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 |
|---|---|---|---|---|---|---|---|---|
94812
|
Kjellaug Nakkim was a member of Moss city council from 1975 to 1988 and served as chair of the Broadcasting Council, an advisory board to the Norwegian Broadcasting Corporation, from 2006 to 2009.
|
{
"evidence": [
"Kjellaug Nakkim_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:\nKjellaug Nakkim was a member of Moss city council from 1975 to 1988 and served as chair of the Broadcasting Council, an advisory board to the Norwegian Broadcasting Corporation, from 2006 to 2009.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Kjellaug Nakkim_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6884,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.602",
"value": "?search="
},
{
"operation": "search",
"time": "8.909",
"value": "Kjellaug Nakkim"
},
{
"operation": "Now on",
"time": "9.868",
"value": "Kjellaug Nakkim"
},
{
"operation": "Highlighting",
"time": "15.077",
"value": "Kjellaug Nakkim_sentence_3"
},
{
"operation": "finish",
"time": "32.657",
"value": "finish"
}
]
}
|
32143
|
Isan people overran the Lao Wang and forced them into corvee Iintermitten unpaid labour expected by landowners or monarchs) labour before they were integrated into Isan farming; much of Isan was populated this way.
|
{
"evidence": [
"Isan people_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:\nIsan people overran the Lao Wang and forced them into corvee Iintermitten unpaid labour expected by landowners or monarchs) labour before they were integrated into Isan farming; much of Isan was populated this way.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Isan people_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6885,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.688",
"value": "?search="
},
{
"operation": "search",
"time": "19.68",
"value": "Isan people"
},
{
"operation": "Now on",
"time": "20.685",
"value": "Isan people"
},
{
"operation": "search",
"time": "55.599",
"value": "lao wang"
},
{
"operation": "Now on",
"time": "57.403",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "63.42",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "63.547",
"value": "Isan people"
},
{
"operation": "Page search",
"time": "69.941",
"value": "lao wang"
},
{
"operation": "search",
"time": "90.143",
"value": "Lao wang"
},
{
"operation": "Now on",
"time": "90.763",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "102.588",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "102.718",
"value": "Isan people"
},
{
"operation": "search",
"time": "114.804",
"value": "overran"
},
{
"operation": "Now on",
"time": "115.534",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "124.556",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "124.699",
"value": "Isan people"
},
{
"operation": "Page search",
"time": "129.652",
"value": "overran"
},
{
"operation": "Highlighting",
"time": "133.105",
"value": "Isan people_sentence_1"
},
{
"operation": "search",
"time": "167.422",
"value": "san people overran the Lao Wang"
},
{
"operation": "Now on",
"time": "168.509",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "175.966",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "176.084",
"value": "Isan people"
},
{
"operation": "search",
"time": "197.322",
"value": "corvee Iintermitten unpaid labour"
},
{
"operation": "Now on",
"time": "198.335",
"value": "undefined"
},
{
"operation": "search",
"time": "211.559",
"value": "Isan people"
},
{
"operation": "Now on",
"time": "212.187",
"value": "Isan people"
},
{
"operation": "finish",
"time": "228.97",
"value": "finish"
}
]
}
|
65534
|
The corrosive class involves acids along with bases, which should be cautioned when shipping to ensure no mixing happens.
|
{
"evidence": [
"HAZMAT Class 8 Corrosive substances_sentence_3",
"HAZMAT Class 8 Corrosive substances_sentence_4"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe corrosive class involves acids along with bases, which should be cautioned when shipping to ensure no mixing happens.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"HAZMAT Class 8 Corrosive substances_sentence_3",
"HAZMAT Class 8 Corrosive substances_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6886,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.133",
"value": "?search="
},
{
"operation": "search",
"time": "70.556",
"value": "HAZMAT Class 8 Corrosive substances"
},
{
"operation": "Now on",
"time": "71.949",
"value": "HAZMAT Class 8 Corrosive substances"
},
{
"operation": "Highlighting",
"time": "79.768",
"value": "HAZMAT Class 8 Corrosive substances_sentence_3"
},
{
"operation": "Highlighting",
"time": "91.398",
"value": "HAZMAT Class 8 Corrosive substances_sentence_4"
},
{
"operation": "finish",
"time": "101.143",
"value": "finish"
}
]
}
|
exfever_1964
|
Kendrick Lamar is an American rapper, songwriter, and record producer who is also known as a member of the supergroup Black Hippy, alongside Ab-Soul, Schoolboy Q, and a rapper from Watts, Texas.
|
{
"evidence": [
"Kendrick Lamar",
"Jay 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:\nKendrick Lamar is an American rapper, songwriter, and record producer who is also known as a member of the supergroup Black Hippy, alongside Ab-Soul, Schoolboy Q, and a rapper from Watts, Texas.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Kendrick Lamar",
"Jay Rock"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6887,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
88123
|
Omar Abdillahi Charmarke competed in four local marathon competitions from 1985 to 1992.
|
{
"evidence": [
"Omar Abdillahi Charmarke_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:\nOmar Abdillahi Charmarke competed in four local marathon competitions from 1985 to 1992.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Omar Abdillahi Charmarke_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6888,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.01",
"value": "?search="
},
{
"operation": "search",
"time": "3.998",
"value": "Omar Abdillahi Charmarke"
},
{
"operation": "Now on",
"time": "4.589",
"value": "Omar Abdillahi Charmarke"
},
{
"operation": "Highlighting",
"time": "26.037",
"value": "Omar Abdillahi Charmarke_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "31.153",
"value": "Omar Abdillahi Charmarke_sentence_0"
},
{
"operation": "Highlighting",
"time": "31.795",
"value": "Omar Abdillahi Charmarke_sentence_1"
},
{
"operation": "finish",
"time": "43.31",
"value": "finish"
}
]
}
|
86590
|
Wolman studied geography and environmental studies at Middlebury College in Vermont, then later received his Master of Arts degree in journalism from Stanford University.
|
{
"evidence": [
"David Wolman_sentence_11",
"David Wolman_sentence_13"
],
"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:\nWolman studied geography and environmental studies at Middlebury College in Vermont, then later received his Master of Arts degree in journalism from Stanford University.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"David Wolman_sentence_11",
"David Wolman_sentence_13"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6889,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.972",
"value": "?search="
},
{
"operation": "search",
"time": "150.235",
"value": "Wolman"
},
{
"operation": "Now on",
"time": "150.808",
"value": "Wolman"
},
{
"operation": "hyperlink",
"time": "166.916",
"value": "M. Gordon Wolman"
},
{
"operation": "Now on",
"time": "167.447",
"value": "M. Gordon Wolman"
},
{
"operation": "hyperlink",
"time": "177.711",
"value": "Abel Wolman"
},
{
"operation": "Now on",
"time": "178.449",
"value": "Abel Wolman"
},
{
"operation": "back-button-clicked",
"time": "188.888",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "188.912",
"value": "M. Gordon Wolman"
},
{
"operation": "back-button-clicked",
"time": "189.505",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "189.507",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "189.529",
"value": "Wolman"
},
{
"operation": "hyperlink",
"time": "204.9",
"value": "William Wolman"
},
{
"operation": "Now on",
"time": "205.453",
"value": "William Wolman"
},
{
"operation": "back-button-clicked",
"time": "210.154",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "210.178",
"value": "Wolman"
},
{
"operation": "search",
"time": "245.863",
"value": "david wolman"
},
{
"operation": "Now on",
"time": "247.031",
"value": "David Wolman"
},
{
"operation": "Highlighting",
"time": "252.393",
"value": "David Wolman_sentence_11"
},
{
"operation": "Highlighting",
"time": "259.846",
"value": "David Wolman_sentence_13"
},
{
"operation": "finish",
"time": "288.986",
"value": "finish"
}
]
}
|
exfever_8957
|
Harold Ramis was an American actor, comedian, director, and writer, best known for his film acting roles in a 1981 American war comedy film directed by Norman Jewison.
|
{
"evidence": [
"Harold Ramis",
"Stripes (film)"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHarold Ramis was an American actor, comedian, director, and writer, best known for his film acting roles in a 1981 American war comedy film directed by Norman Jewison.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Harold Ramis",
"Stripes (film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6890,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
12654
|
John Clark started to work on the Eureka in 1930 as a way to automate Latin verse creation; he exhibited his machine in 1845 in the Egyptian Hall, an exhibition hall in London which eventually became a venue for exhibiting major works of art.
|
{
"evidence": [
"The Eureka_sentence_0",
"The Eureka_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:\nJohn Clark started to work on the Eureka in 1930 as a way to automate Latin verse creation; he exhibited his machine in 1845 in the Egyptian Hall, an exhibition hall in London which eventually became a venue for exhibiting major works of art.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"The Eureka_sentence_0",
"The Eureka_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6891,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.6",
"value": "?search="
},
{
"operation": "search",
"time": "7.865",
"value": "John Clark"
},
{
"operation": "Now on",
"time": "8.857",
"value": "John Clark"
},
{
"operation": "search",
"time": "70.654",
"value": "John Clark Eureka"
},
{
"operation": "Now on",
"time": "72.071",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "75.273",
"value": "Euphonia (device)"
},
{
"operation": "Now on",
"time": "75.819",
"value": "Euphonia (device)"
},
{
"operation": "search",
"time": "92.934",
"value": "John Clark"
},
{
"operation": "Now on",
"time": "93.691",
"value": "John Clark"
},
{
"operation": "back-button-clicked",
"time": "97.19",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "97.289",
"value": "Euphonia (device)"
},
{
"operation": "Page search",
"time": "101.798",
"value": "John Clark"
},
{
"operation": "search",
"time": "124.991",
"value": "Eureka machine"
},
{
"operation": "Now on",
"time": "126.539",
"value": "undefined"
},
{
"operation": "search",
"time": "165.371",
"value": "Eureka"
},
{
"operation": "Now on",
"time": "166.167",
"value": "Eureka"
},
{
"operation": "hyperlink",
"time": "211.873",
"value": "The Eureka"
},
{
"operation": "Now on",
"time": "212.423",
"value": "The Eureka"
},
{
"operation": "Highlighting",
"time": "217.482",
"value": "The Eureka_sentence_0"
},
{
"operation": "Highlighting",
"time": "223.252",
"value": "The Eureka_sentence_3"
},
{
"operation": "finish",
"time": "230.796",
"value": "finish"
}
]
}
|
exfever_1861
|
Wynonna Judd is a multi-award-winning American singer whose genre of popular music originated with the blues, including a music, which is the emblematic music of Louisiana.
|
{
"evidence": [
"Wynonna Judd",
"Country music",
"Cajun music"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWynonna Judd is a multi-award-winning American singer whose genre of popular music originated with the blues, including a music, which is the emblematic music of Louisiana.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Wynonna Judd",
"Country music",
"Cajun music"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6892,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5951
|
The Joker, created by Bill Finger, is a supervillain appearing in Japanese comic books published by an American comic book publisher and the flagship unit of DC Entertainment, a subsidiary of Warner Bros. Discovery.
|
{
"evidence": [
"Joker (character)",
"DC 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:\nThe Joker, created by Bill Finger, is a supervillain appearing in Japanese comic books published by an American comic book publisher and the flagship unit of DC Entertainment, a subsidiary of Warner Bros. Discovery.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Joker (character)",
"DC Comics"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6893,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4939
|
Spider-Man 3 is the final installment in Raimi's Spider-Man trilogy and the sequel to Spider-Man and the 2005 American superhero film directed by Sam Raimi and written by Alvin Sargent.
|
{
"evidence": [
"Spider-Man 3",
"Spider-Man 2"
],
"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:\nSpider-Man 3 is the final installment in Raimi's Spider-Man trilogy and the sequel to Spider-Man and the 2005 American superhero film directed by Sam Raimi and written by Alvin Sargent.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Spider-Man 3",
"Spider-Man 2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6894,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
1229
|
The RAHI Foundation is a support center based in New Delhi and is the only center of its kind that specializes in working with adult women survivors of incest and childhood sexual abuse.
|
{
"evidence": [
"RAHI Foundation_sentence_0",
"RAHI Foundation_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:\nThe RAHI Foundation is a support center based in New Delhi and is the only center of its kind that specializes in working with adult women survivors of incest and childhood sexual abuse.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"RAHI Foundation_sentence_0",
"RAHI Foundation_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6895,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "11.881",
"value": "RAHI Foundation"
},
{
"operation": "Highlighting",
"time": "31.663",
"value": "RAHI Foundation_sentence_0"
},
{
"operation": "Highlighting",
"time": "35.083",
"value": "RAHI Foundation_sentence_1"
},
{
"operation": "finish",
"time": "92.805",
"value": "finish"
}
]
}
|
exfever_4352
|
Wayne Rooney spent much of his career on the field between his team's defenders and forwards.
|
{
"evidence": [
"Wayne Rooney",
"Midfielder"
],
"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:\nWayne Rooney spent much of his career on the field between his team's defenders and forwards.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Wayne Rooney",
"Midfielder"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6896,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3413
|
Elise Demetria Neal's final film appearance is in Logan.
|
{
"evidence": [
"Elise Neal",
"UPN"
],
"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:\nElise Demetria Neal's final film appearance is in Logan.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Elise Neal",
"UPN"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6897,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
56697
|
School Profiles, produced at least once in each academic year including standard data information and summary of school budget, were required to be created by all maintained schools in England, except for maintained nursery schools.
|
{
"evidence": [
"School profile_sentence_5",
"School profile_sentence_3",
"School profile_sentence_7"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSchool Profiles, produced at least once in each academic year including standard data information and summary of school budget, were required to be created by all maintained schools in England, except for maintained nursery schools.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"School profile_sentence_5",
"School profile_sentence_3",
"School profile_sentence_7"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6898,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.736",
"value": "?search="
},
{
"operation": "search",
"time": "10.439",
"value": "School profile"
},
{
"operation": "Now on",
"time": "11.151",
"value": "School profile"
},
{
"operation": "Highlighting",
"time": "31.316",
"value": "School profile_sentence_5"
},
{
"operation": "Page search",
"time": "67.798",
"value": "budget"
},
{
"operation": "Page search",
"time": "75.805",
"value": "data"
},
{
"operation": "Highlighting",
"time": "131.531",
"value": "School profile_sentence_2"
},
{
"operation": "Highlighting",
"time": "134.511",
"value": "School profile_sentence_3"
},
{
"operation": "Page search",
"time": "163.163",
"value": "academic year"
},
{
"operation": "Highlighting",
"time": "168.215",
"value": "School profile_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "196.259",
"value": "School profile_sentence_2"
},
{
"operation": "hyperlink",
"time": "198.813",
"value": "National Curriculum assessments"
},
{
"operation": "Now on",
"time": "199.593",
"value": "National Curriculum assessment"
},
{
"operation": "Page search",
"time": "203.116",
"value": "budget"
},
{
"operation": "back-button-clicked",
"time": "207.02",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "207.092",
"value": "School profile"
},
{
"operation": "finish",
"time": "213.015",
"value": "finish"
}
]
}
|
10873
|
There were more seats in the 1962 Lok Sabha general elections than the one in 1967.
|
{
"evidence": [
"1962 Indian general election_sentence_3",
"1967 Indian general election_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:\nThere were more seats in the 1962 Lok Sabha general elections than the one in 1967.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"1962 Indian general election_sentence_3",
"1967 Indian general election_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6899,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.836",
"value": "?search="
},
{
"operation": "search",
"time": "41.063",
"value": "lok sabha general elections"
},
{
"operation": "Now on",
"time": "42.823",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "56.907",
"value": "17th Lok Sabha"
},
{
"operation": "Now on",
"time": "57.915",
"value": "Kokrajhar (Lok Sabha constituency)#17th Lok Sabha"
},
{
"operation": "hyperlink",
"time": "71.416",
"value": "Lok Sabha"
},
{
"operation": "Now on",
"time": "72.586",
"value": "Lok Sabha"
},
{
"operation": "hyperlink",
"time": "92.014",
"value": "6 Lok Sabha general elections Lok Sabha_section_22"
},
{
"operation": "hyperlink",
"time": "98.952",
"value": "1962 Indian general election"
},
{
"operation": "Now on",
"time": "99.956",
"value": "1962 Indian general election"
},
{
"operation": "Highlighting",
"time": "114.631",
"value": "1962 Indian general election_sentence_3"
},
{
"operation": "back-button-clicked",
"time": "116.654",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "116.946",
"value": "Lok Sabha#Lok Sabha general elections Lok Sabha section 22"
},
{
"operation": "hyperlink",
"time": "120.477",
"value": "6 Lok Sabha general elections Lok Sabha_section_22"
},
{
"operation": "hyperlink",
"time": "126.191",
"value": "1967 Indian general election"
},
{
"operation": "Now on",
"time": "126.931",
"value": "1967 Indian general election"
},
{
"operation": "Highlighting",
"time": "160.673",
"value": "1967 Indian general election_sentence_1"
},
{
"operation": "finish",
"time": "183.47",
"value": "finish"
}
]
}
|
16866
|
The Webster–Ashburton Treaty allowed Canada to complete construction of a partially finished fort on contested lands.
|
{
"evidence": [
"Webster–Ashburton Treaty_sentence_9",
"Webster–Ashburton Treaty_sentence_10",
"Webster–Ashburton Treaty_sentence_11"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Webster–Ashburton Treaty allowed Canada to complete construction of a partially finished fort on contested lands.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Webster–Ashburton Treaty_sentence_9",
"Webster–Ashburton Treaty_sentence_10",
"Webster–Ashburton Treaty_sentence_11"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6900,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.189",
"value": "?search="
},
{
"operation": "search",
"time": "37.414",
"value": "Webster–Ashburton Treaty"
},
{
"operation": "Now on",
"time": "38.42",
"value": "Webster–Ashburton Treaty"
},
{
"operation": "Page search",
"time": "114.165",
"value": "fort"
},
{
"operation": "Highlighting",
"time": "127.829",
"value": "Webster–Ashburton Treaty_sentence_9"
},
{
"operation": "Highlighting",
"time": "133.042",
"value": "Webster–Ashburton Treaty_sentence_10"
},
{
"operation": "Highlighting",
"time": "139.456",
"value": "Webster–Ashburton Treaty_sentence_11"
},
{
"operation": "Highlighting",
"time": "153.95",
"value": "Webster–Ashburton Treaty_sentence_19"
},
{
"operation": "Highlighting deleted",
"time": "178.388",
"value": "Webster–Ashburton Treaty_sentence_19"
},
{
"operation": "finish",
"time": "182.085",
"value": "finish"
}
]
}
|
95267
|
Tin Hau Food Square, which featured Hong Kong's characteristic street foods, was where Mr. Wong opened his Chinese dessert shop called Ching Ching Dessert and has since became a potential location for people to start their businesses.
|
{
"evidence": [
"Tin Hau Food Square_sentence_20",
"Tin Hau Food Square_sentence_6",
"Tin Hau Food Square_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:\nTin Hau Food Square, which featured Hong Kong's characteristic street foods, was where Mr. Wong opened his Chinese dessert shop called Ching Ching Dessert and has since became a potential location for people to start their businesses.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Tin Hau Food Square_sentence_20",
"Tin Hau Food Square_sentence_6",
"Tin Hau Food Square_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6901,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.194",
"value": "?search="
},
{
"operation": "search",
"time": "7.556",
"value": "Tin Hau Food Square"
},
{
"operation": "Now on",
"time": "8.496",
"value": "Tin Hau Food Square"
},
{
"operation": "Highlighting",
"time": "70.152",
"value": "Tin Hau Food Square_sentence_13"
},
{
"operation": "Highlighting",
"time": "70.154",
"value": "Tin Hau Food Square_section_4"
},
{
"operation": "Highlighting",
"time": "94.537",
"value": "Tin Hau Food Square_sentence_20"
},
{
"operation": "Highlighting",
"time": "101.697",
"value": "Tin Hau Food Square_sentence_6"
},
{
"operation": "Highlighting",
"time": "131.837",
"value": "Tin Hau Food Square_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "150.921",
"value": "Tin Hau Food Square_sentence_13"
},
{
"operation": "Highlighting deleted",
"time": "150.924",
"value": "Tin Hau Food Square_section_4"
},
{
"operation": "finish",
"time": "192.411",
"value": "finish"
}
]
}
|
76970
|
In addition to Resident Representative, Pedro Agulto Tenorio was the second Senator of the Northern Mariana Islands a position that takes over for the governor upon his death, removal, or resignation.
|
{
"evidence": [
"Pedro Agulto Tenorio_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:\nIn addition to Resident Representative, Pedro Agulto Tenorio was the second Senator of the Northern Mariana Islands a position that takes over for the governor upon his death, removal, or resignation.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Pedro Agulto Tenorio_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6902,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.165",
"value": "?search="
},
{
"operation": "search",
"time": "101.437",
"value": "Pedro Agulto Tenorio"
},
{
"operation": "Now on",
"time": "102.157",
"value": "Pedro Agulto Tenorio"
},
{
"operation": "Highlighting",
"time": "104.713",
"value": "Pedro Agulto Tenorio_sentence_1"
},
{
"operation": "Highlighting",
"time": "276.658",
"value": "Pedro Agulto Tenorio_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "277.846",
"value": "Pedro Agulto Tenorio_sentence_1"
},
{
"operation": "finish",
"time": "334.832",
"value": "finish"
}
]
}
|
18726
|
Pascal Gastien was appointed as manager of Niort in January 2005 and in June 2009.
|
{
"evidence": [
"Pascal Gastien_sentence_27",
"Pascal Gastien_sentence_30"
],
"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:\nPascal Gastien was appointed as manager of Niort in January 2005 and in June 2009.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Pascal Gastien_sentence_27",
"Pascal Gastien_sentence_30"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6903,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.48",
"value": "?search="
},
{
"operation": "search",
"time": "11.3",
"value": "Pascal Gastien"
},
{
"operation": "Now on",
"time": "13.139",
"value": "Pascal Gastien"
},
{
"operation": "Page search",
"time": "31.446",
"value": "manager"
},
{
"operation": "Highlighting",
"time": "47.435",
"value": "Pascal Gastien_sentence_25"
},
{
"operation": "Highlighting deleted",
"time": "59.858",
"value": "Pascal Gastien_sentence_25"
},
{
"operation": "Highlighting",
"time": "63.007",
"value": "Pascal Gastien_sentence_27"
},
{
"operation": "Highlighting",
"time": "66.608",
"value": "Pascal Gastien_sentence_30"
},
{
"operation": "Highlighting deleted",
"time": "75.608",
"value": "Pascal Gastien_sentence_27"
},
{
"operation": "Highlighting",
"time": "76.693",
"value": "Pascal Gastien_sentence_26"
},
{
"operation": "Highlighting deleted",
"time": "84.517",
"value": "Pascal Gastien_sentence_30"
},
{
"operation": "Highlighting deleted",
"time": "89.442",
"value": "Pascal Gastien_sentence_26"
},
{
"operation": "Highlighting",
"time": "91.062",
"value": "Pascal Gastien_sentence_27"
},
{
"operation": "Highlighting",
"time": "99.833",
"value": "Pascal Gastien_sentence_30"
},
{
"operation": "finish",
"time": "110.925",
"value": "finish"
}
]
}
|
60770
|
Simonin, born on July 1, 1991, spent the first two seasons of his college career at Lander University; then helped lead the Bearcats to a Peach Belt Conference regular season title in 2012.
|
{
"evidence": [
"Clément Simonin_sentence_0",
"Clément Simonin_sentence_3",
"Clément Simonin_sentence_4"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSimonin, born on July 1, 1991, spent the first two seasons of his college career at Lander University; then helped lead the Bearcats to a Peach Belt Conference regular season title in 2012.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Clément Simonin_sentence_0",
"Clément Simonin_sentence_3",
"Clément Simonin_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6904,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.25",
"value": "?search="
},
{
"operation": "search",
"time": "68.635",
"value": "Simonin"
},
{
"operation": "Now on",
"time": "69.416",
"value": "Albert Simonin"
},
{
"operation": "search",
"time": "94.273",
"value": "Contains...Simonin"
},
{
"operation": "Now on",
"time": "95.3",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "122.166",
"value": "Clément Simonin"
},
{
"operation": "Now on",
"time": "122.898",
"value": "Clément Simonin"
},
{
"operation": "Highlighting",
"time": "126.728",
"value": "Clément Simonin_sentence_0"
},
{
"operation": "Highlighting",
"time": "155.811",
"value": "Clément Simonin_sentence_3"
},
{
"operation": "Highlighting",
"time": "171.706",
"value": "Clément Simonin_sentence_4"
},
{
"operation": "finish",
"time": "344.933",
"value": "finish"
}
]
}
|
44923
|
No Basta received a Billboard Music nomination, but was never rated by Latin single of the year which charts Top Latin Songs.
|
{
"evidence": [
"No Basta_sentence_7"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nNo Basta received a Billboard Music nomination, but was never rated by Latin single of the year which charts Top Latin Songs.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"No Basta_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6905,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.914",
"value": "?search="
},
{
"operation": "search",
"time": "7.52",
"value": "No Basta"
},
{
"operation": "Now on",
"time": "8.454",
"value": "No Basta"
},
{
"operation": "Highlighting",
"time": "15.53",
"value": "No Basta_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "24.35",
"value": "No Basta_sentence_4"
},
{
"operation": "hyperlink",
"time": "67.022",
"value": "eighth best performing Latin single of the year"
},
{
"operation": "Now on",
"time": "70.342",
"value": "Billboard Top Latin Songs Year-End Chart"
},
{
"operation": "Highlighting",
"time": "76.663",
"value": "Billboard Top Latin Songs Year-End Chart_sentence_0"
},
{
"operation": "Highlighting",
"time": "76.664",
"value": "Billboard Top Latin Songs Year-End Chart_title"
},
{
"operation": "back-button-clicked",
"time": "81.716",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "81.765",
"value": "No Basta"
},
{
"operation": "Highlighting",
"time": "90.422",
"value": "No Basta_sentence_7"
},
{
"operation": "finish",
"time": "114.115",
"value": "finish"
}
]
}
|
85464
|
Lifson–Roig model is a helix-coil transition model applied to the alpha helix-random coil transition of polypeptides, a longer, continuous, unbranched peptide, and is one of the first two in a series analogous transfer-matrix methods in polymer science that have also been applied to nucleic acids and branched polymers.
|
{
"evidence": [
"Lifson–Roig model_sentence_0",
"Peptide_sentence_4",
"Lifson–Roig model_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:\nLifson–Roig model is a helix-coil transition model applied to the alpha helix-random coil transition of polypeptides, a longer, continuous, unbranched peptide, and is one of the first two in a series analogous transfer-matrix methods in polymer science that have also been applied to nucleic acids and branched polymers.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Lifson–Roig model_sentence_0",
"Peptide_sentence_4",
"Lifson–Roig model_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6906,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.822",
"value": "?search="
},
{
"operation": "search",
"time": "13.882",
"value": "Lifson–Roig model"
},
{
"operation": "Now on",
"time": "14.788",
"value": "Lifson–Roig model"
},
{
"operation": "Highlighting",
"time": "24.699",
"value": "Lifson–Roig model_sentence_0"
},
{
"operation": "hyperlink",
"time": "40.863",
"value": "random coil"
},
{
"operation": "Now on",
"time": "41.54",
"value": "Random coil"
},
{
"operation": "back-button-clicked",
"time": "56.89",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "56.996",
"value": "Lifson–Roig model"
},
{
"operation": "hyperlink",
"time": "59.315",
"value": "alpha helix"
},
{
"operation": "Now on",
"time": "60.266",
"value": "Alpha helix"
},
{
"operation": "search",
"time": "90.269",
"value": "Contains...longer, continuous, unbranched peptide"
},
{
"operation": "Now on",
"time": "91.746",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "106.112",
"value": "Antimicrobial peptides Peptide_section_3"
},
{
"operation": "Now on",
"time": "107.012",
"value": "Peptide#Antimicrobial peptides Peptide section 3"
},
{
"operation": "Page search",
"time": "125.192",
"value": "longer, continuous, unbranched peptide"
},
{
"operation": "Highlighting",
"time": "128.765",
"value": "Peptide_sentence_4"
},
{
"operation": "search",
"time": "155.965",
"value": "Lifson–Roig model"
},
{
"operation": "Now on",
"time": "156.784",
"value": "Lifson–Roig model"
},
{
"operation": "Highlighting",
"time": "187.906",
"value": "Lifson–Roig model_sentence_3"
},
{
"operation": "finish",
"time": "223.223",
"value": "finish"
}
]
}
|
exfever_5589
|
Christopher Nolan's 2015 film Batman Begins focuses on the actions of individuals who usually possess extraordinary abilities.
|
{
"evidence": [
"Batman Begins",
"Superhero film"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nChristopher Nolan's 2015 film Batman Begins focuses on the actions of individuals who usually possess extraordinary abilities.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Batman Begins",
"Superhero film"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6907,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_916
|
Mark Alan Ruffalo is an American actor who appeared in a 2000 American drama film written and directed by an American film director who co-wrote the film Gangs of New York.
|
{
"evidence": [
"Mark Ruffalo",
"You Can Count On Me",
"Kenneth Lonergan"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMark Alan Ruffalo is an American actor who appeared in a 2000 American drama film written and directed by an American film director who co-wrote the film Gangs of New York.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mark Ruffalo",
"You Can Count On Me",
"Kenneth Lonergan"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6908,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2019
|
The Fifth Element was written by two screenwriters, including the creator of The Karate Kid franchise, who collaborated with a Belgian filmmaker associated with the Cinéma du Look film movement.
|
{
"evidence": [
"The Fifth Element",
"Robert Mark Kamen",
"Luc Besson"
],
"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 Fifth Element was written by two screenwriters, including the creator of The Karate Kid franchise, who collaborated with a Belgian filmmaker associated with the Cinéma du Look film movement.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Fifth Element",
"Robert Mark Kamen",
"Luc Besson"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6909,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
77270
|
In the Czech Republic in the Eurovision Song Contest 2007, "Malá dáma," written and performed by Petr Bende (a hard rock band hailing from Teplice, Czech Republic) came in first place with 28,343 televotes.
|
{
"evidence": [
"Czech Republic in the Eurovision Song Contest 2007_sentence_13"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn the Czech Republic in the Eurovision Song Contest 2007, \"Malá dáma,\" written and performed by Petr Bende (a hard rock band hailing from Teplice, Czech Republic) came in first place with 28,343 televotes.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Czech Republic in the Eurovision Song Contest 2007_sentence_13"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6910,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.021",
"value": "?search="
},
{
"operation": "search",
"time": "26.536",
"value": "2007 Eurovision Song Contest"
},
{
"operation": "Now on",
"time": "28.506",
"value": "Eurovision Song Contest 2007"
},
{
"operation": "Page search",
"time": "32.68",
"value": "dama"
},
{
"operation": "Page search",
"time": "68.874",
"value": "petr bend"
},
{
"operation": "Page search",
"time": "172.554",
"value": "win"
},
{
"operation": "Highlighting",
"time": "175.202",
"value": "Eurovision Song Contest 2007_sentence_3"
},
{
"operation": "Highlighting",
"time": "175.203",
"value": "Eurovision Song Contest 2007_title"
},
{
"operation": "Highlighting deleted",
"time": "182.239",
"value": "Eurovision Song Contest 2007_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "182.241",
"value": "Eurovision Song Contest 2007_title"
},
{
"operation": "Highlighting",
"time": "189.534",
"value": "Eurovision Song Contest 2007_sentence_13"
},
{
"operation": "Highlighting",
"time": "189.536",
"value": "Eurovision Song Contest 2007_title"
},
{
"operation": "Page search",
"time": "213.047",
"value": "czech"
},
{
"operation": "hyperlink",
"time": "223.418",
"value": "Czech Republic"
},
{
"operation": "Now on",
"time": "224.258",
"value": "Czech Republic in the Eurovision Song Contest"
},
{
"operation": "Page search",
"time": "229.556",
"value": "dama"
},
{
"operation": "Page search",
"time": "260.624",
"value": "2007"
},
{
"operation": "Highlighting",
"time": "263.063",
"value": "Czech Republic in the Eurovision Song Contest_sentence_0"
},
{
"operation": "Highlighting",
"time": "263.063",
"value": "Czech Republic in the Eurovision Song Contest_title"
},
{
"operation": "hyperlink",
"time": "270.219",
"value": "2007"
},
{
"operation": "Now on",
"time": "271.912",
"value": "Eurovision Song Contest 2007"
},
{
"operation": "back-button-clicked",
"time": "273.611",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "273.822",
"value": "Czech Republic in the Eurovision Song Contest"
},
{
"operation": "Highlighting deleted",
"time": "277.417",
"value": "Czech Republic in the Eurovision Song Contest_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "277.419",
"value": "Czech Republic in the Eurovision Song Contest_title"
},
{
"operation": "Page search",
"time": "289.447",
"value": "2007"
},
{
"operation": "Highlighting",
"time": "297.517",
"value": "Czech Republic in the Eurovision Song Contest_sentence_13"
},
{
"operation": "Highlighting",
"time": "297.518",
"value": "Czech Republic in the Eurovision Song Contest_title"
},
{
"operation": "Highlighting",
"time": "297.519",
"value": "Czech Republic in the Eurovision Song Contest_section_1"
},
{
"operation": "Highlighting",
"time": "297.52",
"value": "Czech Republic in the Eurovision Song Contest_section_2"
},
{
"operation": "hyperlink",
"time": "301.679",
"value": "Czech Republic in the Eurovision Song Contest 2007"
},
{
"operation": "Now on",
"time": "302.401",
"value": "Czech Republic in the Eurovision Song Contest 2007"
},
{
"operation": "Page search",
"time": "306.711",
"value": "dama"
},
{
"operation": "Highlighting",
"time": "342.346",
"value": "Czech Republic in the Eurovision Song Contest 2007_sentence_13"
},
{
"operation": "finish",
"time": "369.83",
"value": "finish"
}
]
}
|
exfever_3242
|
Hamlet is a 1990 drama film directed by Franco Zeffirelli and starring talented artists, including an English actor who came to prominence in the 1960s.
|
{
"evidence": [
"Hamlet (1990 film)",
"Alan Bates"
],
"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:\nHamlet is a 1990 drama film directed by Franco Zeffirelli and starring talented artists, including an English actor who came to prominence in the 1960s.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hamlet (1990 film)",
"Alan Bates"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6911,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2898
|
Kitti's hog-nosed bat, also known as the bumblebee bat, is found in eastern Siam, a Southeast Asian country.
|
{
"evidence": [
"Kitti's hog-nosed bat",
"Thailand"
],
"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:\nKitti's hog-nosed bat, also known as the bumblebee bat, is found in eastern Siam, a Southeast Asian country.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Kitti's hog-nosed bat",
"Thailand"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6912,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1452
|
Luxembourg's capital is Luxembourg City, one of the three official capitals of the political and economic union of 27 member states that are located primarily in Europe.
|
{
"evidence": [
"Luxembourg",
"European Union"
],
"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:\nLuxembourg's capital is Luxembourg City, one of the three official capitals of the political and economic union of 27 member states that are located primarily in Europe.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Luxembourg",
"European Union"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6913,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8358
|
Daniel Robert Elfman is the first American composer who received more than five Saturn Awards for Best Music.
|
{
"evidence": [
"Danny Elfman",
"Film score"
],
"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:\nDaniel Robert Elfman is the first American composer who received more than five Saturn Awards for Best Music.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Danny Elfman",
"Film score"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6914,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6122
|
Benjamin Géza Affleck-Boldt is an American actor and filmmaker who appeared in a 1997 American comedy-drama film and focuses on passion, emotion, and the affectionate romantic involvement of the main characters.
|
{
"evidence": [
"Ben Affleck",
"Chasing Amy",
"Romance 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:\nBenjamin Géza Affleck-Boldt is an American actor and filmmaker who appeared in a 1997 American comedy-drama film and focuses on passion, emotion, and the affectionate romantic involvement of the main characters.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ben Affleck",
"Chasing Amy",
"Romance film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6915,
"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": 6916,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7889
|
New Horizons was launched as a part of a series of space exploration missions being conducted by NASA.
|
{
"evidence": [
"New Horizons",
"New Frontiers program"
],
"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:\nNew Horizons was launched as a part of a series of space exploration missions being conducted by NASA.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"New Horizons",
"New Frontiers program"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6917,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
28809
|
A Pass met Lukyamuzi Bashir, a music and film director and producer from Uganda, who recognized his talent and took him to various studios and introduced him to influential people in the music business to help with his career.
|
{
"evidence": [
"A Pass_sentence_6",
"Lukyamuzi Bashir_sentence_0",
"A Pass_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:\nA Pass met Lukyamuzi Bashir, a music and film director and producer from Uganda, who recognized his talent and took him to various studios and introduced him to influential people in the music business to help with his career.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"A Pass_sentence_6",
"Lukyamuzi Bashir_sentence_0",
"A Pass_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6918,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.675",
"value": "undefined"
},
{
"operation": "search",
"time": "9.331",
"value": "Contains...Pass met Lukyamuzi Bashir"
},
{
"operation": "Now on",
"time": "10.359",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "19.639",
"value": "A Pass"
},
{
"operation": "Now on",
"time": "20.453",
"value": "A Pass"
},
{
"operation": "Highlighting",
"time": "35.998",
"value": "A Pass_sentence_6"
},
{
"operation": "hyperlink",
"time": "38.499",
"value": "Lukyamuzi Bashir"
},
{
"operation": "Now on",
"time": "39.232",
"value": "Lukyamuzi Bashir"
},
{
"operation": "Highlighting",
"time": "42.825",
"value": "Lukyamuzi Bashir_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "45.05",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "45.148",
"value": "A Pass"
},
{
"operation": "Highlighting",
"time": "55.168",
"value": "A Pass_sentence_7"
},
{
"operation": "finish",
"time": "71.851",
"value": "finish"
}
]
}
|
54753
|
Slim Amamou was against the Tunisian Revolution.
|
{
"evidence": [
"Slim Amamou_sentence_4",
"Slim Amamou_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:\nSlim Amamou was against the Tunisian Revolution.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Slim Amamou_sentence_4",
"Slim Amamou_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6919,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.23",
"value": "?search="
},
{
"operation": "search",
"time": "184.357",
"value": "Slim Amamou"
},
{
"operation": "Now on",
"time": "185.456",
"value": "Slim Amamou"
},
{
"operation": "Highlighting",
"time": "451.202",
"value": "Slim Amamou_sentence_5"
},
{
"operation": "hyperlink",
"time": "452.033",
"value": "Tunisian Revolution"
},
{
"operation": "Now on",
"time": "454.8",
"value": "Tunisian Revolution"
},
{
"operation": "search",
"time": "493.756",
"value": "Slim Amamou"
},
{
"operation": "Now on",
"time": "494.694",
"value": "Slim Amamou"
},
{
"operation": "Highlighting",
"time": "517.835",
"value": "Slim Amamou_sentence_4"
},
{
"operation": "hyperlink",
"time": "518.561",
"value": "Tunisian Revolution"
},
{
"operation": "Now on",
"time": "519.159",
"value": "Tunisian Revolution"
},
{
"operation": "search",
"time": "524.253",
"value": "Slim Amamou"
},
{
"operation": "Now on",
"time": "524.912",
"value": "Slim Amamou"
},
{
"operation": "Highlighting deleted",
"time": "526.625",
"value": "Slim Amamou_sentence_5"
},
{
"operation": "Highlighting",
"time": "527.016",
"value": "Slim Amamou_sentence_5"
},
{
"operation": "finish",
"time": "540.913",
"value": "finish"
}
]
}
|
80821
|
Lucas Höler is currently a player for SC Freiburg.
|
{
"evidence": [
"Lucas Höler_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:\nLucas Höler is currently a player for SC Freiburg.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Lucas Höler_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6920,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.037",
"value": "?search="
},
{
"operation": "search",
"time": "8.327",
"value": "Lucas Höler"
},
{
"operation": "Now on",
"time": "10.103",
"value": "Lucas Höler"
},
{
"operation": "Highlighting",
"time": "50.468",
"value": "Lucas Höler_sentence_0"
},
{
"operation": "finish",
"time": "129.433",
"value": "finish"
}
]
}
|
exfever_1562
|
The 2002 Royal Rumble was the most viewed pay-per-view event in history.
|
{
"evidence": [
"Royal Rumble (2002)",
"WWE brand extension"
],
"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 2002 Royal Rumble was the most viewed pay-per-view event in history.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Royal Rumble (2002)",
"WWE brand extension"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6921,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
60257
|
The 1626 Sadeya is a stony Phocaea asteroid discovered in 10 January 1927 by J. Comas Solà in the Fabra Observatory.
|
{
"evidence": [
"1626 Sadeya_sentence_0",
"1626 Sadeya_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:\nThe 1626 Sadeya is a stony Phocaea asteroid discovered in 10 January 1927 by J. Comas Solà in the Fabra Observatory.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"1626 Sadeya_sentence_0",
"1626 Sadeya_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6922,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.302",
"value": "?search="
},
{
"operation": "Now on",
"time": "7.196",
"value": "?search="
},
{
"operation": "search",
"time": "382",
"value": "1626 Sadeya"
},
{
"operation": "Now on",
"time": "383.364",
"value": "1626 Sadeya"
},
{
"operation": "Highlighting",
"time": "458.053",
"value": "1626 Sadeya_sentence_0"
},
{
"operation": "Highlighting",
"time": "507.195",
"value": "1626 Sadeya_sentence_1"
},
{
"operation": "finish",
"time": "589.747",
"value": "finish"
}
]
}
|
27307
|
Anren County is in Hunan Province, China under administration of Chenzhou but has no residents.
|
{
"evidence": [
"Anren County_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:\nAnren County is in Hunan Province, China under administration of Chenzhou but has no residents.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Anren County_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6923,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.29",
"value": "?search="
},
{
"operation": "search",
"time": "19.457",
"value": "Anren County"
},
{
"operation": "Now on",
"time": "20.554",
"value": "Anren County"
},
{
"operation": "Highlighting",
"time": "32.903",
"value": "Anren County_sentence_3"
},
{
"operation": "finish",
"time": "44.022",
"value": "finish"
}
]
}
|
52160
|
Yuanhui District, an administrative divisions of Luohe (a prefecture-level city in central Hubei province, China), has a total population of 342,300 and a population density of 1,694.55 per sq. km.
|
{
"evidence": [
"Yuanhui 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:\nYuanhui District, an administrative divisions of Luohe (a prefecture-level city in central Hubei province, China), has a total population of 342,300 and a population density of 1,694.55 per sq. km.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Yuanhui District_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6924,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.809",
"value": "?search="
},
{
"operation": "search",
"time": "16.479",
"value": "Contains...Yuanhui District,"
},
{
"operation": "Now on",
"time": "17.932",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "19.847",
"value": "Yuanhui District"
},
{
"operation": "Now on",
"time": "20.376",
"value": "Yuanhui District"
},
{
"operation": "Highlighting",
"time": "34.39",
"value": "Yuanhui District_sentence_0"
},
{
"operation": "finish",
"time": "48.856",
"value": "finish"
}
]
}
|
65587
|
Rich Baum functioned as Spitzer's top aide and chief of staff and was involved in the Eliot Spitzer political surveillance controversy, also known as the Troopergate controversy, in 2007.
|
{
"evidence": [
"Rich Baum_sentence_3",
"Rich Baum_sentence_4",
"Eliot Spitzer political surveillance controversy_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:\nRich Baum functioned as Spitzer's top aide and chief of staff and was involved in the Eliot Spitzer political surveillance controversy, also known as the Troopergate controversy, in 2007.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Rich Baum_sentence_3",
"Rich Baum_sentence_4",
"Eliot Spitzer political surveillance controversy_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6925,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "6.728",
"value": "?search="
},
{
"operation": "search",
"time": "72.133",
"value": "Rich Baum"
},
{
"operation": "Now on",
"time": "73.044",
"value": "Rich Baum"
},
{
"operation": "Highlighting",
"time": "88.475",
"value": "Rich Baum_sentence_1"
},
{
"operation": "Highlighting",
"time": "89.634",
"value": "Rich Baum_sentence_2"
},
{
"operation": "Highlighting",
"time": "97.611",
"value": "Rich Baum_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "101.13",
"value": "Rich Baum_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "102.139",
"value": "Rich Baum_sentence_1"
},
{
"operation": "Highlighting",
"time": "106.602",
"value": "Rich Baum_sentence_4"
},
{
"operation": "hyperlink",
"time": "112.306",
"value": "Eliot Spitzer political surveillance controversy"
},
{
"operation": "Now on",
"time": "112.958",
"value": "Eliot Spitzer political surveillance controversy"
},
{
"operation": "Highlighting",
"time": "114.674",
"value": "Eliot Spitzer political surveillance controversy_sentence_0"
},
{
"operation": "finish",
"time": "125.977",
"value": "finish"
}
]
}
|
40969
|
Tomás Mac Síomóin, presently living and working in Catalonia, has poems, stories, articles and translations from Catalan—the official language of Andorra which has a semi-official status in the Italian commune of Alghero—and Spanish which appeared in various publications; however, these are the only languages his works have been translated into.
|
{
"evidence": [
"Tomás Mac Síomóin_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:\nTomás Mac Síomóin, presently living and working in Catalonia, has poems, stories, articles and translations from Catalan—the official language of Andorra which has a semi-official status in the Italian commune of Alghero—and Spanish which appeared in various publications; however, these are the only languages his works have been translated into.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Tomás Mac Síomóin_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6926,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.093",
"value": "?search="
},
{
"operation": "search",
"time": "20.92",
"value": "Tomás Mac Síomóin"
},
{
"operation": "Now on",
"time": "22.291",
"value": "Tomás Mac Síomóin"
},
{
"operation": "Highlighting",
"time": "29.048",
"value": "Tomás Mac Síomóin_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "32.448",
"value": "Tomás Mac Síomóin_sentence_2"
},
{
"operation": "hyperlink",
"time": "36.692",
"value": "2.6 Translations Tomás Mac Síomóin_section_7"
},
{
"operation": "Highlighting",
"time": "56.998",
"value": "Tomás Mac Síomóin_sentence_6"
},
{
"operation": "finish",
"time": "87.744",
"value": "finish"
}
]
}
|
71976
|
Eldar served as spokesperson for the former mayor of Jerusalem, Teddy Kollek, and was a reporter and editor at the Kol Yisrael (India's public domestic and international radio service).
|
{
"evidence": [
"Kol Yisrael_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:\nEldar served as spokesperson for the former mayor of Jerusalem, Teddy Kollek, and was a reporter and editor at the Kol Yisrael (India's public domestic and international radio service).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Kol Yisrael_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6927,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.739",
"value": "?search="
},
{
"operation": "search",
"time": "27.34",
"value": "Kol Yisrael"
},
{
"operation": "Now on",
"time": "28.069",
"value": "Kol Yisrael"
},
{
"operation": "Highlighting",
"time": "30.024",
"value": "Kol Yisrael_sentence_2"
},
{
"operation": "finish",
"time": "41.699",
"value": "finish"
}
]
}
|
56128
|
ART Holdings is a company in Northern America and has yet to expand to other continents.
|
{
"evidence": [
"ART Holdings_sentence_0",
"ART Holdings_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:\nART Holdings is a company in Northern America and has yet to expand to other continents.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"ART Holdings_sentence_0",
"ART Holdings_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6928,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.987",
"value": "?search="
},
{
"operation": "search",
"time": "3.89",
"value": "ART Holdings"
},
{
"operation": "Now on",
"time": "5.363",
"value": "ART Holdings"
},
{
"operation": "Highlighting",
"time": "8.668",
"value": "ART Holdings_sentence_0"
},
{
"operation": "Highlighting",
"time": "19.323",
"value": "ART Holdings_sentence_2"
},
{
"operation": "finish",
"time": "23.489",
"value": "finish"
}
]
}
|
exfever_7179
|
After studying acting at the Drama Centre London, Edward Thomas Hardy made his film debut in Black Hawk Down (2001) by an English film director and producer.
|
{
"evidence": [
"Tom Hardy",
"Ridley Scott"
],
"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:\nAfter studying acting at the Drama Centre London, Edward Thomas Hardy made his film debut in Black Hawk Down (2001) by an English film director and producer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tom Hardy",
"Ridley Scott"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6929,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
47062
|
Xanthium strumarium or Marula Matangi in Telugu, has been used in traditional medicine.
|
{
"evidence": [
"Xanthium strumarium_sentence_8",
"Xanthium strumarium_sentence_9"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nXanthium strumarium or Marula Matangi in Telugu, has been used in traditional medicine.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Xanthium strumarium_sentence_8",
"Xanthium strumarium_sentence_9"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6930,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.023",
"value": "?search="
},
{
"operation": "search",
"time": "13.649",
"value": "Xanthium strumarium"
},
{
"operation": "Now on",
"time": "14.325",
"value": "Xanthium strumarium"
},
{
"operation": "Highlighting",
"time": "26.151",
"value": "Xanthium strumarium_sentence_8"
},
{
"operation": "Highlighting",
"time": "33.045",
"value": "Xanthium strumarium_sentence_9"
},
{
"operation": "finish",
"time": "44.072",
"value": "finish"
}
]
}
|
95097
|
Walter Calverley (eldest son of William Calverley) murdered two of his sons, William and Walter.
|
{
"evidence": [
"Walter Calverley_sentence_3",
"Walter Calverley_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:\nWalter Calverley (eldest son of William Calverley) murdered two of his sons, William and Walter.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Walter Calverley_sentence_3",
"Walter Calverley_sentence_14"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6931,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.242",
"value": "?search="
},
{
"operation": "search",
"time": "8.801",
"value": "Walter Calverley"
},
{
"operation": "Now on",
"time": "9.411",
"value": "Walter Calverley"
},
{
"operation": "Page search",
"time": "17.812",
"value": "william"
},
{
"operation": "Highlighting",
"time": "21.12",
"value": "Walter Calverley_sentence_3"
},
{
"operation": "Page search",
"time": "31.551",
"value": "sons"
},
{
"operation": "Highlighting",
"time": "74.928",
"value": "Walter Calverley_sentence_14"
},
{
"operation": "finish",
"time": "184.319",
"value": "finish"
}
]
}
|
exfever_3446
|
Little Miss Sunshine is a 2006 American tragicomedy road series whose stars include an American actor.
|
{
"evidence": [
"Little Miss Sunshine",
"Greg Kinnear"
],
"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:\nLittle Miss Sunshine is a 2006 American tragicomedy road series whose stars include an American actor.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Little Miss Sunshine",
"Greg Kinnear"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6932,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2275
|
Omar Sharif is generally regarded as Africa's greatest person who portrays a character in an artwork or art exhibition created through actions executed by the artist or other participants.
|
{
"evidence": [
"Omar Sharif",
"Actor",
"Performance art"
],
"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:\nOmar Sharif is generally regarded as Africa's greatest person who portrays a character in an artwork or art exhibition created through actions executed by the artist or other participants.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Omar Sharif",
"Actor",
"Performance art"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6933,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
36124
|
Programming by demonstration, an end-user development technique for teaching a computer or a robot new behaviors by demonstrating the task, believes that a humanoid robot is supposed by its nature to adapt to new environments and that the robot requires the ability to continuously learn new skills and adapt the existing skills to new contexts.
|
{
"evidence": [
"Programming by demonstration_sentence_0",
"Programming by demonstration_sentence_16",
"Programming by demonstration_sentence_17"
],
"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:\nProgramming by demonstration, an end-user development technique for teaching a computer or a robot new behaviors by demonstrating the task, believes that a humanoid robot is supposed by its nature to adapt to new environments and that the robot requires the ability to continuously learn new skills and adapt the existing skills to new contexts.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Programming by demonstration_sentence_0",
"Programming by demonstration_sentence_16",
"Programming by demonstration_sentence_17"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6934,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.693",
"value": "?search="
},
{
"operation": "search",
"time": "21.71",
"value": "Programming by demonstration"
},
{
"operation": "Now on",
"time": "23.318",
"value": "Programming by demonstration"
},
{
"operation": "Highlighting",
"time": "55.929",
"value": "Programming by demonstration_sentence_0"
},
{
"operation": "Page search",
"time": "69.118",
"value": "humanoid"
},
{
"operation": "Highlighting",
"time": "83.667",
"value": "Programming by demonstration_sentence_16"
},
{
"operation": "Highlighting",
"time": "192.445",
"value": "Programming by demonstration_sentence_17"
},
{
"operation": "finish",
"time": "210.143",
"value": "finish"
}
]
}
|
88843
|
Antipathy is based on the positive character of others.
|
{
"evidence": [
"Antipathy_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:\nAntipathy is based on the positive character of others.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Antipathy_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6935,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.229",
"value": "?search="
},
{
"operation": "search",
"time": "10.543",
"value": "Antipathy"
},
{
"operation": "Now on",
"time": "11.25",
"value": "Antipathy"
},
{
"operation": "Page search",
"time": "20.211",
"value": "character"
},
{
"operation": "Highlighting",
"time": "31.084",
"value": "Antipathy_sentence_13"
},
{
"operation": "Highlighting",
"time": "31.086",
"value": "Antipathy_section_0"
},
{
"operation": "Highlighting",
"time": "31.086",
"value": "Antipathy_section_2"
},
{
"operation": "Highlighting deleted",
"time": "43.037",
"value": "Antipathy_sentence_13"
},
{
"operation": "Highlighting deleted",
"time": "43.041",
"value": "Antipathy_section_0"
},
{
"operation": "Highlighting deleted",
"time": "43.043",
"value": "Antipathy_section_2"
},
{
"operation": "Highlighting",
"time": "62.52",
"value": "Antipathy_sentence_0"
},
{
"operation": "finish",
"time": "68.987",
"value": "finish"
}
]
}
|
exfever_8841
|
Legion is a fictional character appearing in American comic books published by the brand name and primary imprint of Marvel Worldwide Inc., which launched other superhero titles created by some of the artists, including the artist and co-creator of the Marvel superheroes Spider-Man and Doctor Strange.
|
{
"evidence": [
"Legion (Marvel Comics)",
"Marvel Comics",
"Steve Ditko"
],
"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:\nLegion is a fictional character appearing in American comic books published by the brand name and primary imprint of Marvel Worldwide Inc., which launched other superhero titles created by some of the artists, including the artist and co-creator of the Marvel superheroes Spider-Man and Doctor Strange.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Legion (Marvel Comics)",
"Marvel Comics",
"Steve Ditko"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6936,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5272
|
Patty Jenkins is the director of the 2003 film that is about serial killer Aileen Wuornos, a street prostitute who murdered seven of her male clients between 1989 and 1990 and was executed in Florida in 2002.
|
{
"evidence": [
"Patty Jenkins",
"Monster (2003 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:\nPatty Jenkins is the director of the 2003 film that is about serial killer Aileen Wuornos, a street prostitute who murdered seven of her male clients between 1989 and 1990 and was executed in Florida in 2002.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Patty Jenkins",
"Monster (2003 film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6937,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5615
|
Estonian Philharmonic Chamber Choir is one of the most famous choirs in the world.
|
{
"evidence": [
"Choir",
"Orchestra",
"Classical music"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nEstonian Philharmonic Chamber Choir is one of the most famous choirs in the world.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Choir",
"Orchestra",
"Classical music"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6938,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2145
|
Dracula is a novel with no single antagonist, but opens with solicitor Jonathan Harker taking a business trip to stay at the castle of a historical region in central Romania.
|
{
"evidence": [
"Dracula",
"Transylvania"
],
"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:\nDracula is a novel with no single antagonist, but opens with solicitor Jonathan Harker taking a business trip to stay at the castle of a historical region in central Romania.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dracula",
"Transylvania"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6939,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
35556
|
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 20–10.
|
{
"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 20–10.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Milan Puskar Stadium_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6940,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.644",
"value": "undefined"
},
{
"operation": "search",
"time": "15.53",
"value": "Milan Puskar Stadium"
},
{
"operation": "Now on",
"time": "16.247",
"value": "Milan Puskar Stadium"
},
{
"operation": "Highlighting",
"time": "27.741",
"value": "Milan Puskar Stadium_sentence_1"
},
{
"operation": "hyperlink",
"time": "54.514",
"value": "West Virginia Mountaineers football"
},
{
"operation": "Now on",
"time": "55.335",
"value": "West Virginia Mountaineers football"
},
{
"operation": "hyperlink",
"time": "62.993",
"value": "West Virginia University"
},
{
"operation": "Now on",
"time": "64.472",
"value": "West Virginia University"
},
{
"operation": "hyperlink",
"time": "157.155",
"value": "West Virginia Mountaineers football"
},
{
"operation": "Now on",
"time": "157.964",
"value": "West Virginia Mountaineers football"
},
{
"operation": "Page search",
"time": "188.047",
"value": "duke"
},
{
"operation": "finish",
"time": "249.861",
"value": "finish"
}
]
}
|
exfever_3668
|
Alpha House is an Australian satire streaming series produced by Amazon Studios, a satire that specializes in gaining entertainment from politics.
|
{
"evidence": [
"Alpha House",
"Political satire"
],
"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:\nAlpha House is an Australian satire streaming series produced by Amazon Studios, a satire that specializes in gaining entertainment from politics.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Alpha House",
"Political satire"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6941,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
47198
|
The Trinidad and Tobago Guardian, first published in September 1917, is now owned by Guardian Media Limited, which is part of ANSA McAL.
|
{
"evidence": [
"Trinidad and Tobago Guardian_sentence_1",
"Trinidad and Tobago Guardian_sentence_2",
"Guardian Media Limited_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Trinidad and Tobago Guardian, first published in September 1917, is now owned by Guardian Media Limited, which is part of ANSA McAL.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Trinidad and Tobago Guardian_sentence_1",
"Trinidad and Tobago Guardian_sentence_2",
"Guardian Media Limited_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6942,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.105",
"value": "?search="
},
{
"operation": "search",
"time": "12.298",
"value": "Trinidad and Tobago Guardian"
},
{
"operation": "Now on",
"time": "13.267",
"value": "Trinidad and Tobago Guardian"
},
{
"operation": "Highlighting",
"time": "18.694",
"value": "Trinidad and Tobago Guardian_sentence_1"
},
{
"operation": "Highlighting",
"time": "26.882",
"value": "Trinidad and Tobago Guardian_sentence_2"
},
{
"operation": "hyperlink",
"time": "32.234",
"value": "Guardian Media Limited"
},
{
"operation": "Now on",
"time": "33.686",
"value": "Guardian Media Limited"
},
{
"operation": "Highlighting",
"time": "38.974",
"value": "Guardian Media Limited_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "46.181",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "46.313",
"value": "Trinidad and Tobago Guardian"
},
{
"operation": "finish",
"time": "70.022",
"value": "finish"
}
]
}
|
exfever_5329
|
The term cult film itself was first used in the 1970s to describe the culture that surrounded underground films that were rooted in the practice that emerged in the 1950s of local television stations around the United States airing low-budget stylistic or thematic categories for motion pictures based on similarities either in the narrative elements or the emotional response to the film.
|
{
"evidence": [
"Cult film",
"Midnight movie",
"Film genre"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe term cult film itself was first used in the 1970s to describe the culture that surrounded underground films that were rooted in the practice that emerged in the 1950s of local television stations around the United States airing low-budget stylistic or thematic categories for motion pictures based on similarities either in the narrative elements or the emotional response to the film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Cult film",
"Midnight movie",
"Film genre"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6943,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
22834
|
Aeritalia was formed out of the merger of two companies, and was a partner of the Boeing 767, as well as the Alfa rocket in the Italian space industry; in 1990 it merged with another company to form Alenia Aeronautica.
|
{
"evidence": [
"Aeritalia_sentence_2",
"Aeritalia_sentence_5",
"Aeritalia_sentence_6"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAeritalia was formed out of the merger of two companies, and was a partner of the Boeing 767, as well as the Alfa rocket in the Italian space industry; in 1990 it merged with another company to form Alenia Aeronautica.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Aeritalia_sentence_2",
"Aeritalia_sentence_5",
"Aeritalia_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6944,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.017",
"value": "?search="
},
{
"operation": "search",
"time": "4.85",
"value": "Aeritalia"
},
{
"operation": "Now on",
"time": "5.631",
"value": "Aeritalia"
},
{
"operation": "Highlighting",
"time": "8.887",
"value": "Aeritalia_sentence_2"
},
{
"operation": "Highlighting",
"time": "16.904",
"value": "Aeritalia_sentence_5"
},
{
"operation": "Highlighting",
"time": "32.38",
"value": "Aeritalia_sentence_6"
},
{
"operation": "finish",
"time": "62.266",
"value": "finish"
}
]
}
|
exfever_6310
|
Berlin is Germany's municipality holding primary status in a country, state, province, department, or other sub-national entity, usually as its seat of the government.
|
{
"evidence": [
"Berlin",
"Capital city"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBerlin is Germany's municipality holding primary status in a country, state, province, department, or other sub-national entity, usually as its seat of the government.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Berlin",
"Capital city"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6945,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
19459
|
The left Front won their most number of seats in Indian general elections in the 14th Lok Sabha elections.
|
{
"evidence": [
"Left Front (West Bengal)_sentence_146",
"Left Front (West Bengal)_sentence_147",
"Left Front (West Bengal)_sentence_148",
"Left Front (West Bengal)_sentence_0",
"14th Lok Sabha_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 left Front won their most number of seats in Indian general elections in the 14th Lok Sabha elections.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Left Front (West Bengal)_sentence_146",
"Left Front (West Bengal)_sentence_147",
"Left Front (West Bengal)_sentence_148",
"Left Front (West Bengal)_sentence_0",
"14th Lok Sabha_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6946,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.216",
"value": "?search="
},
{
"operation": "search",
"time": "12.996",
"value": "Contains...14th Lok Sabha elections"
},
{
"operation": "Now on",
"time": "14.505",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "18.528",
"value": "14th lok sabha election"
},
{
"operation": "Now on",
"time": "19.247",
"value": "undefined"
},
{
"operation": "search",
"time": "48.403",
"value": "14th lok sabha Indian general election"
},
{
"operation": "Now on",
"time": "49.552",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "52.449",
"value": "14th Lok Sabha Manmohan Singh_section_6"
},
{
"operation": "Now on",
"time": "54.195",
"value": "Manmohan Singh#14th Lok Sabha Manmohan Singh section 6"
},
{
"operation": "back-button-clicked",
"time": "64.274",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "65.361",
"value": "undefined"
},
{
"operation": "search",
"time": "77.071",
"value": "\"left fron\""
},
{
"operation": "Now on",
"time": "78.307",
"value": "undefined"
},
{
"operation": "search",
"time": "90.806",
"value": "14th Lok Sabha"
},
{
"operation": "Now on",
"time": "92.048",
"value": "14th Lok Sabha"
},
{
"operation": "Page search",
"time": "98.604",
"value": "front"
},
{
"operation": "search",
"time": "142.348",
"value": "MNF"
},
{
"operation": "Now on",
"time": "143.266",
"value": "MNF"
},
{
"operation": "hyperlink",
"time": "146.761",
"value": "3 Political organizations MNF_section_2"
},
{
"operation": "Highlighting",
"time": "155.578",
"value": "MNF_item_2_2"
},
{
"operation": "Highlighting",
"time": "155.58",
"value": "MNF_title"
},
{
"operation": "Highlighting",
"time": "155.581",
"value": "MNF_section_2"
},
{
"operation": "back-button-clicked",
"time": "167.491",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "167.624",
"value": "14th Lok Sabha"
},
{
"operation": "Page search",
"time": "180.614",
"value": "left front"
},
{
"operation": "Page search",
"time": "190.458",
"value": "front"
},
{
"operation": "search",
"time": "200.689",
"value": "NPF"
},
{
"operation": "Now on",
"time": "202.64",
"value": "NPF"
},
{
"operation": "Highlighting",
"time": "213.546",
"value": "NPF_item_0_0"
},
{
"operation": "Highlighting",
"time": "213.548",
"value": "NPF_title"
},
{
"operation": "back-button-clicked",
"time": "215.122",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "215.248",
"value": "14th Lok Sabha"
},
{
"operation": "Page search",
"time": "224.281",
"value": "front"
},
{
"operation": "search",
"time": "234.46",
"value": "SDF"
},
{
"operation": "Now on",
"time": "235.938",
"value": "SDF"
},
{
"operation": "hyperlink",
"time": "238.436",
"value": "3 Political parties SDF_section_2"
},
{
"operation": "Highlighting",
"time": "242.322",
"value": "SDF_item_2_17"
},
{
"operation": "Highlighting",
"time": "242.324",
"value": "SDF_title"
},
{
"operation": "Highlighting",
"time": "242.324",
"value": "SDF_section_2"
},
{
"operation": "back-button-clicked",
"time": "253.143",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "253.253",
"value": "14th Lok Sabha"
},
{
"operation": "search",
"time": "265.291",
"value": "Left Front"
},
{
"operation": "Now on",
"time": "265.97",
"value": "Left Front"
},
{
"operation": "hyperlink",
"time": "283.085",
"value": "1.2 India Left Front_section_2"
},
{
"operation": "hyperlink",
"time": "286.719",
"value": "Left Front (West Bengal)"
},
{
"operation": "Now on",
"time": "287.977",
"value": "Left Front (West Bengal)"
},
{
"operation": "hyperlink",
"time": "310.133",
"value": "9.2 2004 Lok Sabha election Left Front (West Bengal)_section_27"
},
{
"operation": "Page search",
"time": "324.8",
"value": "lok sabha"
},
{
"operation": "back-button-clicked",
"time": "334.938",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "335.045",
"value": "Left Front#India Left Front section 2"
},
{
"operation": "back-button-clicked",
"time": "335.572",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "336.535",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "336.649",
"value": "14th Lok Sabha"
},
{
"operation": "search",
"time": "352.131",
"value": "left front"
},
{
"operation": "Now on",
"time": "353.636",
"value": "Left front"
},
{
"operation": "hyperlink",
"time": "356.077",
"value": "1.2 India Left Front_section_2"
},
{
"operation": "hyperlink",
"time": "358.608",
"value": "Left Front (West Bengal)"
},
{
"operation": "Now on",
"time": "359.395",
"value": "Left Front (West Bengal)"
},
{
"operation": "hyperlink",
"time": "369.488",
"value": "9.2 2004 Lok Sabha election Left Front (West Bengal)_section_27"
},
{
"operation": "Highlighting",
"time": "374.061",
"value": "Left Front (West Bengal)_sentence_146"
},
{
"operation": "Highlighting",
"time": "375.307",
"value": "Left Front (West Bengal)_sentence_147"
},
{
"operation": "Highlighting",
"time": "377.724",
"value": "Left Front (West Bengal)_sentence_148"
},
{
"operation": "Highlighting",
"time": "408.084",
"value": "Left Front (West Bengal)_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "409.264",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "409.468",
"value": "Left Front#India Left Front section 2"
},
{
"operation": "back-button-clicked",
"time": "410.368",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "411.918",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "412.026",
"value": "14th Lok Sabha"
},
{
"operation": "Page search",
"time": "432.447",
"value": "west bengal"
},
{
"operation": "Highlighting",
"time": "454.621",
"value": "14th Lok Sabha_sentence_0"
},
{
"operation": "finish",
"time": "489.673",
"value": "finish"
}
]
}
|
60042
|
Texas earned the right to represent the Southwest Conference in the Cotton Bowl, where their opponent was eighth-ranked Irish, which represented Notre Dame's 1st bowl game in 40 years, featuring Joe Theismann and the University of Texas at Austin.
|
{
"evidence": [
"Cotton Bowl Classic_sentence_74"
],
"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:\nTexas earned the right to represent the Southwest Conference in the Cotton Bowl, where their opponent was eighth-ranked Irish, which represented Notre Dame's 1st bowl game in 40 years, featuring Joe Theismann and the University of Texas at Austin.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Cotton Bowl Classic_sentence_74"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6947,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.22",
"value": "?search="
},
{
"operation": "search",
"time": "38.202",
"value": "texas in the cotton bowl"
},
{
"operation": "Now on",
"time": "40.008",
"value": "undefined"
},
{
"operation": "search",
"time": "72.523",
"value": "tesas and notre dame in the cotton bowl"
},
{
"operation": "Now on",
"time": "74.181",
"value": "undefined"
},
{
"operation": "search",
"time": "102.004",
"value": "joe theismann"
},
{
"operation": "Now on",
"time": "103.588",
"value": "Joe Theismann"
},
{
"operation": "hyperlink",
"time": "135.409",
"value": "1970 Cotton Bowl Classic"
},
{
"operation": "Now on",
"time": "143.803",
"value": "Cotton Bowl Classic#Game results"
},
{
"operation": "back-button-clicked",
"time": "144.321",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "144.454",
"value": "Joe Theismann"
},
{
"operation": "hyperlink",
"time": "150.786",
"value": "1970 Cotton Bowl Classic"
},
{
"operation": "Now on",
"time": "151.558",
"value": "Cotton Bowl Classic#Game results"
},
{
"operation": "Page search",
"time": "159.653",
"value": "texas"
},
{
"operation": "Highlighting",
"time": "170.774",
"value": "Cotton Bowl Classic_sentence_17"
},
{
"operation": "Highlighting",
"time": "170.776",
"value": "Cotton Bowl Classic_section_2"
},
{
"operation": "Highlighting deleted",
"time": "190.156",
"value": "Cotton Bowl Classic_sentence_17"
},
{
"operation": "Highlighting deleted",
"time": "190.163",
"value": "Cotton Bowl Classic_section_2"
},
{
"operation": "back-button-clicked",
"time": "227.541",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "227.704",
"value": "Joe Theismann"
},
{
"operation": "hyperlink",
"time": "252.203",
"value": "1971 Cotton Bowl Classic"
},
{
"operation": "Now on",
"time": "252.75",
"value": "Cotton Bowl Classic#Game results"
},
{
"operation": "back-button-clicked",
"time": "258.847",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "258.987",
"value": "Joe Theismann"
},
{
"operation": "hyperlink",
"time": "260.971",
"value": "1970 Cotton Bowl Classic"
},
{
"operation": "Now on",
"time": "261.524",
"value": "Cotton Bowl Classic#Game results"
},
{
"operation": "back-button-clicked",
"time": "266.562",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "266.706",
"value": "Joe Theismann"
},
{
"operation": "hyperlink",
"time": "270.496",
"value": "1971 Cotton Bowl Classic"
},
{
"operation": "Now on",
"time": "271.038",
"value": "Cotton Bowl Classic#Game results"
},
{
"operation": "back-button-clicked",
"time": "278.842",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "279.059",
"value": "Joe Theismann"
},
{
"operation": "hyperlink",
"time": "285.966",
"value": "1970 Cotton Bowl Classic"
},
{
"operation": "Now on",
"time": "286.501",
"value": "Cotton Bowl Classic#Game results"
},
{
"operation": "Page search",
"time": "292.208",
"value": "1970"
},
{
"operation": "Highlighting",
"time": "315.197",
"value": "Cotton Bowl Classic_sentence_76"
},
{
"operation": "hyperlink",
"time": "322.249",
"value": "Texas"
},
{
"operation": "Now on",
"time": "327.013",
"value": "1969 Texas Longhorns football team"
},
{
"operation": "Page search",
"time": "332.117",
"value": "southwest"
},
{
"operation": "Highlighting",
"time": "334.509",
"value": "1969 Texas Longhorns football team_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "334.51",
"value": "1969 Texas Longhorns football team_title"
},
{
"operation": "back-button-clicked",
"time": "336.791",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "337.141",
"value": "Cotton Bowl Classic#Game results"
},
{
"operation": "hyperlink",
"time": "386.736",
"value": "Notre Dame"
},
{
"operation": "Now on",
"time": "390.048",
"value": "Notre Dame Fighting Irish football"
},
{
"operation": "back-button-clicked",
"time": "392.59",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "392.925",
"value": "Cotton Bowl Classic#Game results"
},
{
"operation": "hyperlink",
"time": "407.877",
"value": "Notre Dame"
},
{
"operation": "Now on",
"time": "408.637",
"value": "Notre Dame Fighting Irish football"
},
{
"operation": "back-button-clicked",
"time": "410.299",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "410.639",
"value": "Cotton Bowl Classic#Game results"
},
{
"operation": "Highlighting",
"time": "412.838",
"value": "Cotton Bowl Classic_sentence_74"
},
{
"operation": "hyperlink",
"time": "416.723",
"value": "Joe Theismann"
},
{
"operation": "Now on",
"time": "417.107",
"value": "Joe Theismann"
},
{
"operation": "back-button-clicked",
"time": "419.014",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "419.327",
"value": "Cotton Bowl Classic#Game results"
},
{
"operation": "Highlighting deleted",
"time": "420.847",
"value": "Cotton Bowl Classic_sentence_76"
},
{
"operation": "finish",
"time": "440.18",
"value": "finish"
}
]
}
|
20274
|
Dmitri Naumkin competed solo bagging 1st place in international events Nebelhorn Trophy, Junior Worlds, and Blue Swords.
|
{
"evidence": [
"Dmitri Naumkin_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:\nDmitri Naumkin competed solo bagging 1st place in international events Nebelhorn Trophy, Junior Worlds, and Blue Swords.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Dmitri Naumkin_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6948,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.991",
"value": "?search="
},
{
"operation": "search",
"time": "17.931",
"value": "Dmitri Naumkin"
},
{
"operation": "Now on",
"time": "18.837",
"value": "Dmitri Naumkin"
},
{
"operation": "Highlighting",
"time": "68.07",
"value": "Dmitri Naumkin_sentence_3"
},
{
"operation": "finish",
"time": "118.598",
"value": "finish"
}
]
}
|
5862
|
Disneyland, operated by Six Flags, opened to the public in June 16, 1967.
|
{
"evidence": [
"Disney Parks, Experiences and Products_sentence_247"
],
"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:\nDisneyland, operated by Six Flags, opened to the public in June 16, 1967.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Disney Parks, Experiences and Products_sentence_247"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6949,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.504",
"value": "?search="
},
{
"operation": "search",
"time": "22.686",
"value": "Disneyland"
},
{
"operation": "Now on",
"time": "23.879",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "36.115",
"value": "Magic Kingdom"
},
{
"operation": "Now on",
"time": "37.047",
"value": "Magic Kingdom"
},
{
"operation": "search",
"time": "79.236",
"value": "disneyland"
},
{
"operation": "Now on",
"time": "79.737",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "85.978",
"value": "Disney Parks, Experiences and Products"
},
{
"operation": "Now on",
"time": "87.215",
"value": "Disney Parks, Experiences and Products"
},
{
"operation": "hyperlink",
"time": "105.016",
"value": "3.1 Disneyland Resort Disney Parks, Experiences and Products_section_8"
},
{
"operation": "Highlighting",
"time": "122.84",
"value": "Disney Parks, Experiences and Products_sentence_247"
},
{
"operation": "finish",
"time": "148.132",
"value": "finish"
}
]
}
|
35157
|
Christianity in Manipur was the second-largest religion in Manipur after Hinduism which is at 41.39% vs 41.29% Christianity according to the 2011 Census of India.
|
{
"evidence": [
"Manipur_sentence_20",
"Manipur_sentence_147"
],
"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:\nChristianity in Manipur was the second-largest religion in Manipur after Hinduism which is at 41.39% vs 41.29% Christianity according to the 2011 Census of India.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Manipur_sentence_20",
"Manipur_sentence_147"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6950,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.161",
"value": "?search="
},
{
"operation": "search",
"time": "11.838",
"value": "Manipur"
},
{
"operation": "Now on",
"time": "13.001",
"value": "Manipur"
},
{
"operation": "Highlighting",
"time": "62.736",
"value": "Manipur_sentence_20"
},
{
"operation": "Highlighting",
"time": "86.968",
"value": "Manipur_sentence_147"
},
{
"operation": "finish",
"time": "99.415",
"value": "finish"
}
]
}
|
35557
|
Samuel Alfred Varley was among the first to make a self-excited dynamo, a type of direct current electric generator in which field winding and armature winding are connected in parallel, in 1873 along other contenders who were Charles Wheatstone and Werner von Siemens.
|
{
"evidence": [
"Samuel Alfred Varley_sentence_8"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSamuel Alfred Varley was among the first to make a self-excited dynamo, a type of direct current electric generator in which field winding and armature winding are connected in parallel, in 1873 along other contenders who were Charles Wheatstone and Werner von Siemens.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Samuel Alfred Varley_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6951,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.708",
"value": "?search="
},
{
"operation": "search",
"time": "11.382",
"value": "Samuel Alfred Varley"
},
{
"operation": "Now on",
"time": "12.267",
"value": "Samuel Alfred Varley"
},
{
"operation": "Highlighting",
"time": "26.047",
"value": "Samuel Alfred Varley_header_cell_0_5_0"
},
{
"operation": "Highlighting",
"time": "29.988",
"value": "Samuel Alfred Varley_cell_0_5_1"
},
{
"operation": "Highlighting",
"time": "29.995",
"value": "Samuel Alfred Varley_header_cell_0_5_0"
},
{
"operation": "Highlighting",
"time": "29.995",
"value": "Samuel Alfred Varley_header_cell_0_0_0"
},
{
"operation": "hyperlink",
"time": "31.273",
"value": "self-excited dynamo"
},
{
"operation": "Now on",
"time": "31.793",
"value": "Shunt generator"
},
{
"operation": "Highlighting",
"time": "40.729",
"value": "Shunt generator_sentence_0"
},
{
"operation": "Highlighting",
"time": "40.73",
"value": "Shunt generator_title"
},
{
"operation": "back-button-clicked",
"time": "48.147",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "48.241",
"value": "Samuel Alfred Varley"
},
{
"operation": "Highlighting",
"time": "60.712",
"value": "Samuel Alfred Varley_sentence_8"
},
{
"operation": "finish",
"time": "85.352",
"value": "finish"
}
]
}
|
52475
|
Valkyrie Drive is a Japanese media franchise created by Marvelous that consists of three projects that revolve around girls who have been infected with a mysterious virus that allows them to transform into weapons.
|
{
"evidence": [
"Valkyrie Drive_sentence_0",
"Valkyrie Drive_sentence_1",
"Valkyrie Drive_sentence_2",
"Valkyrie Drive_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:\nValkyrie Drive is a Japanese media franchise created by Marvelous that consists of three projects that revolve around girls who have been infected with a mysterious virus that allows them to transform into weapons.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Valkyrie Drive_sentence_0",
"Valkyrie Drive_sentence_1",
"Valkyrie Drive_sentence_2",
"Valkyrie Drive_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6952,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.159",
"value": "?search="
},
{
"operation": "search",
"time": "10.336",
"value": "Valkyrie Drive"
},
{
"operation": "Now on",
"time": "12.362",
"value": "Valkyrie Drive"
},
{
"operation": "Highlighting",
"time": "15.406",
"value": "Valkyrie Drive_sentence_0"
},
{
"operation": "Highlighting",
"time": "23.454",
"value": "Valkyrie Drive_sentence_1"
},
{
"operation": "Highlighting",
"time": "39.996",
"value": "Valkyrie Drive_sentence_2"
},
{
"operation": "Highlighting",
"time": "103.738",
"value": "Valkyrie Drive_sentence_3"
},
{
"operation": "finish",
"time": "132.909",
"value": "finish"
}
]
}
|
70651
|
Andrew Mazzei moved with his family to London in 1891, but had settled in Blackpool within ten years.
|
{
"evidence": [
"Andrew Mazzei_sentence_9",
"Andrew Mazzei_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:\nAndrew Mazzei moved with his family to London in 1891, but had settled in Blackpool within ten years.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Andrew Mazzei_sentence_9",
"Andrew Mazzei_sentence_10"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6953,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.84",
"value": "?search="
},
{
"operation": "search",
"time": "9.058",
"value": "Andrew Mazzei"
},
{
"operation": "Now on",
"time": "27.158",
"value": "Andrew Mazzei"
},
{
"operation": "Highlighting",
"time": "65.302",
"value": "Andrew Mazzei_sentence_9"
},
{
"operation": "Highlighting",
"time": "77.641",
"value": "Andrew Mazzei_sentence_10"
},
{
"operation": "finish",
"time": "96.65",
"value": "finish"
}
]
}
|
exfever_3515
|
Bhagat Singh was an Indian revolutionary derived from the charisma of a follower.
|
{
"evidence": [
"Bhagat Singh",
"Charismatic authority"
],
"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:\nBhagat Singh was an Indian revolutionary derived from the charisma of a follower.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Bhagat Singh",
"Charismatic authority"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6954,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
38521
|
Amrithadhare is a 2005 Kannada film directed and co-produced by Nagathihalli Chandrashekhar and written by actor Ramesh Aravind with a soundtrack that includes ten songs such as Mane Katti Nodu by Raju Ananthaswamy and Nanditha that has a duration of 04:43, Giliyu Panjaradolila by C. Ashwath (duration 03:43), and Huduga Huduga by Chaitra. H. G. (duration 04:45).
|
{
"evidence": [
"Amrithadhare_sentence_12"
],
"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:\nAmrithadhare is a 2005 Kannada film directed and co-produced by Nagathihalli Chandrashekhar and written by actor Ramesh Aravind with a soundtrack that includes ten songs such as Mane Katti Nodu by Raju Ananthaswamy and Nanditha that has a duration of 04:43, Giliyu Panjaradolila by C. Ashwath (duration 03:43), and Huduga Huduga by Chaitra. H. G. (duration 04:45).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Amrithadhare_sentence_12"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6955,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.014",
"value": "?search="
},
{
"operation": "search",
"time": "3.514",
"value": "Amrithadhare"
},
{
"operation": "Now on",
"time": "4.291",
"value": "Amrithadhare"
},
{
"operation": "Highlighting",
"time": "5.744",
"value": "Amrithadhare_sentence_0"
},
{
"operation": "Highlighting",
"time": "21.988",
"value": "Amrithadhare_sentence_12"
},
{
"operation": "finish",
"time": "45.579",
"value": "finish"
}
]
}
|
exfever_5911
|
Toy Story is a 1995 American computer-animated comedy film that was written by an Australian screenwriter and voice actor.
|
{
"evidence": [
"Toy Story",
"Andrew Stanton"
],
"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:\nToy Story is a 1995 American computer-animated comedy film that was written by an Australian screenwriter and voice actor.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Toy Story",
"Andrew Stanton"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6956,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5253
|
Susan Sarandon is an American actress and activist who received an honor for the best British and international contributions to film and broadcast first on the BBC in 1957 with Vivien Leigh. Vivien Leigh was a British actress who played on stage in London's West End in 1949.
|
{
"evidence": [
"Susan Sarandon",
"British Academy Film Awards",
"Vivien Leigh"
],
"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:\nSusan Sarandon is an American actress and activist who received an honor for the best British and international contributions to film and broadcast first on the BBC in 1957 with Vivien Leigh. Vivien Leigh was a British actress who played on stage in London's West End in 1949.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Susan Sarandon",
"British Academy Film Awards",
"Vivien Leigh"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6957,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
74792
|
Kamilla Rytter Juhl paired with at least two people.
|
{
"evidence": [
"Kamilla Rytter Juhl_sentence_5",
"Kamilla Rytter Juhl_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:\nKamilla Rytter Juhl paired with at least two people.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Kamilla Rytter Juhl_sentence_5",
"Kamilla Rytter Juhl_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6958,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.976",
"value": "?search="
},
{
"operation": "search",
"time": "10.935",
"value": "Kamilla Rytter Juhl"
},
{
"operation": "Now on",
"time": "15.908",
"value": "Kamilla Rytter Juhl"
},
{
"operation": "Page search",
"time": "33.253",
"value": "paired"
},
{
"operation": "Highlighting",
"time": "49.164",
"value": "Kamilla Rytter Juhl_sentence_7"
},
{
"operation": "Highlighting",
"time": "90.166",
"value": "Kamilla Rytter Juhl_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "91.237",
"value": "Kamilla Rytter Juhl_sentence_7"
},
{
"operation": "Highlighting",
"time": "92.197",
"value": "Kamilla Rytter Juhl_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "96.773",
"value": "Kamilla Rytter Juhl_sentence_7"
},
{
"operation": "Highlighting",
"time": "98.333",
"value": "Kamilla Rytter Juhl_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "114.807",
"value": "Kamilla Rytter Juhl_sentence_3"
},
{
"operation": "Highlighting",
"time": "115.719",
"value": "Kamilla Rytter Juhl_sentence_7"
},
{
"operation": "finish",
"time": "122.206",
"value": "finish"
}
]
}
|
exfever_4437
|
What Them Girls Like is the first single by Ludacris.
|
{
"evidence": [
"Ludacris",
"Grammy Award"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWhat Them Girls Like is the first single by Ludacris.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ludacris",
"Grammy Award"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6959,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
2639
|
Jackie Blanchflower helped his team win the league twice.
|
{
"evidence": [
"Jackie Blanchflower_sentence_1",
"Jackie Blanchflower_sentence_6"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJackie Blanchflower helped his team win the league twice.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Jackie Blanchflower_sentence_1",
"Jackie Blanchflower_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6960,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "50.898",
"value": "Jackie Blanchflower"
},
{
"operation": "Highlighting",
"time": "121.06",
"value": "Jackie Blanchflower_sentence_1"
},
{
"operation": "Highlighting",
"time": "295.423",
"value": "Jackie Blanchflower_sentence_6"
},
{
"operation": "finish",
"time": "310.631",
"value": "finish"
}
]
}
|
exfever_4460
|
The 79th Academy Awards ceremony was hosted for the first time by a Mexican comedian who provided the voice of Dory in the film by an American computer animation studio based in Emeryville, California.
|
{
"evidence": [
"79th Academy Awards",
"Ellen DeGeneres",
"Pixar"
],
"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 79th Academy Awards ceremony was hosted for the first time by a Mexican comedian who provided the voice of Dory in the film by an American computer animation studio based in Emeryville, California.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"79th Academy Awards",
"Ellen DeGeneres",
"Pixar"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6961,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_525
|
Tropic Thunder is a 2008 action-comedy film directed by Ben Stiller and starring the American actor and producer who portrayed Iron Man in 2008.
|
{
"evidence": [
"Tropic Thunder",
"Robert Downey Jr."
],
"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:\nTropic Thunder is a 2008 action-comedy film directed by Ben Stiller and starring the American actor and producer who portrayed Iron Man in 2008.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tropic Thunder",
"Robert Downey Jr."
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6962,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1462
|
The David Geffen School of Drama is a graduate professional school of a private research university in New Haven that belongs to an British collegiate athletic conference.
|
{
"evidence": [
"Yale School of Drama",
"Yale University",
"Ivy League"
],
"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 David Geffen School of Drama is a graduate professional school of a private research university in New Haven that belongs to an British collegiate athletic conference.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Yale School of Drama",
"Yale University",
"Ivy League"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6963,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
41234
|
Shyne Factory was a Tennessee progressive rock and power pop band from Halifax, Nova Scotia, it was formed in 1992 by Shawn Scott and twin brothers Mike and Matt Johnson.
|
{
"evidence": [
"Shyne Factory_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:\nShyne Factory was a Tennessee progressive rock and power pop band from Halifax, Nova Scotia, it was formed in 1992 by Shawn Scott and twin brothers Mike and Matt Johnson.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Shyne Factory_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6964,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.647",
"value": "?search="
},
{
"operation": "search",
"time": "12.815",
"value": "Shyne Factory"
},
{
"operation": "Now on",
"time": "13.542",
"value": "Shyne Factory"
},
{
"operation": "Highlighting",
"time": "22.952",
"value": "Shyne Factory_sentence_0"
},
{
"operation": "finish",
"time": "32.018",
"value": "finish"
}
]
}
|
59806
|
The 2014 PDC Pro Tour was a series of non-televised darts tournaments organised by the Professional Darts Corporation (PDC) is a professional darts organisation in the United Kingdom, established in 1992 when a group of leading players split from the British Darts Organisation to form what was initially called the World Darts Council (WDC).
|
{
"evidence": [
"2014 PDC Pro Tour_sentence_0",
"Professional Darts Corporation_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe 2014 PDC Pro Tour was a series of non-televised darts tournaments organised by the Professional Darts Corporation (PDC) is a professional darts organisation in the United Kingdom, established in 1992 when a group of leading players split from the British Darts Organisation to form what was initially called the World Darts Council (WDC).\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"2014 PDC Pro Tour_sentence_0",
"Professional Darts Corporation_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6965,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.184",
"value": "?search="
},
{
"operation": "search",
"time": "14.41",
"value": "2014 PDC Pro Tour"
},
{
"operation": "Now on",
"time": "16.489",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "18.714",
"value": "2014 PDC Pro Tour"
},
{
"operation": "Now on",
"time": "20.314",
"value": "2014 PDC Pro Tour"
},
{
"operation": "Highlighting",
"time": "21.89",
"value": "2014 PDC Pro Tour_sentence_0"
},
{
"operation": "Highlighting",
"time": "33.544",
"value": "2014 PDC Pro Tour_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "100.767",
"value": "2014 PDC Pro Tour_sentence_1"
},
{
"operation": "hyperlink",
"time": "108.787",
"value": "Professional Darts Corporation"
},
{
"operation": "Now on",
"time": "113.143",
"value": "Professional Darts Corporation"
},
{
"operation": "Highlighting",
"time": "114.323",
"value": "Professional Darts Corporation_sentence_0"
},
{
"operation": "finish",
"time": "151.739",
"value": "finish"
}
]
}
|
94966
|
Luc Dionne is a Canadian screenwriter and director who won the 2005 Genie for Best Cinematography.
|
{
"evidence": [
"Luc Dionne_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:\nLuc Dionne is a Canadian screenwriter and director who won the 2005 Genie for Best Cinematography.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Luc Dionne_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6966,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.118",
"value": "?search="
},
{
"operation": "search",
"time": "504.262",
"value": "Luc Dionne"
},
{
"operation": "Now on",
"time": "505.185",
"value": "Luc Dionne"
},
{
"operation": "Highlighting",
"time": "507.908",
"value": "Luc Dionne_sentence_0"
},
{
"operation": "Highlighting",
"time": "526.685",
"value": "Luc Dionne_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "528.248",
"value": "Luc Dionne_sentence_0"
},
{
"operation": "hyperlink",
"time": "542.498",
"value": "Genie"
},
{
"operation": "Now on",
"time": "543.025",
"value": "Genie Awards"
},
{
"operation": "back-button-clicked",
"time": "557.924",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "558.015",
"value": "Luc Dionne"
},
{
"operation": "finish",
"time": "561.364",
"value": "finish"
}
]
}
|
74784
|
Alban Towers is an apartment hotel with many features such as dog walkers, 24-hour maid service, a public dining room, a beauty shop, and a grocery store.
|
{
"evidence": [
"Alban Towers_sentence_14",
"Alban Towers_sentence_15"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAlban Towers is an apartment hotel with many features such as dog walkers, 24-hour maid service, a public dining room, a beauty shop, and a grocery store.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Alban Towers_sentence_14",
"Alban Towers_sentence_15"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6967,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.171",
"value": "?search="
},
{
"operation": "search",
"time": "245.138",
"value": "Alban Towers"
},
{
"operation": "Now on",
"time": "245.927",
"value": "Alban Towers"
},
{
"operation": "Highlighting",
"time": "264.353",
"value": "Alban Towers_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "279.943",
"value": "Alban Towers_sentence_0"
},
{
"operation": "Highlighting",
"time": "481.277",
"value": "Alban Towers_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "500.661",
"value": "Alban Towers_sentence_5"
},
{
"operation": "Highlighting",
"time": "604.349",
"value": "Alban Towers_sentence_24"
},
{
"operation": "Highlighting deleted",
"time": "893.419",
"value": "Alban Towers_sentence_24"
},
{
"operation": "Highlighting",
"time": "895.499",
"value": "Alban Towers_sentence_14"
},
{
"operation": "Highlighting",
"time": "912.635",
"value": "Alban Towers_sentence_15"
},
{
"operation": "finish",
"time": "1135.995",
"value": "finish"
}
]
}
|
9889
|
Hart to Hart had only two seasons before it stopped airing.
|
{
"evidence": [
"Stefanie Powers_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:\nHart to Hart had only two seasons before it stopped airing.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Stefanie Powers_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6968,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.238",
"value": "?search="
},
{
"operation": "search",
"time": "10.658",
"value": "Contains...Hart to Hart"
},
{
"operation": "Now on",
"time": "12.362",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "15.8",
"value": "Stefanie Powers"
},
{
"operation": "Now on",
"time": "16.923",
"value": "Stefanie Powers"
},
{
"operation": "Highlighting",
"time": "32.488",
"value": "Stefanie Powers_sentence_0"
},
{
"operation": "hyperlink",
"time": "42.152",
"value": "Robert Wagner"
},
{
"operation": "Now on",
"time": "43.462",
"value": "Robert Wagner"
},
{
"operation": "search",
"time": "65.881",
"value": "Hart to Hart"
},
{
"operation": "Now on",
"time": "66.814",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "86.73",
"value": "Stefanie Powers"
},
{
"operation": "Now on",
"time": "87.973",
"value": "Stefanie Powers"
},
{
"operation": "Highlighting deleted",
"time": "90.74",
"value": "Stefanie Powers_sentence_0"
},
{
"operation": "Highlighting",
"time": "99.513",
"value": "Stefanie Powers_sentence_0"
},
{
"operation": "finish",
"time": "123.25",
"value": "finish"
}
]
}
|
58819
|
The OSS Watch's briefing notes have some bias because of the influence of commercial need.
|
{
"evidence": [
"OSS Watch_sentence_17"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe OSS Watch's briefing notes have some bias because of the influence of commercial need.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"OSS Watch_sentence_17"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6969,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.873",
"value": "?search="
},
{
"operation": "search",
"time": "9.201",
"value": "OSS Watch"
},
{
"operation": "Now on",
"time": "10.011",
"value": "OSS Watch"
},
{
"operation": "Highlighting",
"time": "28.283",
"value": "OSS Watch_sentence_17"
},
{
"operation": "finish",
"time": "47.206",
"value": "finish"
}
]
}
|
exfever_4256
|
Karl Urban is a New Zealand actor who portrayed "Bones", a character in the American science-fiction franchise Star Trek.
|
{
"evidence": [
"Karl Urban",
"Leonard McCoy"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nKarl Urban is a New Zealand actor who portrayed \"Bones\", a character in the American science-fiction franchise Star Trek.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Karl Urban",
"Leonard McCoy"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6970,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
66795
|
Called the second Apostle of the North, Johannes Bugenhagen (born in June 24, 1485 in Wollin, Pomerania, Holy Roman Empire (now Wolin, Poland))was a Muslim priest and theologian.
|
{
"evidence": [
"Johannes Bugenhagen_sentence_0",
"Reformation_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:\nCalled the second Apostle of the North, Johannes Bugenhagen (born in June 24, 1485 in Wollin, Pomerania, Holy Roman Empire (now Wolin, Poland))was a Muslim priest and theologian.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Johannes Bugenhagen_sentence_0",
"Reformation_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6971,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.85",
"value": "?search="
},
{
"operation": "search",
"time": "14.531",
"value": "Johannes Bugenhagen"
},
{
"operation": "Now on",
"time": "15.485",
"value": "Johannes Bugenhagen"
},
{
"operation": "Highlighting",
"time": "42.492",
"value": "Johannes Bugenhagen_sentence_2"
},
{
"operation": "Highlighting",
"time": "54.677",
"value": "Johannes Bugenhagen_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "54.682",
"value": "Johannes Bugenhagen_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "54.682",
"value": "Johannes Bugenhagen_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "115.612",
"value": "Johannes Bugenhagen_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "119.371",
"value": "Johannes Bugenhagen_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "120.235",
"value": "Johannes Bugenhagen_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "120.24",
"value": "Johannes Bugenhagen_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "120.241",
"value": "Johannes Bugenhagen_header_cell_0_0_0"
},
{
"operation": "hyperlink",
"time": "126.131",
"value": "Protestant Reformation"
},
{
"operation": "Now on",
"time": "127.08",
"value": "Reformation"
},
{
"operation": "Highlighting",
"time": "134.58",
"value": "Reformation_sentence_2"
},
{
"operation": "finish",
"time": "141.779",
"value": "finish"
}
]
}
|
exfever_4871
|
The Doctor was the most famous fictional character in Europe.
|
{
"evidence": [
"The Doctor (Doctor Who)",
"TARDIS"
],
"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 Doctor was the most famous fictional character in Europe.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Doctor (Doctor Who)",
"TARDIS"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6972,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8276
|
Karan Johar is the first Indian filmmaker to win the National Film Award for Best Popular Film Providing Wholesome Entertainment.
|
{
"evidence": [
"Karan Johar",
"Dharma Productions"
],
"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:\nKaran Johar is the first Indian filmmaker to win the National Film Award for Best Popular Film Providing Wholesome Entertainment.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Karan Johar",
"Dharma Productions"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6973,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
19163
|
Jacob De Haas, a Jewish journalist who was born in London in 1937, worked as a secretary of the First Zionist Congress and as a member of Hovevei Zion to promote the Zionist program of Theodor Herzl.
|
{
"evidence": [
"Jacob de Haas_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:\nJacob De Haas, a Jewish journalist who was born in London in 1937, worked as a secretary of the First Zionist Congress and as a member of Hovevei Zion to promote the Zionist program of Theodor Herzl.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Jacob de Haas_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6974,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.491",
"value": "?search="
},
{
"operation": "search",
"time": "11.429",
"value": "Jacob de Haas"
},
{
"operation": "Now on",
"time": "11.942",
"value": "Jacob de Haas"
},
{
"operation": "Highlighting",
"time": "28.96",
"value": "Jacob de Haas_sentence_0"
},
{
"operation": "finish",
"time": "94.048",
"value": "finish"
}
]
}
|
81566
|
The Angola Ferry was a small and little-known and difficult to access ferry service that crossed the Mississippi River connecting Lettsworth, Arkansas with the Louisiana State Penitentiary and served mainly prison personnel traveling to and from work until it ceased operation in 2016.
|
{
"evidence": [
"Angola Ferry_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 Angola Ferry was a small and little-known and difficult to access ferry service that crossed the Mississippi River connecting Lettsworth, Arkansas with the Louisiana State Penitentiary and served mainly prison personnel traveling to and from work until it ceased operation in 2016.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Angola Ferry_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6975,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.889",
"value": "?search="
},
{
"operation": "search",
"time": "33.118",
"value": "Contains...The Angola Ferry"
},
{
"operation": "Now on",
"time": "34.51",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "36.246",
"value": "Angola Ferry"
},
{
"operation": "Now on",
"time": "36.786",
"value": "Angola Ferry"
},
{
"operation": "Highlighting",
"time": "42.63",
"value": "Angola Ferry_sentence_0"
},
{
"operation": "finish",
"time": "60.692",
"value": "finish"
}
]
}
|
exfever_4051
|
Winnipeg is the capital and largest city of the province of Manitoba in Canada that is centred on the confluence of the river in the north-central United States and central Canada.
|
{
"evidence": [
"Winnipeg",
"Red River of the North"
],
"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:\nWinnipeg is the capital and largest city of the province of Manitoba in Canada that is centred on the confluence of the river in the north-central United States and central Canada.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Winnipeg",
"Red River of the North"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6976,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8426
|
How to Train Your Dragon is a 2010 American computer-animated fantasy film with a film genre in which the protagonist is thrust into a series of events that typically involve violence and physical feats.
|
{
"evidence": [
"How to Train Your Dragon (film)",
"Action film"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHow to Train Your Dragon is a 2010 American computer-animated fantasy film with a film genre in which the protagonist is thrust into a series of events that typically involve violence and physical feats.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"How to Train Your Dragon (film)",
"Action film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6977,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2741
|
Con Man is an Australian comedy web series that was acquired by an American basic cable channel owned by the NBCUniversal Television and Streaming division of Comcast through NBCUniversal Cable Entertainment.
|
{
"evidence": [
"Con Man (web series)",
"Syfy"
],
"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:\nCon Man is an Australian comedy web series that was acquired by an American basic cable channel owned by the NBCUniversal Television and Streaming division of Comcast through NBCUniversal Cable Entertainment.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Con Man (web series)",
"Syfy"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6978,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
68434
|
Zhangjiang Subdistrict with an area of 80.38 km is a subdistrict and the county seat of Taoyuan County incorporated from a part of the former Zhangjiang Town
|
{
"evidence": [
"Zhangjiang Subdistrict_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:\nZhangjiang Subdistrict with an area of 80.38 km is a subdistrict and the county seat of Taoyuan County incorporated from a part of the former Zhangjiang Town\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Zhangjiang Subdistrict_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6979,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.091",
"value": "?search="
},
{
"operation": "search",
"time": "11.466",
"value": "Zhangjiang Subdistrict"
},
{
"operation": "Now on",
"time": "12.613",
"value": "Zhangjiang Subdistrict"
},
{
"operation": "Highlighting",
"time": "72.561",
"value": "Zhangjiang Subdistrict_sentence_2"
},
{
"operation": "Highlighting",
"time": "81.774",
"value": "Zhangjiang Subdistrict_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "83.103",
"value": "Zhangjiang Subdistrict_sentence_0"
},
{
"operation": "finish",
"time": "92",
"value": "finish"
}
]
}
|
18292
|
The Pakistani team players outnumbered the Indian cricket team in the test matches and the One Day Internationals in 2007-08.
|
{
"evidence": [
"India national cricket team_sentence_0",
"India national cricket team_sentence_1",
"India national cricket team_sentence_11"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Pakistani team players outnumbered the Indian cricket team in the test matches and the One Day Internationals in 2007-08.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"India national cricket team_sentence_0",
"India national cricket team_sentence_1",
"India national cricket team_sentence_11"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6980,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.812",
"value": "?search="
},
{
"operation": "search",
"time": "19.763",
"value": "One Day Internationals 2007-08"
},
{
"operation": "Now on",
"time": "21.142",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "36.066",
"value": "One-Day International status (2007–present) Ireland cricket team_section_4"
},
{
"operation": "Now on",
"time": "36.947",
"value": "Ireland cricket team#One-Day International status .282007.E2.80.93present.29 Ireland cricket team section 4"
},
{
"operation": "back-button-clicked",
"time": "56.65",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "57.275",
"value": "undefined"
},
{
"operation": "search",
"time": "107.09",
"value": "Pakistani One Day Internationals 2007-08"
},
{
"operation": "Now on",
"time": "108.051",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "110.177",
"value": "pakistan one day internationals 2007 08"
},
{
"operation": "Now on",
"time": "110.935",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "120.929",
"value": "One Day Internationals Ajantha Mendis_section_7"
},
{
"operation": "Now on",
"time": "121.502",
"value": "Ajantha Mendis#One Day Internationals Ajantha Mendis section 7"
},
{
"operation": "search",
"time": "149.274",
"value": "One Day Internationals cricket"
},
{
"operation": "Now on",
"time": "150.07",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "168.825",
"value": "Maharashtra Cricket Association Stadium"
},
{
"operation": "Now on",
"time": "169.314",
"value": "Maharashtra Cricket Association Stadium"
},
{
"operation": "Page search",
"time": "175.25",
"value": "one day internationas"
},
{
"operation": "Page search",
"time": "178.297",
"value": "one day internationals"
},
{
"operation": "hyperlink",
"time": "182.753",
"value": "4.3 One Day Internationals Maharashtra Cricket Association Stadium_section_8"
},
{
"operation": "search",
"time": "220.665",
"value": "test matches one day internationals 2007-08"
},
{
"operation": "Now on",
"time": "221.806",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "242.281",
"value": "West Indian cricket team in South Africa in 2007–08"
},
{
"operation": "Now on",
"time": "242.726",
"value": "West Indian cricket team in South Africa in 2007–08"
},
{
"operation": "hyperlink",
"time": "270.001",
"value": "2 Test series West Indian cricket team in South Africa in 2007–08_section_3"
},
{
"operation": "hyperlink",
"time": "273.616",
"value": "West Indies cricket team"
},
{
"operation": "Now on",
"time": "274.468",
"value": "West Indies cricket team"
},
{
"operation": "search",
"time": "296.809",
"value": "Indian cricket 2007-08"
},
{
"operation": "Now on",
"time": "297.708",
"value": "undefined"
},
{
"operation": "search",
"time": "322.016",
"value": "Indian Cricket team test matches"
},
{
"operation": "Now on",
"time": "323.324",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "328.009",
"value": "India national cricket team"
},
{
"operation": "Now on",
"time": "328.909",
"value": "India national cricket team"
},
{
"operation": "hyperlink",
"time": "337.432",
"value": "2007 ICC World Twenty20"
},
{
"operation": "Now on",
"time": "338.007",
"value": "2007 ICC World Twenty20"
},
{
"operation": "back-button-clicked",
"time": "340.656",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "340.846",
"value": "India national cricket team"
},
{
"operation": "Page search",
"time": "349.787",
"value": "one day internationa;s"
},
{
"operation": "Page search",
"time": "351.489",
"value": "one day internationals"
},
{
"operation": "Highlighting",
"time": "362.129",
"value": "India national cricket team_sentence_0"
},
{
"operation": "Highlighting",
"time": "362.743",
"value": "India national cricket team_sentence_1"
},
{
"operation": "Highlighting",
"time": "370.431",
"value": "India national cricket team_sentence_11"
},
{
"operation": "finish",
"time": "378.159",
"value": "finish"
}
]
}
|
exfever_7028
|
David Dhawan is an American director who has directed several films, including a 1997 remake of the Telugu film Hello Brother.
|
{
"evidence": [
"David Dhawan",
"Judwaa"
],
"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:\nDavid Dhawan is an American director who has directed several films, including a 1997 remake of the Telugu film Hello Brother.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"David Dhawan",
"Judwaa"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6981,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
60231
|
DVCAM is an XDCAM format which uses an MXF container, a format in professional digital video defined by a set of SMPTE standards.
|
{
"evidence": [
"DV_sentence_1",
"DV_sentence_42"
],
"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:\nDVCAM is an XDCAM format which uses an MXF container, a format in professional digital video defined by a set of SMPTE standards.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"DV_sentence_1",
"DV_sentence_42"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6982,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.908",
"value": "?search="
},
{
"operation": "search",
"time": "12.505",
"value": "DVCAM"
},
{
"operation": "Now on",
"time": "14.372",
"value": "DV"
},
{
"operation": "Highlighting",
"time": "130.054",
"value": "DV_sentence_42"
},
{
"operation": "hyperlink",
"time": "225.585",
"value": "DV (disambiguation)"
},
{
"operation": "Now on",
"time": "226.856",
"value": "DV (disambiguation)"
},
{
"operation": "hyperlink",
"time": "245.465",
"value": "DV"
},
{
"operation": "Now on",
"time": "246.41",
"value": "DV"
},
{
"operation": "Highlighting",
"time": "250.032",
"value": "DV_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "335.486",
"value": "DV_sentence_42"
},
{
"operation": "Highlighting",
"time": "336.917",
"value": "DV_sentence_42"
},
{
"operation": "finish",
"time": "363.012",
"value": "finish"
}
]
}
|
exfever_1853
|
Ava Marie DuVernay directed her second feature film, a 2012 independent feature film winner at the film festival that took place from January 19 until January 29, 2012 in Park City, Utah.
|
{
"evidence": [
"Ava DuVernay",
"Middle of Nowhere (2012 film)",
"2012 Sundance Film Festival"
],
"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:\nAva Marie DuVernay directed her second feature film, a 2012 independent feature film winner at the film festival that took place from January 19 until January 29, 2012 in Park City, Utah.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ava DuVernay",
"Middle of Nowhere (2012 film)",
"2012 Sundance Film Festival"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6983,
"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.