton singles discography"
},
{
"operation": "Now on",
"time": "492.827",
"value": "Eric Clapton singles discography"
},
{
"operation": "back-button-clicked",
"time": "500.993",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "501.173",
"value": "Eric Clapton"
},
{
"operation": "Page search",
"time": "515.371",
"value": "alchohol"
},
{
"operation": "Page search",
"time": "520.106",
"value": "alcohol"
},
{
"operation": "Highlighting",
"time": "539.816",
"value": "Eric Clapton_sentence_177"
},
{
"operation": "Highlighting",
"time": "564.231",
"value": "Eric Clapton_sentence_170"
},
{
"operation": "Highlighting",
"time": "567.119",
"value": "Eric Clapton_sentence_171"
},
{
"operation": "finish",
"time": "614.853",
"value": "finish"
}
]
}
|
48629
|
Highly Evolved is an album by The Vines that included the debut single "Highly Evolved", and placed 4th in the best album of the year race by NME (a British music, film and culture website and brand which has been around since 1952), as well as being included in the book 1001 Albums You Must Hear Before You Die.
|
{
"evidence": [
"Highly Evolved_sentence_0",
"Highly Evolved_sentence_5",
"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:\nHighly Evolved is an album by The Vines that included the debut single \"Highly Evolved\", and placed 4th in the best album of the year race by NME (a British music, film and culture website and brand which has been around since 1952), as well as being included in the book 1001 Albums You Must Hear Before You Die.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Highly Evolved_sentence_0",
"Highly Evolved_sentence_5",
"NME_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4615,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.892",
"value": "?search="
},
{
"operation": "search",
"time": "10.186",
"value": "Highly Evolved"
},
{
"operation": "Now on",
"time": "11.007",
"value": "Highly Evolved"
},
{
"operation": "Highlighting",
"time": "15.156",
"value": "Highly Evolved_sentence_0"
},
{
"operation": "Highlighting",
"time": "25.117",
"value": "Highly Evolved_sentence_5"
},
{
"operation": "hyperlink",
"time": "89.063",
"value": "NME"
},
{
"operation": "Now on",
"time": "89.653",
"value": "NME"
},
{
"operation": "Highlighting",
"time": "101.159",
"value": "NME_sentence_2"
},
{
"operation": "Page search",
"time": "122.016",
"value": "year"
},
{
"operation": "hyperlink",
"time": "182.13",
"value": "NME Awards"
},
{
"operation": "Now on",
"time": "182.805",
"value": "NME Awards"
},
{
"operation": "Page search",
"time": "207.097",
"value": "vines"
},
{
"operation": "finish",
"time": "224.163",
"value": "finish"
}
]
}
|
exfever_2743
|
Assassin's Creed is an open-world action-adventure stealth video game whose creators include an American video game writer.
|
{
"evidence": [
"Assassin's Creed",
"Corey May"
],
"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:\nAssassin's Creed is an open-world action-adventure stealth video game whose creators include an American video game writer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Assassin's Creed",
"Corey May"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4616,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2511
|
Ronald Reagan also reshaped the Republican party.
|
{
"evidence": [
"Ronald Reagan",
"California"
],
"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:\nRonald Reagan also reshaped the Republican party.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ronald Reagan",
"California"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4617,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3081
|
Sir Alfred Joseph Hitchcock never won the award presented annually by the Academy of Motion Picture Arts and Sciences which was awarded to Lewis Milestone for an American comedy film produced by Howard Hughes.
|
{
"evidence": [
"Alfred Hitchcock",
"Academy Award for Best Director",
"Two Arabian Knights"
],
"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:\nSir Alfred Joseph Hitchcock never won the award presented annually by the Academy of Motion Picture Arts and Sciences which was awarded to Lewis Milestone for an American comedy film produced by Howard Hughes.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Alfred Hitchcock",
"Academy Award for Best Director",
"Two Arabian Knights"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4618,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5790
|
Pot distilled gin represents the earliest style of gin.
|
{
"evidence": [
"Gin",
"Juniperus communis",
"Circumpolar 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:\nPot distilled gin represents the earliest style of gin.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Gin",
"Juniperus communis",
"Circumpolar distribution"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4619,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1104
|
Grease is a 1978 American film where songs by the characters are interwoven into the narrative.
|
{
"evidence": [
"Grease (film)",
"Musical 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:\nGrease is a 1978 American film where songs by the characters are interwoven into the narrative.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Grease (film)",
"Musical film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4620,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8298
|
Elsa Lafuente Medianu, known professionally as Elsa Pataky, is a Spanish model and actress who appeared in the 2009 American neo-noir crime thriller movie directed by Thomas Jane.
|
{
"evidence": [
"Elsa Pataky",
"Give 'Em Hell, Malone"
],
"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:\nElsa Lafuente Medianu, known professionally as Elsa Pataky, is a Spanish model and actress who appeared in the 2009 American neo-noir crime thriller movie directed by Thomas Jane.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Elsa Pataky",
"Give 'Em Hell, Malone"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4621,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3297
|
Manatees are large, fully aquatic, mostly herbivorous aquatic animals that rely on the ocean for their existence.
|
{
"evidence": [
"Manatee",
"Marine mammal"
],
"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:\nManatees are large, fully aquatic, mostly herbivorous aquatic animals that rely on the ocean for their existence.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Manatee",
"Marine mammal"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4622,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4135
|
Mammals are a group of vertebrates constituting the class Mammalia, characterized by the presence of mammary glands in females that produce milk for feeding their young and a set of layers of the mammalian cerebral cortex involved in higher-order brain functions such as sensory perception, cognition, generation of motor commands, spatial reasoning, and language.
|
{
"evidence": [
"Mammal",
"Neocortex"
],
"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:\nMammals are a group of vertebrates constituting the class Mammalia, characterized by the presence of mammary glands in females that produce milk for feeding their young and a set of layers of the mammalian cerebral cortex involved in higher-order brain functions such as sensory perception, cognition, generation of motor commands, spatial reasoning, and language.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mammal",
"Neocortex"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4623,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2883
|
Home for the Holidays is a 1995 film directed by Jodie Foster whose narrative follows a young woman who departs Chicago to spend her national holiday celebrated in the United Kingdom.
|
{
"evidence": [
"Home for the Holidays (1995 film)",
"Thanksgiving"
],
"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:\nHome for the Holidays is a 1995 film directed by Jodie Foster whose narrative follows a young woman who departs Chicago to spend her national holiday celebrated in the United Kingdom.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Home for the Holidays (1995 film)",
"Thanksgiving"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4624,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
26571
|
77 Bombay Street is a folk rock group consisting of brothers Matt, Joe, Simri-Ramon and Esra Buchli from Switzerland that released four studio albums Dead Bird, Up in the Sky, Oko Town and Seven Mountains with Gadget Records, two of which platinum in Switzerland.
|
{
"evidence": [
"77 Bombay Street_sentence_0",
"77 Bombay Street_sentence_1",
"77 Bombay Street_sentence_3",
"77 Bombay Street_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:\n77 Bombay Street is a folk rock group consisting of brothers Matt, Joe, Simri-Ramon and Esra Buchli from Switzerland that released four studio albums Dead Bird, Up in the Sky, Oko Town and Seven Mountains with Gadget Records, two of which platinum in Switzerland.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"77 Bombay Street_sentence_0",
"77 Bombay Street_sentence_1",
"77 Bombay Street_sentence_3",
"77 Bombay Street_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4625,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.713",
"value": "undefined"
},
{
"operation": "search",
"time": "13.676",
"value": "77 Bombay Street"
},
{
"operation": "Now on",
"time": "15.043",
"value": "77 Bombay Street"
},
{
"operation": "Highlighting",
"time": "25.565",
"value": "77 Bombay Street_sentence_0"
},
{
"operation": "Highlighting",
"time": "31.01",
"value": "77 Bombay Street_sentence_1"
},
{
"operation": "Highlighting",
"time": "43.077",
"value": "77 Bombay Street_sentence_3"
},
{
"operation": "Highlighting",
"time": "57.201",
"value": "77 Bombay Street_sentence_4"
},
{
"operation": "finish",
"time": "126.05",
"value": "finish"
}
]
}
|
86251
|
Sara Headley is a nurse and an American former professional racing cyclist who was part of the Podium Ambition Pro Cycling (a UCI Women's cycling team based in the United Kingdom).
|
{
"evidence": [
"Sara Headley_sentence_1",
"Sara Headley_sentence_2",
"Sara Headley_sentence_4",
"Podium Ambition Pro Cycling_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:\nSara Headley is a nurse and an American former professional racing cyclist who was part of the Podium Ambition Pro Cycling (a UCI Women's cycling team based in the United Kingdom).\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Sara Headley_sentence_1",
"Sara Headley_sentence_2",
"Sara Headley_sentence_4",
"Podium Ambition Pro Cycling_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4626,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.589",
"value": "?search="
},
{
"operation": "search",
"time": "137.214",
"value": "Sara Headley"
},
{
"operation": "Now on",
"time": "138.344",
"value": "Sara Headley"
},
{
"operation": "Highlighting",
"time": "143.422",
"value": "Sara Headley_sentence_1"
},
{
"operation": "Highlighting",
"time": "154.841",
"value": "Sara Headley_sentence_2"
},
{
"operation": "Highlighting",
"time": "163.788",
"value": "Sara Headley_sentence_4"
},
{
"operation": "hyperlink",
"time": "183.838",
"value": "Podium Ambition Pro Cycling"
},
{
"operation": "Now on",
"time": "184.512",
"value": "Podium Ambition Pro Cycling"
},
{
"operation": "Highlighting",
"time": "187.383",
"value": "Podium Ambition Pro Cycling_sentence_0"
},
{
"operation": "finish",
"time": "193.784",
"value": "finish"
}
]
}
|
exfever_66
|
Jennifer Jason Leigh is an American actress who has received critical praise for her performances in a 1980 American black comedy crime film directed by George Armitage.
|
{
"evidence": [
"Jennifer Jason Leigh",
"Miami Blues"
],
"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:\nJennifer Jason Leigh is an American actress who has received critical praise for her performances in a 1980 American black comedy crime film directed by George Armitage.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jennifer Jason Leigh",
"Miami Blues"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4627,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
77392
|
The announcement of the proposal revealed that the preferred transport option was a maglev system, able to reach Shepparton from Brisbane in 30 minutes and Sydney in under two hours, however, financial details of the plan were not released, with the company's directors describing the complete funding model as commercial-in-confidence.
|
{
"evidence": [
"Consolidated Land and Rail Australia_sentence_15"
],
"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 announcement of the proposal revealed that the preferred transport option was a maglev system, able to reach Shepparton from Brisbane in 30 minutes and Sydney in under two hours, however, financial details of the plan were not released, with the company's directors describing the complete funding model as commercial-in-confidence.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Consolidated Land and Rail Australia_sentence_15"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4628,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.171",
"value": "?search="
},
{
"operation": "search",
"time": "98.775",
"value": "Consolidated Land and Rail Australia"
},
{
"operation": "Now on",
"time": "99.594",
"value": "Consolidated Land and Rail Australia"
},
{
"operation": "Page search",
"time": "108.699",
"value": "maglev"
},
{
"operation": "Highlighting",
"time": "111.262",
"value": "Consolidated Land and Rail Australia_sentence_15"
},
{
"operation": "finish",
"time": "129.461",
"value": "finish"
}
]
}
|
exfever_3292
|
Macbeth is a tragedy by William Shakespeare, first published in the Comedies, Histories, & Tragedies, a collection of plays.
|
{
"evidence": [
"Macbeth",
"First Folio"
],
"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:\nMacbeth is a tragedy by William Shakespeare, first published in the Comedies, Histories, & Tragedies, a collection of plays.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Macbeth",
"First Folio"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4629,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5926
|
X-Men is a 2000 American superhero film directed by Bryan Singer that features an ensemble cast, including the Australian actor who has headlined films in various genres, including the 2004 action gothic horror film written and directed by Bruce Davison.
|
{
"evidence": [
"X-Men (film)",
"Hugh Jackman",
"Van Helsing (film)"
],
"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:\nX-Men is a 2000 American superhero film directed by Bryan Singer that features an ensemble cast, including the Australian actor who has headlined films in various genres, including the 2004 action gothic horror film written and directed by Bruce Davison.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"X-Men (film)",
"Hugh Jackman",
"Van Helsing (film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4630,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_346
|
Mark Millar is an American director, screenwriter, and producer who earned a nomination for the Academy Award for Best Adapted Screenplay for his 2017 American superhero film inspired by an alternative version of the Marvel Comics character Wolverine.
|
{
"evidence": [
"James Mangold",
"Logan (film)",
"Old Man Logan"
],
"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:\nMark Millar is an American director, screenwriter, and producer who earned a nomination for the Academy Award for Best Adapted Screenplay for his 2017 American superhero film inspired by an alternative version of the Marvel Comics character Wolverine.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"James Mangold",
"Logan (film)",
"Old Man Logan"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4631,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8226
|
Taylor Schilling is known for her role as the antagonist of an American comedy-drama streaming television series created by Jenji Kohan.
|
{
"evidence": [
"Taylor Schilling",
"Piper Chapman",
"Orange Is the New Black"
],
"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:\nTaylor Schilling is known for her role as the antagonist of an American comedy-drama streaming television series created by Jenji Kohan.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Taylor Schilling",
"Piper Chapman",
"Orange Is the New Black"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4632,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
13503
|
Paul Gerrie was searching for suitable to open a winery, he visited France and Oregon before hiring a consultant to look for an appropriate location.
|
{
"evidence": [
"Cristom Vineyards_sentence_2",
"Cristom Vineyards_sentence_3",
"Cristom Vineyards_sentence_4",
"Cristom Vineyards_sentence_5"
],
"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:\nPaul Gerrie was searching for suitable to open a winery, he visited France and Oregon before hiring a consultant to look for an appropriate location.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Cristom Vineyards_sentence_2",
"Cristom Vineyards_sentence_3",
"Cristom Vineyards_sentence_4",
"Cristom Vineyards_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4633,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.125",
"value": "?search="
},
{
"operation": "search",
"time": "9.224",
"value": "Contains...Paul Gerrie"
},
{
"operation": "Now on",
"time": "10.7",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "16.64",
"value": "Cristom Vineyards"
},
{
"operation": "Now on",
"time": "17.998",
"value": "Cristom Vineyards"
},
{
"operation": "Highlighting",
"time": "20.96",
"value": "Cristom Vineyards_sentence_2"
},
{
"operation": "Highlighting",
"time": "37.874",
"value": "Cristom Vineyards_sentence_3"
},
{
"operation": "Highlighting",
"time": "39.27",
"value": "Cristom Vineyards_sentence_4"
},
{
"operation": "Highlighting",
"time": "44.764",
"value": "Cristom Vineyards_sentence_5"
},
{
"operation": "finish",
"time": "77.323",
"value": "finish"
}
]
}
|
37234
|
Hugh McCracken played for the band The Funatics and the experimental jazz-rock White Elephant Orchestra, otherwise known as Mike Mainieri & Friends.
|
{
"evidence": [
"Hugh McCracken_sentence_6",
"Hugh McCracken_sentence_8",
"White Elephant (band)_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:\nHugh McCracken played for the band The Funatics and the experimental jazz-rock White Elephant Orchestra, otherwise known as Mike Mainieri & Friends.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Hugh McCracken_sentence_6",
"Hugh McCracken_sentence_8",
"White Elephant (band)_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4634,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.336",
"value": "?search="
},
{
"operation": "search",
"time": "7.407",
"value": "Hugh McCracken"
},
{
"operation": "Now on",
"time": "8.729",
"value": "Hugh McCracken"
},
{
"operation": "Page search",
"time": "17.787",
"value": "funatics"
},
{
"operation": "Highlighting",
"time": "21.551",
"value": "Hugh McCracken_sentence_6"
},
{
"operation": "Highlighting",
"time": "32.755",
"value": "Hugh McCracken_sentence_8"
},
{
"operation": "hyperlink",
"time": "38.907",
"value": "White Elephant Orchestra"
},
{
"operation": "Now on",
"time": "39.536",
"value": "White Elephant (band)"
},
{
"operation": "Highlighting",
"time": "42.692",
"value": "White Elephant (band)_sentence_0"
},
{
"operation": "finish",
"time": "48.697",
"value": "finish"
}
]
}
|
36222
|
The Palace Hotel is a future project that will be the tallest building in Hobart (a city in Australia).
|
{
"evidence": [
"Hobart Building_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:\nThe Palace Hotel is a future project that will be the tallest building in Hobart (a city in Australia).\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Hobart Building_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4635,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.021",
"value": "?search="
},
{
"operation": "search",
"time": "18.51",
"value": "The Palace Hotel hobart"
},
{
"operation": "Now on",
"time": "19.798",
"value": "undefined"
},
{
"operation": "search",
"time": "30.953",
"value": "The Palace Hotel"
},
{
"operation": "Now on",
"time": "31.822",
"value": "Kimpton Clocktower Hotel"
},
{
"operation": "search",
"time": "98.172",
"value": "palace hotel hobart"
},
{
"operation": "Now on",
"time": "98.899",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "106.681",
"value": "Willis Polk"
},
{
"operation": "Now on",
"time": "107.372",
"value": "Willis Polk"
},
{
"operation": "Page search",
"time": "125.805",
"value": "tallest"
},
{
"operation": "hyperlink",
"time": "145.626",
"value": "Hobart Building"
},
{
"operation": "Now on",
"time": "146.229",
"value": "Hobart Building"
},
{
"operation": "Highlighting",
"time": "337.224",
"value": "Hobart Building_sentence_0"
},
{
"operation": "finish",
"time": "348.314",
"value": "finish"
}
]
}
|
exfever_8911
|
Rashida Jones grows up with two famous parents.
|
{
"evidence": [
"Andy Dwyer",
"Rashida Jones"
],
"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:\nRashida Jones grows up with two famous parents.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Andy Dwyer",
"Rashida Jones"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4636,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7876
|
Jackie Evancho gained wide notice with her 2nd-place finish in the fifth season of a televised American talent show competition created by an English television personality, entrepreneur, and record executive.
|
{
"evidence": [
"America's Got Talent",
"Simon Cowell"
],
"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:\nJackie Evancho gained wide notice with her 2nd-place finish in the fifth season of a televised American talent show competition created by an English television personality, entrepreneur, and record executive.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"America's Got Talent",
"Simon Cowell"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4637,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
51179
|
Dorr was cast in two Hollywood comedy shorts, both released by Pathé Exchange (RKO), one of the Big Five studios of Hollywood's Golden Age, in 1930: All Stuck Up and Ride 'em Cowboy; then for the remainder of the 1930s, he demonstrated his abilities at portraying an array of characters such as hotel clerks, police officers, reporters, and many more.
|
{
"evidence": [
"Stegath Dorr_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:\nDorr was cast in two Hollywood comedy shorts, both released by Pathé Exchange (RKO), one of the Big Five studios of Hollywood's Golden Age, in 1930: All Stuck Up and Ride 'em Cowboy; then for the remainder of the 1930s, he demonstrated his abilities at portraying an array of characters such as hotel clerks, police officers, reporters, and many more.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Stegath Dorr_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4638,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.715",
"value": "?search="
},
{
"operation": "search",
"time": "34.944",
"value": "Dorr"
},
{
"operation": "Now on",
"time": "35.658",
"value": "Dorr"
},
{
"operation": "hyperlink",
"time": "45.695",
"value": "Dorr (surname)"
},
{
"operation": "Now on",
"time": "46.098",
"value": "Dorr (surname)"
},
{
"operation": "Page search",
"time": "74.746",
"value": "actor"
},
{
"operation": "search",
"time": "97.55",
"value": "All Stuck Up"
},
{
"operation": "Now on",
"time": "100.555",
"value": "undefined"
},
{
"operation": "search",
"time": "121.217",
"value": "Ride 'Em Cowboy (1936 film)"
},
{
"operation": "Now on",
"time": "121.996",
"value": "Ride 'Em Cowboy (1936 film)"
},
{
"operation": "Page search",
"time": "136.017",
"value": "dorr"
},
{
"operation": "search",
"time": "154.4",
"value": "Ride 'Em Cowboy (1942 film)"
},
{
"operation": "Now on",
"time": "155.165",
"value": "Ride 'Em Cowboy (1942 film)"
},
{
"operation": "Page search",
"time": "158.912",
"value": "dorr"
},
{
"operation": "search",
"time": "206.878",
"value": "Dorr"
},
{
"operation": "Now on",
"time": "207.392",
"value": "Dorr"
},
{
"operation": "hyperlink",
"time": "209.67",
"value": "Dorr (surname)"
},
{
"operation": "Now on",
"time": "210.029",
"value": "Dorr (surname)"
},
{
"operation": "hyperlink",
"time": "212.936",
"value": "Stegath Dorr"
},
{
"operation": "Now on",
"time": "213.386",
"value": "Stegath Dorr"
},
{
"operation": "Highlighting",
"time": "219.893",
"value": "Stegath Dorr_sentence_0"
},
{
"operation": "finish",
"time": "233.503",
"value": "finish"
}
]
}
|
42860
|
John Ostell's first work in Montreal was the city's original Custom House; then he refused to submit designs for the new St. James Cathedral in Toronto.
|
{
"evidence": [
"John Ostell_sentence_7"
],
"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:\nJohn Ostell's first work in Montreal was the city's original Custom House; then he refused to submit designs for the new St. James Cathedral in Toronto.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"John Ostell_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4639,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.756",
"value": "?search="
},
{
"operation": "Now on",
"time": "8.687",
"value": "?search="
},
{
"operation": "search",
"time": "24.504",
"value": "John Ostell"
},
{
"operation": "Now on",
"time": "25.046",
"value": "John Ostell"
},
{
"operation": "Page search",
"time": "33.324",
"value": "montreal"
},
{
"operation": "Highlighting",
"time": "41.063",
"value": "John Ostell_sentence_5"
},
{
"operation": "Page search",
"time": "52.064",
"value": "refused"
},
{
"operation": "Page search",
"time": "62.485",
"value": "St. James"
},
{
"operation": "Highlighting",
"time": "68.674",
"value": "John Ostell_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "78.734",
"value": "John Ostell_sentence_5"
},
{
"operation": "finish",
"time": "94.068",
"value": "finish"
}
]
}
|
exfever_4632
|
Another One Bites the Dust is a movie by the British rock band Queen that reached number seven on listing the top-selling singles in the United Kingdom-based on physical sales, paid-for downloads, and streaming.
|
{
"evidence": [
"Another One Bites the Dust",
"UK Singles Chart"
],
"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:\nAnother One Bites the Dust is a movie by the British rock band Queen that reached number seven on listing the top-selling singles in the United Kingdom-based on physical sales, paid-for downloads, and streaming.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Another One Bites the Dust",
"UK Singles Chart"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4640,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5527
|
Mumtaz Mahal was the only wife of Shah Jahan.
|
{
"evidence": [
"Taj Mahal",
"Yamuna"
],
"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:\nMumtaz Mahal was the only wife of Shah Jahan.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Taj Mahal",
"Yamuna"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4641,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1158
|
A fictional character appearing in American comic books who is introduced as a Russian spy, created by an American comic book writer-artist.
|
{
"evidence": [
"Black Widow (Natasha Romanova)",
"Don Rico",
"Black Widow (Natasha Romanova)"
],
"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 fictional character appearing in American comic books who is introduced as a Russian spy, created by an American comic book writer-artist.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Black Widow (Natasha Romanova)",
"Don Rico",
"Black Widow (Natasha Romanova)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4642,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
27613
|
Tim Mudde, the older brother of the political scientist Cas Mudde, was one of the two founders of Brigade M.
|
{
"evidence": [
"Tim Mudde_sentence_4",
"Tim Mudde_sentence_5"
],
"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:\nTim Mudde, the older brother of the political scientist Cas Mudde, was one of the two founders of Brigade M.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Tim Mudde_sentence_4",
"Tim Mudde_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4643,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.603",
"value": "?search="
},
{
"operation": "search",
"time": "11.908",
"value": "Tim Mudde"
},
{
"operation": "Now on",
"time": "12.633",
"value": "Tim Mudde"
},
{
"operation": "Highlighting",
"time": "61.842",
"value": "Tim Mudde_sentence_4"
},
{
"operation": "Highlighting",
"time": "68.689",
"value": "Tim Mudde_sentence_5"
},
{
"operation": "finish",
"time": "90.201",
"value": "finish"
}
]
}
|
85344
|
Kristjánsson took charge of Icelandic premier league team Fram on May 30, 2004, replacing Ion Geolgău; then he was in charge again for the 2005 season, but left the club after relegation.
|
{
"evidence": [
"Ólafur Kristjánsson_sentence_6"
],
"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:\nKristjánsson took charge of Icelandic premier league team Fram on May 30, 2004, replacing Ion Geolgău; then he was in charge again for the 2005 season, but left the club after relegation.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Ólafur Kristjánsson_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4644,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.995",
"value": "?search="
},
{
"operation": "search",
"time": "16.589",
"value": "Kristjánsson"
},
{
"operation": "Now on",
"time": "17.269",
"value": "Kristjánsson"
},
{
"operation": "search",
"time": "43.573",
"value": "Fram"
},
{
"operation": "Now on",
"time": "44.238",
"value": "Fram"
},
{
"operation": "search",
"time": "59.908",
"value": "kristjansson fram"
},
{
"operation": "Now on",
"time": "61.13",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "65.981",
"value": "Ólafur Kristjánsson"
},
{
"operation": "Now on",
"time": "66.429",
"value": "Ólafur Kristjánsson"
},
{
"operation": "Highlighting",
"time": "74.493",
"value": "Ólafur Kristjánsson_sentence_6"
},
{
"operation": "finish",
"time": "233.523",
"value": "finish"
}
]
}
|
exfever_4772
|
Sharon Marie Tate Polanski was an American actress who appeared in the British mystery horror film directed by J. Lee Thompson.
|
{
"evidence": [
"Sharon Tate",
"Eye of the Devil"
],
"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:\nSharon Marie Tate Polanski was an American actress who appeared in the British mystery horror film directed by J. Lee Thompson.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sharon Tate",
"Eye of the Devil"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4645,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
71146
|
John King of Pollard Ball took accepted the position of chairman of the board at Dennis Specialist Vehicles, and was joined by Julian Amery, a former Minister of Aviation, as well as a new CEO and a new finance director.
|
{
"evidence": [
"Dennis Specialist Vehicles_sentence_94",
"Dennis Specialist Vehicles_sentence_95"
],
"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:\nJohn King of Pollard Ball took accepted the position of chairman of the board at Dennis Specialist Vehicles, and was joined by Julian Amery, a former Minister of Aviation, as well as a new CEO and a new finance director.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Dennis Specialist Vehicles_sentence_94",
"Dennis Specialist Vehicles_sentence_95"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4646,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.482",
"value": "?search="
},
{
"operation": "search",
"time": "30.082",
"value": "Dennis Specialist Vehicles"
},
{
"operation": "Now on",
"time": "32.35",
"value": "Dennis Specialist Vehicles"
},
{
"operation": "search",
"time": "82.045",
"value": "Contains...John King of Pollard Ball"
},
{
"operation": "Now on",
"time": "84.151",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "89.596",
"value": "Dennis Specialist Vehicles"
},
{
"operation": "Now on",
"time": "90.188",
"value": "Dennis Specialist Vehicles"
},
{
"operation": "back-button-clicked",
"time": "93.144",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "94.241",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "103.649",
"value": "1948 Nationalisation of road haulage Dennis Specialist Vehicles_section_10"
},
{
"operation": "Now on",
"time": "104.226",
"value": "Dennis Specialist Vehicles#1948 Nationalisation of road haulage Dennis Specialist Vehicles section 10"
},
{
"operation": "Highlighting",
"time": "155.543",
"value": "Dennis Specialist Vehicles_sentence_94"
},
{
"operation": "Highlighting",
"time": "158.619",
"value": "Dennis Specialist Vehicles_sentence_95"
},
{
"operation": "finish",
"time": "189.631",
"value": "finish"
}
]
}
|
exfever_5742
|
In 2016, the Government of India honoured the Indian actress, model, and singer, Priyanka Chopra, with the fourth-highest civilian award in the Republic of America.
|
{
"evidence": [
"Priyanka Chopra",
"Padma Shri"
],
"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:\nIn 2016, the Government of India honoured the Indian actress, model, and singer, Priyanka Chopra, with the fourth-highest civilian award in the Republic of America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Priyanka Chopra",
"Padma Shri"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4647,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3215
|
Toy Story is an American computer-animated comedy film produced by Pixar Animation Studios and released by an American film production studio and subsidiary of the corporate headquarters for The Walt Disney Company media conglomerate.
|
{
"evidence": [
"Toy Story",
"Walt Disney Pictures",
"Walt Disney Studios (Burbank)"
],
"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:\nToy Story is an American computer-animated comedy film produced by Pixar Animation Studios and released by an American film production studio and subsidiary of the corporate headquarters for The Walt Disney Company media conglomerate.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Toy Story",
"Walt Disney Pictures",
"Walt Disney Studios (Burbank)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4648,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4771
|
Starbucks is the first American coffeehouse headquartered in Seattle, Washington.
|
{
"evidence": [
"Starbucks",
"Espresso"
],
"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:\nStarbucks is the first American coffeehouse headquartered in Seattle, Washington.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Starbucks",
"Espresso"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4649,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1420
|
Henri Christophe was Haitan's most prominent revolutionary.
|
{
"evidence": [
"Citadelle Laferrière",
"Haitian Revolution"
],
"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:\nHenri Christophe was Haitan's most prominent revolutionary.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Citadelle Laferrière",
"Haitian Revolution"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4650,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4206
|
Death in Paradise is a British–French crime drama television series created by Robert Thorogood?and filmed on the French Caribbean islands of an archipelago, an overseas department, and a region of France in the Caribbean.
|
{
"evidence": [
"Death in Paradise (TV series)",
"Guadeloupe"
],
"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:\nDeath in Paradise is a British–French crime drama television series created by Robert Thorogood?and filmed on the French Caribbean islands of an archipelago, an overseas department, and a region of France in the Caribbean.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Death in Paradise (TV series)",
"Guadeloupe"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4651,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
47041
|
Schmitz (former head coach at Eastern Illinois) earned his first head coaching job at his alma mater Wilmington in Wilmington, Ohio in 1984; before he served three seasons and led the program to its first 20 win season and first NAIA Tournament appearance in 1986.
|
{
"evidence": [
"Jim Schmitz_sentence_0",
"Jim Schmitz_sentence_1",
"Jim Schmitz_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:\nSchmitz (former head coach at Eastern Illinois) earned his first head coaching job at his alma mater Wilmington in Wilmington, Ohio in 1984; before he served three seasons and led the program to its first 20 win season and first NAIA Tournament appearance in 1986.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Jim Schmitz_sentence_0",
"Jim Schmitz_sentence_1",
"Jim Schmitz_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4652,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.235",
"value": "?search="
},
{
"operation": "search",
"time": "16.882",
"value": "Schmitz"
},
{
"operation": "Now on",
"time": "18.25",
"value": "Schmitz"
},
{
"operation": "hyperlink",
"time": "56.726",
"value": "Danny Schmitz"
},
{
"operation": "Now on",
"time": "58.033",
"value": "Danny Schmitz"
},
{
"operation": "back-button-clicked",
"time": "93.782",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "93.821",
"value": "Schmitz"
},
{
"operation": "hyperlink",
"time": "106.818",
"value": "Jim Schmitz"
},
{
"operation": "Now on",
"time": "107.84",
"value": "Jim Schmitz"
},
{
"operation": "Highlighting",
"time": "115.864",
"value": "Jim Schmitz_sentence_0"
},
{
"operation": "Highlighting",
"time": "136.898",
"value": "Jim Schmitz_sentence_1"
},
{
"operation": "Highlighting",
"time": "156.767",
"value": "Jim Schmitz_sentence_2"
},
{
"operation": "finish",
"time": "219.284",
"value": "finish"
}
]
}
|
exfever_8915
|
Master of None is a comedy-drama streaming television series created by Alan Yang and an American actor, comedian, writer, director, and producer.
|
{
"evidence": [
"Master of None",
"Aziz Ansari"
],
"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:\nMaster of None is a comedy-drama streaming television series created by Alan Yang and an American actor, comedian, writer, director, and producer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Master of None",
"Aziz Ansari"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4653,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_831
|
Lauren Graham's film work includes roles in a 2001 American romantic drama film.
|
{
"evidence": [
"Lauren Graham",
"Sweet November (2001 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:\nLauren Graham's film work includes roles in a 2001 American romantic drama film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lauren Graham",
"Sweet November (2001 film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4654,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
19577
|
Janica Kostelic of Croatia was overall winner with 5 competition wins at the 2002–03 FIS Alpine Ski World Cup but had surgeries which kept her out of competition until 2005 when she won 3 gold medals.
|
{
"evidence": [
"Janica Kostelić_sentence_28",
"Janica Kostelić_sentence_30"
],
"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:\nJanica Kostelic of Croatia was overall winner with 5 competition wins at the 2002–03 FIS Alpine Ski World Cup but had surgeries which kept her out of competition until 2005 when she won 3 gold medals.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Janica Kostelić_sentence_28",
"Janica Kostelić_sentence_30"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4655,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.573",
"value": "?search="
},
{
"operation": "search",
"time": "29.319",
"value": "Janica Kostelić"
},
{
"operation": "Now on",
"time": "30.224",
"value": "Janica Kostelić"
},
{
"operation": "Highlighting",
"time": "34.291",
"value": "Janica Kostelić_sentence_0"
},
{
"operation": "Highlighting",
"time": "129.133",
"value": "Janica Kostelić_sentence_28"
},
{
"operation": "Highlighting",
"time": "138.931",
"value": "Janica Kostelić_sentence_30"
},
{
"operation": "Highlighting",
"time": "144.546",
"value": "Janica Kostelić_sentence_31"
},
{
"operation": "Highlighting deleted",
"time": "164.536",
"value": "Janica Kostelić_sentence_31"
},
{
"operation": "Highlighting deleted",
"time": "173.052",
"value": "Janica Kostelić_sentence_0"
},
{
"operation": "finish",
"time": "188.351",
"value": "finish"
}
]
}
|
89910
|
Eastern striped bandicoot is a plant.
|
{
"evidence": [
"Eastern striped bandicoot_sentence_0",
"Marsupial_sentence_2",
"Mammal_sentence_6"
],
"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:\nEastern striped bandicoot is a plant.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Eastern striped bandicoot_sentence_0",
"Marsupial_sentence_2",
"Mammal_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4656,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.252",
"value": "?search="
},
{
"operation": "search",
"time": "10.398",
"value": "Eastern striped bandicoot"
},
{
"operation": "Now on",
"time": "11.267",
"value": "Eastern striped bandicoot"
},
{
"operation": "Highlighting",
"time": "19.412",
"value": "Eastern striped bandicoot_sentence_0"
},
{
"operation": "hyperlink",
"time": "25.284",
"value": "marsupial"
},
{
"operation": "Now on",
"time": "26.338",
"value": "Marsupial"
},
{
"operation": "Highlighting",
"time": "51.662",
"value": "Marsupial_sentence_2"
},
{
"operation": "hyperlink",
"time": "53.276",
"value": "mammalian"
},
{
"operation": "Now on",
"time": "55.233",
"value": "Mammal"
},
{
"operation": "Highlighting",
"time": "56.884",
"value": "Mammal_sentence_6"
},
{
"operation": "finish",
"time": "69.317",
"value": "finish"
}
]
}
|
exfever_1975
|
The Summertime 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 an American lyricist who collaborated with his younger brother, to create some of the most memorable songs in the English language of the 20th century.
|
{
"evidence": [
"Summertime (George Gershwin song)",
"Ira Gershwin"
],
"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:\nThe Summertime 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 an American lyricist who collaborated with his younger brother, to create some of the most memorable songs in the English language of the 20th century.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Summertime (George Gershwin song)",
"Ira Gershwin"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4657,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
34040
|
KZY station featured professional singers as well as prominent speakers, but the main source of programming was the daily matinee concerts, over 1,500 of which were broadcast by the end of 2010.
|
{
"evidence": [
"KZY_sentence_19"
],
"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:\nKZY station featured professional singers as well as prominent speakers, but the main source of programming was the daily matinee concerts, over 1,500 of which were broadcast by the end of 2010.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"KZY_sentence_19"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4658,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.64",
"value": "?search="
},
{
"operation": "search",
"time": "7.555",
"value": "KZY station"
},
{
"operation": "Now on",
"time": "9.188",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "12.918",
"value": "KZY"
},
{
"operation": "Now on",
"time": "13.337",
"value": "KZY"
},
{
"operation": "Page search",
"time": "18.13",
"value": "singer"
},
{
"operation": "Highlighting",
"time": "20.42",
"value": "KZY_sentence_17"
},
{
"operation": "Highlighting",
"time": "47.291",
"value": "KZY_sentence_16"
},
{
"operation": "Highlighting",
"time": "53.592",
"value": "KZY_sentence_19"
},
{
"operation": "Highlighting deleted",
"time": "70.908",
"value": "KZY_sentence_17"
},
{
"operation": "Highlighting deleted",
"time": "71.392",
"value": "KZY_sentence_16"
},
{
"operation": "finish",
"time": "83.691",
"value": "finish"
}
]
}
|
91095
|
In 2016 the Nebraska–Kearney Lopers football team played in 111th season with 2nd year coach Josh Lamberson had a 1-10 record and their third game had the highest attendance.
|
{
"evidence": [
"2016 Nebraska–Kearney Lopers football team_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:\nIn 2016 the Nebraska–Kearney Lopers football team played in 111th season with 2nd year coach Josh Lamberson had a 1-10 record and their third game had the highest attendance.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"2016 Nebraska–Kearney Lopers football team_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4659,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.612",
"value": "?search="
},
{
"operation": "search",
"time": "17.283",
"value": "2016 Nebraska–Kearney Lopers football team"
},
{
"operation": "Now on",
"time": "18.266",
"value": "2016 Nebraska–Kearney Lopers football team"
},
{
"operation": "Highlighting",
"time": "29.158",
"value": "2016 Nebraska–Kearney Lopers football team_sentence_2"
},
{
"operation": "Highlighting",
"time": "36.837",
"value": "2016 Nebraska–Kearney Lopers football team_sentence_3"
},
{
"operation": "Highlighting",
"time": "77.124",
"value": "2016 Nebraska–Kearney Lopers football team_sentence_5"
},
{
"operation": "Highlighting",
"time": "79.324",
"value": "2016 Nebraska–Kearney Lopers football team_sentence_5"
},
{
"operation": "Now on",
"time": "103.285",
"value": "2016 Nebraska–Kearney Lopers football team"
},
{
"operation": "finish",
"time": "107.66",
"value": "finish"
}
]
}
|
40707
|
Following the American Civil War, the Army and Navy were unsupportive of each other at either the planning or operational level; thus the Joint Board, acting as an "advisory committee", was created to plan joint operations and resolve problems of common rivalry between the two services.
|
{
"evidence": [
"American Civil War_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:\nFollowing the American Civil War, the Army and Navy were unsupportive of each other at either the planning or operational level; thus the Joint Board, acting as an \"advisory committee\", was created to plan joint operations and resolve problems of common rivalry between the two services.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"American Civil War_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4660,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.309",
"value": "?search="
},
{
"operation": "search",
"time": "30.862",
"value": "American Civil War"
},
{
"operation": "Now on",
"time": "33.077",
"value": "American Civil War"
},
{
"operation": "Page search",
"time": "45.273",
"value": "supportive"
},
{
"operation": "Page search",
"time": "45.771",
"value": "supportive"
},
{
"operation": "Page search",
"time": "45.993",
"value": "supportive"
},
{
"operation": "Page search",
"time": "46.161",
"value": "supportive"
},
{
"operation": "Page search",
"time": "81.464",
"value": "navy"
},
{
"operation": "Page search",
"time": "81.657",
"value": "navy"
},
{
"operation": "search",
"time": "198.445",
"value": "Following the American Civil War, the Army and Navy were unsupportive"
},
{
"operation": "Now on",
"time": "200.939",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "217.407",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "218.534",
"value": "American Civil War"
},
{
"operation": "Page search",
"time": "232.426",
"value": "Army and Navy"
},
{
"operation": "Page search",
"time": "232.589",
"value": "Army and Navy"
},
{
"operation": "Page search",
"time": "232.756",
"value": "Army and Navy"
},
{
"operation": "Page search",
"time": "232.927",
"value": "Army and Navy"
},
{
"operation": "search",
"time": "666.26",
"value": "thus the Joint Board, acting as an \"advisory committee\""
},
{
"operation": "Now on",
"time": "668.708",
"value": "undefined"
},
{
"operation": "search",
"time": "683.973",
"value": "was created to plan joint operations and resolve problems of common rivalry between the two services."
},
{
"operation": "Now on",
"time": "687.054",
"value": "undefined"
},
{
"operation": "search",
"time": "704.365",
"value": "Following the American Civil War"
},
{
"operation": "Now on",
"time": "706.649",
"value": "undefined"
},
{
"operation": "search",
"time": "750.319",
"value": "\"joint board\" \" American Civil War\""
},
{
"operation": "Now on",
"time": "751.255",
"value": "undefined"
},
{
"operation": "search",
"time": "778.639",
"value": "American Civil War"
},
{
"operation": "Now on",
"time": "780.71",
"value": "American Civil War"
},
{
"operation": "Highlighting",
"time": "784.128",
"value": "American Civil War_sentence_1"
},
{
"operation": "finish",
"time": "789.151",
"value": "finish"
}
]
}
|
87222
|
Connecticut Route 229 nominally begins at the end of the eastbound Exit 31 off-ramp of I-84 (an east–west Interstate highway across the state of Connecticut through Danbury, Waterbury, Hartford and Union) and West Street is four lanes wide in the vicinity
|
{
"evidence": [
"Connecticut Route 229_sentence_5",
"Connecticut Route 229_sentence_7",
"Interstate 84 in Connecticut_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:\nConnecticut Route 229 nominally begins at the end of the eastbound Exit 31 off-ramp of I-84 (an east–west Interstate highway across the state of Connecticut through Danbury, Waterbury, Hartford and Union) and West Street is four lanes wide in the vicinity\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Connecticut Route 229_sentence_5",
"Connecticut Route 229_sentence_7",
"Interstate 84 in Connecticut_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4661,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.049",
"value": "?search="
},
{
"operation": "search",
"time": "55.966",
"value": "Connecticut Route 229"
},
{
"operation": "Now on",
"time": "56.659",
"value": "Connecticut Route 229"
},
{
"operation": "Highlighting",
"time": "65.054",
"value": "Connecticut Route 229_sentence_5"
},
{
"operation": "Highlighting",
"time": "70.468",
"value": "Connecticut Route 229_sentence_7"
},
{
"operation": "hyperlink",
"time": "71.996",
"value": "I-84"
},
{
"operation": "Now on",
"time": "73.543",
"value": "Interstate 84 in Connecticut"
},
{
"operation": "Highlighting",
"time": "78.436",
"value": "Interstate 84 in Connecticut_sentence_2"
},
{
"operation": "finish",
"time": "82.82",
"value": "finish"
}
]
}
|
exfever_1238
|
Cara Jocelyn Delevingne's most notable roles include Margo Roth Spiegelman in a 2015 American romantic mystery comedy-drama series.
|
{
"evidence": [
"Cara Delevingne",
"Paper Towns (film)"
],
"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:\nCara Jocelyn Delevingne's most notable roles include Margo Roth Spiegelman in a 2015 American romantic mystery comedy-drama series.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Cara Delevingne",
"Paper Towns (film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4662,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
64441
|
In the 2014–15 season, Kam began competing in ice dancing with Lee Ho-jung, and they eventually placed at 20th at the 2016 World Junior Championships (hosted by Hungarian Skating Union) in Debrecen, Hungary.
|
{
"evidence": [
"Richard Kam_sentence_11"
],
"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:\nIn the 2014–15 season, Kam began competing in ice dancing with Lee Ho-jung, and they eventually placed at 20th at the 2016 World Junior Championships (hosted by Hungarian Skating Union) in Debrecen, Hungary.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Richard Kam_sentence_11"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4663,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.228",
"value": "undefined"
},
{
"operation": "search",
"time": "35.779",
"value": "Contains... 2016 World Junior Championships"
},
{
"operation": "Now on",
"time": "37.476",
"value": "undefined"
},
{
"operation": "search",
"time": "68.365",
"value": "Kam Mickolio"
},
{
"operation": "Now on",
"time": "70.135",
"value": "Kam Mickolio"
},
{
"operation": "search",
"time": "78.543",
"value": "Kam Nai-wai"
},
{
"operation": "Now on",
"time": "79.836",
"value": "Kam Nai-wai"
},
{
"operation": "search",
"time": "90.615",
"value": "kam lee ho-jung"
},
{
"operation": "Now on",
"time": "94.347",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "100.177",
"value": "Lee Ho-jung"
},
{
"operation": "Now on",
"time": "116.26",
"value": "Richard Kam"
},
{
"operation": "Highlighting",
"time": "128.933",
"value": "Richard Kam_sentence_8"
},
{
"operation": "Highlighting",
"time": "148.231",
"value": "Richard Kam_sentence_11"
},
{
"operation": "Highlighting deleted",
"time": "151.026",
"value": "Richard Kam_sentence_8"
},
{
"operation": "finish",
"time": "174.586",
"value": "finish"
}
]
}
|
exfever_7689
|
Lounge music is the most popular in the United States.
|
{
"evidence": [
"Sister Act",
"Lounge music",
"Beautiful music"
],
"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:\nLounge music is the most popular in the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sister Act",
"Lounge music",
"Beautiful music"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4664,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
26397
|
A Distributed Bragg Reflector laser has diffraction grating (which splits and diffracts light into several beams travelling in different directions) on one or both end mirrors, which reflect a narrow band of light back in the cavity which help them be more spectrally stable.
|
{
"evidence": [
"Diffraction grating_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:\nA Distributed Bragg Reflector laser has diffraction grating (which splits and diffracts light into several beams travelling in different directions) on one or both end mirrors, which reflect a narrow band of light back in the cavity which help them be more spectrally stable.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Diffraction grating_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4665,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.602",
"value": "?search="
},
{
"operation": "search",
"time": "25.256",
"value": "Distributed Bragg reflector laser"
},
{
"operation": "Now on",
"time": "26.253",
"value": "Distributed Bragg reflector laser"
},
{
"operation": "search",
"time": "56.206",
"value": "grating"
},
{
"operation": "Now on",
"time": "57.574",
"value": "Grating"
},
{
"operation": "back-button-clicked",
"time": "59.459",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "59.497",
"value": "Distributed Bragg reflector laser"
},
{
"operation": "Page search",
"time": "62.924",
"value": "grating"
},
{
"operation": "search",
"time": "116.74",
"value": "Diffraction grating"
},
{
"operation": "Now on",
"time": "118.071",
"value": "Diffraction grating"
},
{
"operation": "Highlighting",
"time": "126.22",
"value": "Diffraction grating_sentence_0"
},
{
"operation": "search",
"time": "188.645",
"value": "Bragg reflector"
},
{
"operation": "Now on",
"time": "189.652",
"value": "Distributed Bragg reflector"
},
{
"operation": "hyperlink",
"time": "267.645",
"value": "distributed bragg reflector (DBR) lasers"
},
{
"operation": "Now on",
"time": "268.608",
"value": "Distributed Bragg reflector laser"
},
{
"operation": "Page search",
"time": "355.588",
"value": "end mirrors"
},
{
"operation": "hyperlink",
"time": "378.131",
"value": "distributed Bragg reflector"
},
{
"operation": "Now on",
"time": "379.472",
"value": "Distributed Bragg reflector"
},
{
"operation": "finish",
"time": "463.558",
"value": "finish"
}
]
}
|
exfever_7913
|
Keeping the Faith is a 2000 American romantic comedy film that was released by an American film distribution label of Walt Disney Studios Motion Pictures, created and owned by The Walt Disney Company.
|
{
"evidence": [
"Keeping the Faith",
"Touchstone Pictures"
],
"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:\nKeeping the Faith is a 2000 American romantic comedy film that was released by an American film distribution label of Walt Disney Studios Motion Pictures, created and owned by The Walt Disney Company.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Keeping the Faith",
"Touchstone Pictures"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4666,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
39416
|
Real Madrid (a Spanish professional football club based in Madrid) qualifed for the Champions League group stage with a score of 62 points.
|
{
"evidence": [
"Real Madrid CF_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:\nReal Madrid (a Spanish professional football club based in Madrid) qualifed for the Champions League group stage with a score of 62 points.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Real Madrid CF_sentence_5"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4667,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.815",
"value": "?search="
},
{
"operation": "search",
"time": "12.923",
"value": "Real Madrid fc"
},
{
"operation": "Now on",
"time": "15.232",
"value": "Real Madrid CF"
},
{
"operation": "Highlighting",
"time": "18.916",
"value": "Real Madrid CF_sentence_5"
},
{
"operation": "hyperlink",
"time": "119.139",
"value": "2017–18"
},
{
"operation": "Now on",
"time": "122.467",
"value": "2017–18 UEFA Champions League"
},
{
"operation": "back-button-clicked",
"time": "123.845",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "125.827",
"value": "Real Madrid CF"
},
{
"operation": "hyperlink",
"time": "136.275",
"value": "2016–17"
},
{
"operation": "hyperlink",
"time": "137.063",
"value": "2015–16"
},
{
"operation": "hyperlink",
"time": "137.601",
"value": "2013–14"
},
{
"operation": "hyperlink",
"time": "354.865",
"value": "1999–2000"
},
{
"operation": "Now on",
"time": "357.694",
"value": "1999–2000 UEFA Champions League"
},
{
"operation": "finish",
"time": "397.921",
"value": "finish"
}
]
}
|
exfever_4181
|
The National Basketball Association is the most profitable association in America.
|
{
"evidence": [
"Magic Johnson",
"Bill Russell NBA Finals Most Valuable Player Award"
],
"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 National Basketball Association is the most profitable association in America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Magic Johnson",
"Bill Russell NBA Finals Most Valuable Player Award"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4668,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2713
|
Dennis Quaid is the first American actor to win a New York Film Critics Circle Award.
|
{
"evidence": [
"Dennis Quaid",
"Breaking Away",
"Bloomington, Indiana"
],
"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:\nDennis Quaid is the first American actor to win a New York Film Critics Circle Award.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dennis Quaid",
"Breaking Away",
"Bloomington, Indiana"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4669,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6099
|
Sarrainodu is produced by an Indian film production and distribution company.
|
{
"evidence": [
"Sarrainodu",
"Geetha Arts"
],
"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:\nSarrainodu is produced by an Indian film production and distribution company.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sarrainodu",
"Geetha Arts"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4670,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_387
|
Kenny Rogers is the oldest country musician in the industry as of now.
|
{
"evidence": [
"Kenny Rogers",
"The New Christy Minstrels"
],
"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:\nKenny Rogers is the oldest country musician in the industry as of now.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Kenny Rogers",
"The New Christy Minstrels"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4671,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3715
|
Prince music incorporated a wide variety of styles including a popular music genre that originated in the African American community.
|
{
"evidence": [
"Prince (musician)",
"Soul music"
],
"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:\nPrince music incorporated a wide variety of styles including a popular music genre that originated in the African American community.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Prince (musician)",
"Soul music"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4672,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
11065
|
Sengalese football team Diambars FC was founded in 2003 and elevated to international status in 2009; they won their only League Cup in 2016.
|
{
"evidence": [
"Diambars FC_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:\nSengalese football team Diambars FC was founded in 2003 and elevated to international status in 2009; they won their only League Cup in 2016.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Diambars FC_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4673,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.704",
"value": "?search="
},
{
"operation": "search",
"time": "10.775",
"value": "Diambars FC"
},
{
"operation": "Now on",
"time": "12.28",
"value": "Diambars FC"
},
{
"operation": "Highlighting",
"time": "14.979",
"value": "Diambars FC_sentence_0"
},
{
"operation": "Highlighting",
"time": "18.312",
"value": "Diambars FC_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "18.315",
"value": "Diambars FC_header_cell_0_1_0"
},
{
"operation": "Page search",
"time": "41.419",
"value": "2009"
},
{
"operation": "Highlighting",
"time": "56.336",
"value": "Diambars FC_sentence_3"
},
{
"operation": "hyperlink",
"time": "90.149",
"value": "CAF Champions League"
},
{
"operation": "Now on",
"time": "92.335",
"value": "CAF Champions League"
},
{
"operation": "back-button-clicked",
"time": "97.979",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "98.153",
"value": "Diambars FC"
},
{
"operation": "Highlighting deleted",
"time": "104.506",
"value": "Diambars FC_sentence_0"
},
{
"operation": "Page search",
"time": "120.409",
"value": "league cup"
},
{
"operation": "Highlighting",
"time": "125.81",
"value": "Diambars FC_sentence_9"
},
{
"operation": "Highlighting deleted",
"time": "159.375",
"value": "Diambars FC_sentence_9"
},
{
"operation": "Highlighting deleted",
"time": "164.091",
"value": "Diambars FC_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "164.096",
"value": "Diambars FC_header_cell_0_1_0"
},
{
"operation": "finish",
"time": "173.375",
"value": "finish"
}
]
}
|
70396
|
Joseph ibn Abitur was from a very prestigious Spanish family, he was a great scholar of the Torah which means teaching in Hebrew.
|
{
"evidence": [
"Joseph ibn Abitur_sentence_2",
"Joseph ibn Abitur_sentence_4",
"Torah_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:\nJoseph ibn Abitur was from a very prestigious Spanish family, he was a great scholar of the Torah which means teaching in Hebrew.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Joseph ibn Abitur_sentence_2",
"Joseph ibn Abitur_sentence_4",
"Torah_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4674,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.068",
"value": "?search="
},
{
"operation": "search",
"time": "11.426",
"value": "Joseph ibn Abitur"
},
{
"operation": "Now on",
"time": "12.367",
"value": "Joseph ibn Abitur"
},
{
"operation": "Highlighting",
"time": "17.718",
"value": "Joseph ibn Abitur_sentence_2"
},
{
"operation": "Highlighting",
"time": "24.612",
"value": "Joseph ibn Abitur_sentence_4"
},
{
"operation": "hyperlink",
"time": "26.235",
"value": "Torah"
},
{
"operation": "Now on",
"time": "27.342",
"value": "Torah"
},
{
"operation": "Highlighting",
"time": "38.798",
"value": "Torah_sentence_7"
},
{
"operation": "finish",
"time": "57.733",
"value": "finish"
}
]
}
|
exfever_2964
|
Semitic languages, which are a branch of the Afroasiatic language family, are spoken by more than 330 million people in much of Asia's westernmost subregion.
|
{
"evidence": [
"Semitic languages",
"Western Asia"
],
"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:\nSemitic languages, which are a branch of the Afroasiatic language family, are spoken by more than 330 million people in much of Asia's westernmost subregion.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Semitic languages",
"Western Asia"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4675,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
76989
|
Jinfeng station closed down before the early 2000s.
|
{
"evidence": [
"Jinfeng station_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:\nJinfeng station closed down before the early 2000s.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Jinfeng station_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4676,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.105",
"value": "?search="
},
{
"operation": "search",
"time": "17.859",
"value": "Jinfeng station"
},
{
"operation": "Now on",
"time": "19.464",
"value": "Jinfeng station"
},
{
"operation": "Highlighting",
"time": "31.03",
"value": "Jinfeng station_sentence_2"
},
{
"operation": "finish",
"time": "63.328",
"value": "finish"
}
]
}
|
exfever_8260
|
Thirteen is a 2003 African teen drama film starring an American actress who is the recipient of a Critics' Choice Television Award as well as three Primetime Emmy Award nominations and three Golden Globe Award nominations for her work in film and television.
|
{
"evidence": [
"Thirteen (2003 film)",
"Evan Rachel Wood"
],
"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:\nThirteen is a 2003 African teen drama film starring an American actress who is the recipient of a Critics' Choice Television Award as well as three Primetime Emmy Award nominations and three Golden Globe Award nominations for her work in film and television.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Thirteen (2003 film)",
"Evan Rachel Wood"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4677,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
74835
|
During the 1988 Bossier City tornado outbreak one of the tornados went through the Central Park division and a power substation.
|
{
"evidence": [
"1978 Bossier City tornado outbreak_sentence_23",
"1978 Bossier City tornado outbreak_sentence_25"
],
"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:\nDuring the 1988 Bossier City tornado outbreak one of the tornados went through the Central Park division and a power substation.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"1978 Bossier City tornado outbreak_sentence_23",
"1978 Bossier City tornado outbreak_sentence_25"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4678,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.714",
"value": "?search="
},
{
"operation": "search",
"time": "19.742",
"value": "1988 Bossier City tornado outbreak"
},
{
"operation": "Now on",
"time": "20.868",
"value": "undefined"
},
{
"operation": "search",
"time": "48.074",
"value": "1988 Bossier City tornado"
},
{
"operation": "Now on",
"time": "48.831",
"value": "undefined"
},
{
"operation": "search",
"time": "67.368",
"value": "Bossier City tornado outbreak"
},
{
"operation": "Now on",
"time": "69.21",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "74.231",
"value": "1978 Bossier City tornado outbreak"
},
{
"operation": "Now on",
"time": "75.773",
"value": "1978 Bossier City tornado outbreak"
},
{
"operation": "Page search",
"time": "81.105",
"value": "Central Park division"
},
{
"operation": "Page search",
"time": "100.712",
"value": "Central Park"
},
{
"operation": "Highlighting",
"time": "108.926",
"value": "1978 Bossier City tornado outbreak_sentence_23"
},
{
"operation": "Page search",
"time": "121.759",
"value": "power substation"
},
{
"operation": "Highlighting",
"time": "131.474",
"value": "1978 Bossier City tornado outbreak_sentence_25"
},
{
"operation": "finish",
"time": "146.282",
"value": "finish"
}
]
}
|
exfever_8986
|
James Joseph Parsons was the first American actor and producer to receive a Golden Globe Award.
|
{
"evidence": [
"Jim Parsons",
"Primetime Emmy Award"
],
"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:\nJames Joseph Parsons was the first American actor and producer to receive a Golden Globe Award.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jim Parsons",
"Primetime Emmy Award"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4679,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
66785
|
District Football Associations are the local governing bodies of association football in Sweden, providing grassroots support by promoting and administering football in their respective districts.
|
{
"evidence": [
"Swedish District Football Associations_sentence_0",
"Swedish District Football Associations_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:\nDistrict Football Associations are the local governing bodies of association football in Sweden, providing grassroots support by promoting and administering football in their respective districts.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Swedish District Football Associations_sentence_0",
"Swedish District Football Associations_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4680,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Highlighting",
"time": "0",
"value": "Miss D_sentence_7"
},
{
"operation": "Highlighting",
"time": "0",
"value": "Miss D_title"
},
{
"operation": "Now on",
"time": "0.001",
"value": "?search="
},
{
"operation": "search",
"time": "0.011",
"value": "Contains...District Football Associations"
},
{
"operation": "Now on",
"time": "0.013",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "0.016",
"value": "F.C. Pedras Rubras"
},
{
"operation": "Now on",
"time": "0.017",
"value": "F.C. Pedras Rubras"
},
{
"operation": "hyperlink",
"time": "0.035",
"value": "Portuguese District Football Associations"
},
{
"operation": "Now on",
"time": "0.035",
"value": "Portuguese District Football Associations"
},
{
"operation": "search",
"time": "0.073",
"value": "Swedish District Football Associations"
},
{
"operation": "Now on",
"time": "0.074",
"value": "Swedish District Football Associations"
},
{
"operation": "Highlighting",
"time": "0.078",
"value": "Swedish District Football Associations_sentence_0"
},
{
"operation": "Highlighting",
"time": "0.084",
"value": "Swedish District Football Associations_sentence_1"
},
{
"operation": "Now on",
"time": "0.103",
"value": "Swedish District Football Associations"
},
{
"operation": "Now on",
"time": "0.115",
"value": "Swedish District Football Associations"
},
{
"operation": "Highlighting",
"time": "0.117",
"value": "Swedish District Football Associations_sentence_0"
},
{
"operation": "Highlighting",
"time": "0.118",
"value": "Swedish District Football Associations_sentence_1"
},
{
"operation": "Now on",
"time": "0.137",
"value": "Swedish District Football Associations"
},
{
"operation": "finish",
"time": "0.151",
"value": "finish"
},
{
"operation": "Now on",
"time": "1621805012.954",
"value": "Miss D"
},
{
"operation": "finish",
"time": "1621805022.068",
"value": "finish"
}
]
}
|
exfever_8978
|
There Will Be Blood is loosely based on the 1927 novel Oil! by the American muckraker who acquired particular fame for his classic blackening novels, which are reform-minded journalists, writers, and photographers in the Progressive Era in the United States.
|
{
"evidence": [
"There Will Be Blood",
"Upton Sinclair",
"Muckraker"
],
"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:\nThere Will Be Blood is loosely based on the 1927 novel Oil! by the American muckraker who acquired particular fame for his classic blackening novels, which are reform-minded journalists, writers, and photographers in the Progressive Era in the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"There Will Be Blood",
"Upton Sinclair",
"Muckraker"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4681,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3780
|
The Fosters is a television series created by an American actor who made his directorial and writing debut with a 2005 black comedy film.
|
{
"evidence": [
"The Fosters (2013 TV series)",
"Peter Paige",
"Say Uncle (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:\nThe Fosters is a television series created by an American actor who made his directorial and writing debut with a 2005 black comedy film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Fosters (2013 TV series)",
"Peter Paige",
"Say Uncle (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4682,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
88644
|
Huvudstabron connects the industrial area Ulvsunda, it was built 1964–67 but has a less than 10 metres horizontal clearance.
|
{
"evidence": [
"Huvudstabron_sentence_4"
],
"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:\nHuvudstabron connects the industrial area Ulvsunda, it was built 1964–67 but has a less than 10 metres horizontal clearance.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Huvudstabron_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4683,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.681",
"value": "?search="
},
{
"operation": "search",
"time": "8.334",
"value": "Huvudstabron connects"
},
{
"operation": "Now on",
"time": "9.342",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "11.354",
"value": "Huvudstabron"
},
{
"operation": "Now on",
"time": "12.167",
"value": "Huvudstabron"
},
{
"operation": "Highlighting",
"time": "13.728",
"value": "Huvudstabron_sentence_1"
},
{
"operation": "Highlighting",
"time": "21.338",
"value": "Huvudstabron_sentence_3"
},
{
"operation": "Highlighting",
"time": "31.711",
"value": "Huvudstabron_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "32.195",
"value": "Huvudstabron_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "32.835",
"value": "Huvudstabron_sentence_1"
},
{
"operation": "finish",
"time": "37.079",
"value": "finish"
}
]
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.