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 |
|---|---|---|---|---|---|---|---|---|
29975
|
The article reported that Charles Dougherty (former Duquesne University President) praised students who live off-campus as the ideal lifestyle for students and described their situation as "a preparation for being an adult." The comments gained popularity in the campus community and received coverage from national outlets such as The New York Times.
|
{
"evidence": [
"Charles J. Dougherty_sentence_11",
"Charles J. Dougherty_sentence_12",
"Charles J. Dougherty_sentence_15"
],
"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 article reported that Charles Dougherty (former Duquesne University President) praised students who live off-campus as the ideal lifestyle for students and described their situation as \"a preparation for being an adult.\" The comments gained popularity in the campus community and received coverage from national outlets such as The New York Times.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Charles J. Dougherty_sentence_11",
"Charles J. Dougherty_sentence_12",
"Charles J. Dougherty_sentence_15"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3684,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.248",
"value": "?search="
},
{
"operation": "search",
"time": "45.584",
"value": "Charles Dougherty"
},
{
"operation": "Now on",
"time": "46.358",
"value": "Charles Dougherty"
},
{
"operation": "hyperlink",
"time": "56.417",
"value": "Charles J. Dougherty"
},
{
"operation": "Now on",
"time": "57.048",
"value": "Charles J. Dougherty"
},
{
"operation": "hyperlink",
"time": "108.11",
"value": "Duquesne University"
},
{
"operation": "Now on",
"time": "109.505",
"value": "Duquesne University"
},
{
"operation": "Page search",
"time": "123.171",
"value": "Charles Dougherty"
},
{
"operation": "search",
"time": "173.607",
"value": "Contains... The comments gained popularity in the campus community and received coverage from national outlets such as The New York Times."
},
{
"operation": "Now on",
"time": "177.001",
"value": "undefined"
},
{
"operation": "search",
"time": "200.75",
"value": "Contains...praised students who live off-campus as the ideal lifestyle for students and described their situation as \"a preparation for being an adult"
},
{
"operation": "Now on",
"time": "202.146",
"value": "undefined"
},
{
"operation": "search",
"time": "242.71",
"value": "Charles Dougherty praised students who live off-campus as the ideal lifestyle for students and described their situation as a preparation for being an adult."
},
{
"operation": "Now on",
"time": "243.453",
"value": "undefined"
},
{
"operation": "search",
"time": "270.542",
"value": "Charles Dougherty"
},
{
"operation": "Now on",
"time": "271.699",
"value": "Charles Dougherty"
},
{
"operation": "hyperlink",
"time": "276.995",
"value": "Charles J. Dougherty"
},
{
"operation": "Now on",
"time": "277.785",
"value": "Charles J. Dougherty"
},
{
"operation": "Highlighting",
"time": "320.478",
"value": "Charles J. Dougherty_sentence_11"
},
{
"operation": "Highlighting",
"time": "321.692",
"value": "Charles J. Dougherty_sentence_12"
},
{
"operation": "Highlighting",
"time": "323.678",
"value": "Charles J. Dougherty_sentence_14"
},
{
"operation": "Highlighting",
"time": "327.381",
"value": "Charles J. Dougherty_sentence_15"
},
{
"operation": "finish",
"time": "346.418",
"value": "finish"
}
]
}
|
exfever_246
|
Jun Ji-Hyun is a South Korean actress and model who rose to fame for her role as The Girl in the 2001 South Korean romantic comedy film directed by Kwak Jae-Yong, starring Jun Ji-Hyun and the South Korean actor, best known for his lead role in Scandal Makers.
|
{
"evidence": [
"Jun Ji-hyun",
"My Sassy Girl",
"Cha Tae-hyun"
],
"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:\nJun Ji-Hyun is a South Korean actress and model who rose to fame for her role as The Girl in the 2001 South Korean romantic comedy film directed by Kwak Jae-Yong, starring Jun Ji-Hyun and the South Korean actor, best known for his lead role in Scandal Makers.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jun Ji-hyun",
"My Sassy Girl",
"Cha Tae-hyun"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3685,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8200
|
Huntington's disease is a neurodegenerative disease that causes movement disorder in which there is an increase in muscular activity that can result in excessive abnormal movements.
|
{
"evidence": [
"Huntington's disease",
"Hyperkinesia"
],
"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:\nHuntington's disease is a neurodegenerative disease that causes movement disorder in which there is an increase in muscular activity that can result in excessive abnormal movements.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Huntington's disease",
"Hyperkinesia"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3686,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3811
|
Fred Savage earned several awards and nominations, such as the Young Artist Awards and an American awards show recognizing people in entertainment, voted online by the general public and fans.
|
{
"evidence": [
"Fred Savage",
"People's Choice Awards"
],
"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:\nFred Savage earned several awards and nominations, such as the Young Artist Awards and an American awards show recognizing people in entertainment, voted online by the general public and fans.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Fred Savage",
"People's Choice Awards"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3687,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
19510
|
Samolus repens, a species of water pimpernel from the order ericales and genus samolus, is endemic to Australia and New Zealand.
|
{
"evidence": [
"Samolus repens_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:\nSamolus repens, a species of water pimpernel from the order ericales and genus samolus, is endemic to Australia and New Zealand.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Samolus repens_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3688,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.667",
"value": "?search="
},
{
"operation": "search",
"time": "6.631",
"value": "Samolus repens"
},
{
"operation": "Now on",
"time": "7.396",
"value": "Samolus repens"
},
{
"operation": "Highlighting",
"time": "10.218",
"value": "Samolus repens_sentence_0"
},
{
"operation": "Highlighting",
"time": "16.094",
"value": "Samolus repens_cell_0_7_0"
},
{
"operation": "Highlighting",
"time": "16.101",
"value": "Samolus repens_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "16.101",
"value": "Samolus repens_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "17.722",
"value": "Samolus repens_cell_0_7_1"
},
{
"operation": "Highlighting",
"time": "19.463",
"value": "Samolus repens_cell_0_9_0"
},
{
"operation": "Highlighting",
"time": "21.245",
"value": "Samolus repens_cell_0_9_1"
},
{
"operation": "Highlighting deleted",
"time": "70.054",
"value": "Samolus repens_cell_0_7_0"
},
{
"operation": "Highlighting deleted",
"time": "70.478",
"value": "Samolus repens_cell_0_9_0"
},
{
"operation": "Highlighting deleted",
"time": "71.384",
"value": "Samolus repens_cell_0_9_1"
},
{
"operation": "Highlighting deleted",
"time": "71.898",
"value": "Samolus repens_cell_0_7_1"
},
{
"operation": "Highlighting deleted",
"time": "71.902",
"value": "Samolus repens_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "71.903",
"value": "Samolus repens_header_cell_0_0_0"
},
{
"operation": "finish",
"time": "90.141",
"value": "finish"
}
]
}
|
16673
|
The main shops of interest in Berchem are Cogels-Osylei, Transvaalstraat, and Waterloostraat where demolition was prevented when the area got monumental status in the 1980s.
|
{
"evidence": [
"Berchem_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 main shops of interest in Berchem are Cogels-Osylei, Transvaalstraat, and Waterloostraat where demolition was prevented when the area got monumental status in the 1980s.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Berchem_sentence_15"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3689,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.357",
"value": "?search="
},
{
"operation": "search",
"time": "41.311",
"value": "berchem"
},
{
"operation": "Now on",
"time": "42.468",
"value": "Berchem"
},
{
"operation": "Highlighting",
"time": "145.814",
"value": "Berchem_sentence_15"
},
{
"operation": "Page search",
"time": "152.938",
"value": "shop"
},
{
"operation": "finish",
"time": "185.329",
"value": "finish"
}
]
}
|
4223
|
SK Rapid Wien is an Austria professional football team that has competed in various national and international competitions though no successful and Wikipedia article has a list of all players.
|
{
"evidence": [
"SK Rapid Wien_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:\nSK Rapid Wien is an Austria professional football team that has competed in various national and international competitions though no successful and Wikipedia article has a list of all players.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"SK Rapid Wien_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3690,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "79.79",
"value": "SK Rapid Wien"
},
{
"operation": "Highlighting",
"time": "85.405",
"value": "SK Rapid Wien_sentence_0"
},
{
"operation": "Highlighting",
"time": "101.866",
"value": "SK Rapid Wien_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "106.281",
"value": "SK Rapid Wien_sentence_2"
},
{
"operation": "Highlighting",
"time": "106.921",
"value": "SK Rapid Wien_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "108.837",
"value": "SK Rapid Wien_sentence_0"
},
{
"operation": "finish",
"time": "123.686",
"value": "finish"
}
]
}
|
exfever_1074
|
Paul Feig is an American actor, director, and writer who has a net worth of $30 million.
|
{
"evidence": [
"Ghostbusters (2016 film)",
"Justin Kirk"
],
"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:\nPaul Feig is an American actor, director, and writer who has a net worth of $30 million.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ghostbusters (2016 film)",
"Justin Kirk"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3691,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
66310
|
The Mayor of Penang Island was a position forst held by Patahiyah Ismail in 2015 an was succeeded by Maimunah Mohd Sharif who later was appointed as Executive Director of the United Nations Human Settlements Programme.
|
{
"evidence": [
"Mayor of Penang Island_sentence_5",
"Mayor of Penang Island_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 Mayor of Penang Island was a position forst held by Patahiyah Ismail in 2015 an was succeeded by Maimunah Mohd Sharif who later was appointed as Executive Director of the United Nations Human Settlements Programme.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Mayor of Penang Island_sentence_5",
"Mayor of Penang Island_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3692,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.758",
"value": "?search="
},
{
"operation": "search",
"time": "10.171",
"value": "Mayor of Penang Island"
},
{
"operation": "Now on",
"time": "11.568",
"value": "Mayor of Penang Island"
},
{
"operation": "Highlighting",
"time": "19.889",
"value": "Mayor of Penang Island_sentence_5"
},
{
"operation": "Highlighting",
"time": "27.686",
"value": "Mayor of Penang Island_sentence_6"
},
{
"operation": "finish",
"time": "32.852",
"value": "finish"
}
]
}
|
exfever_2754
|
Willow Smith, known mononymously as Willow, is an American singer and actress who is the granddaughter of Jada Pinkett Smith and an American actor, rapper, and producer.
|
{
"evidence": [
"Willow Smith",
"Will Smith"
],
"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:\nWillow Smith, known mononymously as Willow, is an American singer and actress who is the granddaughter of Jada Pinkett Smith and an American actor, rapper, and producer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Willow Smith",
"Will Smith"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3693,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2922
|
Heartlight is a song written by Neil Diamon, reportedly inspired by the 1982 Canadian science fiction film starring?an American actress, director, producer, and talk show host.
|
{
"evidence": [
"Heartlight (song)",
"E.T. the Extra-Terrestrial",
"Drew Barrymore"
],
"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:\nHeartlight is a song written by Neil Diamon, reportedly inspired by the 1982 Canadian science fiction film starring?an American actress, director, producer, and talk show host.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Heartlight (song)",
"E.T. the Extra-Terrestrial",
"Drew Barrymore"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3694,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5609
|
Mary-Louise Parker is an American actress who came to prominence for a role in a 1991 American comedy-drama film directed by Jon Avnet and based on Fannie Flagg's 1987 novel.
|
{
"evidence": [
"Mary-Louise Parker",
"Fried Green Tomatoes"
],
"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:\nMary-Louise Parker is an American actress who came to prominence for a role in a 1991 American comedy-drama film directed by Jon Avnet and based on Fannie Flagg's 1987 novel.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mary-Louise Parker",
"Fried Green Tomatoes"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3695,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6487
|
Audrey Horne is a fictional character from the television series of an American multinational commercial broadcast television network that serves as the flagship property of Walt Disney Television.
|
{
"evidence": [
"Audrey Horne",
"American Broadcasting Company"
],
"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:\nAudrey Horne is a fictional character from the television series of an American multinational commercial broadcast television network that serves as the flagship property of Walt Disney Television.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Audrey Horne",
"American Broadcasting Company"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3696,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4393
|
Heroin is used as a recreational drug for its euphoric effects classified as a substance that acts on opioid receptors to produce morphine-like effects.
|
{
"evidence": [
"Heroin",
"Opioid"
],
"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:\nHeroin is used as a recreational drug for its euphoric effects classified as a substance that acts on opioid receptors to produce morphine-like effects.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Heroin",
"Opioid"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3697,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1826
|
Yusuke Murata's other major work is his illustration of Eyeshield, a manga about a team sport also known as "mintonette."
|
{
"evidence": [
"Yusuke Murata",
"American football"
],
"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:\nYusuke Murata's other major work is his illustration of Eyeshield, a manga about a team sport also known as \"mintonette.\"\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Yusuke Murata",
"American football"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3698,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
18910
|
Charmaine Nelson was an associate member of Concordia University's Gail and Stephen A. Jarislowsky Institute for Studies in Canadian Art, then three years later she was named a Fellow of the Royal Society of Canada's College of New Scholars, Artists, and Scientists.
|
{
"evidence": [
"Charmaine Nelson_sentence_8",
"Charmaine Nelson_sentence_10"
],
"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:\nCharmaine Nelson was an associate member of Concordia University's Gail and Stephen A. Jarislowsky Institute for Studies in Canadian Art, then three years later she was named a Fellow of the Royal Society of Canada's College of New Scholars, Artists, and Scientists.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Charmaine Nelson_sentence_8",
"Charmaine Nelson_sentence_10"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3699,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.577",
"value": "?search="
},
{
"operation": "search",
"time": "14.077",
"value": "Charmaine Nelson"
},
{
"operation": "Now on",
"time": "14.797",
"value": "Charmaine Nelson"
},
{
"operation": "Highlighting",
"time": "60.688",
"value": "Charmaine Nelson_sentence_8"
},
{
"operation": "Highlighting",
"time": "98.83",
"value": "Charmaine Nelson_sentence_10"
},
{
"operation": "finish",
"time": "112.582",
"value": "finish"
}
]
}
|
16540
|
KRML began regular operations earlier than 1958 as KTEE and was established by Seaside Electronic Associates.
|
{
"evidence": [
"KRML_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:\nKRML began regular operations earlier than 1958 as KTEE and was established by Seaside Electronic Associates.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"KRML_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3700,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.727",
"value": "?search="
},
{
"operation": "search",
"time": "60.912",
"value": "KRML"
},
{
"operation": "Now on",
"time": "62.317",
"value": "KRML"
},
{
"operation": "Highlighting",
"time": "76.886",
"value": "KRML_sentence_1"
},
{
"operation": "Page search",
"time": "93.73",
"value": "regular"
},
{
"operation": "Highlighting",
"time": "103.334",
"value": "KRML_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "106.38",
"value": "KRML_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "108.331",
"value": "KRML_sentence_5"
},
{
"operation": "Highlighting",
"time": "108.891",
"value": "KRML_sentence_5"
},
{
"operation": "finish",
"time": "114.247",
"value": "finish"
}
]
}
|
exfever_5836
|
Davis Guggenheim directed a 30-minute documentary The Dream now.
|
{
"evidence": [
"Davis Guggenheim",
"The Shield",
"Sony Pictures Television"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDavis Guggenheim directed a 30-minute documentary The Dream now.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Davis Guggenheim",
"The Shield",
"Sony Pictures Television"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3701,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
13064
|
The Yoshizawa–Randlett system was first described in Samuel Randlett's Art of Origami in 1960 after Samuel Randlett and Robert Harbin added a few symbols to Akira Yoshizawa's diagramming.
|
{
"evidence": [
"Yoshizawa–Randlett system_sentence_13"
],
"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 Yoshizawa–Randlett system was first described in Samuel Randlett's Art of Origami in 1960 after Samuel Randlett and Robert Harbin added a few symbols to Akira Yoshizawa's diagramming.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Yoshizawa–Randlett system_sentence_13"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3702,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.755",
"value": "?search="
},
{
"operation": "search",
"time": "38.687",
"value": "Yoshizawa–Randlett system"
},
{
"operation": "Now on",
"time": "39.574",
"value": "Yoshizawa–Randlett system"
},
{
"operation": "Highlighting",
"time": "87.224",
"value": "Yoshizawa–Randlett system_sentence_10"
},
{
"operation": "Highlighting",
"time": "97.054",
"value": "Yoshizawa–Randlett system_sentence_13"
},
{
"operation": "Highlighting deleted",
"time": "98.791",
"value": "Yoshizawa–Randlett system_sentence_10"
},
{
"operation": "finish",
"time": "114.228",
"value": "finish"
}
]
}
|
exfever_4356
|
Nick Nolte is an Australian actor who received an Academy Award nomination for the 2011 American sports drama film directed by Gavin O'Connor.
|
{
"evidence": [
"Nick Nolte",
"Warrior (2011 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:\nNick Nolte is an Australian actor who received an Academy Award nomination for the 2011 American sports drama film directed by Gavin O'Connor.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Nick Nolte",
"Warrior (2011 film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3703,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7322
|
My Sweet Lord is a song by George Harrison who originally gave the song to his fellow Apple Records artist whose work encompassed a music genre that originated in African American communities in the mid-1960s.
|
{
"evidence": [
"My Sweet Lord",
"Billy Preston",
"Funk"
],
"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:\nMy Sweet Lord is a song by George Harrison who originally gave the song to his fellow Apple Records artist whose work encompassed a music genre that originated in African American communities in the mid-1960s.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"My Sweet Lord",
"Billy Preston",
"Funk"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3704,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_778
|
Michelle Obama is a former first lady of the United States who delivered a keynote address during her husband's campaign at the quadrennial presidential nominating convention of the Democratic Party held in Aspen, Colorado.
|
{
"evidence": [
"Michelle Obama",
"2008 Democratic National Convention"
],
"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:\nMichelle Obama is a former first lady of the United States who delivered a keynote address during her husband's campaign at the quadrennial presidential nominating convention of the Democratic Party held in Aspen, Colorado.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Michelle Obama",
"2008 Democratic National Convention"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3705,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7819
|
Li Lianjie is a Korean film actor who is a naturalized citizen in a sovereign island city-state in maritime Southeast Asia that has four official languages, including a language that is classified into two distinct dialects.
|
{
"evidence": [
"Jet Li",
"Singapore",
"Singaporean Mandarin"
],
"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:\nLi Lianjie is a Korean film actor who is a naturalized citizen in a sovereign island city-state in maritime Southeast Asia that has four official languages, including a language that is classified into two distinct dialects.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jet Li",
"Singapore",
"Singaporean Mandarin"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3706,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7170
|
Bonnie Lynn Hunt is an actress who starred in a 2005 American family comedy film starring an American actor born August 14, 1945.
|
{
"evidence": [
"Bonnie Hunt",
"Cheaper by the Dozen 2",
"Steve Martin"
],
"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:\nBonnie Lynn Hunt is an actress who starred in a 2005 American family comedy film starring an American actor born August 14, 1945.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Bonnie Hunt",
"Cheaper by the Dozen 2",
"Steve Martin"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3707,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4906
|
Taylor Swift is an American singer-songwriter who signed a recording deal in 2005 with an American independent record label and a joint venture between Borchetta and an American country music singer and record producer whose breakthrough single was "How Do You Like Me Now?!"
|
{
"evidence": [
"Taylor Swift",
"Big Machine Records",
"Toby Keith"
],
"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:\nTaylor Swift is an American singer-songwriter who signed a recording deal in 2005 with an American independent record label and a joint venture between Borchetta and an American country music singer and record producer whose breakthrough single was \"How Do You Like Me Now?!\"\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Taylor Swift",
"Big Machine Records",
"Toby Keith"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3708,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
61590
|
Linear grammar indicates that all linear languages are context-free and that regular languages are a proper subset of the linear languages.
|
{
"evidence": [
"Linear grammar_sentence_11",
"Linear grammar_sentence_12"
],
"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:\nLinear grammar indicates that all linear languages are context-free and that regular languages are a proper subset of the linear languages.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Linear grammar_sentence_11",
"Linear grammar_sentence_12"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3709,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.702",
"value": "?search="
},
{
"operation": "search",
"time": "12.715",
"value": "Linear grammar"
},
{
"operation": "Now on",
"time": "13.375",
"value": "Linear grammar"
},
{
"operation": "Highlighting",
"time": "52.784",
"value": "Linear grammar_sentence_0"
},
{
"operation": "Highlighting",
"time": "54.135",
"value": "Linear grammar_sentence_1"
},
{
"operation": "Highlighting",
"time": "85.983",
"value": "Linear grammar_sentence_5"
},
{
"operation": "Highlighting",
"time": "85.985",
"value": "Linear grammar_section_1"
},
{
"operation": "Highlighting",
"time": "87.535",
"value": "Linear grammar_sentence_4"
},
{
"operation": "Highlighting",
"time": "112.445",
"value": "Linear grammar_sentence_11"
},
{
"operation": "Highlighting",
"time": "117.162",
"value": "Linear grammar_sentence_12"
},
{
"operation": "Highlighting deleted",
"time": "119.76",
"value": "Linear grammar_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "120.721",
"value": "Linear grammar_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "120.724",
"value": "Linear grammar_section_1"
},
{
"operation": "Highlighting deleted",
"time": "123.833",
"value": "Linear grammar_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "124.733",
"value": "Linear grammar_sentence_1"
},
{
"operation": "finish",
"time": "140.579",
"value": "finish"
}
]
}
|
87001
|
Mahakörgis Khan was given the title Ükegtü, they fought Tümeds and defeated Oirads who are part of Mongols.
|
{
"evidence": [
"Mahakörgis Khan_sentence_0",
"Mahakörgis Khan_sentence_8",
"Mahakörgis Khan_sentence_6",
"Tumed_sentence_13",
"Tumed_sentence_14",
"Mahakörgis Khan_sentence_5",
"Oirats_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:\nMahakörgis Khan was given the title Ükegtü, they fought Tümeds and defeated Oirads who are part of Mongols.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Mahakörgis Khan_sentence_0",
"Mahakörgis Khan_sentence_8",
"Mahakörgis Khan_sentence_6",
"Tumed_sentence_13",
"Tumed_sentence_14",
"Mahakörgis Khan_sentence_5",
"Oirats_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3710,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.115",
"value": "undefined"
},
{
"operation": "search",
"time": "10.368",
"value": "Mahakörgis Khan"
},
{
"operation": "Now on",
"time": "11.398",
"value": "Mahakörgis Khan"
},
{
"operation": "Page search",
"time": "23.644",
"value": "Ükegtü"
},
{
"operation": "Highlighting",
"time": "39.546",
"value": "Mahakörgis Khan_sentence_0"
},
{
"operation": "Highlighting",
"time": "40.815",
"value": "Mahakörgis Khan_sentence_8"
},
{
"operation": "Highlighting",
"time": "51.472",
"value": "Mahakörgis Khan_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "57.456",
"value": "Mahakörgis Khan_sentence_5"
},
{
"operation": "Page search",
"time": "67.628",
"value": "Tümeds"
},
{
"operation": "Highlighting",
"time": "103.45",
"value": "Mahakörgis Khan_sentence_6"
},
{
"operation": "hyperlink",
"time": "106.698",
"value": "Tümeds"
},
{
"operation": "Now on",
"time": "108.649",
"value": "Tümed"
},
{
"operation": "Highlighting",
"time": "154.934",
"value": "Tumed_sentence_13"
},
{
"operation": "Highlighting",
"time": "159.275",
"value": "Tumed_sentence_14"
},
{
"operation": "back-button-clicked",
"time": "164.709",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "164.998",
"value": "Mahakörgis Khan"
},
{
"operation": "Page search",
"time": "194.344",
"value": "fought"
},
{
"operation": "Highlighting",
"time": "211.751",
"value": "Mahakörgis Khan_sentence_5"
},
{
"operation": "hyperlink",
"time": "213.99",
"value": "Oirads"
},
{
"operation": "Now on",
"time": "215.169",
"value": "Oirats"
},
{
"operation": "Highlighting",
"time": "221.975",
"value": "Oirats_sentence_2"
},
{
"operation": "back-button-clicked",
"time": "229.354",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "229.632",
"value": "Mahakörgis Khan"
},
{
"operation": "hyperlink",
"time": "260.152",
"value": "Tümeds"
},
{
"operation": "Now on",
"time": "260.763",
"value": "Tümed"
},
{
"operation": "back-button-clicked",
"time": "270.105",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "270.372",
"value": "Mahakörgis Khan"
},
{
"operation": "hyperlink",
"time": "305.882",
"value": "Tümeds"
},
{
"operation": "Now on",
"time": "306.815",
"value": "Tümed"
},
{
"operation": "Page search",
"time": "315.417",
"value": "Mahakörgis Khan"
},
{
"operation": "finish",
"time": "333.443",
"value": "finish"
}
]
}
|
exfever_617
|
Yusuke Murata is a Japanese manga animator best known for illustrating One's One-Punch Man, serialized in the monthly Japanese seinen manga magazine published by a Japanese company established in 1925 as the entertainment-related publishing division of Japanese publisher Shogakukan.
|
{
"evidence": [
"Yusuke Murata",
"Weekly Young Jump",
"Shueisha"
],
"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:\nYusuke Murata is a Japanese manga animator best known for illustrating One's One-Punch Man, serialized in the monthly Japanese seinen manga magazine published by a Japanese company established in 1925 as the entertainment-related publishing division of Japanese publisher Shogakukan.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Yusuke Murata",
"Weekly Young Jump",
"Shueisha"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3711,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
8323
|
Dean Detton, who was born in Richmond, Utah, a city with a total area of 3.4 square miles, was a football player and a wrestler.
|
{
"evidence": [
"Dean Detton_sentence_2",
"Richmond, Utah_sentence_17",
"Dean Detton_sentence_0",
"Dean Detton_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:\nDean Detton, who was born in Richmond, Utah, a city with a total area of 3.4 square miles, was a football player and a wrestler.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Dean Detton_sentence_2",
"Richmond, Utah_sentence_17",
"Dean Detton_sentence_0",
"Dean Detton_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3712,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.273",
"value": "?search="
},
{
"operation": "search",
"time": "50.197",
"value": "Dean Detton"
},
{
"operation": "Now on",
"time": "50.841",
"value": "Dean Detton"
},
{
"operation": "Highlighting",
"time": "67.824",
"value": "Dean Detton_cell_0_2_1"
},
{
"operation": "Highlighting",
"time": "67.862",
"value": "Dean Detton_header_cell_0_2_0"
},
{
"operation": "Highlighting",
"time": "67.871",
"value": "Dean Detton_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "75.973",
"value": "Dean Detton_cell_0_2_1"
},
{
"operation": "Highlighting deleted",
"time": "76.019",
"value": "Dean Detton_header_cell_0_2_0"
},
{
"operation": "Highlighting deleted",
"time": "76.025",
"value": "Dean Detton_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "77.886",
"value": "Dean Detton_sentence_2"
},
{
"operation": "hyperlink",
"time": "80.603",
"value": "Richmond, Utah"
},
{
"operation": "Now on",
"time": "81.58",
"value": "Richmond, Utah"
},
{
"operation": "Highlighting",
"time": "109.444",
"value": "Richmond, Utah_sentence_17"
},
{
"operation": "back-button-clicked",
"time": "110.841",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "111.077",
"value": "Dean Detton"
},
{
"operation": "Highlighting",
"time": "112.319",
"value": "Dean Detton_sentence_0"
},
{
"operation": "Highlighting",
"time": "116.295",
"value": "Dean Detton_sentence_1"
},
{
"operation": "finish",
"time": "123.065",
"value": "finish"
}
]
}
|
37892
|
Nicklas Strunck, a Danish professional footballer who was born in 1999, debuted for FC Nordsjælland at the age of 18 and was promoted into the first team of the FC in 2018.
|
{
"evidence": [
"Nicklas Strunck_sentence_0",
"Nicklas Strunck_sentence_6",
"Nicklas Strunck_sentence_9"
],
"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:\nNicklas Strunck, a Danish professional footballer who was born in 1999, debuted for FC Nordsjælland at the age of 18 and was promoted into the first team of the FC in 2018.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Nicklas Strunck_sentence_0",
"Nicklas Strunck_sentence_6",
"Nicklas Strunck_sentence_9"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3713,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.74",
"value": "?search="
},
{
"operation": "search",
"time": "21.257",
"value": "Nicklas Strunck"
},
{
"operation": "Now on",
"time": "21.852",
"value": "Nicklas Strunck"
},
{
"operation": "Highlighting",
"time": "38.516",
"value": "Nicklas Strunck_sentence_0"
},
{
"operation": "Highlighting",
"time": "61.956",
"value": "Nicklas Strunck_sentence_6"
},
{
"operation": "Highlighting",
"time": "98.522",
"value": "Nicklas Strunck_sentence_9"
},
{
"operation": "finish",
"time": "126.036",
"value": "finish"
}
]
}
|
exfever_3675
|
Caryn Elaine Johnson is an American actor who is one of 16 entertainers to win the EGOT, which includes an award for artistic and technical merit in the television industry.
|
{
"evidence": [
"Whoopi Goldberg",
"Emmy Award"
],
"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:\nCaryn Elaine Johnson is an American actor who is one of 16 entertainers to win the EGOT, which includes an award for artistic and technical merit in the television industry.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Whoopi Goldberg",
"Emmy Award"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3714,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
77971
|
Dunga, a Brazilian football manager and former professional player who played as a defensive midfielder, was appointed as coach of the Brazil national football team twice, in 2006 and 2012.
|
{
"evidence": [
"Dunga_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:\nDunga, a Brazilian football manager and former professional player who played as a defensive midfielder, was appointed as coach of the Brazil national football team twice, in 2006 and 2012.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Dunga_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3715,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.836",
"value": "?search="
},
{
"operation": "search",
"time": "777.056",
"value": "Dunga"
},
{
"operation": "Now on",
"time": "778.829",
"value": "Dunga"
},
{
"operation": "Page search",
"time": "941.469",
"value": "manager"
},
{
"operation": "Page search",
"time": "995.56",
"value": "coach"
},
{
"operation": "Highlighting",
"time": "1026.097",
"value": "Dunga_sentence_6"
},
{
"operation": "Highlighting",
"time": "1037.239",
"value": "Dunga_sentence_7"
},
{
"operation": "Highlighting",
"time": "1038.302",
"value": "Dunga_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "1039.649",
"value": "Dunga_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "1040.284",
"value": "Dunga_sentence_6"
},
{
"operation": "Highlighting",
"time": "1041.171",
"value": "Dunga_sentence_9"
},
{
"operation": "Highlighting deleted",
"time": "1042.974",
"value": "Dunga_sentence_9"
},
{
"operation": "finish",
"time": "1077.112",
"value": "finish"
}
]
}
|
exfever_2046
|
The Mummy is a 2017 American fantasy action-adventure film co-written by an American film and television writer who played the part of Richard Cameron in the 1989 film Dead Poets Society as well as Dr. Allen Painter in The Way of the Gun in 2000.
|
{
"evidence": [
"The Mummy (2017 film)",
"Dylan Kussman"
],
"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 Mummy is a 2017 American fantasy action-adventure film co-written by an American film and television writer who played the part of Richard Cameron in the 1989 film Dead Poets Society as well as Dr. Allen Painter in The Way of the Gun in 2000.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Mummy (2017 film)",
"Dylan Kussman"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3716,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5059
|
Stephon Xavier Marbury is a Canadian professional player of a team sport in which two teams, most commonly of five players each, opposing one another on a rectangular court, compete with the primary objective of shooting the ball at the ring.
|
{
"evidence": [
"Stephon Marbury",
"Basketball"
],
"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:\nStephon Xavier Marbury is a Canadian professional player of a team sport in which two teams, most commonly of five players each, opposing one another on a rectangular court, compete with the primary objective of shooting the ball at the ring.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Stephon Marbury",
"Basketball"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3717,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6399
|
In 1972, Jackie Robinson was inducted into the Museum that was operated by private interests in Cooperstown, New York.
|
{
"evidence": [
"Jackie Robinson",
"National Baseball Hall of Fame and Museum"
],
"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 1972, Jackie Robinson was inducted into the Museum that was operated by private interests in Cooperstown, New York.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jackie Robinson",
"National Baseball Hall of Fame and Museum"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3718,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
65045
|
George Phippen is a sculptor and painter who made the bronze sculture Cowboy in a Storm and co-founded the Cowboy Artists of America, an exclusive organization dedicated to portraying the lifestyles of the cowboy and the American West.
|
{
"evidence": [
"George Phippen_sentence_9",
"George Phippen_sentence_11",
"Cowboy Artists of America_sentence_2",
"George Phippen_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:\nGeorge Phippen is a sculptor and painter who made the bronze sculture Cowboy in a Storm and co-founded the Cowboy Artists of America, an exclusive organization dedicated to portraying the lifestyles of the cowboy and the American West.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"George Phippen_sentence_9",
"George Phippen_sentence_11",
"Cowboy Artists of America_sentence_2",
"George Phippen_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3719,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.079",
"value": "?search="
},
{
"operation": "search",
"time": "12.123",
"value": "George Phippen"
},
{
"operation": "Now on",
"time": "13.285",
"value": "George Phippen"
},
{
"operation": "Highlighting",
"time": "63.704",
"value": "George Phippen_sentence_9"
},
{
"operation": "Highlighting",
"time": "93.624",
"value": "George Phippen_sentence_11"
},
{
"operation": "Highlighting deleted",
"time": "144.855",
"value": "George Phippen_sentence_11"
},
{
"operation": "Highlighting",
"time": "149.417",
"value": "George Phippen_sentence_11"
},
{
"operation": "hyperlink",
"time": "162.516",
"value": "Cowboy Artists of America"
},
{
"operation": "Now on",
"time": "163.257",
"value": "Cowboy Artists of America"
},
{
"operation": "Highlighting",
"time": "167.123",
"value": "Cowboy Artists of America_sentence_0"
},
{
"operation": "Highlighting",
"time": "179.471",
"value": "Cowboy Artists of America_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "202.763",
"value": "Cowboy Artists of America_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "205.608",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "205.765",
"value": "George Phippen"
},
{
"operation": "Highlighting",
"time": "208.207",
"value": "George Phippen_sentence_0"
},
{
"operation": "finish",
"time": "266.099",
"value": "finish"
}
]
}
|
exfever_8618
|
Jerome was a Christian priest, confessor, theologian, and historian born at Stridon, a village near the Roman castrum where the navigable Ljubljanica river came closest to Malabar Hill.
|
{
"evidence": [
"Jerome",
"Emona"
],
"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:\nJerome was a Christian priest, confessor, theologian, and historian born at Stridon, a village near the Roman castrum where the navigable Ljubljanica river came closest to Malabar Hill.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jerome",
"Emona"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3720,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7003
|
Moscovium is one of 23 known chemical elements that do not occur naturally on Earth with the symbol Mc and atomic number 115.
|
{
"evidence": [
"Moscovium",
"Synthetic element"
],
"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:\nMoscovium is one of 23 known chemical elements that do not occur naturally on Earth with the symbol Mc and atomic number 115.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Moscovium",
"Synthetic element"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3721,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8072
|
Keturah was a wife and a concubine of the common patriarch of the Abrahamic religions, including Judaism, Christianity, and an Abrahamic monotheistic religion whose text, the Quran, is considered by followers, known as Muslims, to be the word of God.
|
{
"evidence": [
"Keturah",
"Abraham",
"Islam"
],
"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:\nKeturah was a wife and a concubine of the common patriarch of the Abrahamic religions, including Judaism, Christianity, and an Abrahamic monotheistic religion whose text, the Quran, is considered by followers, known as Muslims, to be the word of God.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Keturah",
"Abraham",
"Islam"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3722,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5085
|
The Crazies is a 2010 American science fiction horror film starring a Polish film actor and singer best known for his work in Across the Universe.
|
{
"evidence": [
"The Crazies (2010 film)",
"Joe Anderson (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:\nThe Crazies is a 2010 American science fiction horror film starring a Polish film actor and singer best known for his work in Across the Universe.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Crazies (2010 film)",
"Joe Anderson (actor)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3723,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_69
|
The Football League Cup Final was played between Leicester City and an amateur association football club based in Birkenhead, Merseyside, England.
|
{
"evidence": [
"2000 Football League Cup Final",
"Tranmere Rovers F.C."
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Football League Cup Final was played between Leicester City and an amateur association football club based in Birkenhead, Merseyside, England.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"2000 Football League Cup Final",
"Tranmere Rovers F.C."
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3724,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3747
|
Channing Tatum is an American actor whose breakout role was as werewolf Alcide Herveaux in the series True Blood, which was shown on an American pay television network launched on November 8, 1972.
|
{
"evidence": [
"Joe Manganiello",
"HBO"
],
"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:\nChanning Tatum is an American actor whose breakout role was as werewolf Alcide Herveaux in the series True Blood, which was shown on an American pay television network launched on November 8, 1972.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Joe Manganiello",
"HBO"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3725,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
75683
|
Phillip Hughes was a left-handed opening batsman who played for two seasons with New South Wales, an Australian men's professional first class cricket team based in Sydney, New South Wales, and made his One Day International Debut in 2014.
|
{
"evidence": [
"Phillip Hughes_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:\nPhillip Hughes was a left-handed opening batsman who played for two seasons with New South Wales, an Australian men's professional first class cricket team based in Sydney, New South Wales, and made his One Day International Debut in 2014.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Phillip Hughes_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3726,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.027",
"value": "?search="
},
{
"operation": "search",
"time": "9.794",
"value": "Phillip Hughes"
},
{
"operation": "Now on",
"time": "10.791",
"value": "Phillip Hughes"
},
{
"operation": "Highlighting",
"time": "25.296",
"value": "Phillip Hughes_sentence_3"
},
{
"operation": "hyperlink",
"time": "31.704",
"value": "New South Wales"
},
{
"operation": "Now on",
"time": "32.734",
"value": "New South Wales cricket team"
},
{
"operation": "Highlighting",
"time": "39.833",
"value": "New South Wales cricket team_sentence_0"
},
{
"operation": "Highlighting",
"time": "39.835",
"value": "New South Wales cricket team_title"
},
{
"operation": "back-button-clicked",
"time": "42.361",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "42.744",
"value": "Phillip Hughes"
},
{
"operation": "Highlighting",
"time": "45.552",
"value": "Phillip Hughes_sentence_4"
},
{
"operation": "Highlighting",
"time": "50.031",
"value": "Phillip Hughes_sentence_4"
},
{
"operation": "finish",
"time": "56.263",
"value": "finish"
}
]
}
|
exfever_4986
|
Emma Stone was the most famous American actress.
|
{
"evidence": [
"Emma Stone",
"The Wind in the Willows",
"The Big Read"
],
"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:\nEmma Stone was the most famous American actress.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Emma Stone",
"The Wind in the Willows",
"The Big Read"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3727,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
27760
|
Generator Hostels are headquartered in London, and employ international travel expert Anwar Mekhayech as their CEO.
|
{
"evidence": [
"Generator Hostels_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:\nGenerator Hostels are headquartered in London, and employ international travel expert Anwar Mekhayech as their CEO.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Generator Hostels_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3728,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.628",
"value": "?search="
},
{
"operation": "search",
"time": "5.735",
"value": "Generator Hostels"
},
{
"operation": "Now on",
"time": "6.398",
"value": "Generator Hostels"
},
{
"operation": "Highlighting",
"time": "13.585",
"value": "Generator Hostels_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "14.03",
"value": "Generator Hostels_cell_0_3_1"
},
{
"operation": "Highlighting",
"time": "14.033",
"value": "Generator Hostels_header_cell_0_3_0"
},
{
"operation": "Page search",
"time": "33.657",
"value": "anwar"
},
{
"operation": "Highlighting",
"time": "34.975",
"value": "Generator Hostels_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "45.504",
"value": "Generator Hostels_cell_0_3_1"
},
{
"operation": "Highlighting deleted",
"time": "45.505",
"value": "Generator Hostels_header_cell_0_3_0"
},
{
"operation": "finish",
"time": "190.737",
"value": "finish"
}
]
}
|
47678
|
Tommy Bennett is a former American football safety in the National Football League who graduated from Samuel F. B. Morse High School, an urban public high school located in southeastern San Diego, in 1990.
|
{
"evidence": [
"Tommy Bennett_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:\nTommy Bennett is a former American football safety in the National Football League who graduated from Samuel F. B. Morse High School, an urban public high school located in southeastern San Diego, in 1990.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Tommy Bennett_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3729,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.194",
"value": "?search="
},
{
"operation": "search",
"time": "10.896",
"value": "Tommy Bennett"
},
{
"operation": "Now on",
"time": "11.575",
"value": "Tommy Bennett"
},
{
"operation": "Highlighting",
"time": "20.259",
"value": "Tommy Bennett_sentence_0"
},
{
"operation": "Highlighting",
"time": "26.515",
"value": "Tommy Bennett_sentence_1"
},
{
"operation": "hyperlink",
"time": "33.961",
"value": "Samuel F. B. Morse High School"
},
{
"operation": "Now on",
"time": "34.71",
"value": "Morse High School (California)"
},
{
"operation": "Highlighting",
"time": "35.89",
"value": "Morse High School (California)_sentence_0"
},
{
"operation": "Highlighting",
"time": "35.892",
"value": "Morse High School (California)_title"
},
{
"operation": "back-button-clicked",
"time": "44.028",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "44.121",
"value": "Tommy Bennett"
},
{
"operation": "Highlighting deleted",
"time": "50.978",
"value": "Tommy Bennett_sentence_0"
},
{
"operation": "finish",
"time": "62.38",
"value": "finish"
}
]
}
|
76421
|
Linda Mcmahon led the Senatorial race in Connecticut gaining 49.44% (60,749) of the total votes.
|
{
"evidence": [
"Linda McMahon_sentence_139",
"Linda McMahon_sentence_148"
],
"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:\nLinda Mcmahon led the Senatorial race in Connecticut gaining 49.44% (60,749) of the total votes.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Linda McMahon_sentence_139",
"Linda McMahon_sentence_148"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3730,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.672",
"value": "?search="
},
{
"operation": "search",
"time": "15.625",
"value": "Linda McMahon"
},
{
"operation": "Now on",
"time": "16.443",
"value": "Linda McMahon"
},
{
"operation": "hyperlink",
"time": "51.801",
"value": "4.2 2010 U.S. Senate campaign Linda McMahon_section_16"
},
{
"operation": "Highlighting",
"time": "92.894",
"value": "Linda McMahon_sentence_139"
},
{
"operation": "Highlighting",
"time": "110.374",
"value": "Linda McMahon_sentence_148"
},
{
"operation": "hyperlink",
"time": "129.626",
"value": "2012 United States Senate election in Connecticut"
},
{
"operation": "Now on",
"time": "130.387",
"value": "2012 United States Senate election in Connecticut"
},
{
"operation": "finish",
"time": "183.366",
"value": "finish"
}
]
}
|
66530
|
Hermosa Beach City School District oversees two elementary schools (comes after preschool), the school district is independent.
|
{
"evidence": [
"Hermosa Beach City School District_sentence_1",
"Primary school_sentence_4"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHermosa Beach City School District oversees two elementary schools (comes after preschool), the school district is independent.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Hermosa Beach City School District_sentence_1",
"Primary school_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3731,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.033",
"value": "?search="
},
{
"operation": "search",
"time": "10.057",
"value": "Hermosa Beach City School District"
},
{
"operation": "Now on",
"time": "12.101",
"value": "Hermosa Beach City School District"
},
{
"operation": "Highlighting",
"time": "15.676",
"value": "Hermosa Beach City School District_sentence_1"
},
{
"operation": "Page search",
"time": "28.338",
"value": "independent"
},
{
"operation": "hyperlink",
"time": "78.951",
"value": "elementary schools"
},
{
"operation": "Now on",
"time": "79.966",
"value": "Primary school"
},
{
"operation": "Highlighting",
"time": "83.081",
"value": "Primary school_sentence_4"
},
{
"operation": "finish",
"time": "103.63",
"value": "finish"
}
]
}
|
exfever_6972
|
Andrea Joy Cook was the most successful Canadian actress.
|
{
"evidence": [
"A. J. Cook",
"The Virgin Suicides (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:\nAndrea Joy Cook was the most successful Canadian actress.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"A. J. Cook",
"The Virgin Suicides (film)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3732,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
78762
|
Vadiz was the daughter of Italian actress and show host Meche Marchand, who wanted to enroll her in the Escuela Libre de Música to begin her art studies, and of Puerto Rican trio singer Jose Luis Vadiz Colon.
|
{
"evidence": [
"Yolanda Vadiz_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:\nVadiz was the daughter of Italian actress and show host Meche Marchand, who wanted to enroll her in the Escuela Libre de Música to begin her art studies, and of Puerto Rican trio singer Jose Luis Vadiz Colon.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Yolanda Vadiz_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3733,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.002",
"value": "?search="
},
{
"operation": "search",
"time": "0.014",
"value": "Contains...Vadiz"
},
{
"operation": "Now on",
"time": "0.019",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "0.03",
"value": "Yolanda Vadiz"
},
{
"operation": "Now on",
"time": "0.032",
"value": "Yolanda Vadiz"
},
{
"operation": "Highlighting",
"time": "0.04",
"value": "Yolanda Vadiz_sentence_6"
},
{
"operation": "finish",
"time": "0.068",
"value": "finish"
},
{
"operation": "Now on",
"time": "1622220677.697",
"value": "Yolanda Vadiz"
},
{
"operation": "finish",
"time": "1622220680.84",
"value": "finish"
}
]
}
|
42582
|
Dimmelsvik was the administrative centre of Kninnherad municipality in Norway, the village has a barrel factory but is a small supplier of gravel, sand, and crushed stone.
|
{
"evidence": [
"Dimmelsvik_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:\nDimmelsvik was the administrative centre of Kninnherad municipality in Norway, the village has a barrel factory but is a small supplier of gravel, sand, and crushed stone.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Dimmelsvik_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3734,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.664",
"value": "?search="
},
{
"operation": "search",
"time": "5.289",
"value": "Dimmelsvik"
},
{
"operation": "Now on",
"time": "5.943",
"value": "Dimmelsvik"
},
{
"operation": "Page search",
"time": "11.17",
"value": "cent"
},
{
"operation": "Page search",
"time": "11.441",
"value": "cent"
},
{
"operation": "Page search",
"time": "11.619",
"value": "cent"
},
{
"operation": "Page search",
"time": "11.803",
"value": "cent"
},
{
"operation": "Page search",
"time": "13.095",
"value": "centre"
},
{
"operation": "Highlighting",
"time": "16.584",
"value": "Dimmelsvik_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "23.071",
"value": "Dimmelsvik_sentence_4"
},
{
"operation": "Highlighting",
"time": "23.609",
"value": "Dimmelsvik_sentence_0"
},
{
"operation": "Highlighting",
"time": "24.217",
"value": "Dimmelsvik_sentence_4"
},
{
"operation": "Highlighting",
"time": "43.251",
"value": "Dimmelsvik_sentence_5"
},
{
"operation": "Highlighting",
"time": "54.63",
"value": "Dimmelsvik_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "60.878",
"value": "Dimmelsvik_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "61.195",
"value": "Dimmelsvik_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "62.101",
"value": "Dimmelsvik_sentence_0"
},
{
"operation": "finish",
"time": "66.864",
"value": "finish"
}
]
}
|
70055
|
Benny Faccone started working in the recording industry in Montreal (the second-most populous city in Canada) even before graduating, and became a freelance engineer.
|
{
"evidence": [
"Benny Faccone_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:\nBenny Faccone started working in the recording industry in Montreal (the second-most populous city in Canada) even before graduating, and became a freelance engineer.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Benny Faccone_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3735,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.245",
"value": "?search="
},
{
"operation": "search",
"time": "6.125",
"value": "Benny Faccone"
},
{
"operation": "Now on",
"time": "7.703",
"value": "Benny Faccone"
},
{
"operation": "Highlighting",
"time": "69.244",
"value": "Benny Faccone_sentence_4"
},
{
"operation": "hyperlink",
"time": "77.589",
"value": "Montreal"
},
{
"operation": "Now on",
"time": "79.151",
"value": "Montreal"
},
{
"operation": "Highlighting",
"time": "80.213",
"value": "Montreal_sentence_2"
},
{
"operation": "Highlighting",
"time": "80.215",
"value": "Montreal_title"
},
{
"operation": "back-button-clicked",
"time": "84.14",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "84.293",
"value": "Benny Faccone"
},
{
"operation": "Highlighting",
"time": "96.427",
"value": "Benny Faccone_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "136.163",
"value": "Benny Faccone_sentence_6"
},
{
"operation": "finish",
"time": "154.875",
"value": "finish"
}
]
}
|
exfever_8172
|
Tropic Thunder is a 2008 action-comedy film directed by an Australian actor, comedian, producer, director, and screenwriter with a 2013 American adventure comedy-drama film credited to him.
|
{
"evidence": [
"Tropic Thunder",
"Ben Stiller",
"The Secret Life of Walter Mitty (2013 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:\nTropic Thunder is a 2008 action-comedy film directed by an Australian actor, comedian, producer, director, and screenwriter with a 2013 American adventure comedy-drama film credited to him.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tropic Thunder",
"Ben Stiller",
"The Secret Life of Walter Mitty (2013 film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3736,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_53
|
Morgan Fairchild takes a lead role in the NBC series based on a 1949 movie scripted by Wilder and starring an American actress and a dancer in traveling theatrical companies.
|
{
"evidence": [
"Morgan Fairchild",
"Joan Crawford"
],
"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:\nMorgan Fairchild takes a lead role in the NBC series based on a 1949 movie scripted by Wilder and starring an American actress and a dancer in traveling theatrical companies.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Morgan Fairchild",
"Joan Crawford"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3737,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
3437
|
2010 kicked off one of the most successful sides of Korea's football sides, which was also the 28th season for South Korean professional football club based in Pohang, North Gyeongsang called the Pohang Steelers, that is the joint-most successful team in Asia with three AFC Champions League titles and was originally called Pohang Steelworks, after the Pohang Iron and Steel Company that owned the team and founded it in 1973.
|
{
"evidence": [
"Pohang Steelers_sentence_4",
"Pohang Steelers_sentence_2",
"Pohang Steelers_sentence_3",
"2010 Pohang Steelers season_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:\n2010 kicked off one of the most successful sides of Korea's football sides, which was also the 28th season for South Korean professional football club based in Pohang, North Gyeongsang called the Pohang Steelers, that is the joint-most successful team in Asia with three AFC Champions League titles and was originally called Pohang Steelworks, after the Pohang Iron and Steel Company that owned the team and founded it in 1973.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Pohang Steelers_sentence_4",
"Pohang Steelers_sentence_2",
"Pohang Steelers_sentence_3",
"2010 Pohang Steelers season_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3738,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "41.17",
"value": "28th season for South Korean professional football club"
},
{
"operation": "search",
"time": "138.647",
"value": "2010 south korean professional football club"
},
{
"operation": "search",
"time": "156.014",
"value": "2010 korean football season"
},
{
"operation": "hyperlink",
"time": "163.246",
"value": "Korean League Cup"
},
{
"operation": "hyperlink",
"time": "193.372",
"value": "2010"
},
{
"operation": "search",
"time": "218.791",
"value": "South Korean professional football club based"
},
{
"operation": "search",
"time": "249.782",
"value": "Pohang Steelers"
},
{
"operation": "Page search",
"time": "290.911",
"value": "2010"
},
{
"operation": "Highlighting",
"time": "336.708",
"value": "Pohang Steelers_sentence_4"
},
{
"operation": "Highlighting",
"time": "342.418",
"value": "Pohang Steelers_sentence_2"
},
{
"operation": "Highlighting",
"time": "359.112",
"value": "Pohang Steelers_sentence_3"
},
{
"operation": "search",
"time": "486.15",
"value": "Contains...Pohang Steelers 2010"
},
{
"operation": "hyperlink",
"time": "497.266",
"value": "2010 Pohang Steelers season"
},
{
"operation": "Highlighting",
"time": "502.123",
"value": "2010 Pohang Steelers season_sentence_0"
},
{
"operation": "hyperlink",
"time": "537.343",
"value": "2010 K-League"
},
{
"operation": "hyperlink",
"time": "582.653",
"value": "2010 in South Korean football"
},
{
"operation": "back-button-clicked",
"time": "604.265",
"value": "back-button-clicked"
},
{
"operation": "hyperlink",
"time": "617.448",
"value": "2010 K League Championship"
},
{
"operation": "back-button-clicked",
"time": "625.958",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "626.975",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "626.978",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "626.983",
"value": "back-button-clicked"
},
{
"operation": "hyperlink",
"time": "641.13",
"value": "K-League"
},
{
"operation": "back-button-clicked",
"time": "655.946",
"value": "back-button-clicked"
},
{
"operation": "search",
"time": "753.24",
"value": "2010 kicked off one of the most successful sides of Korea's football sides"
},
{
"operation": "search",
"time": "784.461",
"value": "Pohang Steelers"
},
{
"operation": "hyperlink",
"time": "942.904",
"value": "2010"
},
{
"operation": "back-button-clicked",
"time": "965.104",
"value": "back-button-clicked"
},
{
"operation": "hyperlink",
"time": "976.479",
"value": "2010"
},
{
"operation": "back-button-clicked",
"time": "986.592",
"value": "back-button-clicked"
},
{
"operation": "hyperlink",
"time": "1024.37",
"value": "2010"
},
{
"operation": "back-button-clicked",
"time": "1031.662",
"value": "back-button-clicked"
},
{
"operation": "search",
"time": "1055.525",
"value": "Pohang Steelers season 2010"
},
{
"operation": "finish",
"time": "1122.929",
"value": "finish"
}
]
}
|
6783
|
LGBT rights in Mali are still challenged and LGBT persons still face stigmatization because of the prevailing cultural and religious beliefs of most Mali citizens that view same-sex sexual activity and non-traditional gender roles as immoral, especially since these are also illegal in Mali.
|
{
"evidence": [
"LGBT rights in Mali_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:\nLGBT rights in Mali are still challenged and LGBT persons still face stigmatization because of the prevailing cultural and religious beliefs of most Mali citizens that view same-sex sexual activity and non-traditional gender roles as immoral, especially since these are also illegal in Mali.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"LGBT rights in Mali_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3739,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.799",
"value": "?search="
},
{
"operation": "search",
"time": "16.818",
"value": "LGBT rights in Mali"
},
{
"operation": "Now on",
"time": "17.765",
"value": "LGBT rights in Mali"
},
{
"operation": "Highlighting",
"time": "80.042",
"value": "LGBT rights in Mali_sentence_3"
},
{
"operation": "finish",
"time": "116.49",
"value": "finish"
}
]
}
|
50163
|
The song begins with a sample of the "put your bodies upon the gears" speech address given by Mario Savio, a key figure in the Berkeley Free Speech Movement (a massive, long-lasting student protest), which bear some traces of a commentary on the abuses of the powerful.
|
{
"evidence": [
"Wretches and Kings_sentence_6",
"Free Speech Movement_sentence_0",
"Wretches and Kings_sentence_7"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe song begins with a sample of the \"put your bodies upon the gears\" speech address given by Mario Savio, a key figure in the Berkeley Free Speech Movement (a massive, long-lasting student protest), which bear some traces of a commentary on the abuses of the powerful.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Wretches and Kings_sentence_6",
"Free Speech Movement_sentence_0",
"Wretches and Kings_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3740,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.714",
"value": "?search="
},
{
"operation": "search",
"time": "26.351",
"value": "\"put your bodies upon the gears\""
},
{
"operation": "Now on",
"time": "27.552",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "39.036",
"value": "Wretches and Kings"
},
{
"operation": "Now on",
"time": "39.586",
"value": "Wretches and Kings"
},
{
"operation": "Page search",
"time": "42.535",
"value": "sample"
},
{
"operation": "Highlighting",
"time": "45.18",
"value": "Wretches and Kings_sentence_6"
},
{
"operation": "hyperlink",
"time": "88.818",
"value": "Berkeley Free Speech Movement"
},
{
"operation": "Now on",
"time": "89.663",
"value": "Free Speech Movement"
},
{
"operation": "Highlighting",
"time": "91.869",
"value": "Free Speech Movement_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "92.918",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "92.992",
"value": "Wretches and Kings"
},
{
"operation": "Highlighting",
"time": "97.93",
"value": "Wretches and Kings_sentence_7"
},
{
"operation": "finish",
"time": "165.072",
"value": "finish"
}
]
}
|
exfever_2377
|
Elisabeth Sladen was an English actress best known in the television series Doctor Who as the investigative journalist dedicated to protecting Venus from alien invasions.
|
{
"evidence": [
"Elisabeth Sladen",
"Sarah Jane Smith"
],
"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:\nElisabeth Sladen was an English actress best known in the television series Doctor Who as the investigative journalist dedicated to protecting Venus from alien invasions.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Elisabeth Sladen",
"Sarah Jane Smith"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3741,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6866
|
Sofia Coppola is an American filmmaker and actress who is the youngest child and only daughter of an Afghan film director, producer, and screenwriter.
|
{
"evidence": [
"Sofia Coppola",
"Francis Ford Coppola"
],
"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:\nSofia Coppola is an American filmmaker and actress who is the youngest child and only daughter of an Afghan film director, producer, and screenwriter.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sofia Coppola",
"Francis Ford Coppola"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3742,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1655
|
Minnesota is a constituent political entity that has shared sovereignty in which Americans are citizens, including a federation of states with a republican form of government.
|
{
"evidence": [
"Minnesota",
"U.S. state",
"Federal republic"
],
"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:\nMinnesota is a constituent political entity that has shared sovereignty in which Americans are citizens, including a federation of states with a republican form of government.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Minnesota",
"U.S. state",
"Federal republic"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3743,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1172
|
Marnie is a 1964 American psychological thriller film with screenplay by Jay Presson Allen was based on the 1961 novel of the same name by an English novelist who was the author's pseudonym until he changed his name by a legal document binding on a single person or several persons acting jointly to express an intention or create an obligation.
|
{
"evidence": [
"Marnie (film)",
"Winston Graham",
"Deed poll"
],
"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:\nMarnie is a 1964 American psychological thriller film with screenplay by Jay Presson Allen was based on the 1961 novel of the same name by an English novelist who was the author's pseudonym until he changed his name by a legal document binding on a single person or several persons acting jointly to express an intention or create an obligation.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Marnie (film)",
"Winston Graham",
"Deed poll"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3744,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2200
|
Humanz is the fifth studio album by a British virtual band, formed in 1998.
|
{
"evidence": [
"Humanz",
"Gorillaz"
],
"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:\nHumanz is the fifth studio album by a British virtual band, formed in 1998.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Humanz",
"Gorillaz"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3745,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
79722
|
In the first ever season of a top football league in South Korea, two players received awards for being the best midfielders.
|
{
"evidence": [
"Cho Kwang-rae_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:\nIn the first ever season of a top football league in South Korea, two players received awards for being the best midfielders.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Cho Kwang-rae_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3746,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.376",
"value": "?search="
},
{
"operation": "search",
"time": "33.837",
"value": "Contains...first ever season of a top football league in South Korea"
},
{
"operation": "Now on",
"time": "44.778",
"value": "undefined"
},
{
"operation": "search",
"time": "67.854",
"value": "Contains...top football league in South Korea"
},
{
"operation": "Now on",
"time": "69.392",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "130.355",
"value": "Korean League Cup"
},
{
"operation": "Now on",
"time": "132.032",
"value": "Korean League Cup"
},
{
"operation": "search",
"time": "264.922",
"value": "South Korea best midfielder"
},
{
"operation": "Now on",
"time": "266.261",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "289.019",
"value": "Yoo Sang-chul"
},
{
"operation": "Now on",
"time": "290.622",
"value": "Yoo Sang-chul"
},
{
"operation": "hyperlink",
"time": "397.766",
"value": "2004 Summer Olympics"
},
{
"operation": "Now on",
"time": "399.659",
"value": "Football at the 2004 Summer Olympics – Men's tournament"
},
{
"operation": "search",
"time": "429.836",
"value": "\"top football leagues in South Korea\""
},
{
"operation": "Now on",
"time": "430.898",
"value": "undefined"
},
{
"operation": "search",
"time": "469.639",
"value": "\"two players received awards for being the best midfielders\""
},
{
"operation": "Now on",
"time": "470.629",
"value": "undefined"
},
{
"operation": "search",
"time": "489.255",
"value": "Contains...two players received awards"
},
{
"operation": "Now on",
"time": "491.332",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "529.42",
"value": "next 20"
},
{
"operation": "Now on",
"time": "530.522",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "543.915",
"value": "Awards FIFA World Cup_section_20"
},
{
"operation": "Now on",
"time": "546.367",
"value": "FIFA World Cup#Awards FIFA World Cup section 20"
},
{
"operation": "search",
"time": "726.751",
"value": "FIFA World Cup 2002"
},
{
"operation": "Now on",
"time": "729.349",
"value": "2002 FIFA World Cup"
},
{
"operation": "search",
"time": "918.632",
"value": "\"in the first ever season\""
},
{
"operation": "Now on",
"time": "924.361",
"value": "undefined"
},
{
"operation": "search",
"time": "1065.251",
"value": "South Korean football league"
},
{
"operation": "Now on",
"time": "1066.679",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "1104.968",
"value": "Football in South Korea"
},
{
"operation": "Now on",
"time": "1106.02",
"value": "Football in South Korea"
},
{
"operation": "search",
"time": "1211.83",
"value": "South Korea Best Midfielder list"
},
{
"operation": "Now on",
"time": "1212.867",
"value": "undefined"
},
{
"operation": "search",
"time": "1355.816",
"value": "List of South Korea midfielders"
},
{
"operation": "Now on",
"time": "1357.636",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "1444.947",
"value": "Cho Kwang-rae"
},
{
"operation": "Now on",
"time": "1446.11",
"value": "Cho Kwang-rae"
},
{
"operation": "Highlighting",
"time": "1460.423",
"value": "Cho Kwang-rae_sentence_3"
},
{
"operation": "finish",
"time": "1507.981",
"value": "finish"
}
]
}
|
exfever_4175
|
Saved is a play that was originally refused a license without severe cuts by the most senior officer of the Royal Household of the United States.
|
{
"evidence": [
"Saved (play)",
"Lord Chamberlain"
],
"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:\nSaved is a play that was originally refused a license without severe cuts by the most senior officer of the Royal Household of the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Saved (play)",
"Lord Chamberlain"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3747,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5696
|
The Silence of the Lambs is a 1993 American film with a genre focus on mental, emotional, and psychological states to frighten, disturb, and/or unsettle its audience.
|
{
"evidence": [
"The Silence of the Lambs (film)",
"Psychological horror"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Silence of the Lambs is a 1993 American film with a genre focus on mental, emotional, and psychological states to frighten, disturb, and/or unsettle its audience.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Silence of the Lambs (film)",
"Psychological horror"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3748,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5288
|
Rage Against the Machine is an American jazz band that consists of talented musicians, including an American drummer who formed the supergroup Audioslave with the founder and frontman of Temple of the Dog.?
|
{
"evidence": [
"Rage Against the Machine",
"Brad Wilk",
"Chris Cornell"
],
"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:\nRage Against the Machine is an American jazz band that consists of talented musicians, including an American drummer who formed the supergroup Audioslave with the founder and frontman of Temple of the Dog.?\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Rage Against the Machine",
"Brad Wilk",
"Chris Cornell"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3749,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6089
|
Water is an inorganic chemical substance vital for all, even though it doesn't provide food or any substance used by an organism to survive.
|
{
"evidence": [
"Water",
"Nutrient"
],
"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:\nWater is an inorganic chemical substance vital for all, even though it doesn't provide food or any substance used by an organism to survive.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Water",
"Nutrient"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3750,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
71416
|
Alfatar is a 248.57 km2 town located in the Danubian Plain, in the area of the South Dobrudzha, Northeastern Bulgaria with a total population of 1,714 as of 2009.
|
{
"evidence": [
"Alfatar Municipality_sentence_0",
"Alfatar Municipality_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:\nAlfatar is a 248.57 km2 town located in the Danubian Plain, in the area of the South Dobrudzha, Northeastern Bulgaria with a total population of 1,714 as of 2009.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Alfatar Municipality_sentence_0",
"Alfatar Municipality_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3751,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.046",
"value": "?search="
},
{
"operation": "search",
"time": "11.532",
"value": "Alfatar"
},
{
"operation": "Now on",
"time": "12.231",
"value": "Alfatar"
},
{
"operation": "search",
"time": "99.892",
"value": "Contains...located in the Danubian Plain"
},
{
"operation": "Now on",
"time": "101.429",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "107.162",
"value": "Glavinitsa"
},
{
"operation": "Now on",
"time": "108.347",
"value": "Glavinitsa"
},
{
"operation": "back-button-clicked",
"time": "172.001",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "172.838",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "176.622",
"value": "Gramada Municipality"
},
{
"operation": "Now on",
"time": "177.314",
"value": "Gramada Municipality"
},
{
"operation": "back-button-clicked",
"time": "186.616",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "187.409",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "193.486",
"value": "Makresh Municipality"
},
{
"operation": "Now on",
"time": "194.233",
"value": "Makresh Municipality"
},
{
"operation": "back-button-clicked",
"time": "197.231",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "198.052",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "203.508",
"value": "Ruzhintsi Municipality"
},
{
"operation": "Now on",
"time": "204.532",
"value": "Ruzhintsi Municipality"
},
{
"operation": "back-button-clicked",
"time": "207.328",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "208.415",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "213.75",
"value": "Loznitsa Municipality"
},
{
"operation": "Now on",
"time": "214.901",
"value": "Loznitsa Municipality"
},
{
"operation": "back-button-clicked",
"time": "219.144",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "219.947",
"value": "undefined"
},
{
"operation": "search",
"time": "235.528",
"value": "Alfatar"
},
{
"operation": "Now on",
"time": "236.771",
"value": "Alfatar"
},
{
"operation": "hyperlink",
"time": "240.723",
"value": "Alfatar Municipality"
},
{
"operation": "Now on",
"time": "243.383",
"value": "Alfatar Municipality"
},
{
"operation": "Highlighting",
"time": "268.622",
"value": "Alfatar Municipality_sentence_0"
},
{
"operation": "Highlighting",
"time": "275.938",
"value": "Alfatar Municipality_sentence_2"
},
{
"operation": "finish",
"time": "287.735",
"value": "finish"
}
]
}
|
62585
|
The 2014 Lamar Cardinals football team had 17 new signers with 7 transfers and 11 from High School.
|
{
"evidence": [
"2014 Lamar Cardinals football team_sentence_10"
],
"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 2014 Lamar Cardinals football team had 17 new signers with 7 transfers and 11 from High School.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"2014 Lamar Cardinals football team_sentence_10"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3752,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.748",
"value": "?search="
},
{
"operation": "search",
"time": "11.977",
"value": "2014 Lamar Cardinals football team"
},
{
"operation": "Now on",
"time": "18.512",
"value": "2014 Lamar Cardinals football team"
},
{
"operation": "Highlighting",
"time": "42.107",
"value": "2014 Lamar Cardinals football team_sentence_10"
},
{
"operation": "finish",
"time": "55.968",
"value": "finish"
}
]
}
|
exfever_3216
|
Nicole Scherzinger found modest success with Eden's Crush, a girl group created through an international reality television franchise that also produced groups such as an English-Irish pop girl group.
|
{
"evidence": [
"Nicole Scherzinger",
"Popstars",
"Girls Aloud"
],
"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:\nNicole Scherzinger found modest success with Eden's Crush, a girl group created through an international reality television franchise that also produced groups such as an English-Irish pop girl group.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Nicole Scherzinger",
"Popstars",
"Girls Aloud"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3753,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1570
|
Hugh Michael Jackman is an Australian actor best known for his films X-Men and the 2004 action gothic horror film written and directed by Stephen Sommers.
|
{
"evidence": [
"Hugh Jackman",
"Van Helsing (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:\nHugh Michael Jackman is an Australian actor best known for his films X-Men and the 2004 action gothic horror film written and directed by Stephen Sommers.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hugh Jackman",
"Van Helsing (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3754,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
95451
|
Hemiscyllium ocellatum species has no ability to venture on land.
|
{
"evidence": [
"Epaulette shark_sentence_93"
],
"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:\nHemiscyllium ocellatum species has no ability to venture on land.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Epaulette shark_sentence_93"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3755,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.311",
"value": "?search="
},
{
"operation": "search",
"time": "10.809",
"value": "Hemiscyllium ocellatum"
},
{
"operation": "Now on",
"time": "12.48",
"value": "Epaulette shark"
},
{
"operation": "Highlighting",
"time": "44.634",
"value": "Epaulette shark_sentence_0"
},
{
"operation": "Highlighting",
"time": "87.579",
"value": "Epaulette shark_sentence_5"
},
{
"operation": "Page search",
"time": "96.496",
"value": "land"
},
{
"operation": "Highlighting",
"time": "151.381",
"value": "Epaulette shark_sentence_93"
},
{
"operation": "finish",
"time": "167.634",
"value": "finish"
}
]
}
|
38551
|
One of the few marathons in the world that starts in one country and finishes in another is the Niagara Falls International Marathon, and in 1980 Peter Pfitzinger (USA) got the shortest time (2:17:10) in this marathon.
|
{
"evidence": [
"Niagara Falls International Marathon_sentence_1",
"Niagara Falls International Marathon_sentence_10"
],
"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:\nOne of the few marathons in the world that starts in one country and finishes in another is the Niagara Falls International Marathon, and in 1980 Peter Pfitzinger (USA) got the shortest time (2:17:10) in this marathon.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Niagara Falls International Marathon_sentence_1",
"Niagara Falls International Marathon_sentence_10"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3756,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.914",
"value": "undefined"
},
{
"operation": "search",
"time": "12.011",
"value": "Niagara Falls International Marathon"
},
{
"operation": "Now on",
"time": "13.646",
"value": "Niagara Falls International Marathon"
},
{
"operation": "Highlighting",
"time": "20.885",
"value": "Niagara Falls International Marathon_sentence_1"
},
{
"operation": "Highlighting",
"time": "28.797",
"value": "Niagara Falls International Marathon_sentence_10"
},
{
"operation": "finish",
"time": "53.71",
"value": "finish"
}
]
}
|
exfever_3200
|
Troy Baker was the most famous American voice actor.
|
{
"evidence": [
"Troy Baker",
"Tales from the Borderlands"
],
"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:\nTroy Baker was the most famous American voice actor.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Troy Baker",
"Tales from the Borderlands"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3757,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
56600
|
Jimmy Durante (born 1893) Presents the Lennon Sisters Hour is an English thriller movie written Don Reo.
|
{
"evidence": [
"Jimmy Durante Presents the Lennon Sisters_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:\nJimmy Durante (born 1893) Presents the Lennon Sisters Hour is an English thriller movie written Don Reo.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Jimmy Durante Presents the Lennon Sisters_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3758,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.183",
"value": "?search="
},
{
"operation": "search",
"time": "33.799",
"value": "Jimmy Durante Presents the Lennon Sisters"
},
{
"operation": "Now on",
"time": "34.556",
"value": "Jimmy Durante Presents the Lennon Sisters"
},
{
"operation": "Highlighting",
"time": "41.137",
"value": "Jimmy Durante Presents the Lennon Sisters_sentence_0"
},
{
"operation": "finish",
"time": "52.028",
"value": "finish"
}
]
}
|
16169
|
The Virtual particle's concept arises in the perturbation theory of quantum field theory, where the calculations are often performed using schematic representations known as Feynman diagrams.
|
{
"evidence": [
"Virtual particle_sentence_3",
"Virtual particle_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:\nThe Virtual particle's concept arises in the perturbation theory of quantum field theory, where the calculations are often performed using schematic representations known as Feynman diagrams.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Virtual particle_sentence_3",
"Virtual particle_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3759,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.599",
"value": "?search="
},
{
"operation": "search",
"time": "15.762",
"value": "Contains...The Virtual particle's"
},
{
"operation": "Now on",
"time": "16.815",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "18.943",
"value": "the virtual particle"
},
{
"operation": "Now on",
"time": "19.778",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "23.319",
"value": "Virtual particle"
},
{
"operation": "Now on",
"time": "24.009",
"value": "Virtual particle"
},
{
"operation": "Highlighting",
"time": "39.64",
"value": "Virtual particle_sentence_3"
},
{
"operation": "Highlighting",
"time": "57.527",
"value": "Virtual particle_sentence_4"
},
{
"operation": "finish",
"time": "76.02",
"value": "finish"
}
]
}
|
exfever_6863
|
Mariah Carey was the first artist to have her first five singles reach number one on the Billboard Hot 100, from "Vision of Love" to a song released on August 13, 1991 that musically it is a genre of Christian music and R&B.
|
{
"evidence": [
"Mariah Carey",
"Gospel 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:\nMariah Carey was the first artist to have her first five singles reach number one on the Billboard Hot 100, from \"Vision of Love\" to a song released on August 13, 1991 that musically it is a genre of Christian music and R&B.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mariah Carey",
"Gospel music"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3760,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5879
|
Mediterranean cuisine is the food and methods of preparation used by the people of the Mediterranean Basin; writers define the three core elements of the cuisine as the olive, wheat, and the grape, yielding olive oil, bread and a type of food typically made from an unleavened dough of wheat flour mixed with water or eggs, formed into sheets or other shapes, then cooked by boiling or baking.
|
{
"evidence": [
"Mediterranean cuisine",
"Pasta"
],
"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:\nMediterranean cuisine is the food and methods of preparation used by the people of the Mediterranean Basin; writers define the three core elements of the cuisine as the olive, wheat, and the grape, yielding olive oil, bread and a type of food typically made from an unleavened dough of wheat flour mixed with water or eggs, formed into sheets or other shapes, then cooked by boiling or baking.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mediterranean cuisine",
"Pasta"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3761,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
78187
|
The Cyprus Regiment, founded in April 1940, included Infantry, Mechanical, Transport and Pack Transport Companies and their mule drivers (a person who transports goods using pack animals, especially mules) served in France, Ethiopia, Palestine and Italy carrying equipment to areas inaccessible to vehicles.
|
{
"evidence": [
"Cyprus Regiment_sentence_3",
"Cyprus Regiment_sentence_4",
"Cyprus Regiment_sentence_5",
"Arriero_sentence_0",
"Cyprus Regiment_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 Cyprus Regiment, founded in April 1940, included Infantry, Mechanical, Transport and Pack Transport Companies and their mule drivers (a person who transports goods using pack animals, especially mules) served in France, Ethiopia, Palestine and Italy carrying equipment to areas inaccessible to vehicles.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Cyprus Regiment_sentence_3",
"Cyprus Regiment_sentence_4",
"Cyprus Regiment_sentence_5",
"Arriero_sentence_0",
"Cyprus Regiment_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3762,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.499",
"value": "undefined"
},
{
"operation": "search",
"time": "11.535",
"value": "Cyprus Regiment"
},
{
"operation": "Now on",
"time": "12.162",
"value": "Cyprus Regiment"
},
{
"operation": "Highlighting",
"time": "29.462",
"value": "Cyprus Regiment_sentence_3"
},
{
"operation": "Highlighting",
"time": "37.611",
"value": "Cyprus Regiment_sentence_4"
},
{
"operation": "Highlighting",
"time": "63.813",
"value": "Cyprus Regiment_sentence_5"
},
{
"operation": "hyperlink",
"time": "67.27",
"value": "mule drivers"
},
{
"operation": "Now on",
"time": "67.862",
"value": "Arriero"
},
{
"operation": "Highlighting",
"time": "69.816",
"value": "Arriero_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "72.296",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "72.363",
"value": "Cyprus Regiment"
},
{
"operation": "Highlighting",
"time": "79.203",
"value": "Cyprus Regiment_sentence_6"
},
{
"operation": "finish",
"time": "189.038",
"value": "finish"
}
]
}
|
11197
|
The United Nations Security Council Resolution 775, which discusses Somalia, was passed and adopted on August 1992 by the United Nations Security Council, which is in charge of insuring international peace and security.
|
{
"evidence": [
"United Nations Security Council Resolution 775_sentence_0",
"United Nations Security Council_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:\nThe United Nations Security Council Resolution 775, which discusses Somalia, was passed and adopted on August 1992 by the United Nations Security Council, which is in charge of insuring international peace and security.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"United Nations Security Council Resolution 775_sentence_0",
"United Nations Security Council_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3763,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.555",
"value": "?search="
},
{
"operation": "search",
"time": "7.245",
"value": "United Nations Security Council Resolution 775"
},
{
"operation": "Now on",
"time": "8.168",
"value": "United Nations Security Council Resolution 775"
},
{
"operation": "Page search",
"time": "12.125",
"value": "Somalia"
},
{
"operation": "page-search-reset",
"time": "28.255",
"value": "page-search-reset"
},
{
"operation": "Highlighting",
"time": "43.314",
"value": "United Nations Security Council Resolution 775_sentence_0"
},
{
"operation": "Page search",
"time": "59.864",
"value": "security"
},
{
"operation": "Page search",
"time": "63.961",
"value": "peace"
},
{
"operation": "Highlighting deleted",
"time": "80.501",
"value": "United Nations Security Council Resolution 775_sentence_0"
},
{
"operation": "Highlighting",
"time": "80.821",
"value": "United Nations Security Council Resolution 775_sentence_0"
},
{
"operation": "Page search",
"time": "90.811",
"value": "adopted"
},
{
"operation": "hyperlink",
"time": "103.672",
"value": "Security Council"
},
{
"operation": "Now on",
"time": "105.133",
"value": "United Nations Security Council"
},
{
"operation": "Page search",
"time": "108.677",
"value": "charge"
},
{
"operation": "Page search",
"time": "127.943",
"value": "international peace"
},
{
"operation": "Highlighting",
"time": "129.37",
"value": "United Nations Security Council_sentence_2"
},
{
"operation": "finish",
"time": "133.989",
"value": "finish"
}
]
}
|
exfever_1967
|
The Hunger Games film series is composed of science fiction dystopian films whose genre can also be combined with other film genres, including those in which the protagonist is thrust into a series of events that typically involve violence and physical feats.
|
{
"evidence": [
"The Hunger Games (film series)",
"Adventure film",
"Action 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 Hunger Games film series is composed of science fiction dystopian films whose genre can also be combined with other film genres, including those in which the protagonist is thrust into a series of events that typically involve violence and physical feats.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Hunger Games (film series)",
"Adventure film",
"Action film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3764,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
44283
|
2016–17 UEFA Champions League knockout phase had six teams, one of which was Atlético Madrid Spain.
|
{
"evidence": [
"2016–17 UEFA Champions League knockout phase_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:\n2016–17 UEFA Champions League knockout phase had six teams, one of which was Atlético Madrid Spain.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"2016–17 UEFA Champions League knockout phase_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3765,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.57",
"value": "?search="
},
{
"operation": "search",
"time": "18.454",
"value": "2016-17 UEFA Champions League"
},
{
"operation": "Now on",
"time": "19.794",
"value": "2016–17 UEFA Champions League"
},
{
"operation": "hyperlink",
"time": "24.174",
"value": "6 Knockout phase 2016–17 UEFA Champions League_section_19"
},
{
"operation": "hyperlink",
"time": "37.693",
"value": "2016–17 UEFA Champions League knockout phase"
},
{
"operation": "Now on",
"time": "38.208",
"value": "2016–17 UEFA Champions League knockout phase"
},
{
"operation": "Highlighting",
"time": "43.262",
"value": "2016–17 UEFA Champions League knockout phase_sentence_1"
},
{
"operation": "finish",
"time": "66.508",
"value": "finish"
}
]
}
|
exfever_8791
|
Dame Jean Iris Murdoch was an Irish and British novelist and philosopher, best known for her songs about the very common dichotomy of two opposing principles but not a reality.
|
{
"evidence": [
"Iris Murdoch",
"Good and evil"
],
"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:\nDame Jean Iris Murdoch was an Irish and British novelist and philosopher, best known for her songs about the very common dichotomy of two opposing principles but not a reality.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Iris Murdoch",
"Good and evil"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3766,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3145
|
Burns rose to fame as Sally Hardesty in The Texas Chain Saw Massacre film which seeks to elicit fear, a stylistic or thematic category for motion pictures based on similarities either in the narrative elements, aesthetic approach, or the emotional response to the film.
|
{
"evidence": [
"Marilyn Burns",
"Horror film",
"Film genre"
],
"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:\nBurns rose to fame as Sally Hardesty in The Texas Chain Saw Massacre film which seeks to elicit fear, a stylistic or thematic category for motion pictures based on similarities either in the narrative elements, aesthetic approach, or the emotional response to the film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Marilyn Burns",
"Horror film",
"Film genre"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3767,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_245
|
Harold Allen Ramis was an Italian director who directed the 1983 American road comedy film starring Chevy Chase, John Candy, and Christie Brinkley
|
{
"evidence": [
"Harold Ramis",
"National Lampoon's Vacation"
],
"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:\nHarold Allen Ramis was an Italian director who directed the 1983 American road comedy film starring Chevy Chase, John Candy, and Christie Brinkley\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Harold Ramis",
"National Lampoon's Vacation"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3768,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
87971
|
String Quartet in Four Parts was completed by Cage in 1950 in Paris dedicating the piece to his parents and was premiered on August 12, 1950 at the Black Mountain College.
|
{
"evidence": [
"String Quartet in Four Parts_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:\nString Quartet in Four Parts was completed by Cage in 1950 in Paris dedicating the piece to his parents and was premiered on August 12, 1950 at the Black Mountain College.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"String Quartet in Four Parts_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3769,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.074",
"value": "?search="
},
{
"operation": "search",
"time": "10.714",
"value": "String Quartet in Four Parts"
},
{
"operation": "Now on",
"time": "11.417",
"value": "String Quartet in Four Parts"
},
{
"operation": "Highlighting",
"time": "14.825",
"value": "String Quartet in Four Parts_sentence_0"
},
{
"operation": "Highlighting",
"time": "19.866",
"value": "String Quartet in Four Parts_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "29.594",
"value": "String Quartet in Four Parts_sentence_0"
},
{
"operation": "finish",
"time": "38.5",
"value": "finish"
}
]
}
|
45936
|
Bavasi competed with Frank Cashen, whose previous experience was with the Baltimore Orioles, for the position of general manager of the expansion Toronto Blue Jays (a member club of the American League East Division); he was appointed on June 18, 1976.
|
{
"evidence": [
"Peter Bavasi_sentence_16",
"Peter Bavasi_sentence_19"
],
"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:\nBavasi competed with Frank Cashen, whose previous experience was with the Baltimore Orioles, for the position of general manager of the expansion Toronto Blue Jays (a member club of the American League East Division); he was appointed on June 18, 1976.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Peter Bavasi_sentence_16",
"Peter Bavasi_sentence_19"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3770,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.72",
"value": "?search="
},
{
"operation": "search",
"time": "21.363",
"value": "Bavasi"
},
{
"operation": "Now on",
"time": "22.086",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "25.54",
"value": "Bill Bavasi"
},
{
"operation": "Now on",
"time": "26.162",
"value": "Bill Bavasi"
},
{
"operation": "Page search",
"time": "43.395",
"value": "Frank Cashen"
},
{
"operation": "Page search",
"time": "63.616",
"value": "Toronto Blue Jays"
},
{
"operation": "search",
"time": "139.319",
"value": "Frank Cashen"
},
{
"operation": "Now on",
"time": "140.05",
"value": "Frank Cashen"
},
{
"operation": "Page search",
"time": "146.001",
"value": "Bavasi"
},
{
"operation": "Page search",
"time": "160.913",
"value": "Toronto Blue Jays"
},
{
"operation": "search",
"time": "198.527",
"value": "Bill Bavasi"
},
{
"operation": "Now on",
"time": "199.279",
"value": "Bill Bavasi"
},
{
"operation": "hyperlink",
"time": "223.02",
"value": "Buzzie Bavasi"
},
{
"operation": "Now on",
"time": "223.838",
"value": "Buzzie Bavasi"
},
{
"operation": "Page search",
"time": "231.294",
"value": "Frank Cashen"
},
{
"operation": "back-button-clicked",
"time": "247.466",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "247.624",
"value": "Bill Bavasi"
},
{
"operation": "hyperlink",
"time": "249.953",
"value": "Peter Bavasi"
},
{
"operation": "Now on",
"time": "250.385",
"value": "Peter Bavasi"
},
{
"operation": "Page search",
"time": "257.837",
"value": "Frank Cashen"
},
{
"operation": "Highlighting",
"time": "264.507",
"value": "Peter Bavasi_sentence_16"
},
{
"operation": "Highlighting",
"time": "285.872",
"value": "Peter Bavasi_sentence_19"
},
{
"operation": "finish",
"time": "294.679",
"value": "finish"
}
]
}
|
exfever_7204
|
Sausage Party is a 2016 adult computer-animated comedy film whose writers include a Cambodian producer.
|
{
"evidence": [
"Sausage Party",
"Evan Goldberg"
],
"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:\nSausage Party is a 2016 adult computer-animated comedy film whose writers include a Cambodian producer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sausage Party",
"Evan Goldberg"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3771,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1859
|
Freddie Highmore is the main Actor of the series The Good Dortor.
|
{
"evidence": [
"In Too Deep (1999 film)",
"Hill Harper"
],
"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:\nFreddie Highmore is the main Actor of the series The Good Dortor.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"In Too Deep (1999 film)",
"Hill Harper"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3772,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3160
|
Robert Oppenheimer was Project Y's first director, succeeded by an American physicist in charge of the final assembly of "The Gadget," detonated in July 1946 for the test of the first detonation of a nuclear weapon.
|
{
"evidence": [
"Project Y",
"Norris Bradbury",
"Trinity (nuclear test)"
],
"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:\nRobert Oppenheimer was Project Y's first director, succeeded by an American physicist in charge of the final assembly of \"The Gadget,\" detonated in July 1946 for the test of the first detonation of a nuclear weapon.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Project Y",
"Norris Bradbury",
"Trinity (nuclear test)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3773,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
44667
|
Sepia kiensis, deposited at The Natural History Museum in France, grows to a mantle length of 37 mm.
|
{
"evidence": [
"Sepia kiensis_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:\nSepia kiensis, deposited at The Natural History Museum in France, grows to a mantle length of 37 mm.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sepia kiensis_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3774,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.189",
"value": "?search="
},
{
"operation": "search",
"time": "20.605",
"value": "Sepia kiensis"
},
{
"operation": "Now on",
"time": "21.726",
"value": "Sepia kiensis"
},
{
"operation": "Highlighting",
"time": "62.623",
"value": "Sepia kiensis_sentence_4"
},
{
"operation": "finish",
"time": "147.448",
"value": "finish"
}
]
}
|
28410
|
Cases of small intestine neuroendocrine tumor can be diagnosed by urine sampling.
|
{
"evidence": [
"Small intestine neuroendocrine tumor_sentence_5",
"Small intestine neuroendocrine tumor_sentence_6",
"Small intestine neuroendocrine tumor_sentence_7"
],
"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:\nCases of small intestine neuroendocrine tumor can be diagnosed by urine sampling.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Small intestine neuroendocrine tumor_sentence_5",
"Small intestine neuroendocrine tumor_sentence_6",
"Small intestine neuroendocrine tumor_sentence_7"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3775,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.218",
"value": "?search="
},
{
"operation": "search",
"time": "15.076",
"value": "Small intestine neuroendocrine tumor"
},
{
"operation": "Now on",
"time": "16.365",
"value": "Small intestine neuroendocrine tumor"
},
{
"operation": "Page search",
"time": "47.539",
"value": "urine"
},
{
"operation": "Highlighting",
"time": "59.157",
"value": "Small intestine neuroendocrine tumor_sentence_5"
},
{
"operation": "Highlighting",
"time": "64.191",
"value": "Small intestine neuroendocrine tumor_sentence_6"
},
{
"operation": "hyperlink",
"time": "127.147",
"value": "Tachykinin peptides"
},
{
"operation": "Now on",
"time": "128.384",
"value": "Tachykinin peptides"
},
{
"operation": "Page search",
"time": "134.984",
"value": "urine"
},
{
"operation": "back-button-clicked",
"time": "148.9",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "149.078",
"value": "Small intestine neuroendocrine tumor"
},
{
"operation": "Highlighting",
"time": "179.69",
"value": "Small intestine neuroendocrine tumor_sentence_7"
},
{
"operation": "finish",
"time": "184.929",
"value": "finish"
}
]
}
|
exfever_4534
|
Transformers is a science fiction action film whose executive producer is a film director, producer, and screenwriter who was born in Cincinnati, in a state in the midwestern region of the United Kingdom.
|
{
"evidence": [
"Transformers (film)",
"Steven Spielberg",
"Ohio"
],
"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:\nTransformers is a science fiction action film whose executive producer is a film director, producer, and screenwriter who was born in Cincinnati, in a state in the midwestern region of the United Kingdom.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Transformers (film)",
"Steven Spielberg",
"Ohio"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3776,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
95246
|
Open Bible Standard Churches can be traced to the Azusa Street Revival, which took place in Los Angeles on April 9, 1906, and the early teachings of Martin Luther.
|
{
"evidence": [
"Open Bible Standard Churches_sentence_8",
"Azusa Street Revival_sentence_0",
"Azusa Street Revival_sentence_2",
"Azusa Street Revival_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:\nOpen Bible Standard Churches can be traced to the Azusa Street Revival, which took place in Los Angeles on April 9, 1906, and the early teachings of Martin Luther.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Open Bible Standard Churches_sentence_8",
"Azusa Street Revival_sentence_0",
"Azusa Street Revival_sentence_2",
"Azusa Street Revival_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3777,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.414",
"value": "?search="
},
{
"operation": "search",
"time": "15.856",
"value": "Open Bible Standard Churches"
},
{
"operation": "Now on",
"time": "16.738",
"value": "Open Bible Standard Churches"
},
{
"operation": "Highlighting",
"time": "39.615",
"value": "Open Bible Standard Churches_sentence_8"
},
{
"operation": "hyperlink",
"time": "66.148",
"value": "Azusa Street Revival"
},
{
"operation": "Now on",
"time": "67.659",
"value": "Azusa Street Revival"
},
{
"operation": "Highlighting",
"time": "73.629",
"value": "Azusa Street Revival_sentence_0"
},
{
"operation": "Highlighting",
"time": "76.312",
"value": "Azusa Street Revival_sentence_2"
},
{
"operation": "back-button-clicked",
"time": "149.667",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "149.738",
"value": "Open Bible Standard Churches"
},
{
"operation": "hyperlink",
"time": "304.321",
"value": "Azusa Street Revival"
},
{
"operation": "Now on",
"time": "304.999",
"value": "Azusa Street Revival"
},
{
"operation": "Highlighting",
"time": "316.605",
"value": "Azusa Street Revival_sentence_1"
},
{
"operation": "finish",
"time": "389.043",
"value": "finish"
}
]
}
|
exfever_4405
|
Klute is a 1978 American genre of film that originally flourished during the post-World War II era in the United States.
|
{
"evidence": [
"Klute",
"Neo-noir"
],
"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:\nKlute is a 1978 American genre of film that originally flourished during the post-World War II era in the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Klute",
"Neo-noir"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3778,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
6849
|
Bertram Goodhue moved to Boston Massachusetts, where he was befriended by a group of young, artistic intellectuals including Charles Eliot Norton of Harvard University and Ernest Fenollosa of the Boston Museum of Fine Arts who were involved in the founding of the Society of Arts and Crafts – Boston in 1897 and it was also through this group that Goodhue met Ralph Adams Cram, who would be his business partner for almost 25 years.
|
{
"evidence": [
"Bertram Goodhue_sentence_9",
"Bertram Goodhue_sentence_10",
"Bertram Goodhue_sentence_11"
],
"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:\nBertram Goodhue moved to Boston Massachusetts, where he was befriended by a group of young, artistic intellectuals including Charles Eliot Norton of Harvard University and Ernest Fenollosa of the Boston Museum of Fine Arts who were involved in the founding of the Society of Arts and Crafts – Boston in 1897 and it was also through this group that Goodhue met Ralph Adams Cram, who would be his business partner for almost 25 years.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bertram Goodhue_sentence_9",
"Bertram Goodhue_sentence_10",
"Bertram Goodhue_sentence_11"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 3779,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.852",
"value": "?search="
},
{
"operation": "search",
"time": "5.442",
"value": "bertram goodhue"
},
{
"operation": "Now on",
"time": "6.663",
"value": "Bertram Goodhue"
},
{
"operation": "Highlighting",
"time": "13.695",
"value": "Bertram Goodhue_sentence_9"
},
{
"operation": "Highlighting",
"time": "18.474",
"value": "Bertram Goodhue_sentence_10"
},
{
"operation": "Highlighting",
"time": "35.744",
"value": "Bertram Goodhue_sentence_11"
},
{
"operation": "finish",
"time": "41.906",
"value": "finish"
}
]
}
|
exfever_8631
|
Womb's" only female star is Eva Green.
|
{
"evidence": [
"Womb (film)",
"Benedek Fliegauf"
],
"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:\nWomb's\" only female star is Eva Green.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Womb (film)",
"Benedek Fliegauf"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3780,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7786
|
The Spanish language is credited as the least divergent from it, together with a Romance language, considered as the language that is closest to Latin among all its genealogical descendants.
|
{
"evidence": [
"Italian language",
"Sardinian language"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Spanish language is credited as the least divergent from it, together with a Romance language, considered as the language that is closest to Latin among all its genealogical descendants.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Italian language",
"Sardinian language"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 3781,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3150
|
Keeping Up with the Joneses was also nominated for the Teen Choice Awards.
|
{
"evidence": [
"Keeping Up with the Joneses (film)",
"20th Century Fox",
"Film studio"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nKeeping Up with the Joneses was also nominated for the Teen Choice Awards.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Keeping Up with the Joneses (film)",
"20th Century Fox",
"Film studio"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3782,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_870
|
Christopher Nolan is the oldest British-American film director.
|
{
"evidence": [
"Christopher Nolan",
"Memento (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:\nChristopher Nolan is the oldest British-American film director.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Christopher Nolan",
"Memento (film)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 3783,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.