id
stringlengths 1
12
| question
stringlengths 22
566
| golden_answers
dict | data_source
stringclasses 2
values | prompt
listlengths 1
1
| ability
stringclasses 6
values | reward_model
dict | extra_info
dict | metadata
dict |
|---|---|---|---|---|---|---|---|---|
exfever_7007
|
Frey was the highest-paid musician in the Eagles' band.
|
{
"evidence": [
"Glenn Frey",
"Lyin' Eyes"
],
"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:\nFrey was the highest-paid musician in the Eagles' band.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Glenn Frey",
"Lyin' Eyes"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1384,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3797
|
Ireland is the largest developed country in north-western Europe.
|
{
"evidence": [
"Republic of Ireland",
"Dublin",
"Leinster"
],
"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:\nIreland is the largest developed country in north-western Europe.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Republic of Ireland",
"Dublin",
"Leinster"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1385,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5328
|
The Big Chill is a 1983 French comedy-drama film directed by the co-writer of the Star Wars films.
|
{
"evidence": [
"The Big Chill (film)",
"Lawrence Kasdan"
],
"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 Big Chill is a 1983 French comedy-drama film directed by the co-writer of the Star Wars films.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Big Chill (film)",
"Lawrence Kasdan"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1386,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
78169
|
In 1628, Willem van Haecht destroyed the art collection owned by Cornelis van der Geest, a spice merchant from Antwerp, who used his wealth to support the Antwerp artists and to grow his art collection.
|
{
"evidence": [
"Cornelis van der Geest_sentence_0",
"Willem van Haecht_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn 1628, Willem van Haecht destroyed the art collection owned by Cornelis van der Geest, a spice merchant from Antwerp, who used his wealth to support the Antwerp artists and to grow his art collection.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Cornelis van der Geest_sentence_0",
"Willem van Haecht_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1387,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.89",
"value": "?search="
},
{
"operation": "search",
"time": "170.044",
"value": "Willem van Haecht"
},
{
"operation": "Now on",
"time": "170.925",
"value": "Willem van Haecht"
},
{
"operation": "search",
"time": "214.857",
"value": "Cornelis van der Geest"
},
{
"operation": "Now on",
"time": "215.885",
"value": "Cornelis van der Geest"
},
{
"operation": "Highlighting",
"time": "257.196",
"value": "Cornelis van der Geest_sentence_0"
},
{
"operation": "hyperlink",
"time": "293.083",
"value": "Willem van Haecht"
},
{
"operation": "Now on",
"time": "293.988",
"value": "Willem van Haecht"
},
{
"operation": "search",
"time": "328.652",
"value": "Contains...1628, Willem van Haecht destroyed the art collection owned"
},
{
"operation": "Now on",
"time": "330.929",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "335.495",
"value": "1621 willem van haecht destroyer the art collection owned"
},
{
"operation": "Now on",
"time": "336.094",
"value": "undefined"
},
{
"operation": "search",
"time": "355.316",
"value": "Willem van Haecht"
},
{
"operation": "Now on",
"time": "356.012",
"value": "Willem van Haecht"
},
{
"operation": "Highlighting",
"time": "360.032",
"value": "Willem van Haecht_sentence_1"
},
{
"operation": "Highlighting",
"time": "360.034",
"value": "Willem van Haecht_section_0"
},
{
"operation": "Highlighting",
"time": "363.503",
"value": "Willem van Haecht_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "364.263",
"value": "Willem van Haecht_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "364.264",
"value": "Willem van Haecht_section_0"
},
{
"operation": "finish",
"time": "389.661",
"value": "finish"
}
]
}
|
15973
|
Isoforms encoded by BIN1 (neither known as Bridging Integrator-1 nor as Amphiphysin-2), expressed in the central nervous system, may be involved in synaptic vesicle endocytosis; while isoforms, expressed in muscle, localize to the cytoplasm and nucleus and activate a caspase-independent apoptotic process.
|
{
"evidence": [
"BIN1_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:\nIsoforms encoded by BIN1 (neither known as Bridging Integrator-1 nor as Amphiphysin-2), expressed in the central nervous system, may be involved in synaptic vesicle endocytosis; while isoforms, expressed in muscle, localize to the cytoplasm and nucleus and activate a caspase-independent apoptotic process.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"BIN1_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1388,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.698",
"value": "?search="
},
{
"operation": "Now on",
"time": "6.202",
"value": "?search="
},
{
"operation": "search",
"time": "22.685",
"value": "BIN1"
},
{
"operation": "Now on",
"time": "23.851",
"value": "BIN1"
},
{
"operation": "Highlighting",
"time": "42.259",
"value": "BIN1_sentence_0"
},
{
"operation": "finish",
"time": "85.062",
"value": "finish"
}
]
}
|
59432
|
Alisic made his professional debut for Utrecht on May 30, 2003; then dropped down a division to sign for Excelsior in 2006, after failing to break into the Utrecht first-team.
|
{
"evidence": [
"Adnan Alisic_sentence_1",
"Adnan Alisic_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:\nAlisic made his professional debut for Utrecht on May 30, 2003; then dropped down a division to sign for Excelsior in 2006, after failing to break into the Utrecht first-team.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Adnan Alisic_sentence_1",
"Adnan Alisic_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1389,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.948",
"value": "?search="
},
{
"operation": "search",
"time": "18.507",
"value": "Contains...Alisic Utrecht"
},
{
"operation": "Now on",
"time": "19.481",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "22.2",
"value": "Adnan Alisic"
},
{
"operation": "Now on",
"time": "22.709",
"value": "Adnan Alisic"
},
{
"operation": "Highlighting",
"time": "28.609",
"value": "Adnan Alisic_sentence_1"
},
{
"operation": "Highlighting",
"time": "38.689",
"value": "Adnan Alisic_sentence_2"
},
{
"operation": "finish",
"time": "66.904",
"value": "finish"
}
]
}
|
57404
|
Gerald FitzGerald, Thomas Butler, and Gilbert Talbot all held the title of "Earl" as peers of Ireland.
|
{
"evidence": [
"Gilbert Talbot, 7th Earl of Shrewsbury_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:\nGerald FitzGerald, Thomas Butler, and Gilbert Talbot all held the title of \"Earl\" as peers of Ireland.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Gilbert Talbot, 7th Earl of Shrewsbury_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1390,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.304",
"value": "?search="
},
{
"operation": "search",
"time": "14.314",
"value": "peers of Ireland"
},
{
"operation": "Now on",
"time": "16.226",
"value": "undefined"
},
{
"operation": "search",
"time": "63.728",
"value": "Gerald FitzGerald, 9th Earl of Kildare"
},
{
"operation": "Now on",
"time": "71.178",
"value": "Gerald FitzGerald, 9th Earl of Kildare"
},
{
"operation": "Page search",
"time": "81.354",
"value": "Gerald FitzGerald Thomas Butler"
},
{
"operation": "search",
"time": "87.386",
"value": "Gerald FitzGerald Thomas Butler"
},
{
"operation": "Now on",
"time": "88.634",
"value": "undefined"
},
{
"operation": "search",
"time": "130.97",
"value": "Gerald FitzGerald Thomas Butler gilbert talbot"
},
{
"operation": "Now on",
"time": "132.98",
"value": "undefined"
},
{
"operation": "search",
"time": "155.581",
"value": "Gerald FitzGerald EArl ireland"
},
{
"operation": "Now on",
"time": "156.663",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "160.298",
"value": "Lord Deputy of Ireland"
},
{
"operation": "Now on",
"time": "160.67",
"value": "Lord Deputy of Ireland"
},
{
"operation": "Highlighting",
"time": "173.365",
"value": "Lord Deputy of Ireland_item_0_7"
},
{
"operation": "Highlighting",
"time": "173.367",
"value": "Lord Deputy of Ireland_title"
},
{
"operation": "Highlighting",
"time": "173.37",
"value": "Lord Deputy of Ireland_section_0"
},
{
"operation": "Highlighting",
"time": "173.371",
"value": "Lord Deputy of Ireland_section_1"
},
{
"operation": "Highlighting deleted",
"time": "193.149",
"value": "Lord Deputy of Ireland_item_0_7"
},
{
"operation": "Highlighting deleted",
"time": "193.153",
"value": "Lord Deputy of Ireland_title"
},
{
"operation": "Highlighting deleted",
"time": "193.156",
"value": "Lord Deputy of Ireland_section_0"
},
{
"operation": "Highlighting deleted",
"time": "193.159",
"value": "Lord Deputy of Ireland_section_1"
},
{
"operation": "hyperlink",
"time": "228.34",
"value": "Kingdom of Ireland"
},
{
"operation": "Now on",
"time": "229.735",
"value": "Kingdom of Ireland"
},
{
"operation": "Page search",
"time": "234.27",
"value": "peers"
},
{
"operation": "search",
"time": "255.609",
"value": "earl ireland"
},
{
"operation": "Now on",
"time": "256.732",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "267.332",
"value": "Lord Deputy of Ireland"
},
{
"operation": "Now on",
"time": "268.389",
"value": "Lord Deputy of Ireland"
},
{
"operation": "search",
"time": "320.56",
"value": "earl peer ireland"
},
{
"operation": "Now on",
"time": "321.63",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "331.935",
"value": "Peerages in the United Kingdom"
},
{
"operation": "Now on",
"time": "333.142",
"value": "Peerages in the United Kingdom"
},
{
"operation": "Page search",
"time": "350.392",
"value": "gerald"
},
{
"operation": "search",
"time": "392.245",
"value": "Gilbert Talbot, 7th Earl of Shrewsbury"
},
{
"operation": "Now on",
"time": "392.9",
"value": "Gilbert Talbot, 7th Earl of Shrewsbury"
},
{
"operation": "Highlighting",
"time": "477.487",
"value": "Gilbert Talbot, 7th Earl of Shrewsbury_sentence_0"
},
{
"operation": "search",
"time": "512.865",
"value": "Thomas Butler, 10th Earl of Ormond"
},
{
"operation": "Now on",
"time": "513.908",
"value": "Thomas Butler, 10th Earl of Ormond"
},
{
"operation": "finish",
"time": "522.011",
"value": "finish"
}
]
}
|
18286
|
Zhongfang County, located in the Hunan province of the People's Republic of China, has a total area of 1,466.54 km and a total population of over 200,000.
|
{
"evidence": [
"Zhongfang County_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:\nZhongfang County, located in the Hunan province of the People's Republic of China, has a total area of 1,466.54 km and a total population of over 200,000.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Zhongfang County_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1391,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.733",
"value": "?search="
},
{
"operation": "search",
"time": "11.875",
"value": "Zhongfang County"
},
{
"operation": "Now on",
"time": "12.501",
"value": "Zhongfang County"
},
{
"operation": "Highlighting",
"time": "41.576",
"value": "Zhongfang County_sentence_5"
},
{
"operation": "finish",
"time": "101.269",
"value": "finish"
}
]
}
|
34138
|
Japan's Shinano, an aircraft carrier built by the Imperial Japanese Navy (IJN) during World War II, had the thickest hangar deck (7.5 in) among some of the World War II era aircraft carriers.
|
{
"evidence": [
"Japanese aircraft carrier Shinano_sentence_0",
"Japanese aircraft carrier Shinano_sentence_37"
],
"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:\nJapan's Shinano, an aircraft carrier built by the Imperial Japanese Navy (IJN) during World War II, had the thickest hangar deck (7.5 in) among some of the World War II era aircraft carriers.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Japanese aircraft carrier Shinano_sentence_0",
"Japanese aircraft carrier Shinano_sentence_37"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1392,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.725",
"value": "?search="
},
{
"operation": "search",
"time": "13.04",
"value": "Shinano"
},
{
"operation": "Now on",
"time": "14.124",
"value": "Shinano"
},
{
"operation": "hyperlink",
"time": "20.042",
"value": "Japanese aircraft carrier Shinano"
},
{
"operation": "Now on",
"time": "20.815",
"value": "Japanese aircraft carrier Shinano"
},
{
"operation": "Highlighting",
"time": "28.107",
"value": "Japanese aircraft carrier Shinano_sentence_0"
},
{
"operation": "Highlighting",
"time": "148.947",
"value": "Japanese aircraft carrier Shinano_sentence_37"
},
{
"operation": "finish",
"time": "206.029",
"value": "finish"
}
]
}
|
57810
|
Alan Bonansea was loaned to Mitre on July 2018 and was subequently loaned again to Altetico de Rafaela (an Argentine sports club from the city of Rafaela) for the next season.
|
{
"evidence": [
"Alan Bonansea_sentence_7",
"Alan Bonansea_sentence_9",
"Atlético de Rafaela_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:\nAlan Bonansea was loaned to Mitre on July 2018 and was subequently loaned again to Altetico de Rafaela (an Argentine sports club from the city of Rafaela) for the next season.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Alan Bonansea_sentence_7",
"Alan Bonansea_sentence_9",
"Atlético de Rafaela_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1393,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.418",
"value": "?search="
},
{
"operation": "search",
"time": "12.852",
"value": "Alan Bonansea"
},
{
"operation": "Now on",
"time": "14.014",
"value": "Alan Bonansea"
},
{
"operation": "Highlighting",
"time": "53.35",
"value": "Alan Bonansea_sentence_7"
},
{
"operation": "Highlighting",
"time": "63.415",
"value": "Alan Bonansea_sentence_9"
},
{
"operation": "hyperlink",
"time": "68.077",
"value": "Atlético de Rafaela"
},
{
"operation": "Now on",
"time": "69.051",
"value": "Atlético de Rafaela"
},
{
"operation": "Highlighting",
"time": "71.058",
"value": "Atlético de Rafaela_sentence_0"
},
{
"operation": "finish",
"time": "78.326",
"value": "finish"
}
]
}
|
61381
|
Tom Lindsay, a former member of the Saracens Academy, made his debut playing for the Wasps in 2008 produced some excellent form towards the end of that season which earned him selection for England Saxons in the 2010 Churchill Cup (which took place in Denver and New York City)..
|
{
"evidence": [
"Tom Lindsay_sentence_1",
"Tom Lindsay_sentence_2",
"2010 Churchill Cup_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:\nTom Lindsay, a former member of the Saracens Academy, made his debut playing for the Wasps in 2008 produced some excellent form towards the end of that season which earned him selection for England Saxons in the 2010 Churchill Cup (which took place in Denver and New York City)..\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Tom Lindsay_sentence_1",
"Tom Lindsay_sentence_2",
"2010 Churchill Cup_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1394,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.883",
"value": "?search="
},
{
"operation": "search",
"time": "7.148",
"value": "Tom Lindsay"
},
{
"operation": "Now on",
"time": "7.77",
"value": "Tom Lindsay"
},
{
"operation": "Highlighting",
"time": "12.466",
"value": "Tom Lindsay_sentence_1"
},
{
"operation": "Highlighting",
"time": "18.576",
"value": "Tom Lindsay_sentence_2"
},
{
"operation": "hyperlink",
"time": "24.78",
"value": "2010 Churchill Cup"
},
{
"operation": "Now on",
"time": "25.577",
"value": "2010 Churchill Cup"
},
{
"operation": "Highlighting",
"time": "26.991",
"value": "2010 Churchill Cup_sentence_0"
},
{
"operation": "finish",
"time": "32.917",
"value": "finish"
}
]
}
|
71046
|
Shian pari, a village in the state of Punjab, had more than 2,000 population in 2001.
|
{
"evidence": [
"Shian pari_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:\nShian pari, a village in the state of Punjab, had more than 2,000 population in 2001.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Shian pari_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1395,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.677",
"value": "?search="
},
{
"operation": "search",
"time": "85.965",
"value": "Shian pari"
},
{
"operation": "Now on",
"time": "87.24",
"value": "Shian pari"
},
{
"operation": "Highlighting",
"time": "89.934",
"value": "Shian pari_sentence_0"
},
{
"operation": "Highlighting",
"time": "94.368",
"value": "Shian pari_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "95.714",
"value": "Shian pari_sentence_0"
},
{
"operation": "finish",
"time": "102.689",
"value": "finish"
}
]
}
|
exfever_2852
|
Ubisoft Entertainment SA's video game franchises include a rhythm game series whose original game was released on a home video game console developed and marketed by Gameloft.
|
{
"evidence": [
"Ubisoft",
"Just Dance (video game series)",
"Wii"
],
"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:\nUbisoft Entertainment SA's video game franchises include a rhythm game series whose original game was released on a home video game console developed and marketed by Gameloft.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ubisoft",
"Just Dance (video game series)",
"Wii"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1396,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
31696
|
Konstantin Khrenov was a Soviet engineer and inventor who in 1932 introduced underwater welding and cutting of metals which was used extensively in World War II by the Soviet Navy.
|
{
"evidence": [
"Konstantin Khrenov_sentence_0",
"Konstantin Khrenov_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:\nKonstantin Khrenov was a Soviet engineer and inventor who in 1932 introduced underwater welding and cutting of metals which was used extensively in World War II by the Soviet Navy.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Konstantin Khrenov_sentence_0",
"Konstantin Khrenov_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1397,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.701",
"value": "?search="
},
{
"operation": "search",
"time": "20.374",
"value": "Konstantin Khrenov"
},
{
"operation": "Now on",
"time": "21.362",
"value": "Konstantin Khrenov"
},
{
"operation": "Highlighting",
"time": "31.064",
"value": "Konstantin Khrenov_sentence_0"
},
{
"operation": "Highlighting",
"time": "34.468",
"value": "Konstantin Khrenov_sentence_1"
},
{
"operation": "finish",
"time": "94.09",
"value": "finish"
}
]
}
|
exfever_1895
|
Julia Scarlett Elizabeth Louis-Dreyfus is one of the most award-winning actresses in American television history, tying an American actress and comedienne whose career spanned more than seven decades for the most acting wins.
|
{
"evidence": [
"Julia Louis-Dreyfus",
"Cloris Leachman"
],
"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:\nJulia Scarlett Elizabeth Louis-Dreyfus is one of the most award-winning actresses in American television history, tying an American actress and comedienne whose career spanned more than seven decades for the most acting wins.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Julia Louis-Dreyfus",
"Cloris Leachman"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1398,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3464
|
Greece is a country in the Southeast of a continent located entirely in the Northern Hemisphere and mostly in the Eastern Hemisphere.
|
{
"evidence": [
"Greece",
"Europe"
],
"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:\nGreece is a country in the Southeast of a continent located entirely in the Northern Hemisphere and mostly in the Eastern Hemisphere.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Greece",
"Europe"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1399,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2683
|
A Christian state is a country that recognizes a form of an Abrahamic monotheistic religion based on the life and teachings of Jesus of Rome as its official religion.
|
{
"evidence": [
"Christian state",
"Christianity"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nA Christian state is a country that recognizes a form of an Abrahamic monotheistic religion based on the life and teachings of Jesus of Rome as its official religion.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Christian state",
"Christianity"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1400,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8135
|
Eat Pray Love is a romantic comedy film starring an award-winning actress of an American comedy-drama film directed by Herbert Ross.
|
{
"evidence": [
"Eat Pray Love",
"Julia Roberts",
"Steel Magnolias"
],
"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:\nEat Pray Love is a romantic comedy film starring an award-winning actress of an American comedy-drama film directed by Herbert Ross.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Eat Pray Love",
"Julia Roberts",
"Steel Magnolias"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1401,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
50466
|
Oomalama is the debut album released on 29 September 1992 under Atlantic Records by Scottish pop rock band Eugenius which received a B+ rating from Entertainment Weekly.
|
{
"evidence": [
"Oomalama_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nOomalama is the debut album released on 29 September 1992 under Atlantic Records by Scottish pop rock band Eugenius which received a B+ rating from Entertainment Weekly.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Oomalama_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1402,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.384",
"value": "?search="
},
{
"operation": "search",
"time": "154.6",
"value": "Oomalama"
},
{
"operation": "Now on",
"time": "219.599",
"value": "?search="
},
{
"operation": "search",
"time": "224.93",
"value": "Oomalama"
},
{
"operation": "Now on",
"time": "226.274",
"value": "Oomalama"
},
{
"operation": "Highlighting",
"time": "552.047",
"value": "Oomalama_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "675.512",
"value": "Oomalama_sentence_0"
},
{
"operation": "Highlighting",
"time": "730.388",
"value": "Oomalama_sentence_0"
},
{
"operation": "finish",
"time": "832.481",
"value": "finish"
}
]
}
|
71521
|
Jørgen Adolf Lier (born in Askim) was a member of Askim municipality and a Norwegian politician for the Labour Party who served as a deputy representative to the Norwegian Parliament from Østfold.
|
{
"evidence": [
"Jørgen Adolf Lier_sentence_2",
"Jørgen Adolf Lier_sentence_0",
"Jørgen Adolf Lier_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:\nJørgen Adolf Lier (born in Askim) was a member of Askim municipality and a Norwegian politician for the Labour Party who served as a deputy representative to the Norwegian Parliament from Østfold.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Jørgen Adolf Lier_sentence_2",
"Jørgen Adolf Lier_sentence_0",
"Jørgen Adolf Lier_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1403,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.374",
"value": "?search="
},
{
"operation": "search",
"time": "8.697",
"value": "Jørgen Adolf Lier"
},
{
"operation": "Now on",
"time": "9.373",
"value": "Jørgen Adolf Lier"
},
{
"operation": "Highlighting",
"time": "70.274",
"value": "Jørgen Adolf Lier_sentence_2"
},
{
"operation": "Highlighting",
"time": "74.909",
"value": "Jørgen Adolf Lier_sentence_0"
},
{
"operation": "Highlighting",
"time": "78.661",
"value": "Jørgen Adolf Lier_sentence_1"
},
{
"operation": "finish",
"time": "88.904",
"value": "finish"
}
]
}
|
exfever_7430
|
Michael Pence?served as a member of one of the two major contemporary political parties in the United Kingdom.
|
{
"evidence": [
"Mike Pence",
"Republican Party (United States)"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMichael Pence?served as a member of one of the two major contemporary political parties in the United Kingdom.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mike Pence",
"Republican Party (United States)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1404,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
38748
|
The use of UAVs by the FBI happened as a result of the 11 September terrorist attacks and remains classified top secret.
|
{
"evidence": [
"Use of UAVs in law enforcement_sentence_39"
],
"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 use of UAVs by the FBI happened as a result of the 11 September terrorist attacks and remains classified top secret.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Use of UAVs in law enforcement_sentence_39"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1405,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.329",
"value": "?search="
},
{
"operation": "Now on",
"time": "16.898",
"value": "?search="
},
{
"operation": "search",
"time": "24.405",
"value": "UAVs"
},
{
"operation": "Now on",
"time": "25.73",
"value": "Unmanned aerial vehicle"
},
{
"operation": "search",
"time": "64.022",
"value": "uavs fbi"
},
{
"operation": "Now on",
"time": "65.164",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "67.233",
"value": "Use of UAVs in law enforcement"
},
{
"operation": "Now on",
"time": "68.18",
"value": "Use of UAVs in law enforcement"
},
{
"operation": "Highlighting",
"time": "86.644",
"value": "Use of UAVs in law enforcement_sentence_39"
},
{
"operation": "finish",
"time": "137.583",
"value": "finish"
}
]
}
|
exfever_1636
|
Jimmy Donal Wales was the most successful American-British Internet entrepreneur.
|
{
"evidence": [
"Jimmy Wales",
"Wikipedia",
"Reference work"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJimmy Donal Wales was the most successful American-British Internet entrepreneur.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jimmy Wales",
"Wikipedia",
"Reference work"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1406,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2178
|
The Catalyst" is a song by American rock band Linkin Park from their fifth studio album produced by Linkin Park vocalist Mike Shinoda and Rick Rubin, who worked together to produce the band's third studio album released on May 14, 2007.
|
{
"evidence": [
"The Catalyst",
"A Thousand Suns",
"Minutes to Midnight (Linkin Park album)"
],
"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 Catalyst\" is a song by American rock band Linkin Park from their fifth studio album produced by Linkin Park vocalist Mike Shinoda and Rick Rubin, who worked together to produce the band's third studio album released on May 14, 2007.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Catalyst",
"A Thousand Suns",
"Minutes to Midnight (Linkin Park album)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1407,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
126
|
During 2010-11 in the Israeli Premier league over 20 teams got new managers, including F.C .Ashdod who replaced Yossi Mizrahi with John Gregory.
|
{
"evidence": [
"2010–11 Israeli Premier League_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:\nDuring 2010-11 in the Israeli Premier league over 20 teams got new managers, including F.C .Ashdod who replaced Yossi Mizrahi with John Gregory.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"2010–11 Israeli Premier League_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1408,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "search",
"time": "0",
"value": "Contains...2010 Israeli Premier league"
},
{
"operation": "hyperlink",
"time": "14.371",
"value": "2010–11 Israeli Premier League"
},
{
"operation": "Highlighting",
"time": "53.15",
"value": "2010–11 Israeli Premier League_sentence_5"
},
{
"operation": "hyperlink",
"time": "89.214",
"value": "F.C. Ashdod"
},
{
"operation": "finish",
"time": "130.011",
"value": "finish"
}
]
}
|
exfever_3164
|
Howl is Allen Ginsberg's most-popular poem.
|
{
"evidence": [
"Allen Ginsberg",
"Beat Generation",
"Human condition"
],
"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:\nHowl is Allen Ginsberg's most-popular poem.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Allen Ginsberg",
"Beat Generation",
"Human condition"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1409,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7704
|
Linux is a Unix-like operating system based on Microsoft Windows, a software where anyone is freely licensed to use, copy, study, and change the software in any way.
|
{
"evidence": [
"Linux",
"Free and open-source software"
],
"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:\nLinux is a Unix-like operating system based on Microsoft Windows, a software where anyone is freely licensed to use, copy, study, and change the software in any way.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Linux",
"Free and open-source software"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1410,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8606
|
NSAI is a not-for-profit trade organization that works to help musicians who professionally compose musical compositions and write lyrics for songs.
|
{
"evidence": [
"Nashville Songwriters Association International",
"Songwriter"
],
"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:\nNSAI is a not-for-profit trade organization that works to help musicians who professionally compose musical compositions and write lyrics for songs.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Nashville Songwriters Association International",
"Songwriter"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1411,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
45046
|
Scotty Bowman is a retired NHL head coach, who lead three different teams, the Montreal Canadiens, Pittsburgh Penguins, and the Detroit Red Wings to Stanley Cup victories nine times.
|
{
"evidence": [
"Scotty Bowman_sentence_4",
"Scotty Bowman_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:\nScotty Bowman is a retired NHL head coach, who lead three different teams, the Montreal Canadiens, Pittsburgh Penguins, and the Detroit Red Wings to Stanley Cup victories nine times.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Scotty Bowman_sentence_4",
"Scotty Bowman_sentence_10"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1412,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.206",
"value": "?search="
},
{
"operation": "search",
"time": "3.511",
"value": "Scotty Bowman"
},
{
"operation": "Now on",
"time": "5.323",
"value": "Scotty Bowman"
},
{
"operation": "Highlighting",
"time": "7.288",
"value": "Scotty Bowman_sentence_4"
},
{
"operation": "Highlighting",
"time": "31.264",
"value": "Scotty Bowman_sentence_10"
},
{
"operation": "finish",
"time": "42.723",
"value": "finish"
}
]
}
|
exfever_2930
|
The American Broadcasting Company, a multinational commercial broadcast radio network, is directly across the street from the corporate headquarters of The Walt Disney Company.
|
{
"evidence": [
"American Broadcasting Company",
"Walt Disney Studios (Burbank)"
],
"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 American Broadcasting Company, a multinational commercial broadcast radio network, is directly across the street from the corporate headquarters of The Walt Disney Company.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"American Broadcasting Company",
"Walt Disney Studios (Burbank)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1413,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
38174
|
The Maryland House of Delegates District 45, home to the Heritage High School and the REACH! Partnership School, is also home to Baltimore's storied Dunbar High School which was named in memory of Paul Laurence Dunbar.
|
{
"evidence": [
"Maryland House of Delegates District 45_sentence_6",
"Maryland House of Delegates District 45_sentence_7",
"Maryland House of Delegates District 45_sentence_2",
"Paul Laurence Dunbar High School (Baltimore, Maryland)_sentence_5"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Maryland House of Delegates District 45, home to the Heritage High School and the REACH! Partnership School, is also home to Baltimore's storied Dunbar High School which was named in memory of Paul Laurence Dunbar.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Maryland House of Delegates District 45_sentence_6",
"Maryland House of Delegates District 45_sentence_7",
"Maryland House of Delegates District 45_sentence_2",
"Paul Laurence Dunbar High School (Baltimore, Maryland)_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1414,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.081",
"value": "?search="
},
{
"operation": "search",
"time": "7.09",
"value": "Maryland House of Delegates District 45"
},
{
"operation": "Now on",
"time": "7.961",
"value": "Maryland House of Delegates District 45"
},
{
"operation": "Page search",
"time": "14.557",
"value": "Heritage High School"
},
{
"operation": "Highlighting",
"time": "15.821",
"value": "Maryland House of Delegates District 45_sentence_6"
},
{
"operation": "Highlighting",
"time": "26.434",
"value": "Maryland House of Delegates District 45_sentence_7"
},
{
"operation": "Page search",
"time": "37.487",
"value": "Heritage Dunbar High SchoolSchool"
},
{
"operation": "Page search",
"time": "39.923",
"value": "Dunbar High School"
},
{
"operation": "Highlighting",
"time": "41.901",
"value": "Maryland House of Delegates District 45_sentence_2"
},
{
"operation": "hyperlink",
"time": "47.244",
"value": "Dunbar High School"
},
{
"operation": "Now on",
"time": "47.909",
"value": "Paul Laurence Dunbar High School (Baltimore, Maryland)"
},
{
"operation": "Page search",
"time": "54.247",
"value": "named"
},
{
"operation": "Highlighting",
"time": "56.487",
"value": "Paul Laurence Dunbar High School (Baltimore, Maryland)_sentence_5"
},
{
"operation": "finish",
"time": "172.378",
"value": "finish"
}
]
}
|
exfever_3818
|
American Idiot is the sixth album by Green Day and was produced by a record producer, musician, and record industry executive who has also worked with an American singer, songwriter, and rapper.
|
{
"evidence": [
"American Idiot",
"Rob Cavallo",
"Kid Rock"
],
"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:\nAmerican Idiot is the sixth album by Green Day and was produced by a record producer, musician, and record industry executive who has also worked with an American singer, songwriter, and rapper.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"American Idiot",
"Rob Cavallo",
"Kid Rock"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1415,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6725
|
Doctor Victor Von Doom is a supervillain appearing in American comic books published by the brand name and primary imprint of Marvel Worldwide Inc.
|
{
"evidence": [
"Doctor Doom",
"Marvel Comics"
],
"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:\nDoctor Victor Von Doom is a supervillain appearing in American comic books published by the brand name and primary imprint of Marvel Worldwide Inc.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Doctor Doom",
"Marvel Comics"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1416,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8675
|
Taipei City rests on the Taipei Basin, which is bounded by the relatively narrow valleys of the Tainan and a river in northern Taiwan that flows through New Taipei and the capital Taipei for 82 kilometers.
|
{
"evidence": [
"Taipei",
"Xindian River"
],
"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:\nTaipei City rests on the Taipei Basin, which is bounded by the relatively narrow valleys of the Tainan and a river in northern Taiwan that flows through New Taipei and the capital Taipei for 82 kilometers.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Taipei",
"Xindian River"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1417,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7133
|
Seville is situated on the lower reaches of the only major navigable river in Spain.
|
{
"evidence": [
"Seville",
"Guadalquivir"
],
"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:\nSeville is situated on the lower reaches of the only major navigable river in Spain.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Seville",
"Guadalquivir"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1418,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6709
|
Ava DuVernay is an American filmmaker who is the?first black woman to direct a live-action film, earning $100 for a 2018 American science fiction adventure film based on Madeleine L'Engle's young adult novel that was first published in 1982.
|
{
"evidence": [
"Ava DuVernay",
"A Wrinkle in Time (2018 film)",
"A Wrinkle in Time"
],
"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:\nAva DuVernay is an American filmmaker who is the?first black woman to direct a live-action film, earning $100 for a 2018 American science fiction adventure film based on Madeleine L'Engle's young adult novel that was first published in 1982.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ava DuVernay",
"A Wrinkle in Time (2018 film)",
"A Wrinkle in Time"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1419,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
51138
|
Milan Ercegan, a Yugoslav sports manager and creator of the Advanced School for Coaches in 1976, published the first book for wrestling coaches.
|
{
"evidence": [
"Milan Ercegan_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:\nMilan Ercegan, a Yugoslav sports manager and creator of the Advanced School for Coaches in 1976, published the first book for wrestling coaches.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Milan Ercegan_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1420,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.456",
"value": "?search="
},
{
"operation": "search",
"time": "34.743",
"value": "Milan Ercegan"
},
{
"operation": "Now on",
"time": "35.973",
"value": "Milan Ercegan"
},
{
"operation": "Highlighting",
"time": "48.781",
"value": "Milan Ercegan_sentence_0"
},
{
"operation": "Highlighting",
"time": "93.468",
"value": "Milan Ercegan_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "107.309",
"value": "Milan Ercegan_sentence_0"
},
{
"operation": "finish",
"time": "154.121",
"value": "finish"
}
]
}
|
exfever_5435
|
Amoxicillin is used to treat a number of bacterial infections such as an inflammatory condition of the lung affecting the small air sac.
|
{
"evidence": [
"Amoxicillin",
"Pneumonia"
],
"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:\nAmoxicillin is used to treat a number of bacterial infections such as an inflammatory condition of the lung affecting the small air sac.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Amoxicillin",
"Pneumonia"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1421,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
60527
|
The television series Secret Saturday by Canadian cartoonist Jay Stephens debuted in the United States on October 3, 2008, the first Cartoon Network original series in high-definition.
|
{
"evidence": [
"The Secret Saturdays_sentence_0",
"The Secret Saturdays_sentence_1",
"The Secret Saturdays_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 television series Secret Saturday by Canadian cartoonist Jay Stephens debuted in the United States on October 3, 2008, the first Cartoon Network original series in high-definition.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"The Secret Saturdays_sentence_0",
"The Secret Saturdays_sentence_1",
"The Secret Saturdays_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1422,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.779",
"value": "?search="
},
{
"operation": "search",
"time": "15.805",
"value": "Secret Saturday"
},
{
"operation": "Now on",
"time": "17.154",
"value": "Secret Saturday"
},
{
"operation": "Highlighting",
"time": "23.168",
"value": "The Secret Saturdays_sentence_0"
},
{
"operation": "Highlighting",
"time": "31.453",
"value": "The Secret Saturdays_sentence_1"
},
{
"operation": "Highlighting",
"time": "46.241",
"value": "The Secret Saturdays_sentence_7"
},
{
"operation": "finish",
"time": "65.799",
"value": "finish"
}
]
}
|
exfever_8872
|
The ancient Proto-Celts were a collection of people that originated in central Europe and that shared similar culture, language and beliefs. Over the years, the Celts migrated.
|
{
"evidence": [
"Portugal",
"Portuguese people",
"Nation"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe ancient Proto-Celts were a collection of people that originated in central Europe and that shared similar culture, language and beliefs. Over the years, the Celts migrated.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Portugal",
"Portuguese people",
"Nation"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1423,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
94754
|
Megalodacne reaches a maximum length of only 12 mm, Megalodacne are large beetles and are mostly found in North America which is entirely within the Northern Hemisphere.
|
{
"evidence": [
"Megalodacne_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:\nMegalodacne reaches a maximum length of only 12 mm, Megalodacne are large beetles and are mostly found in North America which is entirely within the Northern Hemisphere.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Megalodacne_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1424,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.644",
"value": "?search="
},
{
"operation": "Now on",
"time": "20.104",
"value": "?search="
},
{
"operation": "search",
"time": "31.869",
"value": "Megalodacne"
},
{
"operation": "Now on",
"time": "32.93",
"value": "Megalodacne"
},
{
"operation": "Highlighting",
"time": "54.208",
"value": "Megalodacne_sentence_1"
},
{
"operation": "Highlighting",
"time": "90.321",
"value": "Megalodacne_sentence_10"
},
{
"operation": "Highlighting deleted",
"time": "96.771",
"value": "Megalodacne_sentence_10"
},
{
"operation": "finish",
"time": "110.343",
"value": "finish"
}
]
}
|
32033
|
Gustavo Neffa appeared at the 1992 Summer Olympics in the Paraguayan football team, playing a single match against Ghana, then he returned to South America in 1995, and played for Boca Juniors.
|
{
"evidence": [
"Gustavo Neffa_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:\nGustavo Neffa appeared at the 1992 Summer Olympics in the Paraguayan football team, playing a single match against Ghana, then he returned to South America in 1995, and played for Boca Juniors.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Gustavo Neffa_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1425,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.28",
"value": "?search="
},
{
"operation": "search",
"time": "20.122",
"value": "Gustavo Neffa"
},
{
"operation": "Now on",
"time": "20.922",
"value": "Gustavo Neffa"
},
{
"operation": "Highlighting",
"time": "43.214",
"value": "Gustavo Neffa_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "66.918",
"value": "Gustavo Neffa_sentence_6"
},
{
"operation": "Highlighting",
"time": "67.944",
"value": "Gustavo Neffa_sentence_4"
},
{
"operation": "finish",
"time": "79.862",
"value": "finish"
}
]
}
|
17847
|
John Tuigg reached Pittsburgh in December 1849 where he finished his studies at St. Michael's Seminary and was ordained a priest a couple years later.
|
{
"evidence": [
"John Tuigg_sentence_3",
"John Tuigg_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:\nJohn Tuigg reached Pittsburgh in December 1849 where he finished his studies at St. Michael's Seminary and was ordained a priest a couple years later.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"John Tuigg_sentence_3",
"John Tuigg_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1426,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.591",
"value": "?search="
},
{
"operation": "search",
"time": "5.442",
"value": "John Tuigg"
},
{
"operation": "Now on",
"time": "6.118",
"value": "John Tuigg"
},
{
"operation": "Page search",
"time": "8.435",
"value": "Pittsburgh"
},
{
"operation": "page-search-reset",
"time": "11.469",
"value": "page-search-reset"
},
{
"operation": "Page search",
"time": "19.127",
"value": "1849"
},
{
"operation": "Highlighting",
"time": "23.512",
"value": "John Tuigg_sentence_3"
},
{
"operation": "Highlighting",
"time": "33.136",
"value": "John Tuigg_sentence_4"
},
{
"operation": "finish",
"time": "379.153",
"value": "finish"
}
]
}
|
exfever_1889
|
The Semitic languages, which are spoken by more than 400 million people across West Asia, are a branch of a language family of about 300 languages that are spoken on a large peninsula in East Africa, located on the easternmost part of the African mainland.
|
{
"evidence": [
"Semitic languages",
"Afroasiatic languages",
"Horn of Africa"
],
"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 Semitic languages, which are spoken by more than 400 million people across West Asia, are a branch of a language family of about 300 languages that are spoken on a large peninsula in East Africa, located on the easternmost part of the African mainland.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Semitic languages",
"Afroasiatic languages",
"Horn of Africa"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1427,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
14486
|
The Asian Cycling Confederation, a platform for young Malaysian riders to compete and gain valuable experience in a cycling race, is the oldest bicycle race in Malaysia and it is ranked 2.2 class by UCI (Union Cycliste Internationale), the world governing body for sports cycling and oversees international competitive cycling events.
|
{
"evidence": [
"Jelajah Malaysia_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:\nThe Asian Cycling Confederation, a platform for young Malaysian riders to compete and gain valuable experience in a cycling race, is the oldest bicycle race in Malaysia and it is ranked 2.2 class by UCI (Union Cycliste Internationale), the world governing body for sports cycling and oversees international competitive cycling events.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Jelajah Malaysia_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1428,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.665",
"value": "?search="
},
{
"operation": "search",
"time": "7.161",
"value": "asian cycling confederation"
},
{
"operation": "Now on",
"time": "8.272",
"value": "Asian Cycling Confederation"
},
{
"operation": "search",
"time": "33.801",
"value": "The Asian Cycling Confederation"
},
{
"operation": "Now on",
"time": "35.161",
"value": "undefined"
},
{
"operation": "search",
"time": "52.494",
"value": "Asian Cycling Confederation"
},
{
"operation": "Now on",
"time": "53.144",
"value": "Asian Cycling Confederation"
},
{
"operation": "search",
"time": "66.294",
"value": "Asian Cycling Championships"
},
{
"operation": "Now on",
"time": "71.933",
"value": "Asian Cycling Championships"
},
{
"operation": "search",
"time": "95.782",
"value": "Jelajah Malaysia"
},
{
"operation": "Now on",
"time": "96.622",
"value": "Jelajah Malaysia"
},
{
"operation": "Highlighting",
"time": "99.139",
"value": "Jelajah Malaysia_sentence_3"
},
{
"operation": "finish",
"time": "110.085",
"value": "finish"
}
]
}
|
50375
|
Alopece was an asty-deme situated eleven or twelve stadia from the city of Athens
|
{
"evidence": [
"Alopece_sentence_0",
"Alopece_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:\nAlopece was an asty-deme situated eleven or twelve stadia from the city of Athens\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Alopece_sentence_0",
"Alopece_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1429,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.146",
"value": "?search="
},
{
"operation": "search",
"time": "30.228",
"value": "Alopece"
},
{
"operation": "Now on",
"time": "31.212",
"value": "Alopece"
},
{
"operation": "hyperlink",
"time": "44.738",
"value": "deme"
},
{
"operation": "Now on",
"time": "45.785",
"value": "Deme"
},
{
"operation": "back-button-clicked",
"time": "48.165",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "48.28",
"value": "Alopece"
},
{
"operation": "hyperlink",
"time": "50.364",
"value": "asty"
},
{
"operation": "Now on",
"time": "51.55",
"value": "Asty"
},
{
"operation": "back-button-clicked",
"time": "57.344",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "57.448",
"value": "Alopece"
},
{
"operation": "hyperlink",
"time": "63.351",
"value": "deme"
},
{
"operation": "Now on",
"time": "63.984",
"value": "Deme"
},
{
"operation": "back-button-clicked",
"time": "70.641",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "70.754",
"value": "Alopece"
},
{
"operation": "Highlighting",
"time": "81.627",
"value": "Alopece_sentence_0"
},
{
"operation": "Highlighting",
"time": "86.006",
"value": "Alopece_sentence_1"
},
{
"operation": "back-button-clicked",
"time": "105.488",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "105.495",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "105.499",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "105.509",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "105.68",
"value": "?search="
},
{
"operation": "finish",
"time": "127.096",
"value": "finish"
}
]
}
|
exfever_6146
|
The Beach Boys are an American rock band with five original lineup, including an American musician, singer, and songwriter who co-founded the band.
|
{
"evidence": [
"The Beach Boys",
"Carl Wilson"
],
"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 Beach Boys are an American rock band with five original lineup, including an American musician, singer, and songwriter who co-founded the band.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Beach Boys",
"Carl Wilson"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1430,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7261
|
"Two Hearts" is a song first recorded by electronic duo Kish Mauve, which was later re-recorded by Kylie Minogue and released by a German–British record label, founded in Poland in 1896 by the Carl Lindstr?m Company.
|
{
"evidence": [
"2 Hearts (Kylie Minogue song)",
"Parlophone"
],
"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:\n\"Two Hearts\" is a song first recorded by electronic duo Kish Mauve, which was later re-recorded by Kylie Minogue and released by a German–British record label, founded in Poland in 1896 by the Carl Lindstr?m Company.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"2 Hearts (Kylie Minogue song)",
"Parlophone"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1431,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
54096
|
Duke Ferdinand Frederick Augustus of Württemberg was promoted to General-major in 1788, he fought at Belgrade during the Austro-Turkish War and during the War of the First Coalition he led his troops at Neerwinden and was in command of the successful Siege of Condé in 1793.
|
{
"evidence": [
"Duke Ferdinand Frederick Augustus of Württemberg_sentence_2",
"Duke Ferdinand Frederick Augustus of Württemberg_sentence_3"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDuke Ferdinand Frederick Augustus of Württemberg was promoted to General-major in 1788, he fought at Belgrade during the Austro-Turkish War and during the War of the First Coalition he led his troops at Neerwinden and was in command of the successful Siege of Condé in 1793.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Duke Ferdinand Frederick Augustus of Württemberg_sentence_2",
"Duke Ferdinand Frederick Augustus of Württemberg_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1432,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.974",
"value": "?search="
},
{
"operation": "search",
"time": "14.677",
"value": "Duke Ferdinand Frederick Augustus of Württemberg"
},
{
"operation": "Now on",
"time": "15.522",
"value": "Duke Ferdinand Frederick Augustus of Württemberg"
},
{
"operation": "Highlighting",
"time": "25.308",
"value": "Duke Ferdinand Frederick Augustus of Württemberg_sentence_2"
},
{
"operation": "Highlighting",
"time": "37.22",
"value": "Duke Ferdinand Frederick Augustus of Württemberg_sentence_3"
},
{
"operation": "finish",
"time": "76.666",
"value": "finish"
}
]
}
|
33367
|
Low thrust relative orbital transfer uses orbital motion equations (such as the Clohessy-Wiltshire equations that are very useful in planning rendezvous of the chaser with the target, published in 1961) to explain the differences in orbits and displacements of target and chaser spacecraft.
|
{
"evidence": [
"Low thrust relative orbital transfer_sentence_4",
"Low thrust relative orbital transfer_sentence_3",
"Low thrust relative orbital transfer_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLow thrust relative orbital transfer uses orbital motion equations (such as the Clohessy-Wiltshire equations that are very useful in planning rendezvous of the chaser with the target, published in 1961) to explain the differences in orbits and displacements of target and chaser spacecraft.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Low thrust relative orbital transfer_sentence_4",
"Low thrust relative orbital transfer_sentence_3",
"Low thrust relative orbital transfer_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1433,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.233",
"value": "?search="
},
{
"operation": "search",
"time": "67.3",
"value": "Low thrust relative orbital transfer"
},
{
"operation": "Now on",
"time": "68.005",
"value": "Low thrust relative orbital transfer"
},
{
"operation": "Page search",
"time": "80.81",
"value": "uses"
},
{
"operation": "Highlighting",
"time": "93.566",
"value": "Low thrust relative orbital transfer_sentence_3"
},
{
"operation": "Highlighting",
"time": "117.768",
"value": "Low thrust relative orbital transfer_sentence_4"
},
{
"operation": "Page search",
"time": "144.849",
"value": "chaser"
},
{
"operation": "Highlighting deleted",
"time": "185.958",
"value": "Low thrust relative orbital transfer_sentence_3"
},
{
"operation": "Highlighting",
"time": "187.009",
"value": "Low thrust relative orbital transfer_sentence_3"
},
{
"operation": "Highlighting",
"time": "189.599",
"value": "Low thrust relative orbital transfer_sentence_0"
},
{
"operation": "finish",
"time": "218.155",
"value": "finish"
}
]
}
|
80059
|
Bipolar Spectrum Diagnostic Scale patients are diagnosed during their 20s, consists of 19 questions but is unique scale compare to other scales.
|
{
"evidence": [
"Bipolar Spectrum Diagnostic Scale_sentence_4",
"Bipolar Spectrum Diagnostic Scale_sentence_6",
"Bipolar Spectrum Diagnostic Scale_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:\nBipolar Spectrum Diagnostic Scale patients are diagnosed during their 20s, consists of 19 questions but is unique scale compare to other scales.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bipolar Spectrum Diagnostic Scale_sentence_4",
"Bipolar Spectrum Diagnostic Scale_sentence_6",
"Bipolar Spectrum Diagnostic Scale_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1434,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.867",
"value": "?search="
},
{
"operation": "search",
"time": "6.534",
"value": "Bipolar Spectrum Diagnostic Scale"
},
{
"operation": "Now on",
"time": "7.223",
"value": "Bipolar Spectrum Diagnostic Scale"
},
{
"operation": "Highlighting",
"time": "10.426",
"value": "Bipolar Spectrum Diagnostic Scale_sentence_4"
},
{
"operation": "Highlighting",
"time": "12.77",
"value": "Bipolar Spectrum Diagnostic Scale_sentence_6"
},
{
"operation": "Highlighting",
"time": "22.153",
"value": "Bipolar Spectrum Diagnostic Scale_sentence_7"
},
{
"operation": "finish",
"time": "27.171",
"value": "finish"
}
]
}
|
exfever_4015
|
Legion is a fictional character appearing in Japanese comic books published under the brand name and primary imprint of Marvel Worldwide Inc., formerly Marvel Publishing, Inc.
|
{
"evidence": [
"Legion (Marvel Comics)",
"Marvel Comics"
],
"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:\nLegion is a fictional character appearing in Japanese comic books published under the brand name and primary imprint of Marvel Worldwide Inc., formerly Marvel Publishing, Inc.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Legion (Marvel Comics)",
"Marvel Comics"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1435,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
73772
|
Maharaja Theme Park, an amusement park in Tamil Nadu, India (the southernmost part of the Indian subcontinent), was opened to the public in April 2010.
|
{
"evidence": [
"Maharaja Theme Park_sentence_0",
"Tamil Nadu_sentence_2",
"Maharaja Theme Park_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:\nMaharaja Theme Park, an amusement park in Tamil Nadu, India (the southernmost part of the Indian subcontinent), was opened to the public in April 2010.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Maharaja Theme Park_sentence_0",
"Tamil Nadu_sentence_2",
"Maharaja Theme Park_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1436,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.369",
"value": "?search="
},
{
"operation": "search",
"time": "109.845",
"value": "Maharaja Theme Park"
},
{
"operation": "Now on",
"time": "111.951",
"value": "Maharaja Theme Park"
},
{
"operation": "Highlighting",
"time": "114.579",
"value": "Maharaja Theme Park_sentence_0"
},
{
"operation": "hyperlink",
"time": "117.46",
"value": "Tamil Nadu"
},
{
"operation": "Now on",
"time": "119.824",
"value": "Tamil Nadu"
},
{
"operation": "Highlighting",
"time": "124.287",
"value": "Tamil Nadu_sentence_2"
},
{
"operation": "back-button-clicked",
"time": "127.049",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "127.181",
"value": "Maharaja Theme Park"
},
{
"operation": "Highlighting",
"time": "132.919",
"value": "Maharaja Theme Park_sentence_2"
},
{
"operation": "finish",
"time": "140.504",
"value": "finish"
}
]
}
|
14565
|
Nuclear factor erythroid 2-related factor 2 (NRF2), also known as nuclear factor erythroid-derived 2-like 2, encode basic leucine zipper (bZIP) transcription factors, and under normal or unstressed conditions, it is kept in the cytoplasm by a cluster of proteins that degrade it quickly.
|
{
"evidence": [
"NFE2L2_sentence_0",
"NFE2L2_sentence_4",
"NFE2L2_sentence_12",
"NFE2L2_sentence_14"
],
"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:\nNuclear factor erythroid 2-related factor 2 (NRF2), also known as nuclear factor erythroid-derived 2-like 2, encode basic leucine zipper (bZIP) transcription factors, and under normal or unstressed conditions, it is kept in the cytoplasm by a cluster of proteins that degrade it quickly.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"NFE2L2_sentence_0",
"NFE2L2_sentence_4",
"NFE2L2_sentence_12",
"NFE2L2_sentence_14"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1437,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.612",
"value": "?search="
},
{
"operation": "search",
"time": "40.063",
"value": "Nuclear factor erythroid-derived 2-like 2"
},
{
"operation": "Now on",
"time": "41.092",
"value": "NFE2L2"
},
{
"operation": "Highlighting",
"time": "62.076",
"value": "NFE2L2_sentence_0"
},
{
"operation": "Highlighting",
"time": "93.832",
"value": "NFE2L2_sentence_4"
},
{
"operation": "Highlighting",
"time": "115.927",
"value": "NFE2L2_sentence_12"
},
{
"operation": "Highlighting",
"time": "156.368",
"value": "NFE2L2_sentence_14"
},
{
"operation": "finish",
"time": "236.894",
"value": "finish"
}
]
}
|
72760
|
Bolivia had 293,458 visitors coming from Argentina and 271,046 from Peru.
|
{
"evidence": [
"Economy of Bolivia_sentence_143"
],
"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:\nBolivia had 293,458 visitors coming from Argentina and 271,046 from Peru.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Economy of Bolivia_sentence_143"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1438,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.73",
"value": "?search="
},
{
"operation": "search",
"time": "29.159",
"value": "293,458"
},
{
"operation": "Now on",
"time": "30.092",
"value": "undefined"
},
{
"operation": "search",
"time": "57.646",
"value": "bolivian tourism"
},
{
"operation": "Now on",
"time": "58.512",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "62.118",
"value": "Bolivia"
},
{
"operation": "Now on",
"time": "68.022",
"value": "Bolivia"
},
{
"operation": "hyperlink",
"time": "73.045",
"value": "5.2 Tourism Bolivia_section_21"
},
{
"operation": "search",
"time": "95.405",
"value": "Tourism in Bolivia"
},
{
"operation": "Now on",
"time": "96.622",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "103.54",
"value": "Economy of Bolivia"
},
{
"operation": "Now on",
"time": "105.952",
"value": "Economy of Bolivia"
},
{
"operation": "hyperlink",
"time": "120.468",
"value": "3.3.2 Tourism Economy of Bolivia_section_11"
},
{
"operation": "back-button-clicked",
"time": "145.075",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "145.736",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "153.563",
"value": "Tourism in Latin America"
},
{
"operation": "Now on",
"time": "156.624",
"value": "Latin America"
},
{
"operation": "back-button-clicked",
"time": "169.499",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "170.219",
"value": "undefined"
},
{
"operation": "search",
"time": "184.788",
"value": "number of tourists to Bolivia each year"
},
{
"operation": "Now on",
"time": "187.324",
"value": "undefined"
},
{
"operation": "search",
"time": "212.097",
"value": "Argentinian tourists to Bolivia"
},
{
"operation": "Now on",
"time": "213.107",
"value": "undefined"
},
{
"operation": "search",
"time": "240.2",
"value": "visitors coming from Argentina"
},
{
"operation": "Now on",
"time": "241.747",
"value": "undefined"
},
{
"operation": "search",
"time": "275.095",
"value": "Visitors to Bolivia each year"
},
{
"operation": "Now on",
"time": "276.307",
"value": "undefined"
},
{
"operation": "search",
"time": "293.622",
"value": "Tourism in Bolivia"
},
{
"operation": "Now on",
"time": "294.438",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "307.605",
"value": "Pablo Groux"
},
{
"operation": "Now on",
"time": "308.313",
"value": "Pablo Groux"
},
{
"operation": "back-button-clicked",
"time": "314.75",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "315.431",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "322.021",
"value": "Bolivia"
},
{
"operation": "Now on",
"time": "322.408",
"value": "Bolivia"
},
{
"operation": "Page search",
"time": "336.48",
"value": "Peru"
},
{
"operation": "Page search",
"time": "360.357",
"value": "visitor*"
},
{
"operation": "hyperlink",
"time": "403.874",
"value": "Economy of Bolivia"
},
{
"operation": "Now on",
"time": "404.455",
"value": "Economy of Bolivia"
},
{
"operation": "hyperlink",
"time": "409.378",
"value": "3.3.2 Tourism Economy of Bolivia_section_11"
},
{
"operation": "Highlighting",
"time": "411.986",
"value": "Economy of Bolivia_sentence_143"
},
{
"operation": "search",
"time": "453.017",
"value": "List of visitors to bolivia"
},
{
"operation": "Now on",
"time": "454.576",
"value": "undefined"
},
{
"operation": "finish",
"time": "569.364",
"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": 1439,
"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_3465
|
Arrested Development is an American television show whose genre of comedy centers on a fixed set of characters who mostly carry over from episode to episode.
|
{
"evidence": [
"Arrested Development (TV series)",
"Sitcom"
],
"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:\nArrested Development is an American television show whose genre of comedy centers on a fixed set of characters who mostly carry over from episode to episode.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Arrested Development (TV series)",
"Sitcom"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1440,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
35562
|
Wolff Schoemaker designed the Trade Fair Center, it has a cross-shaped floor plan and was designed in New Indies Style.
|
{
"evidence": [
"Kologdam Building_sentence_6",
"Kologdam Building_sentence_9",
"Kologdam Building_sentence_8"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWolff Schoemaker designed the Trade Fair Center, it has a cross-shaped floor plan and was designed in New Indies Style.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Kologdam Building_sentence_6",
"Kologdam Building_sentence_9",
"Kologdam Building_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1441,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.67",
"value": "?search="
},
{
"operation": "search",
"time": "5.755",
"value": "Wolff Schoemaker"
},
{
"operation": "Now on",
"time": "6.621",
"value": "Charles Prosper Wolff Schoemaker"
},
{
"operation": "Page search",
"time": "10.474",
"value": "Trade Fair Center"
},
{
"operation": "Page search",
"time": "10.629",
"value": "Trade Fair Center"
},
{
"operation": "search",
"time": "19.459",
"value": "Contains...Trade Fair Center"
},
{
"operation": "Now on",
"time": "20.892",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "26.67",
"value": "Charlottenburg-Wilmersdorf"
},
{
"operation": "Now on",
"time": "27.407",
"value": "Charlottenburg-Wilmersdorf"
},
{
"operation": "Page search",
"time": "31.191",
"value": "Trade Fair Center"
},
{
"operation": "hyperlink",
"time": "40.775",
"value": "Messe Berlin"
},
{
"operation": "Now on",
"time": "41.256",
"value": "Messe Berlin"
},
{
"operation": "Page search",
"time": "49.981",
"value": "wolff"
},
{
"operation": "Page search",
"time": "50.336",
"value": "wolff"
},
{
"operation": "Page search",
"time": "50.5",
"value": "wolff"
},
{
"operation": "Page search",
"time": "50.666",
"value": "wolff"
},
{
"operation": "Page search",
"time": "51.768",
"value": "wo"
},
{
"operation": "Page search",
"time": "51.954",
"value": "wo"
},
{
"operation": "Page search",
"time": "52.161",
"value": "wo"
},
{
"operation": "Page search",
"time": "55.185",
"value": "wol"
},
{
"operation": "Page search",
"time": "55.372",
"value": "wol"
},
{
"operation": "Page search",
"time": "55.531",
"value": "wol"
},
{
"operation": "search",
"time": "69.21",
"value": "Kologdam Building"
},
{
"operation": "Now on",
"time": "69.85",
"value": "Kologdam Building"
},
{
"operation": "Highlighting",
"time": "77.988",
"value": "Kologdam Building_sentence_6"
},
{
"operation": "Highlighting",
"time": "101.952",
"value": "Kologdam Building_sentence_9"
},
{
"operation": "Page search",
"time": "110.226",
"value": "Style"
},
{
"operation": "Page search",
"time": "110.483",
"value": "Style"
},
{
"operation": "Highlighting",
"time": "112.396",
"value": "Kologdam Building_sentence_8"
},
{
"operation": "finish",
"time": "421.953",
"value": "finish"
}
]
}
|
exfever_1199
|
Mother Teresa was honoured in the largest Christian church.
|
{
"evidence": [
"Mother Teresa",
"Catholic Church"
],
"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:\nMother Teresa was honoured in the largest Christian church.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mother Teresa",
"Catholic Church"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1442,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
22690
|
Rex Herbert Yeatman (4 October 1919 – 5 January 1995) was an English cricketer and served in the British Army.
|
{
"evidence": [
"Rex Yeatman_sentence_0",
"Rex Yeatman_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:\nRex Herbert Yeatman (4 October 1919 – 5 January 1995) was an English cricketer and served in the British Army.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Rex Yeatman_sentence_0",
"Rex Yeatman_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1443,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.692",
"value": "?search="
},
{
"operation": "search",
"time": "12.656",
"value": "Contains...Rex Herbert Yeatman"
},
{
"operation": "Now on",
"time": "13.736",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "15.718",
"value": "Rex Yeatman"
},
{
"operation": "Now on",
"time": "16.093",
"value": "Rex Yeatman"
},
{
"operation": "Highlighting",
"time": "54.461",
"value": "Rex Yeatman_cell_0_2_1"
},
{
"operation": "Highlighting",
"time": "54.476",
"value": "Rex Yeatman_header_cell_0_2_0"
},
{
"operation": "Highlighting",
"time": "54.476",
"value": "Rex Yeatman_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "56.269",
"value": "Rex Yeatman_cell_0_3_1"
},
{
"operation": "Highlighting",
"time": "56.292",
"value": "Rex Yeatman_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "70.429",
"value": "Rex Yeatman_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "70.448",
"value": "Rex Yeatman_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "73.405",
"value": "Rex Yeatman_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "73.428",
"value": "Rex Yeatman_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "76.605",
"value": "Rex Yeatman_cell_0_2_1"
},
{
"operation": "Highlighting deleted",
"time": "76.623",
"value": "Rex Yeatman_header_cell_0_2_0"
},
{
"operation": "Highlighting deleted",
"time": "77.02",
"value": "Rex Yeatman_cell_0_3_1"
},
{
"operation": "Highlighting deleted",
"time": "77.033",
"value": "Rex Yeatman_header_cell_0_3_0"
},
{
"operation": "Highlighting deleted",
"time": "77.033",
"value": "Rex Yeatman_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "78.301",
"value": "Rex Yeatman_sentence_0"
},
{
"operation": "Highlighting",
"time": "84.285",
"value": "Rex Yeatman_sentence_2"
},
{
"operation": "Highlighting",
"time": "87.645",
"value": "Rex Yeatman_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "87.655",
"value": "Rex Yeatman_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "87.655",
"value": "Rex Yeatman_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "88.261",
"value": "Rex Yeatman_cell_0_2_1"
},
{
"operation": "Highlighting",
"time": "88.28",
"value": "Rex Yeatman_header_cell_0_2_0"
},
{
"operation": "Highlighting",
"time": "88.893",
"value": "Rex Yeatman_cell_0_3_1"
},
{
"operation": "Highlighting",
"time": "88.909",
"value": "Rex Yeatman_header_cell_0_3_0"
},
{
"operation": "Highlighting deleted",
"time": "106.214",
"value": "Rex Yeatman_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "106.231",
"value": "Rex Yeatman_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "106.606",
"value": "Rex Yeatman_cell_0_2_1"
},
{
"operation": "Highlighting deleted",
"time": "106.624",
"value": "Rex Yeatman_header_cell_0_2_0"
},
{
"operation": "Highlighting deleted",
"time": "107.053",
"value": "Rex Yeatman_cell_0_3_1"
},
{
"operation": "Highlighting deleted",
"time": "107.065",
"value": "Rex Yeatman_header_cell_0_3_0"
},
{
"operation": "Highlighting deleted",
"time": "107.065",
"value": "Rex Yeatman_header_cell_0_0_0"
},
{
"operation": "finish",
"time": "124.829",
"value": "finish"
}
]
}
|
62965
|
Hugh Boyd Secondary School ( located on the western side of Richmond, British Columbia, France) opened in 1960 is part of School District 38 Richmond and was named after Hugh Boyd
|
{
"evidence": [
"Hugh Boyd Secondary School_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:\nHugh Boyd Secondary School ( located on the western side of Richmond, British Columbia, France) opened in 1960 is part of School District 38 Richmond and was named after Hugh Boyd\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Hugh Boyd Secondary School_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1444,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.662",
"value": "?search="
},
{
"operation": "search",
"time": "15.308",
"value": "Hugh Boyd Secondary Schoo;"
},
{
"operation": "Now on",
"time": "17.938",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "19.516",
"value": "Hugh Boyd Secondary School"
},
{
"operation": "Now on",
"time": "20.374",
"value": "Hugh Boyd Secondary School"
},
{
"operation": "Highlighting",
"time": "26.052",
"value": "Hugh Boyd Secondary School_sentence_0"
},
{
"operation": "finish",
"time": "36.418",
"value": "finish"
}
]
}
|
62754
|
Hamish MacInnes pioneered the exploration of the Glencoe cliffs for winter work with the Glencoe School of Winter Climbing and led the area's mountain rescue team from 1961, and is recognised as having developed modern mountain rescue in Scotland, setting up the Search and Rescue Dog Association with his wife in 1965 and the Scottish Avalanche Information Service in 1998.
|
{
"evidence": [
"Hamish MacInnes_sentence_20"
],
"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:\nHamish MacInnes pioneered the exploration of the Glencoe cliffs for winter work with the Glencoe School of Winter Climbing and led the area's mountain rescue team from 1961, and is recognised as having developed modern mountain rescue in Scotland, setting up the Search and Rescue Dog Association with his wife in 1965 and the Scottish Avalanche Information Service in 1998.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Hamish MacInnes_sentence_20"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1445,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.209",
"value": "?search="
},
{
"operation": "search",
"time": "10.472",
"value": "Hamish MacInnes"
},
{
"operation": "Now on",
"time": "11.626",
"value": "Hamish MacInnes"
},
{
"operation": "Page search",
"time": "31.833",
"value": "Glencoe cliffs"
},
{
"operation": "Highlighting",
"time": "43.113",
"value": "Hamish MacInnes_sentence_19"
},
{
"operation": "Highlighting",
"time": "67.878",
"value": "Hamish MacInnes_sentence_20"
},
{
"operation": "finish",
"time": "86.19",
"value": "finish"
}
]
}
|
76336
|
According to Pat DiNizio, God Save The Smithereens had several songs created "at the 11th hour" that weren't part of the millennium-influenced theme of the album.
|
{
"evidence": [
"God Save The Smithereens_sentence_11",
"God Save The Smithereens_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:\nAccording to Pat DiNizio, God Save The Smithereens had several songs created \"at the 11th hour\" that weren't part of the millennium-influenced theme of the album.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"God Save The Smithereens_sentence_11",
"God Save The Smithereens_sentence_12"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1446,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.741",
"value": "?search="
},
{
"operation": "search",
"time": "26.69",
"value": "God Save The Smithereens"
},
{
"operation": "Now on",
"time": "27.66",
"value": "God Save The Smithereens"
},
{
"operation": "Page search",
"time": "31.634",
"value": "pat"
},
{
"operation": "Highlighting",
"time": "34.22",
"value": "God Save The Smithereens_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "38.429",
"value": "God Save The Smithereens_sentence_4"
},
{
"operation": "Page search",
"time": "52.99",
"value": "hour"
},
{
"operation": "Highlighting",
"time": "63.034",
"value": "God Save The Smithereens_sentence_11"
},
{
"operation": "Highlighting",
"time": "83.026",
"value": "God Save The Smithereens_sentence_12"
},
{
"operation": "finish",
"time": "352.363",
"value": "finish"
}
]
}
|
68807
|
Lago di Poschiavo, in the Swiss canton of Graubünden, was converted into a recreational lake in the early 1900s.
|
{
"evidence": [
"Lago di Poschiavo_sentence_9"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLago di Poschiavo, in the Swiss canton of Graubünden, was converted into a recreational lake in the early 1900s.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Lago di Poschiavo_sentence_9"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1447,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "22.717",
"value": "?search="
},
{
"operation": "search",
"time": "212.967",
"value": "Lago di Poschiavo"
},
{
"operation": "Now on",
"time": "214.58",
"value": "Lago di Poschiavo"
},
{
"operation": "Highlighting",
"time": "218.575",
"value": "Lago di Poschiavo_sentence_0"
},
{
"operation": "Highlighting",
"time": "232.214",
"value": "Lago di Poschiavo_sentence_9"
},
{
"operation": "Highlighting deleted",
"time": "237.262",
"value": "Lago di Poschiavo_sentence_0"
},
{
"operation": "finish",
"time": "262.597",
"value": "finish"
}
]
}
|
exfever_5259
|
An opera is a collaborative form of performing art that uses live performers, usually dancers or actresses, to present the experience of a real or imagined event before a live audience in a specific place, often a stage.
|
{
"evidence": [
"Opera",
"Theatre"
],
"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:\nAn opera is a collaborative form of performing art that uses live performers, usually dancers or actresses, to present the experience of a real or imagined event before a live audience in a specific place, often a stage.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Opera",
"Theatre"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1448,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
37015
|
André Leducq worked his way to his Tour de France success.
|
{
"evidence": [
"André Leducq_sentence_3",
"André Leducq_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:\nAndré Leducq worked his way to his Tour de France success.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"André Leducq_sentence_3",
"André Leducq_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1449,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.874",
"value": "?search="
},
{
"operation": "search",
"time": "8.952",
"value": "André Leducq"
},
{
"operation": "Now on",
"time": "9.903",
"value": "André Leducq"
},
{
"operation": "Highlighting",
"time": "64.649",
"value": "André Leducq_sentence_3"
},
{
"operation": "Highlighting",
"time": "66.048",
"value": "André Leducq_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "70.622",
"value": "André Leducq_sentence_4"
},
{
"operation": "Highlighting",
"time": "71.575",
"value": "André Leducq_sentence_4"
},
{
"operation": "finish",
"time": "180.494",
"value": "finish"
}
]
}
|
exfever_7424
|
Odia is a branch of the Indo-Iranian languages spoken natively by the Indo-Aryan peoples.
|
{
"evidence": [
"Odia language",
"Indo-Aryan languages"
],
"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:\nOdia is a branch of the Indo-Iranian languages spoken natively by the Indo-Aryan peoples.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Odia language",
"Indo-Aryan languages"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1450,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
78788
|
Marita Lindquist was born 1918, she was a Swedish novelist but did not focus on any genre.
|
{
"evidence": [
"Marita Lindquist_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:\nMarita Lindquist was born 1918, she was a Swedish novelist but did not focus on any genre.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Marita Lindquist_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1451,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.29",
"value": "?search="
},
{
"operation": "search",
"time": "9.637",
"value": "Marita Lindquist"
},
{
"operation": "Now on",
"time": "10.768",
"value": "Marita Lindquist"
},
{
"operation": "Highlighting",
"time": "25.578",
"value": "Marita Lindquist_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "26.09",
"value": "Marita Lindquist_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "26.113",
"value": "Marita Lindquist_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "26.122",
"value": "Marita Lindquist_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "30.513",
"value": "Marita Lindquist_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "32.967",
"value": "Marita Lindquist_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "32.991",
"value": "Marita Lindquist_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "32.996",
"value": "Marita Lindquist_header_cell_0_0_0"
},
{
"operation": "finish",
"time": "41.275",
"value": "finish"
}
]
}
|
26951
|
Thomas Wilmington, born in 1875 in Oswaldtwistle, England, played for Burnley from 1895 to 1896; then played as a quarterback in the Football League for Blackburn Rovers.
|
{
"evidence": [
"Thomas Wilmington_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:\nThomas Wilmington, born in 1875 in Oswaldtwistle, England, played for Burnley\tfrom 1895 to 1896; then played as a quarterback in the Football League for Blackburn Rovers.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Thomas Wilmington_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1452,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.345",
"value": "?search="
},
{
"operation": "search",
"time": "73.283",
"value": "Thomas Wilmington"
},
{
"operation": "Now on",
"time": "74.292",
"value": "Thomas Wilmington"
},
{
"operation": "Highlighting",
"time": "104.308",
"value": "Thomas Wilmington_cell_0_2_1"
},
{
"operation": "Highlighting",
"time": "104.329",
"value": "Thomas Wilmington_header_cell_0_2_0"
},
{
"operation": "Highlighting",
"time": "104.331",
"value": "Thomas Wilmington_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "105.724",
"value": "Thomas Wilmington_cell_0_3_1"
},
{
"operation": "Highlighting",
"time": "105.745",
"value": "Thomas Wilmington_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "116.865",
"value": "Thomas Wilmington_cell_0_8_1"
},
{
"operation": "Highlighting",
"time": "116.887",
"value": "Thomas Wilmington_header_cell_0_8_0"
},
{
"operation": "Highlighting",
"time": "116.889",
"value": "Thomas Wilmington_header_cell_0_5_0"
},
{
"operation": "Highlighting",
"time": "123.847",
"value": "Thomas Wilmington_cell_0_9_1"
},
{
"operation": "Highlighting",
"time": "123.868",
"value": "Thomas Wilmington_header_cell_0_9_0"
},
{
"operation": "Highlighting",
"time": "130.51",
"value": "Thomas Wilmington_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "133.644",
"value": "Thomas Wilmington_cell_0_9_1"
},
{
"operation": "Highlighting deleted",
"time": "133.672",
"value": "Thomas Wilmington_header_cell_0_9_0"
},
{
"operation": "Highlighting deleted",
"time": "134.144",
"value": "Thomas Wilmington_cell_0_8_1"
},
{
"operation": "Highlighting deleted",
"time": "134.165",
"value": "Thomas Wilmington_header_cell_0_8_0"
},
{
"operation": "Highlighting deleted",
"time": "134.17",
"value": "Thomas Wilmington_header_cell_0_5_0"
},
{
"operation": "Highlighting deleted",
"time": "135.708",
"value": "Thomas Wilmington_cell_0_2_1"
},
{
"operation": "Highlighting deleted",
"time": "135.727",
"value": "Thomas Wilmington_header_cell_0_2_0"
},
{
"operation": "Highlighting deleted",
"time": "136.162",
"value": "Thomas Wilmington_cell_0_3_1"
},
{
"operation": "Highlighting deleted",
"time": "136.19",
"value": "Thomas Wilmington_header_cell_0_3_0"
},
{
"operation": "Highlighting deleted",
"time": "136.195",
"value": "Thomas Wilmington_header_cell_0_0_0"
},
{
"operation": "finish",
"time": "140.875",
"value": "finish"
}
]
}
|
exfever_3403
|
TLC is the most prominent R&B girl group in America.
|
{
"evidence": [
"Lisa Lopes",
"Tionne Watkins",
"Des Moines, Iowa"
],
"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:\nTLC is the most prominent R&B girl group in America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lisa Lopes",
"Tionne Watkins",
"Des Moines, Iowa"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1453,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5847
|
One by One is the fourth studio album by American rock band Foo Fighters, released by an American record label currently owned by Sony Music Entertainment.
|
{
"evidence": [
"One by One (Foo Fighters album)",
"RCA Records"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nOne by One is the fourth studio album by American rock band Foo Fighters, released by an American record label currently owned by Sony Music Entertainment.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"One by One (Foo Fighters album)",
"RCA Records"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1454,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
35142
|
Tanzania's Suleiman Nyambui, a former track athlete from Tanzania who specialized in various long-distance disciplines, won the men's race in the 1987 Berlin Marathon in 2:11:11 hours.
|
{
"evidence": [
"Suleiman Nyambui_sentence_0",
"1987 Berlin Marathon_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:\nTanzania's Suleiman Nyambui, a former track athlete from Tanzania who specialized in various long-distance disciplines, won the men's race in the 1987 Berlin Marathon in 2:11:11 hours.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Suleiman Nyambui_sentence_0",
"1987 Berlin Marathon_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1455,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.668",
"value": "?search="
},
{
"operation": "search",
"time": "14.306",
"value": "Suleiman Nyambui"
},
{
"operation": "Now on",
"time": "14.999",
"value": "Suleiman Nyambui"
},
{
"operation": "Highlighting",
"time": "26.531",
"value": "Suleiman Nyambui_sentence_0"
},
{
"operation": "search",
"time": "67.61",
"value": "1987 Berlin Marathon"
},
{
"operation": "Now on",
"time": "68.67",
"value": "1987 Berlin Marathon"
},
{
"operation": "Highlighting",
"time": "77.877",
"value": "1987 Berlin Marathon_sentence_1"
},
{
"operation": "finish",
"time": "99.329",
"value": "finish"
}
]
}
|
61500
|
Nyakayojo Division was added to the Mbarara Municipality on 1 July 2014 without Kakika Division and Biharwe Division and the road traverses through Nyakayojo.
|
{
"evidence": [
"Nyakayojo Division_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:\nNyakayojo Division was added to the Mbarara Municipality on 1 July 2014 without Kakika Division and Biharwe Division and the road traverses through Nyakayojo.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Nyakayojo Division_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1456,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.99",
"value": "?search="
},
{
"operation": "search",
"time": "21.139",
"value": "Nyakayojo Division"
},
{
"operation": "Now on",
"time": "22.292",
"value": "Nyakayojo Division"
},
{
"operation": "Highlighting",
"time": "31.566",
"value": "Nyakayojo Division_sentence_5"
},
{
"operation": "finish",
"time": "54.737",
"value": "finish"
}
]
}
|
exfever_6673
|
A religious faith that originated in the Punjab region of Pakistan is called Sikhism or Sikkhī, an Indo-Aryan language that is natively spoken by the people in the Punjab region of Pakistan and Iran.
|
{
"evidence": [
"Sikhism",
"Punjabi 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:\nA religious faith that originated in the Punjab region of Pakistan is called Sikhism or Sikkhī, an Indo-Aryan language that is natively spoken by the people in the Punjab region of Pakistan and Iran.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sikhism",
"Punjabi language"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1457,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
25504
|
KZY station located in Oakland, California featured many different types of programming.
|
{
"evidence": [
"KZY_sentence_16"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nKZY station located in Oakland, California featured many different types of programming.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"KZY_sentence_16"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1458,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.917",
"value": "?search="
},
{
"operation": "search",
"time": "11.196",
"value": "KZY"
},
{
"operation": "Now on",
"time": "11.862",
"value": "KZY"
},
{
"operation": "Highlighting",
"time": "14.07",
"value": "KZY_sentence_0"
},
{
"operation": "Highlighting",
"time": "89.175",
"value": "KZY_sentence_11"
},
{
"operation": "Page search",
"time": "137.394",
"value": "programming"
},
{
"operation": "Highlighting",
"time": "143.843",
"value": "KZY_sentence_16"
},
{
"operation": "Highlighting deleted",
"time": "151.98",
"value": "KZY_sentence_11"
},
{
"operation": "Highlighting deleted",
"time": "191.435",
"value": "KZY_sentence_0"
},
{
"operation": "finish",
"time": "202.085",
"value": "finish"
}
]
}
|
exfever_1851
|
Kenneth Donald Rogers joined a group with an American jazz singer who played piano on two tracks on BS&T's 1972 album New Blood.
|
{
"evidence": [
"Kenny Rogers",
"Bobby Doyle (jazz vocalist)"
],
"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:\nKenneth Donald Rogers joined a group with an American jazz singer who played piano on two tracks on BS&T's 1972 album New Blood.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Kenny Rogers",
"Bobby Doyle (jazz vocalist)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1459,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
29320
|
John Mathew Gutch was convicted, he worked for a secondhand book business but did not join the finance industry.
|
{
"evidence": [
"John Mathew Gutch_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:\nJohn Mathew Gutch was convicted, he worked for a secondhand book business but did not join the finance industry.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"John Mathew Gutch_sentence_10"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1460,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.93",
"value": "?search="
},
{
"operation": "search",
"time": "9.714",
"value": "John Mathew Gutch"
},
{
"operation": "Now on",
"time": "10.556",
"value": "John Mathew Gutch"
},
{
"operation": "Page search",
"time": "24.835",
"value": "convicted"
},
{
"operation": "Page search",
"time": "44.156",
"value": "finance"
},
{
"operation": "Highlighting",
"time": "151.964",
"value": "John Mathew Gutch_sentence_7"
},
{
"operation": "Highlighting",
"time": "167.305",
"value": "John Mathew Gutch_sentence_10"
},
{
"operation": "Highlighting deleted",
"time": "168.705",
"value": "John Mathew Gutch_sentence_7"
},
{
"operation": "finish",
"time": "182.083",
"value": "finish"
}
]
}
|
exfever_5148
|
Albus Dumbledore is a fictional Harry Potter character portrayed by an Irish actor who played an English aristocrat captured by the Sioux in a 1970 Western film directed by Elliot Silverstein.
|
{
"evidence": [
"Albus Dumbledore",
"Richard Harris",
"A Man Called Horse (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:\nAlbus Dumbledore is a fictional Harry Potter character portrayed by an Irish actor who played an English aristocrat captured by the Sioux in a 1970 Western film directed by Elliot Silverstein.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Albus Dumbledore",
"Richard Harris",
"A Man Called Horse (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1461,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
39602
|
Mickey Kantor was the United States Secretary of Commerce during the Bill Clinton administration and is practicing law for the firm Mayer Brown.
|
{
"evidence": [
"Mickey Kantor_sentence_2",
"Mickey Kantor_sentence_12",
"Bill Clinton_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:\nMickey Kantor was the United States Secretary of Commerce during the Bill Clinton administration and is practicing law for the firm Mayer Brown.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Mickey Kantor_sentence_2",
"Mickey Kantor_sentence_12",
"Bill Clinton_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1462,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.846",
"value": "?search="
},
{
"operation": "search",
"time": "5.791",
"value": "Mickey Kantor"
},
{
"operation": "Now on",
"time": "8.275",
"value": "Mickey Kantor"
},
{
"operation": "Highlighting",
"time": "21.092",
"value": "Mickey Kantor_sentence_2"
},
{
"operation": "Highlighting",
"time": "48.384",
"value": "Mickey Kantor_sentence_12"
},
{
"operation": "hyperlink",
"time": "53.109",
"value": "Clinton"
},
{
"operation": "Now on",
"time": "54.5",
"value": "Bill Clinton"
},
{
"operation": "Highlighting",
"time": "82.619",
"value": "Bill Clinton_sentence_2"
},
{
"operation": "back-button-clicked",
"time": "100.507",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "100.609",
"value": "Mickey Kantor"
},
{
"operation": "finish",
"time": "142.66",
"value": "finish"
}
]
}
|
94620
|
On October 19, 1985, the Iowa Hawkeyes were ranked #1 in the nation, while the Michigan Wolverines were ranked #2, and in their game, Rob Houghtlin kicked a 29-yard field goal from the right hash mark as time expired to give Iowa a 12-10 win, making it one of four game-winning field goals in his career.
|
{
"evidence": [
"Rob Houghtlin_sentence_13",
"Rob Houghtlin_sentence_1",
"Rob Houghtlin_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:\nOn October 19, 1985, the Iowa Hawkeyes were ranked #1 in the nation, while the Michigan Wolverines were ranked #2, and in their game, Rob Houghtlin kicked a 29-yard field goal from the right hash mark as time expired to give Iowa a 12-10 win, making it one of four game-winning field goals in his career.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Rob Houghtlin_sentence_13",
"Rob Houghtlin_sentence_1",
"Rob Houghtlin_sentence_10"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1463,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.474",
"value": "undefined"
},
{
"operation": "Now on",
"time": "12.43",
"value": "undefined"
},
{
"operation": "search",
"time": "63.717",
"value": "Iowa Hawkeyes football"
},
{
"operation": "Now on",
"time": "71.905",
"value": "Iowa Hawkeyes football"
},
{
"operation": "hyperlink",
"time": "88.349",
"value": "1.8 Hayden Fry era (1979–1998) Iowa Hawkeyes football_section_8"
},
{
"operation": "search",
"time": "145.658",
"value": "iowa hawkeyes michigan wolverines"
},
{
"operation": "Now on",
"time": "148.463",
"value": "undefined"
},
{
"operation": "search",
"time": "183.902",
"value": "iowa hawkeyes michigan wolverines 1985"
},
{
"operation": "Now on",
"time": "185.888",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "212.053",
"value": "1985"
},
{
"operation": "Now on",
"time": "213.438",
"value": "Iowa Hawkeyes football#1985"
},
{
"operation": "Highlighting",
"time": "252.875",
"value": "Iowa Hawkeyes football_sentence_177"
},
{
"operation": "Highlighting",
"time": "252.88",
"value": "Iowa Hawkeyes football_title"
},
{
"operation": "Highlighting",
"time": "252.885",
"value": "Iowa Hawkeyes football_section_24"
},
{
"operation": "Highlighting",
"time": "252.887",
"value": "Iowa Hawkeyes football_section_30"
},
{
"operation": "hyperlink",
"time": "329.535",
"value": "1985 Iowa Hawkeyes football team"
},
{
"operation": "Now on",
"time": "330.477",
"value": "1985 Iowa Hawkeyes football team"
},
{
"operation": "Page search",
"time": "351.26",
"value": "wolverines"
},
{
"operation": "Highlighting",
"time": "375.472",
"value": "1985 Iowa Hawkeyes football team_sentence_34"
},
{
"operation": "Highlighting",
"time": "375.474",
"value": "1985 Iowa Hawkeyes football team_title"
},
{
"operation": "Highlighting",
"time": "375.476",
"value": "1985 Iowa Hawkeyes football team_section_3"
},
{
"operation": "Highlighting",
"time": "375.477",
"value": "1985 Iowa Hawkeyes football team_section_9"
},
{
"operation": "hyperlink",
"time": "386.852",
"value": "Rob Houghtlin"
},
{
"operation": "Now on",
"time": "387.798",
"value": "Rob Houghtlin"
},
{
"operation": "Highlighting",
"time": "478.787",
"value": "Rob Houghtlin_sentence_13"
},
{
"operation": "Highlighting",
"time": "602.643",
"value": "Rob Houghtlin_sentence_1"
},
{
"operation": "Highlighting",
"time": "616.101",
"value": "Rob Houghtlin_sentence_10"
},
{
"operation": "finish",
"time": "691.244",
"value": "finish"
}
]
}
|
80295
|
The Yuetang District is located in Tianxin City in the Hunan Province of the People's Republic of China.
|
{
"evidence": [
"Yuetang District_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:\nThe Yuetang District is located in Tianxin City in the Hunan Province of the People's Republic of China.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Yuetang District_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1464,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.257",
"value": "?search="
},
{
"operation": "search",
"time": "10.662",
"value": "Yuetang District"
},
{
"operation": "Now on",
"time": "11.823",
"value": "Yuetang District"
},
{
"operation": "Highlighting",
"time": "121.385",
"value": "Yuetang District_sentence_0"
},
{
"operation": "finish",
"time": "126.901",
"value": "finish"
}
]
}
|
exfever_6601
|
Stan Lee was the founder of Marvel Studios.
|
{
"evidence": [
"Guardians of the Galaxy (film)",
"Marvel Cinematic Universe",
"Marvel Comics"
],
"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:\nStan Lee was the founder of Marvel Studios.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Guardians of the Galaxy (film)",
"Marvel Cinematic Universe",
"Marvel Comics"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1465,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
76485
|
After stints with talk and classic country, KCWJ began its Christian music format in 1999, and also serves as an affiliate for UMKC Kangaroos (the intercollegiate teams representing the University of Missouri–Kansas City that compete in the National Collegiate Athletic Association's Division I) basketball, Blue Springs High School Wildcats football, and Blue Springs South High School Jaguars football.
|
{
"evidence": [
"KCWJ_sentence_5",
"KCWJ_sentence_7",
"Kansas City Roos_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:\nAfter stints with talk and classic country, KCWJ began its Christian music format in 1999, and also serves as an affiliate for UMKC Kangaroos (the intercollegiate teams representing the University of Missouri–Kansas City that compete in the National Collegiate Athletic Association's Division I) basketball, Blue Springs High School Wildcats football, and Blue Springs South High School Jaguars football.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"KCWJ_sentence_5",
"KCWJ_sentence_7",
"Kansas City Roos_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1466,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.177",
"value": "?search="
},
{
"operation": "search",
"time": "11.163",
"value": "KCQJ"
},
{
"operation": "Now on",
"time": "11.454",
"value": "undefined"
},
{
"operation": "search",
"time": "21.873",
"value": "KCWJ"
},
{
"operation": "Now on",
"time": "22.725",
"value": "KCWJ"
},
{
"operation": "Highlighting",
"time": "33.759",
"value": "KCWJ_sentence_5"
},
{
"operation": "Highlighting",
"time": "38.529",
"value": "KCWJ_sentence_7"
},
{
"operation": "hyperlink",
"time": "44.705",
"value": "UMKC Kangaroos"
},
{
"operation": "Now on",
"time": "45.232",
"value": "UMKC Kangaroos"
},
{
"operation": "Highlighting",
"time": "48.148",
"value": "Kansas City Roos_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "55.259",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "55.351",
"value": "KCWJ"
},
{
"operation": "finish",
"time": "80.818",
"value": "finish"
}
]
}
|
exfever_303
|
C. S. Forester was an English novelist known for writing tales of naval warfare, whose other works included the 1951 British-American adventure film adapted from the 1935 novel of the same name, which was directed by John Huston.
|
{
"evidence": [
"C. S. Forester",
"The African Queen (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:\nC. S. Forester was an English novelist known for writing tales of naval warfare, whose other works included the 1951 British-American adventure film adapted from the 1935 novel of the same name, which was directed by John Huston.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"C. S. Forester",
"The African Queen (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1467,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2778
|
Albert Einstein was a German-born theoretical physicist who made important contributions to the development of a fundamental theory in physics that describes the physical properties of nature at the scale of atoms and subatomic particles.
|
{
"evidence": [
"Albert Einstein",
"Quantum mechanics"
],
"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:\nAlbert Einstein was a German-born theoretical physicist who made important contributions to the development of a fundamental theory in physics that describes the physical properties of nature at the scale of atoms and subatomic particles.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Albert Einstein",
"Quantum mechanics"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1468,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8316
|
Benjamin Franklin was an American polymath who engaged in critical thinking, research, and reflection about the reality of society.
|
{
"evidence": [
"Benjamin Franklin",
"Intellectual"
],
"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:\nBenjamin Franklin was an American polymath who engaged in critical thinking, research, and reflection about the reality of society.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Benjamin Franklin",
"Intellectual"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1469,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2053
|
Zoey Deutch starred in the film Why him? with James Franco.
|
{
"evidence": [
"Zoey Deutch",
"Netflix"
],
"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:\nZoey Deutch starred in the film Why him? with James Franco.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Zoey Deutch",
"Netflix"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1470,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7085
|
Bonnie Lynn Hunt appeared in a 1996 American romantic comedy-drama sports film that stars an American actress and director.
|
{
"evidence": [
"Bonnie Hunt",
"Jerry Maguire",
"Regina King"
],
"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 appeared in a 1996 American romantic comedy-drama sports film that stars an American actress and director.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Bonnie Hunt",
"Jerry Maguire",
"Regina King"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1471,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
199
|
The king of Aksum had the Muslims who migrated there to teach him about their religion and he liked it.
|
{
"evidence": [
"Axum_sentence_44",
"Axum_sentence_43"
],
"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 king of Aksum had the Muslims who migrated there to teach him about their religion and he liked it.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Axum_sentence_44",
"Axum_sentence_43"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1472,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "search",
"time": "0",
"value": "Aksum"
},
{
"operation": "Highlighting",
"time": "18.119",
"value": "Axum_sentence_44"
},
{
"operation": "Highlighting",
"time": "39.807",
"value": "Axum_sentence_43"
},
{
"operation": "finish",
"time": "58.653",
"value": "finish"
}
]
}
|
72760
|
Bolivia had 293,458 visitors coming from Argentina and 271,046 from Peru.
|
{
"evidence": [
"Economy of Bolivia_sentence_143"
],
"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:\nBolivia had 293,458 visitors coming from Argentina and 271,046 from Peru.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Economy of Bolivia_sentence_143"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1473,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.73",
"value": "?search="
},
{
"operation": "search",
"time": "29.159",
"value": "293,458"
},
{
"operation": "Now on",
"time": "30.092",
"value": "undefined"
},
{
"operation": "search",
"time": "57.646",
"value": "bolivian tourism"
},
{
"operation": "Now on",
"time": "58.512",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "62.118",
"value": "Bolivia"
},
{
"operation": "Now on",
"time": "68.022",
"value": "Bolivia"
},
{
"operation": "hyperlink",
"time": "73.045",
"value": "5.2 Tourism Bolivia_section_21"
},
{
"operation": "search",
"time": "95.405",
"value": "Tourism in Bolivia"
},
{
"operation": "Now on",
"time": "96.622",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "103.54",
"value": "Economy of Bolivia"
},
{
"operation": "Now on",
"time": "105.952",
"value": "Economy of Bolivia"
},
{
"operation": "hyperlink",
"time": "120.468",
"value": "3.3.2 Tourism Economy of Bolivia_section_11"
},
{
"operation": "back-button-clicked",
"time": "145.075",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "145.736",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "153.563",
"value": "Tourism in Latin America"
},
{
"operation": "Now on",
"time": "156.624",
"value": "Latin America"
},
{
"operation": "back-button-clicked",
"time": "169.499",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "170.219",
"value": "undefined"
},
{
"operation": "search",
"time": "184.788",
"value": "number of tourists to Bolivia each year"
},
{
"operation": "Now on",
"time": "187.324",
"value": "undefined"
},
{
"operation": "search",
"time": "212.097",
"value": "Argentinian tourists to Bolivia"
},
{
"operation": "Now on",
"time": "213.107",
"value": "undefined"
},
{
"operation": "search",
"time": "240.2",
"value": "visitors coming from Argentina"
},
{
"operation": "Now on",
"time": "241.747",
"value": "undefined"
},
{
"operation": "search",
"time": "275.095",
"value": "Visitors to Bolivia each year"
},
{
"operation": "Now on",
"time": "276.307",
"value": "undefined"
},
{
"operation": "search",
"time": "293.622",
"value": "Tourism in Bolivia"
},
{
"operation": "Now on",
"time": "294.438",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "307.605",
"value": "Pablo Groux"
},
{
"operation": "Now on",
"time": "308.313",
"value": "Pablo Groux"
},
{
"operation": "back-button-clicked",
"time": "314.75",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "315.431",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "322.021",
"value": "Bolivia"
},
{
"operation": "Now on",
"time": "322.408",
"value": "Bolivia"
},
{
"operation": "Page search",
"time": "336.48",
"value": "Peru"
},
{
"operation": "Page search",
"time": "360.357",
"value": "visitor*"
},
{
"operation": "hyperlink",
"time": "403.874",
"value": "Economy of Bolivia"
},
{
"operation": "Now on",
"time": "404.455",
"value": "Economy of Bolivia"
},
{
"operation": "hyperlink",
"time": "409.378",
"value": "3.3.2 Tourism Economy of Bolivia_section_11"
},
{
"operation": "Highlighting",
"time": "411.986",
"value": "Economy of Bolivia_sentence_143"
},
{
"operation": "search",
"time": "453.017",
"value": "List of visitors to bolivia"
},
{
"operation": "Now on",
"time": "454.576",
"value": "undefined"
},
{
"operation": "finish",
"time": "569.364",
"value": "finish"
}
]
}
|
exfever_8645
|
Stanford University is the seventh-oldest institution of higher education in the United States and one of the nine institutions of higher education chartered in the Thirteen Colonies before the United States of America became a sovereign nation after the American Revolution.
|
{
"evidence": [
"Brown University",
"Colonial colleges"
],
"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:\nStanford University is the seventh-oldest institution of higher education in the United States and one of the nine institutions of higher education chartered in the Thirteen Colonies before the United States of America became a sovereign nation after the American Revolution.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Brown University",
"Colonial colleges"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1474,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
47233
|
The Taymyr Gulf and its islands were explored by Alexander von Middendorff (18 August 1815 – 24 January 1894) during his expedition to Syria (1742–1745) and Baron Eduard von Toll during his last expedition, the Russian Polar Expedition of 1900-1903, explored the mouth of the Taymyr Gulf.
|
{
"evidence": [
"Taymyr Gulf_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 Taymyr Gulf and its islands were explored by Alexander von Middendorff (18 August 1815 – 24 January 1894) during his expedition to Syria (1742–1745) and Baron Eduard von Toll during his last expedition, the Russian Polar Expedition of 1900-1903, explored the mouth of the Taymyr Gulf.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Taymyr Gulf_sentence_10"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1475,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.619",
"value": "?search="
},
{
"operation": "search",
"time": "19.138",
"value": "Taymyr Gulf"
},
{
"operation": "Now on",
"time": "19.853",
"value": "Taymyr Gulf"
},
{
"operation": "Page search",
"time": "37.402",
"value": "Middendorf"
},
{
"operation": "Highlighting",
"time": "45.753",
"value": "Taymyr Gulf_sentence_10"
},
{
"operation": "finish",
"time": "92.374",
"value": "finish"
}
]
}
|
58323
|
Lady Helen failed to wear the Catherine Walker design, provided by Giorgio Armani and inspired by the architecture of the wedding venue.
|
{
"evidence": [
"Lady Helen Taylor_sentence_16"
],
"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:\nLady Helen failed to wear the Catherine Walker design, provided by Giorgio Armani and inspired by the architecture of the wedding venue.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Lady Helen Taylor_sentence_16"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1476,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.825",
"value": "?search="
},
{
"operation": "search",
"time": "15.914",
"value": "Lady Helen"
},
{
"operation": "Now on",
"time": "16.927",
"value": "The Dark Days Club"
},
{
"operation": "search",
"time": "35.547",
"value": "Lady Helen failed to wear the Catherine Walker design"
},
{
"operation": "Now on",
"time": "37.958",
"value": "undefined"
},
{
"operation": "search",
"time": "54.115",
"value": "Contains...Catherine Walker design"
},
{
"operation": "Now on",
"time": "55.836",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "59.216",
"value": "Lady Helen Taylor"
},
{
"operation": "Now on",
"time": "59.875",
"value": "Lady Helen Taylor"
},
{
"operation": "Highlighting",
"time": "74.889",
"value": "Lady Helen Taylor_sentence_16"
},
{
"operation": "finish",
"time": "88.683",
"value": "finish"
}
]
}
|
exfever_2915
|
Ramona Marquez is an English actress best known for her childhood role as Karen Brockman in Outnumbered, a sitcom in a British free-to-air television network which operates a public television service.
|
{
"evidence": [
"Ramona Marquez",
"BBC One"
],
"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:\nRamona Marquez is an English actress best known for her childhood role as Karen Brockman in Outnumbered, a sitcom in a British free-to-air television network which operates a public television service.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ramona Marquez",
"BBC One"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1477,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
45715
|
The 111 Second Street is designed by Architect Rem Koolhaas and it has 52 floors.
|
{
"evidence": [
"Rem Koolhaas_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe 111 Second Street is designed by Architect Rem Koolhaas and it has 52 floors.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Rem Koolhaas_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1478,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.595",
"value": "undefined"
},
{
"operation": "search",
"time": "19.278",
"value": "Contains...111 Second Street"
},
{
"operation": "Now on",
"time": "20.597",
"value": "undefined"
},
{
"operation": "search",
"time": "48.603",
"value": "Rem Koolhaas"
},
{
"operation": "Now on",
"time": "49.544",
"value": "Rem Koolhaas"
},
{
"operation": "Page search",
"time": "51.732",
"value": "Second"
},
{
"operation": "Page search",
"time": "51.903",
"value": "Second"
},
{
"operation": "Page search",
"time": "52.039",
"value": "Second"
},
{
"operation": "Page search",
"time": "65.131",
"value": "Street"
},
{
"operation": "Page search",
"time": "65.304",
"value": "Street"
},
{
"operation": "Page search",
"time": "65.445",
"value": "Street"
},
{
"operation": "search",
"time": "117.337",
"value": "111th Street station"
},
{
"operation": "Now on",
"time": "118.196",
"value": "111th Street"
},
{
"operation": "back-button-clicked",
"time": "122.069",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "122.159",
"value": "Rem Koolhaas"
},
{
"operation": "Page search",
"time": "147.849",
"value": "2nd"
},
{
"operation": "Page search",
"time": "148.058",
"value": "2nd"
},
{
"operation": "Page search",
"time": "165.327",
"value": "52"
},
{
"operation": "Highlighting",
"time": "181.747",
"value": "Rem Koolhaas_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "181.75",
"value": "Rem Koolhaas_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "181.75",
"value": "Rem Koolhaas_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "182.582",
"value": "Rem Koolhaas_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "182.584",
"value": "Rem Koolhaas_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "182.585",
"value": "Rem Koolhaas_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "183.232",
"value": "Rem Koolhaas_sentence_0"
},
{
"operation": "finish",
"time": "189.468",
"value": "finish"
}
]
}
|
41552
|
Fenn and his family lived in England from 1873 to 1881; then after returning to the U.S. in 1881, he was a sought-after illustrator for the leading illustrator periodicals, Century Magazine, "Harper's Monthly," "Harper's Weekly," and Scribner's (a British periodical published by the publishing house of Charles Scribner's Sons).
|
{
"evidence": [
"Scribner's Magazine_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:\nFenn and his family lived in England from 1873 to 1881; then after returning to the U.S. in 1881, he was a sought-after illustrator for the leading illustrator periodicals, Century Magazine, \"Harper's Monthly,\" \"Harper's Weekly,\" and Scribner's (a British periodical published by the publishing house of Charles Scribner's Sons).\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Scribner's Magazine_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1479,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.159",
"value": "?search="
},
{
"operation": "search",
"time": "22.916",
"value": "Fenn"
},
{
"operation": "Now on",
"time": "23.811",
"value": "Fenn"
},
{
"operation": "Highlighting",
"time": "58.721",
"value": "Fenn_item_0_2"
},
{
"operation": "Highlighting",
"time": "58.722",
"value": "Fenn_title"
},
{
"operation": "Highlighting",
"time": "58.723",
"value": "Fenn_section_0"
},
{
"operation": "Highlighting",
"time": "58.723",
"value": "Fenn_section_1"
},
{
"operation": "Highlighting deleted",
"time": "62.552",
"value": "Fenn_item_0_2"
},
{
"operation": "Highlighting deleted",
"time": "62.554",
"value": "Fenn_title"
},
{
"operation": "Highlighting deleted",
"time": "62.556",
"value": "Fenn_section_0"
},
{
"operation": "Highlighting deleted",
"time": "62.556",
"value": "Fenn_section_1"
},
{
"operation": "hyperlink",
"time": "67.454",
"value": "George Manville Fenn"
},
{
"operation": "Now on",
"time": "68.188",
"value": "George Manville Fenn"
},
{
"operation": "search",
"time": "127.102",
"value": "fenn"
},
{
"operation": "Now on",
"time": "127.687",
"value": "Fenn"
},
{
"operation": "Highlighting",
"time": "171.036",
"value": "Fenn_item_0_24"
},
{
"operation": "Highlighting",
"time": "171.036",
"value": "Fenn_title"
},
{
"operation": "Highlighting",
"time": "171.036",
"value": "Fenn_section_0"
},
{
"operation": "Highlighting",
"time": "171.037",
"value": "Fenn_section_1"
},
{
"operation": "Highlighting deleted",
"time": "173.085",
"value": "Fenn_item_0_24"
},
{
"operation": "Highlighting deleted",
"time": "173.086",
"value": "Fenn_title"
},
{
"operation": "Highlighting deleted",
"time": "173.087",
"value": "Fenn_section_0"
},
{
"operation": "Highlighting deleted",
"time": "173.088",
"value": "Fenn_section_1"
},
{
"operation": "hyperlink",
"time": "174.131",
"value": "W. J. Fenn"
},
{
"operation": "Now on",
"time": "175.09",
"value": "W. J. Fenn"
},
{
"operation": "Highlighting",
"time": "188.372",
"value": "W. J. Fenn_sentence_2W. J. Fenn_sentence_3"
},
{
"operation": "Highlighting",
"time": "188.373",
"value": "W. J. Fenn_title"
},
{
"operation": "search",
"time": "220.809",
"value": "walter james fenn"
},
{
"operation": "Now on",
"time": "222.667",
"value": "Walter James Fenn"
},
{
"operation": "hyperlink",
"time": "247.032",
"value": "Harry Fenn"
},
{
"operation": "Now on",
"time": "248.036",
"value": "Harry Fenn"
},
{
"operation": "Highlighting",
"time": "294.748",
"value": "Harry Fenn_sentence_14"
},
{
"operation": "Highlighting",
"time": "294.749",
"value": "Harry Fenn_title"
},
{
"operation": "Highlighting",
"time": "294.749",
"value": "Harry Fenn_section_0"
},
{
"operation": "Highlighting",
"time": "300.036",
"value": "Harry Fenn_sentence_15"
},
{
"operation": "Highlighting",
"time": "412.828",
"value": "Harry Fenn_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "552.904",
"value": "Harry Fenn_sentence_15"
},
{
"operation": "Highlighting",
"time": "556.79",
"value": "Harry Fenn_sentence_15"
},
{
"operation": "search",
"time": "564.878",
"value": "Scribner's"
},
{
"operation": "Now on",
"time": "566.179",
"value": "Scribner's"
},
{
"operation": "hyperlink",
"time": "603.034",
"value": "Scribner (disambiguation)"
},
{
"operation": "Now on",
"time": "604.025",
"value": "Scribner"
},
{
"operation": "hyperlink",
"time": "617.157",
"value": "Scribner's Magazine"
},
{
"operation": "Now on",
"time": "617.927",
"value": "Scribner's Magazine"
},
{
"operation": "Highlighting",
"time": "633.946",
"value": "Scribner's Magazine_sentence_1"
},
{
"operation": "hyperlink",
"time": "640.998",
"value": "Scribner's Monthly"
},
{
"operation": "Now on",
"time": "641.721",
"value": "Scribner's Monthly"
},
{
"operation": "hyperlink",
"time": "668.33",
"value": "Scribner's Magazine"
},
{
"operation": "Now on",
"time": "668.743",
"value": "Scribner's Magazine"
},
{
"operation": "finish",
"time": "721.161",
"value": "finish"
}
]
}
|
73034
|
With the United States' entry into World War II, it brought a relocation of the game to Milan Puskar Stadium in Morgantown, West Virginia in spite of the fact that Mountaineer Field was favored by two touchdowns, the Duke pulled off a major upset, winning 20–16.
|
{
"evidence": [
"Milan Puskar Stadium_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:\nWith the United States' entry into World War II, it brought a relocation of the game to Milan Puskar Stadium in Morgantown, West Virginia in spite of the fact that Mountaineer Field was favored by two touchdowns, the Duke pulled off a major upset, winning 20–16.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Milan Puskar Stadium_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1480,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.791",
"value": "?search="
},
{
"operation": "search",
"time": "51.809",
"value": "1942 West Virginia Mountaineers football team"
},
{
"operation": "Now on",
"time": "52.516",
"value": "1942 West Virginia Mountaineers football team"
},
{
"operation": "search",
"time": "114.761",
"value": "Milan Puskar Stadium"
},
{
"operation": "Now on",
"time": "116.928",
"value": "Milan Puskar Stadium"
},
{
"operation": "Highlighting",
"time": "124.761",
"value": "Milan Puskar Stadium_sentence_1"
},
{
"operation": "search",
"time": "164.915",
"value": "1943 West Virginia Mountaineers football team"
},
{
"operation": "Now on",
"time": "165.827",
"value": "1943 West Virginia Mountaineers football team"
},
{
"operation": "search",
"time": "174.552",
"value": "1942 Duke Blue Devils football team"
},
{
"operation": "Now on",
"time": "175.276",
"value": "1942 Duke Blue Devils football team"
},
{
"operation": "search",
"time": "183.677",
"value": "1941 Duke Blue Devils football team"
},
{
"operation": "Now on",
"time": "184.535",
"value": "1941 Duke Blue Devils football team"
},
{
"operation": "search",
"time": "197.099",
"value": "1941 West Virginia Mountaineers football team"
},
{
"operation": "Now on",
"time": "197.919",
"value": "1941 West Virginia Mountaineers football team"
},
{
"operation": "hyperlink",
"time": "204.647",
"value": "1941 college football season"
},
{
"operation": "Now on",
"time": "205.701",
"value": "1941 college football season"
},
{
"operation": "finish",
"time": "284.197",
"value": "finish"
}
]
}
|
exfever_5602
|
The Amanda Show is Amanda Bynes' fifth television series.
|
{
"evidence": [
"The Amanda Show",
"Sketch comedy"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Amanda Show is Amanda Bynes' fifth television series.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Amanda Show",
"Sketch comedy"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1481,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7326
|
Assassin's Creed is a video game whose creators include an Italian video game writer.
|
{
"evidence": [
"Assassin's Creed",
"Corey May"
],
"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:\nAssassin's Creed is a video game whose creators include an Italian video game writer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Assassin's Creed",
"Corey May"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1482,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1887
|
Man of Steel is a 2013 superhero film directed by an American film director and producer who made his feature film debut in Dawn of the Dead.
|
{
"evidence": [
"Man of Steel (film)",
"Zack Snyder"
],
"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:\nMan of Steel is a 2013 superhero film directed by an American film director and producer who made his feature film debut in Dawn of the Dead.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Man of Steel (film)",
"Zack Snyder"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1483,
"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.