id
stringlengths 1
12
| question
stringlengths 22
566
| golden_answers
dict | data_source
stringclasses 2
values | prompt
listlengths 1
1
| ability
stringclasses 6
values | reward_model
dict | extra_info
dict | metadata
dict |
|---|---|---|---|---|---|---|---|---|
exfever_4614
|
Terminator Genisys is a 2015 American science fiction action film written by Laeta Kalogridis and a Canadian-American filmmaker and editor.
|
{
"evidence": [
"Terminator Genisys",
"Patrick Lussier"
],
"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:\nTerminator Genisys is a 2015 American science fiction action film written by Laeta Kalogridis and a Canadian-American filmmaker and editor.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Terminator Genisys",
"Patrick Lussier"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5384,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
50672
|
The Nicaraguan public's excitement for baseball grew, the league had to shut down in 1967 but was re-established in 2004.
|
{
"evidence": [
"Nicaraguan Professional Baseball League_sentence_3",
"Nicaraguan Professional Baseball League_sentence_4",
"Nicaraguan Professional Baseball League_sentence_6"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Nicaraguan public's excitement for baseball grew, the league had to shut down in 1967 but was re-established in 2004.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Nicaraguan Professional Baseball League_sentence_3",
"Nicaraguan Professional Baseball League_sentence_4",
"Nicaraguan Professional Baseball League_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5385,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.784",
"value": "?search="
},
{
"operation": "Now on",
"time": "621.953",
"value": "?search="
},
{
"operation": "search",
"time": "689.665",
"value": "Nicaraguan Professional Baseball League"
},
{
"operation": "Now on",
"time": "690.538",
"value": "Nicaraguan Professional Baseball League"
},
{
"operation": "Page search",
"time": "695.872",
"value": "1967"
},
{
"operation": "Highlighting",
"time": "697.872",
"value": "Nicaraguan Professional Baseball League_sentence_3"
},
{
"operation": "Highlighting",
"time": "701.365",
"value": "Nicaraguan Professional Baseball League_sentence_4"
},
{
"operation": "Highlighting",
"time": "705.561",
"value": "Nicaraguan Professional Baseball League_sentence_6"
},
{
"operation": "finish",
"time": "730.006",
"value": "finish"
}
]
}
|
23843
|
Commitment Hour is a science fiction book published in 1998, and part of the series League of Peoples, written by James Alan Gardner.
|
{
"evidence": [
"Commitment Hour_sentence_0",
"Commitment Hour_sentence_1",
"League of Peoples_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:\nCommitment Hour is a science fiction book published in 1998, and part of the series League of Peoples, written by James Alan Gardner.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Commitment Hour_sentence_0",
"Commitment Hour_sentence_1",
"League of Peoples_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5386,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.623",
"value": "?search="
},
{
"operation": "search",
"time": "11.323",
"value": "Commitment Hour"
},
{
"operation": "Now on",
"time": "12.212",
"value": "Commitment Hour"
},
{
"operation": "Highlighting",
"time": "18.195",
"value": "Commitment Hour_sentence_0"
},
{
"operation": "Highlighting",
"time": "27.808",
"value": "Commitment Hour_sentence_1"
},
{
"operation": "hyperlink",
"time": "32.113",
"value": "League of Peoples"
},
{
"operation": "Now on",
"time": "32.607",
"value": "League of Peoples"
},
{
"operation": "Highlighting",
"time": "38.401",
"value": "League of Peoples_sentence_0"
},
{
"operation": "finish",
"time": "50.544",
"value": "finish"
}
]
}
|
exfever_7526
|
Heartbreak Ridge is a 1986 American war film produced and directed by Clint Eastwood with co-stars including an American actor, who first rose to prominence for his portrayal of a caveman in Quest for Fire.
|
{
"evidence": [
"Heartbreak Ridge",
"Everett McGill"
],
"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:\nHeartbreak Ridge is a 1986 American war film produced and directed by Clint Eastwood with co-stars including an American actor, who first rose to prominence for his portrayal of a caveman in Quest for Fire.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Heartbreak Ridge",
"Everett McGill"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5387,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2200
|
Humanz is the fifth studio album by a British virtual band, formed in 1998.
|
{
"evidence": [
"Humanz",
"Gorillaz"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHumanz is the fifth studio album by a British virtual band, formed in 1998.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Humanz",
"Gorillaz"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5388,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7631
|
Shenzhen is a city and special administrative region of China in the eastern region of one of the most densely urbanized regions in the world and is often considered a megacity in South China.
|
{
"evidence": [
"Hong Kong",
"Pearl River Delta"
],
"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:\nShenzhen is a city and special administrative region of China in the eastern region of one of the most densely urbanized regions in the world and is often considered a megacity in South China.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hong Kong",
"Pearl River Delta"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5389,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3656
|
Alvin and the Chipmunks is a 2007 American live-action/computer-animated comedy film whose genre in which songs by the characters interweave to the narrative, directed by Tim Hill.
|
{
"evidence": [
"Alvin and the Chipmunks (film)",
"Musical film"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAlvin and the Chipmunks is a 2007 American live-action/computer-animated comedy film whose genre in which songs by the characters interweave to the narrative, directed by Tim Hill.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Alvin and the Chipmunks (film)",
"Musical film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5390,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
66890
|
Selling fewer than 80,000 copies, The Papercut Chronicles II did not rank on the US Billboard Top Rap Albums, let alone on the UK R&B Albums Chart, in 2011.
|
{
"evidence": [
"The Papercut Chronicles II_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:\nSelling fewer than 80,000 copies, The Papercut Chronicles II did not rank on the US Billboard Top Rap Albums, let alone on the UK R&B Albums Chart, in 2011.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"The Papercut Chronicles II_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5391,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.602",
"value": "?search="
},
{
"operation": "search",
"time": "13.909",
"value": "The Papercut Chronicles II"
},
{
"operation": "Now on",
"time": "14.793",
"value": "The Papercut Chronicles II"
},
{
"operation": "Highlighting",
"time": "20.957",
"value": "The Papercut Chronicles II_sentence_3"
},
{
"operation": "finish",
"time": "37.256",
"value": "finish"
}
]
}
|
75952
|
Doman trained at the Nottingham School of Art from 1897 to 1901, where he was a pupil of Joseph Else; then he worked for his father and studied at the Royal College of Art, a public research university in South Kensington, London from 1905 to 1908.
|
{
"evidence": [
"Charles Doman_sentence_2",
"Charles Doman_sentence_3",
"Royal College of Art_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:\nDoman trained at the Nottingham School of Art from 1897 to 1901, where he was a pupil of Joseph Else; then he worked for his father and studied at the Royal College of Art, a public research university in South Kensington, London from 1905 to 1908.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Charles Doman_sentence_2",
"Charles Doman_sentence_3",
"Royal College of Art_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5392,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.496",
"value": "undefined"
},
{
"operation": "search",
"time": "15.726",
"value": "Doman"
},
{
"operation": "Now on",
"time": "17.031",
"value": "Doman"
},
{
"operation": "hyperlink",
"time": "25.493",
"value": "Doman (surname)"
},
{
"operation": "Now on",
"time": "26.004",
"value": "Doman (surname)"
},
{
"operation": "search",
"time": "38.36",
"value": "Joseph Else"
},
{
"operation": "Now on",
"time": "38.992",
"value": "Joseph Else"
},
{
"operation": "hyperlink",
"time": "77.626",
"value": "Royal College of Art"
},
{
"operation": "Now on",
"time": "78.803",
"value": "Royal College of Art"
},
{
"operation": "Highlighting",
"time": "81.981",
"value": "Royal College of Art_sentence_0"
},
{
"operation": "search",
"time": "106.087",
"value": "Duman"
},
{
"operation": "Now on",
"time": "106.799",
"value": "Duman"
},
{
"operation": "search",
"time": "119.924",
"value": "dumas"
},
{
"operation": "Now on",
"time": "120.622",
"value": "Dumas"
},
{
"operation": "search",
"time": "151.356",
"value": "Nottingham School of Art"
},
{
"operation": "Now on",
"time": "154.743",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "158.053",
"value": "Charles Doman"
},
{
"operation": "Now on",
"time": "158.42",
"value": "Charles Doman"
},
{
"operation": "Highlighting",
"time": "165.208",
"value": "Charles Doman_sentence_2"
},
{
"operation": "Highlighting",
"time": "173.139",
"value": "Charles Doman_sentence_3"
},
{
"operation": "back-button-clicked",
"time": "210.58",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "212.077",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "214.765",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "214.77",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "214.82",
"value": "Dumas"
},
{
"operation": "back-button-clicked",
"time": "217.162",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "217.166",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "217.206",
"value": "Duman"
},
{
"operation": "back-button-clicked",
"time": "218.309",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "218.314",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "218.375",
"value": "Royal College of Art"
},
{
"operation": "Highlighting deleted",
"time": "220.986",
"value": "Royal College of Art_sentence_0"
},
{
"operation": "Highlighting",
"time": "221.706",
"value": "Royal College of Art_sentence_0"
},
{
"operation": "finish",
"time": "232.652",
"value": "finish"
}
]
}
|
exfever_8637
|
The Woman in Red is a 1984 American romantic comedy film with a co-star, an American actress and model who also starred in a 1985 American science fantasy comedy film.
|
{
"evidence": [
"The Woman in Red (1984 film)",
"Kelly LeBrock",
"Weird Science (film)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Woman in Red is a 1984 American romantic comedy film with a co-star, an American actress and model who also starred in a 1985 American science fantasy comedy film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Woman in Red (1984 film)",
"Kelly LeBrock",
"Weird Science (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5393,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
14261
|
The London Thrashers were the first team formed in the Forect City Roller Derby, but did not play any games that year; they never played the Thames Fatales before they disbanded in 2008 to consolidate with other players.
|
{
"evidence": [
"Forest City Roller Derby_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:\nThe London Thrashers were the first team formed in the Forect City Roller Derby, but did not play any games that year; they never played the Thames Fatales before they disbanded in 2008 to consolidate with other players.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Forest City Roller Derby_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5394,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.769",
"value": "?search="
},
{
"operation": "search",
"time": "8.882",
"value": "the london thrashers"
},
{
"operation": "Now on",
"time": "10.172",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "12.051",
"value": "Forest City Roller Derby"
},
{
"operation": "Now on",
"time": "12.745",
"value": "Forest City Roller Derby"
},
{
"operation": "Highlighting",
"time": "17.605",
"value": "Forest City Roller Derby_sentence_4"
},
{
"operation": "Highlighting",
"time": "29.591",
"value": "Forest City Roller Derby_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "30.327",
"value": "Forest City Roller Derby_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "61.139",
"value": "Forest City Roller Derby_sentence_7"
},
{
"operation": "Highlighting",
"time": "70.395",
"value": "Forest City Roller Derby_sentence_6"
},
{
"operation": "finish",
"time": "81.431",
"value": "finish"
}
]
}
|
exfever_1031
|
Leicester City Football Club is a professional football club that competes in the Premier League, the highest level of a series of interconnected leagues for men's association football clubs in a country that is part of the United Kingdom.
|
{
"evidence": [
"Leicester City F.C.",
"English football league system",
"England"
],
"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:\nLeicester City Football Club is a professional football club that competes in the Premier League, the highest level of a series of interconnected leagues for men's association football clubs in a country that is part of the United Kingdom.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Leicester City F.C.",
"English football league system",
"England"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5395,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
27433
|
William Church Moir never employed workers in his small bakery, where, he made over 500 types of confectionery.
|
{
"evidence": [
"William Church Moir_sentence_4"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWilliam Church Moir never employed workers in his small bakery, where, he made over 500 types of confectionery.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"William Church Moir_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5396,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.219",
"value": "?search="
},
{
"operation": "search",
"time": "12.522",
"value": "William Church Moir"
},
{
"operation": "Now on",
"time": "13.288",
"value": "William Church Moir"
},
{
"operation": "Highlighting",
"time": "21.434",
"value": "William Church Moir_sentence_4"
},
{
"operation": "finish",
"time": "35.189",
"value": "finish"
}
]
}
|
54495
|
Usually seen covered with sponge which they apply themselves, scorpion spider crabs resemble the closely related species Inachus phalangium, a species of crabs from the north-eastern Atlantic Ocean and Mediterranean Sea.
|
{
"evidence": [
"Scorpion spider crab_sentence_2",
"Scorpion spider crab_sentence_4",
"Inachus phalangium_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:\nUsually seen covered with sponge which they apply themselves, scorpion spider crabs resemble the closely related species Inachus phalangium, a species of crabs from the north-eastern Atlantic Ocean and Mediterranean Sea.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Scorpion spider crab_sentence_2",
"Scorpion spider crab_sentence_4",
"Inachus phalangium_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5397,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.18",
"value": "?search="
},
{
"operation": "search",
"time": "29.894",
"value": "Scorpion spider crab"
},
{
"operation": "Now on",
"time": "31.758",
"value": "Scorpion spider crab"
},
{
"operation": "Highlighting",
"time": "44.015",
"value": "Scorpion spider crab_sentence_2"
},
{
"operation": "Highlighting",
"time": "62.166",
"value": "Scorpion spider crab_sentence_4"
},
{
"operation": "hyperlink",
"time": "65.074",
"value": "Inachus phalangium"
},
{
"operation": "Now on",
"time": "65.812",
"value": "Inachus phalangium"
},
{
"operation": "Highlighting",
"time": "69.283",
"value": "Inachus phalangium_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "71.9",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "72.06",
"value": "Scorpion spider crab"
},
{
"operation": "finish",
"time": "313.178",
"value": "finish"
}
]
}
|
67157
|
Sepia mestus ejects water from the mantle which allows it to move through the environment using jet propulsion.
|
{
"evidence": [
"Sepia mestus_sentence_18",
"Sepia mestus_sentence_19"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSepia mestus ejects water from the mantle which allows it to move through the environment using jet propulsion.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sepia mestus_sentence_18",
"Sepia mestus_sentence_19"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5398,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.736",
"value": "?search="
},
{
"operation": "Now on",
"time": "6.604",
"value": "?search="
},
{
"operation": "search",
"time": "15.174",
"value": "sepia mestus"
},
{
"operation": "Now on",
"time": "15.917",
"value": "Sepia mestus"
},
{
"operation": "Highlighting",
"time": "58.113",
"value": "Sepia mestus_sentence_18"
},
{
"operation": "Highlighting",
"time": "59.927",
"value": "Sepia mestus_sentence_19"
},
{
"operation": "finish",
"time": "69.741",
"value": "finish"
}
]
}
|
81771
|
Harold Jarman(born 4 May 1939) father was permanently blinded, he attended Hotwells Primary School and he also showed an aptitude for both cricket and football
|
{
"evidence": [
"Harold Jarman_sentence_0",
"Harold Jarman_sentence_7",
"Harold Jarman_sentence_8",
"Harold Jarman_sentence_5"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHarold Jarman(born 4 May 1939) father was permanently blinded, he attended Hotwells Primary School and he also showed an aptitude for both cricket and football\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Harold Jarman_sentence_0",
"Harold Jarman_sentence_7",
"Harold Jarman_sentence_8",
"Harold Jarman_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5399,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.139",
"value": "?search="
},
{
"operation": "search",
"time": "18.012",
"value": "Harold Jarman"
},
{
"operation": "Now on",
"time": "19.336",
"value": "Harold Jarman"
},
{
"operation": "Highlighting",
"time": "23.257",
"value": "Harold Jarman_sentence_0"
},
{
"operation": "Page search",
"time": "36.993",
"value": "blind"
},
{
"operation": "Highlighting",
"time": "43.769",
"value": "Harold Jarman_sentence_7"
},
{
"operation": "Highlighting",
"time": "52.697",
"value": "Harold Jarman_sentence_8"
},
{
"operation": "Page search",
"time": "70.322",
"value": "cricket"
},
{
"operation": "Highlighting",
"time": "76.189",
"value": "Harold Jarman_sentence_5"
},
{
"operation": "finish",
"time": "317.088",
"value": "finish"
}
]
}
|
exfever_4086
|
Robert J. O'Neill participated in May 2011's Operation Neptune Spear with the United States Navy component of the Joint Special Operations Command.
|
{
"evidence": [
"Robert J. O'Neill (U.S. Navy SEAL)",
"SEAL Team Six"
],
"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:\nRobert J. O'Neill participated in May 2011's Operation Neptune Spear with the United States Navy component of the Joint Special Operations Command.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Robert J. O'Neill (U.S. Navy SEAL)",
"SEAL Team Six"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5400,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6671
|
Salman Khan is the first Indian actor to win a Filmfare Award.
|
{
"evidence": [
"Salman Khan",
"Forbes",
"Forbes Global 2000"
],
"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:\nSalman Khan is the first Indian actor to win a Filmfare Award.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Salman Khan",
"Forbes",
"Forbes Global 2000"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5401,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
41246
|
Mailloux, born on November 2, 1983, participated in the popular Quebec talk show Tout le Monde en Parle, then participated in the 2012 Army Run taking place in Ottawa, Canada.
|
{
"evidence": [
"Simon Mailloux_sentence_0",
"Simon Mailloux_sentence_21",
"Simon Mailloux_sentence_22"
],
"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:\nMailloux, born on November 2, 1983, participated in the popular Quebec talk show Tout le Monde en Parle, then participated in the 2012 Army Run taking place in Ottawa, Canada.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Simon Mailloux_sentence_0",
"Simon Mailloux_sentence_21",
"Simon Mailloux_sentence_22"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5402,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.185",
"value": "?search="
},
{
"operation": "search",
"time": "12.045",
"value": "Mailloux"
},
{
"operation": "Now on",
"time": "13.339",
"value": "Mailloux"
},
{
"operation": "hyperlink",
"time": "39.146",
"value": "Mailloux River"
},
{
"operation": "Now on",
"time": "39.812",
"value": "Mailloux River"
},
{
"operation": "back-button-clicked",
"time": "47.591",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "47.629",
"value": "Mailloux"
},
{
"operation": "search",
"time": "94.508",
"value": "\"Mailloux\" \"1983\" \"quebec\""
},
{
"operation": "Now on",
"time": "95.836",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "106.185",
"value": "Simon Mailloux"
},
{
"operation": "Now on",
"time": "107.232",
"value": "Simon Mailloux"
},
{
"operation": "Highlighting",
"time": "110.248",
"value": "Simon Mailloux_sentence_0"
},
{
"operation": "Page search",
"time": "120.036",
"value": "quebec"
},
{
"operation": "Highlighting",
"time": "144.213",
"value": "Simon Mailloux_sentence_21"
},
{
"operation": "Highlighting",
"time": "155.321",
"value": "Simon Mailloux_sentence_22"
},
{
"operation": "finish",
"time": "189.598",
"value": "finish"
}
]
}
|
exfever_7267
|
Election is a 1979 American black comedy film directed by Alexander Payne and starring an American actress and producer who has received various accolades, including an Academy Award.
|
{
"evidence": [
"Election (1999 film)",
"Reese Witherspoon"
],
"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:\nElection is a 1979 American black comedy film directed by Alexander Payne and starring an American actress and producer who has received various accolades, including an Academy Award.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Election (1999 film)",
"Reese Witherspoon"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5403,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
17365
|
Spanish comedy, Casa de mi padre starring Will Ferrell opened in 2012 and debuted in first place; it receieved mixed reviews from critics.
|
{
"evidence": [
"Casa de mi padre_sentence_27"
],
"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:\nSpanish comedy, Casa de mi padre starring Will Ferrell opened in 2012 and debuted in first place; it receieved mixed reviews from critics.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Casa de mi padre_sentence_27"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5404,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.056",
"value": "?search="
},
{
"operation": "search",
"time": "11.488",
"value": "Casa de mi padre"
},
{
"operation": "Now on",
"time": "12.355",
"value": "Casa de mi padre"
},
{
"operation": "Highlighting",
"time": "52.46",
"value": "Casa de mi padre_sentence_0"
},
{
"operation": "hyperlink",
"time": "69.706",
"value": "6 Reception Casa de mi padre_section_5"
},
{
"operation": "Highlighting",
"time": "73.528",
"value": "Casa de mi padre_sentence_27"
},
{
"operation": "Highlighting deleted",
"time": "86.157",
"value": "Casa de mi padre_sentence_0"
},
{
"operation": "finish",
"time": "97.542",
"value": "finish"
}
]
}
|
exfever_8526
|
Aamir_Khan was known for consistent performances and intelligent choice of scripts.
|
{
"evidence": [
"Dangal (film)",
"Aamir Khan"
],
"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:\nAamir_Khan was known for consistent performances and intelligent choice of scripts.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dangal (film)",
"Aamir Khan"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5405,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
88796
|
The 1968 United States presidential election in Nebraska, which took place on November 5, 1968, resulted to a win for Republican candidate Richard Nixon who gathered almost 60% of the votes.
|
{
"evidence": [
"1968 United States presidential election in Nebraska_sentence_1",
"1968 United States presidential election in Nebraska_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:\nThe 1968 United States presidential election in Nebraska, which took place on November 5, 1968, resulted to a win for Republican candidate Richard Nixon who gathered almost 60% of the votes.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"1968 United States presidential election in Nebraska_sentence_1",
"1968 United States presidential election in Nebraska_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5406,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.446",
"value": "?search="
},
{
"operation": "search",
"time": "7.946",
"value": "1968 United States presidential election"
},
{
"operation": "Now on",
"time": "11.502",
"value": "1968 United States presidential election"
},
{
"operation": "search",
"time": "33.976",
"value": "1968 United States presidential election in Nebraska"
},
{
"operation": "Now on",
"time": "36.234",
"value": "1968 United States presidential election in Nebraska"
},
{
"operation": "Highlighting",
"time": "38.056",
"value": "1968 United States presidential election in Nebraska_sentence_1"
},
{
"operation": "Highlighting",
"time": "44.365",
"value": "1968 United States presidential election in Nebraska_sentence_3"
},
{
"operation": "finish",
"time": "136.764",
"value": "finish"
}
]
}
|
73978
|
"Fall in Line" is centered around lyrical themes of female empowerment, the song received positive reviews from music critics but made minor impact on charts.
|
{
"evidence": [
"Fall in Line_sentence_3",
"Fall in Line_sentence_5",
"Fall in Line_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:\n\"Fall in Line\" is centered around lyrical themes of female empowerment, the song received positive reviews from music critics but made minor impact on charts.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Fall in Line_sentence_3",
"Fall in Line_sentence_5",
"Fall in Line_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5407,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.39",
"value": "?search="
},
{
"operation": "search",
"time": "18.48",
"value": "Fall in Line"
},
{
"operation": "Now on",
"time": "20.142",
"value": "Fall in Line"
},
{
"operation": "Highlighting",
"time": "44.426",
"value": "Fall in Line_sentence_3"
},
{
"operation": "Highlighting",
"time": "49.542",
"value": "Fall in Line_sentence_5"
},
{
"operation": "Highlighting",
"time": "52.162",
"value": "Fall in Line_sentence_4"
},
{
"operation": "finish",
"time": "76.02",
"value": "finish"
}
]
}
|
exfever_5943
|
Sheryl Lee was cast as Laura Palmer and Maddy Ferguson in a mystery-horror serial drama television series that premiered on a Canadian multinational commercial broadcast television network.
|
{
"evidence": [
"Sheryl Lee",
"Twin Peaks",
"American Broadcasting Company"
],
"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:\nSheryl Lee was cast as Laura Palmer and Maddy Ferguson in a mystery-horror serial drama television series that premiered on a Canadian multinational commercial broadcast television network.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sheryl Lee",
"Twin Peaks",
"American Broadcasting Company"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5408,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
60581
|
It was written that John Taylor Caroline believed that men "were pure evil" and that "self interest was the only realy constant in human interaction;" and he believed that "the principles of a society may be virtuous, though the individuals composing it are vicious."
|
{
"evidence": [
"John Taylor of Caroline_sentence_33"
],
"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:\nIt was written that John Taylor Caroline believed that men \"were pure evil\" and that \"self interest was the only realy constant in human interaction;\" and he believed that \"the principles of a society may be virtuous, though the individuals composing it are vicious.\"\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"John Taylor of Caroline_sentence_33"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5409,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.491",
"value": "undefined"
},
{
"operation": "search",
"time": "34.334",
"value": "John Taylor Caroline"
},
{
"operation": "Now on",
"time": "35.962",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "40.606",
"value": "John Taylor of Caroline"
},
{
"operation": "Now on",
"time": "41.213",
"value": "John Taylor of Caroline"
},
{
"operation": "Highlighting",
"time": "45.27",
"value": "John Taylor of Caroline_sentence_0"
},
{
"operation": "Page search",
"time": "52.847",
"value": "evil"
},
{
"operation": "Highlighting",
"time": "61.553",
"value": "John Taylor of Caroline_sentence_33"
},
{
"operation": "Highlighting deleted",
"time": "121.622",
"value": "John Taylor of Caroline_sentence_0"
},
{
"operation": "finish",
"time": "162.441",
"value": "finish"
}
]
}
|
11911
|
Syracuse is a city in the county seat of Onodaga County, New York and Henninger High School is the only high school within the Syracuse School District.
|
{
"evidence": [
"Syracuse, New York_sentence_2",
"Henninger High School_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:\nSyracuse is a city in the county seat of Onodaga County, New York and Henninger High School is the only high school within the Syracuse School District.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Syracuse, New York_sentence_2",
"Henninger High School_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5410,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.602",
"value": "?search="
},
{
"operation": "search",
"time": "7.4",
"value": "Syracuse"
},
{
"operation": "Now on",
"time": "8.518",
"value": "Syracuse"
},
{
"operation": "Highlighting",
"time": "21",
"value": "Syracuse_item_1_2"
},
{
"operation": "Highlighting",
"time": "21.001",
"value": "Syracuse_title"
},
{
"operation": "Highlighting",
"time": "21.001",
"value": "Syracuse_section_0"
},
{
"operation": "Highlighting",
"time": "21.001",
"value": "Syracuse_section_2"
},
{
"operation": "Highlighting deleted",
"time": "22.719",
"value": "Syracuse_item_1_2"
},
{
"operation": "Highlighting deleted",
"time": "22.72",
"value": "Syracuse_title"
},
{
"operation": "Highlighting deleted",
"time": "22.723",
"value": "Syracuse_section_0"
},
{
"operation": "Highlighting deleted",
"time": "22.725",
"value": "Syracuse_section_2"
},
{
"operation": "hyperlink",
"time": "29.465",
"value": "Syracuse, New York"
},
{
"operation": "Now on",
"time": "32.434",
"value": "Syracuse, New York"
},
{
"operation": "Highlighting",
"time": "33.358",
"value": "Syracuse, New York_sentence_2"
},
{
"operation": "Page search",
"time": "44.707",
"value": "enninger High School"
},
{
"operation": "search",
"time": "71.659",
"value": "enninger High School"
},
{
"operation": "Now on",
"time": "74.144",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "81.34",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "81.655",
"value": "Syracuse, New York"
},
{
"operation": "Page search",
"time": "83.855",
"value": "Henninger High School"
},
{
"operation": "search",
"time": "99.793",
"value": "Henninger High School"
},
{
"operation": "Now on",
"time": "100.865",
"value": "Henninger High School"
},
{
"operation": "search",
"time": "118.442",
"value": "Syracuse School District"
},
{
"operation": "Now on",
"time": "119.981",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "128.973",
"value": "Schools North Syracuse Central School District_section_2"
},
{
"operation": "Now on",
"time": "130.036",
"value": "North Syracuse Central School District#Schools North Syracuse Central School District section 2"
},
{
"operation": "back-button-clicked",
"time": "142.1",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "143.267",
"value": "undefined"
},
{
"operation": "search",
"time": "147.813",
"value": "Syracuse city School District"
},
{
"operation": "Now on",
"time": "149.259",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "159.604",
"value": "Henninger High School"
},
{
"operation": "Now on",
"time": "160.125",
"value": "Henninger High School"
},
{
"operation": "Highlighting",
"time": "164.516",
"value": "Henninger High School_cell_0_4_1"
},
{
"operation": "Highlighting",
"time": "164.523",
"value": "Henninger High School_header_cell_0_4_0"
},
{
"operation": "Highlighting",
"time": "164.524",
"value": "Henninger High School_header_cell_0_2_0"
},
{
"operation": "Highlighting",
"time": "164.524",
"value": "Henninger High School_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "164.525",
"value": "Henninger High School_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "169.092",
"value": "Henninger High School_cell_0_4_1"
},
{
"operation": "Highlighting deleted",
"time": "169.098",
"value": "Henninger High School_header_cell_0_4_0"
},
{
"operation": "Highlighting deleted",
"time": "169.1",
"value": "Henninger High School_header_cell_0_2_0"
},
{
"operation": "Highlighting deleted",
"time": "169.102",
"value": "Henninger High School_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "169.104",
"value": "Henninger High School_header_cell_0_0_0"
},
{
"operation": "search",
"time": "171.094",
"value": "Syracuse City School District"
},
{
"operation": "Now on",
"time": "171.951",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "194.006",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "194.141",
"value": "Henninger High School"
},
{
"operation": "back-button-clicked",
"time": "195.778",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "195.783",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "197.01",
"value": "undefined"
},
{
"operation": "search",
"time": "203.191",
"value": "Syracuse City School District"
},
{
"operation": "Now on",
"time": "204.375",
"value": "undefined"
},
{
"operation": "search",
"time": "216.525",
"value": "Syracuse school District"
},
{
"operation": "Now on",
"time": "218.263",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "221.239",
"value": "North Syracuse Central School District"
},
{
"operation": "Now on",
"time": "221.558",
"value": "North Syracuse Central School District"
},
{
"operation": "hyperlink",
"time": "225.983",
"value": "Syracuse, New York"
},
{
"operation": "Now on",
"time": "226.673",
"value": "Syracuse, New York"
},
{
"operation": "Page search",
"time": "235.911",
"value": "school"
},
{
"operation": "Page search",
"time": "241.488",
"value": "school districy"
},
{
"operation": "Page search",
"time": "242.598",
"value": "school district"
},
{
"operation": "search",
"time": "281.415",
"value": "Henninger High School"
},
{
"operation": "Now on",
"time": "282.341",
"value": "Henninger High School"
},
{
"operation": "Highlighting",
"time": "322.327",
"value": "Henninger High School_sentence_0"
},
{
"operation": "finish",
"time": "328.443",
"value": "finish"
}
]
}
|
exfever_5543
|
Andrew Geoffrey Kaufman was an American actor and performance artist who had performed in comedy clubs and theaters as an absurdly foul-mouthed and domineering lounge singer claiming to hail from Las Vegas.
|
{
"evidence": [
"Andy Kaufman",
"Tony Clifton"
],
"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:\nAndrew Geoffrey Kaufman was an American actor and performance artist who had performed in comedy clubs and theaters as an absurdly foul-mouthed and domineering lounge singer claiming to hail from Las Vegas.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Andy Kaufman",
"Tony Clifton"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5411,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3757
|
As the World Turns is Annie Parisse's fifth appearance in a soap opera.
|
{
"evidence": [
"Annie Parisse",
"Daytime Emmy Award"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAs the World Turns is Annie Parisse's fifth appearance in a soap opera.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Annie Parisse",
"Daytime Emmy Award"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5412,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
5294
|
The Cecil Kelley criticality accident involved an inexperienced 28-year-old Kelley who was permitted to operate a large capacity mixing tank of potentially dangerous chemicals.
|
{
"evidence": [
"Cecil Kelley criticality accident_sentence_3"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Cecil Kelley criticality accident involved an inexperienced 28-year-old Kelley who was permitted to operate a large capacity mixing tank of potentially dangerous chemicals.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Cecil Kelley criticality accident_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5413,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.936",
"value": "?search="
},
{
"operation": "search",
"time": "13.118",
"value": "Cecil Kelley criticality accident"
},
{
"operation": "Now on",
"time": "14.308",
"value": "Cecil Kelley criticality accident"
},
{
"operation": "Highlighting",
"time": "33.339",
"value": "Cecil Kelley criticality accident_sentence_3"
},
{
"operation": "finish",
"time": "59.238",
"value": "finish"
}
]
}
|
21377
|
Found in China, Drepanidae chinoea has a wingspan of 16-20.5 mm for males, with lustrous white fore- and hindwings.
|
{
"evidence": [
"Drepanidae_sentence_3"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nFound in China, Drepanidae chinoea has a wingspan of 16-20.5 mm for males, with lustrous white fore- and hindwings.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Drepanidae_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5414,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.619",
"value": "?search="
},
{
"operation": "search",
"time": "27.18",
"value": "Drepanidae chinoea"
},
{
"operation": "Now on",
"time": "27.845",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "91.159",
"value": "drepanidae chingola"
},
{
"operation": "Now on",
"time": "91.716",
"value": "undefined"
},
{
"operation": "search",
"time": "112.482",
"value": "Contains...Drepanidae chinoe"
},
{
"operation": "Now on",
"time": "112.938",
"value": "undefined"
},
{
"operation": "search",
"time": "131.282",
"value": "Drepanidae"
},
{
"operation": "Now on",
"time": "132.134",
"value": "Drepanidae"
},
{
"operation": "Highlighting",
"time": "148.513",
"value": "Drepanidae_sentence_3"
},
{
"operation": "finish",
"time": "163.631",
"value": "finish"
}
]
}
|
exfever_1385
|
No Escape is a 2015 American action thriller film starring an English actor with a long association with filmmaker Wes Anderson.
|
{
"evidence": [
"No Escape (2015 film)",
"Owen Wilson"
],
"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:\nNo Escape is a 2015 American action thriller film starring an English actor with a long association with filmmaker Wes Anderson.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"No Escape (2015 film)",
"Owen Wilson"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5415,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8053
|
The Blue Lagoon is a novel that inspired film adaptations, most notably The Blue Lagoon, starring an American actress and model who gained critical acclaim for her role in a 1978 historical drama film.
|
{
"evidence": [
"The Blue Lagoon (novel)",
"Brooke Shields",
"Pretty Baby (1978 film)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Blue Lagoon is a novel that inspired film adaptations, most notably The Blue Lagoon, starring an American actress and model who gained critical acclaim for her role in a 1978 historical drama film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Blue Lagoon (novel)",
"Brooke Shields",
"Pretty Baby (1978 film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5416,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3732
|
A glacier forms where the accumulation of rain exceeds its destruction, which includes ablative materials like a component of a structure that slows the spread of the effects of smoke without system activation.
|
{
"evidence": [
"Glacier",
"Ablation",
"Passive fire protection"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nA glacier forms where the accumulation of rain exceeds its destruction, which includes ablative materials like a component of a structure that slows the spread of the effects of smoke without system activation.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Glacier",
"Ablation",
"Passive fire protection"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5417,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7588
|
Lee Daniels is an English film producer and screenwriter who directed a 2009 American drama film, based on the debut novel by American author Sapphire and was made into a film, which won numerous accolades.
|
{
"evidence": [
"Lee Daniels",
"Precious (film)",
"Push (novel)"
],
"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:\nLee Daniels is an English film producer and screenwriter who directed a 2009 American drama film, based on the debut novel by American author Sapphire and was made into a film, which won numerous accolades.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lee Daniels",
"Precious (film)",
"Push (novel)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5418,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3711
|
Romelu Lukaku Bolingoli is a Canadian professional footballer who plays as a striker for the top level of the football pyramid, a series of interconnected leagues for men's association football clubs in England.
|
{
"evidence": [
"Romelu Lukaku",
"Premier League",
"English football league system"
],
"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:\nRomelu Lukaku Bolingoli is a Canadian professional footballer who plays as a striker for the top level of the football pyramid, a series of interconnected leagues for men's association football clubs in England.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Romelu Lukaku",
"Premier League",
"English football league system"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5419,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
768
|
Irwin Kula (featured in the 2005 film Time for a New God) is a blogger for the Huffington Post "On Faith" column and an adjunct member at United theological Seminary.
|
{
"evidence": [
"Irwin Kula_sentence_6",
"Irwin Kula_sentence_7",
"Irwin Kula_sentence_9"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIrwin Kula (featured in the 2005 film Time for a New God) is a blogger for the Huffington Post \"On Faith\" column and an adjunct member at United theological Seminary.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Irwin Kula_sentence_6",
"Irwin Kula_sentence_7",
"Irwin Kula_sentence_9"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5420,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "16.761",
"value": "Irwin Kula"
},
{
"operation": "Highlighting",
"time": "29.916",
"value": "Irwin Kula_sentence_6"
},
{
"operation": "Highlighting",
"time": "37.132",
"value": "Irwin Kula_sentence_7"
},
{
"operation": "Highlighting",
"time": "51.645",
"value": "Irwin Kula_sentence_9"
},
{
"operation": "search",
"time": "64.743",
"value": "Contains...time for a new god"
},
{
"operation": "search",
"time": "79.166",
"value": "Irwin Kula"
},
{
"operation": "finish",
"time": "105.356",
"value": "finish"
}
]
}
|
63493
|
Lundell's work was a combination of conservation and economics, he was interested Maya culture and discovered many living things.
|
{
"evidence": [
"Cyrus Longworth Lundell_sentence_6",
"Cyrus Longworth Lundell_sentence_7",
"Cyrus Longworth Lundell_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:\nLundell's work was a combination of conservation and economics, he was interested Maya culture and discovered many living things.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Cyrus Longworth Lundell_sentence_6",
"Cyrus Longworth Lundell_sentence_7",
"Cyrus Longworth Lundell_sentence_9"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5421,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.16",
"value": "?search="
},
{
"operation": "search",
"time": "187.524",
"value": "Lundell"
},
{
"operation": "Now on",
"time": "188.76",
"value": "Lundell"
},
{
"operation": "hyperlink",
"time": "215.122",
"value": "Ulf Lundell"
},
{
"operation": "Now on",
"time": "221.028",
"value": "Ulf Lundell"
},
{
"operation": "back-button-clicked",
"time": "284.211",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "284.292",
"value": "Lundell"
},
{
"operation": "hyperlink",
"time": "293.155",
"value": "Johan August Lundell"
},
{
"operation": "Now on",
"time": "293.936",
"value": "Johan August Lundell"
},
{
"operation": "back-button-clicked",
"time": "306.33",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "306.443",
"value": "Lundell"
},
{
"operation": "Highlighting",
"time": "310.689",
"value": "Lundell_item_0_3"
},
{
"operation": "Highlighting",
"time": "310.69",
"value": "Lundell_title"
},
{
"operation": "Highlighting deleted",
"time": "312.048",
"value": "Lundell_item_0_3"
},
{
"operation": "Highlighting deleted",
"time": "312.049",
"value": "Lundell_title"
},
{
"operation": "hyperlink",
"time": "317.537",
"value": "Cyrus Longworth Lundell"
},
{
"operation": "Now on",
"time": "318.211",
"value": "Cyrus Longworth Lundell"
},
{
"operation": "Highlighting",
"time": "345.301",
"value": "Cyrus Longworth Lundell_sentence_6"
},
{
"operation": "Highlighting",
"time": "350.599",
"value": "Cyrus Longworth Lundell_sentence_7"
},
{
"operation": "Highlighting",
"time": "359.369",
"value": "Cyrus Longworth Lundell_sentence_9"
},
{
"operation": "finish",
"time": "381.589",
"value": "finish"
}
]
}
|
exfever_3787
|
Sir Richard Branson is a British billionaire that founded the Virgin Rail Group to bid for passenger rail franchises during the process by which the government passed control into the private hands of the operation of the newest railway in the world.
|
{
"evidence": [
"Richard Branson",
"Privatisation of British Rail",
"Rail transport in Great Britain"
],
"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:\nSir Richard Branson is a British billionaire that founded the Virgin Rail Group to bid for passenger rail franchises during the process by which the government passed control into the private hands of the operation of the newest railway in the world.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Richard Branson",
"Privatisation of British Rail",
"Rail transport in Great Britain"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5422,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2874
|
Thomas Ruggles Pynchon Jr. is a novelist who is also a songwriter.
|
{
"evidence": [
"Thomas Pynchon",
"Science"
],
"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:\nThomas Ruggles Pynchon Jr. is a novelist who is also a songwriter.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Thomas Pynchon",
"Science"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5423,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6305
|
NSYNC was an American boy band consisting of a British singer, dancer, actor, and television personality who came in second place on the ABC reality show Dancing with the Stars.
|
{
"evidence": [
"NSYNC",
"Joey Fatone"
],
"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:\nNSYNC was an American boy band consisting of a British singer, dancer, actor, and television personality who came in second place on the ABC reality show Dancing with the Stars.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"NSYNC",
"Joey Fatone"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5424,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
35188
|
The flow process chart was introduced by Frank and Lillian Gilbreth to members of ASME in 1921, then in the 1930s, an industrial engineer, Allan H. Mogensen ( American industrial engineer, and industry consultant) began training business people in the use of some of the tools of industrial engineering at his Work Simplification Conferences in Lake Placid, New York.
|
{
"evidence": [
"Flow process chart_sentence_1",
"Flow process chart_sentence_3",
"Allan H. Mogensen_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 flow process chart was introduced by Frank and Lillian Gilbreth to members of ASME in 1921, then in the 1930s, an industrial engineer, Allan H. Mogensen ( American industrial engineer, and industry consultant) began training business people in the use of some of the tools of industrial engineering at his Work Simplification Conferences in Lake Placid, New York.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Flow process chart_sentence_1",
"Flow process chart_sentence_3",
"Allan H. Mogensen_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5425,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.026",
"value": "?search="
},
{
"operation": "search",
"time": "13.321",
"value": "Flow process chart"
},
{
"operation": "Now on",
"time": "14.003",
"value": "Flow process chart"
},
{
"operation": "Highlighting",
"time": "26.454",
"value": "Flow process chart_sentence_1"
},
{
"operation": "Highlighting",
"time": "37.158",
"value": "Flow process chart_sentence_3"
},
{
"operation": "hyperlink",
"time": "47.397",
"value": "Allan H. Mogensen"
},
{
"operation": "Now on",
"time": "47.919",
"value": "Allan H. Mogensen"
},
{
"operation": "Highlighting",
"time": "50.114",
"value": "Allan H. Mogensen_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "56.142",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "56.196",
"value": "Flow process chart"
},
{
"operation": "finish",
"time": "110.974",
"value": "finish"
}
]
}
|
7448
|
Paul Gerrie was searching for suitable to open a winery, he visited France whose capital is Paris and United States before hiring a consultant to look for an appropriate location.
|
{
"evidence": [
"Cristom Vineyards_sentence_2",
"Cristom Vineyards_sentence_3",
"Cristom Vineyards_sentence_4",
"Cristom Vineyards_sentence_5"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPaul Gerrie was searching for suitable to open a winery, he visited France whose capital is Paris and United States before hiring a consultant to look for an appropriate location.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Cristom Vineyards_sentence_2",
"Cristom Vineyards_sentence_3",
"Cristom Vineyards_sentence_4",
"Cristom Vineyards_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5426,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.501",
"value": "?search="
},
{
"operation": "search",
"time": "15.035",
"value": "Contains...Paul Gerrie"
},
{
"operation": "Now on",
"time": "16.548",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "20.889",
"value": "Cristom Vineyards"
},
{
"operation": "Now on",
"time": "21.61",
"value": "Cristom Vineyards"
},
{
"operation": "Page search",
"time": "29.72",
"value": "Paul Gerrie"
},
{
"operation": "Highlighting",
"time": "32.984",
"value": "Cristom Vineyards_sentence_2"
},
{
"operation": "Highlighting",
"time": "34.767",
"value": "Cristom Vineyards_sentence_3"
},
{
"operation": "Highlighting",
"time": "49.073",
"value": "Cristom Vineyards_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "61.173",
"value": "Cristom Vineyards_sentence_5"
},
{
"operation": "Highlighting",
"time": "61.855",
"value": "Cristom Vineyards_sentence_4"
},
{
"operation": "Highlighting",
"time": "62.367",
"value": "Cristom Vineyards_sentence_5"
},
{
"operation": "finish",
"time": "68.707",
"value": "finish"
}
]
}
|
exfever_389
|
Alec Sokolow is a screenwriter who has worked on computer-animated comics written by the screenwriter behind the movie Cheaper by the Dozen.
|
{
"evidence": [
"Alec Sokolow",
"Toy Story",
"Joel Cohen (writer)"
],
"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:\nAlec Sokolow is a screenwriter who has worked on computer-animated comics written by the screenwriter behind the movie Cheaper by the Dozen.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Alec Sokolow",
"Toy Story",
"Joel Cohen (writer)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5427,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_470
|
Don Simpson was an American film producer who co-produced hit films, including a 1984 American buddy cop action comedy film directed by Martin Brest.
|
{
"evidence": [
"Don Simpson",
"Beverly Hills Cop"
],
"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:\nDon Simpson was an American film producer who co-produced hit films, including a 1984 American buddy cop action comedy film directed by Martin Brest.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Don Simpson",
"Beverly Hills Cop"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5428,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
1365
|
Tosio Kato aworked in the field of mathematics where he became known for "Kato's conjecture", a mathematical problem posed in 1953 which remains unsolved today.
|
{
"evidence": [
"Tosio Kato_sentence_0",
"Kato's conjecture_sentence_0",
"Kato's conjecture_sentence_1",
"Kato's conjecture_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:\nTosio Kato aworked in the field of mathematics where he became known for \"Kato's conjecture\", a mathematical problem posed in 1953 which remains unsolved today.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Tosio Kato_sentence_0",
"Kato's conjecture_sentence_0",
"Kato's conjecture_sentence_1",
"Kato's conjecture_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5429,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "9.643",
"value": "Tosio Kato"
},
{
"operation": "Highlighting",
"time": "16.202",
"value": "Tosio Kato_sentence_0"
},
{
"operation": "hyperlink",
"time": "53.107",
"value": "Asahi Prize"
},
{
"operation": "back-button-clicked",
"time": "166.668",
"value": "back-button-clicked"
},
{
"operation": "search",
"time": "176.632",
"value": "Kato's conjecture"
},
{
"operation": "Highlighting",
"time": "182.092",
"value": "Kato's conjecture_sentence_0"
},
{
"operation": "Highlighting",
"time": "184.258",
"value": "Kato's conjecture_sentence_1"
},
{
"operation": "Highlighting",
"time": "187.602",
"value": "Kato's conjecture_sentence_2"
},
{
"operation": "finish",
"time": "216.567",
"value": "finish"
}
]
}
|
exfever_4780
|
A wide variety of arguments for and against the existence of God or deities can be categorized as a way of losing knowledge by means of direct and indirect observation or conscious events.
|
{
"evidence": [
"Existence of God",
"Empirical research",
"Experience"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nA wide variety of arguments for and against the existence of God or deities can be categorized as a way of losing knowledge by means of direct and indirect observation or conscious events.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Existence of God",
"Empirical research",
"Experience"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5430,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
79219
|
Chethiya Wadugodapitiya was a Sri Lankan flanker (forward position in Rugby) who played for Kingswood college at 16 and joined Sri Lanka national rugby union team at 19.
|
{
"evidence": [
"Chethiya Wadugodapitiya_sentence_0",
"Chethiya Wadugodapitiya_sentence_4",
"Chethiya Wadugodapitiya_sentence_5"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nChethiya Wadugodapitiya was a Sri Lankan flanker (forward position in Rugby) who played for Kingswood college at 16 and joined Sri Lanka national rugby union team at 19.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Chethiya Wadugodapitiya_sentence_0",
"Chethiya Wadugodapitiya_sentence_4",
"Chethiya Wadugodapitiya_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5431,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.999",
"value": "?search="
},
{
"operation": "search",
"time": "10.461",
"value": "Chethiya Wadugodapitiya"
},
{
"operation": "Now on",
"time": "11.425",
"value": "Chethiya Wadugodapitiya"
},
{
"operation": "Highlighting",
"time": "30.254",
"value": "Chethiya Wadugodapitiya_sentence_0"
},
{
"operation": "Highlighting",
"time": "51.196",
"value": "Chethiya Wadugodapitiya_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "58.405",
"value": "Chethiya Wadugodapitiya_sentence_2"
},
{
"operation": "Highlighting",
"time": "61.556",
"value": "Chethiya Wadugodapitiya_sentence_4"
},
{
"operation": "Highlighting",
"time": "75.991",
"value": "Chethiya Wadugodapitiya_sentence_5"
},
{
"operation": "finish",
"time": "159.815",
"value": "finish"
}
]
}
|
exfever_1425
|
Enemy is a Canadian psychological comedy film starring an Italian actress who is the daughter of the Swedish actress Ingrid Bergman and the Italian painter Roberto Rossellini.
|
{
"evidence": [
"Enemy (2013 film)",
"Isabella Rossellini"
],
"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:\nEnemy is a Canadian psychological comedy film starring an Italian actress who is the daughter of the Swedish actress Ingrid Bergman and the Italian painter Roberto Rossellini.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Enemy (2013 film)",
"Isabella Rossellini"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5432,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
53929
|
Ron Kent was a bankrupt politician.
|
{
"evidence": [
"Ron Kent_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:\nRon Kent was a bankrupt politician.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ron Kent_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5433,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.329",
"value": "?search="
},
{
"operation": "search",
"time": "10.46",
"value": "Ron Kent"
},
{
"operation": "Now on",
"time": "86.122",
"value": "Ron Kent"
},
{
"operation": "search",
"time": "337.425",
"value": "Ron Kent"
},
{
"operation": "Now on",
"time": "338.186",
"value": "Ron Kent"
},
{
"operation": "Highlighting",
"time": "372.992",
"value": "Ron Kent_sentence_5"
},
{
"operation": "Highlighting",
"time": "376.706",
"value": "Ron Kent_sentence_7"
},
{
"operation": "Highlighting",
"time": "382.273",
"value": "Ron Kent_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "385.099",
"value": "Ron Kent_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "386.291",
"value": "Ron Kent_sentence_7"
},
{
"operation": "finish",
"time": "401.308",
"value": "finish"
}
]
}
|
39622
|
Bartolomeo della Gatta, an Italian (Florentine) painter and son of a goldsmith, was an atheist.
|
{
"evidence": [
"Bartolomeo della Gatta_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:\nBartolomeo della Gatta, an Italian (Florentine) painter and son of a goldsmith, was an atheist.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bartolomeo della Gatta_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5434,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.081",
"value": "?search="
},
{
"operation": "search",
"time": "12.598",
"value": "Bartolomeo della Gatta"
},
{
"operation": "Now on",
"time": "13.502",
"value": "Bartolomeo della Gatta"
},
{
"operation": "Highlighting",
"time": "15.641",
"value": "Bartolomeo della Gatta_sentence_0"
},
{
"operation": "Highlighting",
"time": "20.254",
"value": "Bartolomeo della Gatta_sentence_1"
},
{
"operation": "Highlighting",
"time": "45.126",
"value": "Bartolomeo della Gatta_sentence_3"
},
{
"operation": "finish",
"time": "49.44",
"value": "finish"
}
]
}
|
exfever_1507
|
Barry Gibb was a British musician who formed a songwriting partnership with fraternal twins that included a British singer who gained worldwide fame as a member of a music group formed in 1958.
|
{
"evidence": [
"Barry Gibb",
"Robin Gibb",
"Bee Gees"
],
"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:\nBarry Gibb was a British musician who formed a songwriting partnership with fraternal twins that included a British singer who gained worldwide fame as a member of a music group formed in 1958.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Barry Gibb",
"Robin Gibb",
"Bee Gees"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5435,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8385
|
Pulp Fiction is a 1994 American black comedy crime film that starred a Canadian actor and singer.
|
{
"evidence": [
"Pulp Fiction",
"John Travolta"
],
"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:\nPulp Fiction is a 1994 American black comedy crime film that starred a Canadian actor and singer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Pulp Fiction",
"John Travolta"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5436,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
31389
|
The California Credit Union, which services the 26,000 member Los Angeles Unified School District, was founded in 1990 and is located in the building at 701 North Brand in Glendale.
|
{
"evidence": [
"California Credit Union_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:\nThe California Credit Union, which services the 26,000 member Los Angeles Unified School District, was founded in 1990 and is located in the building at 701 North Brand in Glendale.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"California Credit Union_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5437,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.746",
"value": "?search="
},
{
"operation": "search",
"time": "12.253",
"value": "California Credit Union"
},
{
"operation": "Now on",
"time": "12.907",
"value": "California Credit Union"
},
{
"operation": "Highlighting",
"time": "37.515",
"value": "California Credit Union_sentence_8"
},
{
"operation": "finish",
"time": "47.542",
"value": "finish"
}
]
}
|
75646
|
The winner in Uva Province, Viyaluwa Polling Division during the 2010 Sri Lankan Parliamentary elections was UPFA (a political alliance in Sri Lanka founded by former Sri Lankan president Chandrika Kumaratunga in 2004).
|
{
"evidence": [
"Viyaluwa Polling Division_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 winner in Uva Province, Viyaluwa Polling Division during the 2010 Sri Lankan Parliamentary elections was UPFA (a political alliance in Sri Lanka founded by former Sri Lankan president Chandrika Kumaratunga in 2004).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Viyaluwa Polling Division_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5438,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "8.379",
"value": "undefined"
},
{
"operation": "search",
"time": "43.313",
"value": "2010 Sri Lankan parliamentary election"
},
{
"operation": "Now on",
"time": "45.957",
"value": "2010 Sri Lankan parliamentary election"
},
{
"operation": "hyperlink",
"time": "56.267",
"value": "5.1 District by district 2010 Sri Lankan parliamentary election_section_5"
},
{
"operation": "search",
"time": "197.092",
"value": "Viyaluwa Polling Divisio"
},
{
"operation": "Now on",
"time": "201.714",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "209.063",
"value": "Viyaluwa Polling Division"
},
{
"operation": "Now on",
"time": "212.396",
"value": "Viyaluwa Polling Division"
},
{
"operation": "Highlighting",
"time": "217.161",
"value": "Viyaluwa Polling Division_sentence_0"
},
{
"operation": "hyperlink",
"time": "226.147",
"value": "2.3 2010 Sri Lankan Parliamentary Election Viyaluwa Polling Division_section_13"
},
{
"operation": "hyperlink",
"time": "256.784",
"value": "2.3 2010 Sri Lankan Parliamentary Election Viyaluwa Polling Division_section_13"
},
{
"operation": "back-button-clicked",
"time": "266.634",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "274.242",
"value": "undefined"
},
{
"operation": "finish",
"time": "280.866",
"value": "finish"
}
]
}
|
39991
|
Margaret Mungherera was one of five siblings, who attended Nakasero Primary School and Gayaza High School for her elementary and secondary education, then Makerere University (Uganda's largest and oldest institution of higher learning, first established as a technical school in 1922) to study human medicine in 1987.
|
{
"evidence": [
"Margaret Mungherera_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:\nMargaret Mungherera was one of five siblings, who attended Nakasero Primary School and Gayaza High School for her elementary and secondary education, then Makerere University (Uganda's largest and oldest institution of higher learning, first established as a technical school in 1922) to study human medicine in 1987.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Margaret Mungherera_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5439,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.124",
"value": "?search="
},
{
"operation": "search",
"time": "15.289",
"value": "Margaret Mungherera"
},
{
"operation": "Now on",
"time": "16.186",
"value": "Margaret Mungherera"
},
{
"operation": "Highlighting",
"time": "26.651",
"value": "Margaret Mungherera_sentence_5"
},
{
"operation": "Highlighting",
"time": "33.402",
"value": "Margaret Mungherera_sentence_6"
},
{
"operation": "Highlighting",
"time": "169.14",
"value": "Margaret Mungherera_sentence_7"
},
{
"operation": "hyperlink",
"time": "171.317",
"value": "Makerere University"
},
{
"operation": "Now on",
"time": "172.073",
"value": "Makerere University"
},
{
"operation": "Highlighting",
"time": "176.507",
"value": "Makerere University_sentence_0"
},
{
"operation": "Highlighting",
"time": "176.509",
"value": "Makerere University_title"
},
{
"operation": "back-button-clicked",
"time": "178.574",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "178.674",
"value": "Margaret Mungherera"
},
{
"operation": "Highlighting",
"time": "192.099",
"value": "Margaret Mungherera_sentence_7"
},
{
"operation": "finish",
"time": "202.784",
"value": "finish"
}
]
}
|
exfever_5837
|
The French language was the first language to be taught in international schools.
|
{
"evidence": [
"French language",
"Vulgar Latin"
],
"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 French language was the first language to be taught in international schools.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"French language",
"Vulgar Latin"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5440,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6412
|
The Belko Experiment is a 2019 American horror thriller film starring an American actor and musician known for playing in American Idiot, a Broadway musical of an American rock band formed in the East Bay of California in 1987 by lead vocalist and guitarist Billie Joe Armstrong.
|
{
"evidence": [
"The Belko Experiment",
"John Gallagher Jr.",
"Green Day"
],
"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 Belko Experiment is a 2019 American horror thriller film starring an American actor and musician known for playing in American Idiot, a Broadway musical of an American rock band formed in the East Bay of California in 1987 by lead vocalist and guitarist Billie Joe Armstrong.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Belko Experiment",
"John Gallagher Jr.",
"Green Day"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5441,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
67742
|
The German Nicaraguan School has grades up to senior high school and was founded in 1957.
|
{
"evidence": [
"German Nicaraguan School_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:\nThe German Nicaraguan School has grades up to senior high school and was founded in 1957.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"German Nicaraguan School_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5442,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.133",
"value": "?search="
},
{
"operation": "search",
"time": "10.938",
"value": "German Nicaraguan School"
},
{
"operation": "Now on",
"time": "12.182",
"value": "German Nicaraguan School"
},
{
"operation": "Highlighting",
"time": "23.136",
"value": "German Nicaraguan School_sentence_2"
},
{
"operation": "finish",
"time": "32.072",
"value": "finish"
}
]
}
|
exfever_482
|
Iron Man is a 2008 American superhero film and the first film in the American media franchise and shared universe centered on a series of superheroes based on characters appearing in Asian comic books published by a publisher of American comic books and related media.
|
{
"evidence": [
"Iron Man (2008 film)",
"Marvel Cinematic Universe",
"Marvel Comics"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIron Man is a 2008 American superhero film and the first film in the American media franchise and shared universe centered on a series of superheroes based on characters appearing in Asian comic books published by a publisher of American comic books and related media.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Iron Man (2008 film)",
"Marvel Cinematic Universe",
"Marvel Comics"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5443,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_261
|
Angelina Jolie is an American actress, filmmaker, and humanitarian who made her screen debut as a child in a 1982 American comedy film directed by Jon Voight.
|
{
"evidence": [
"Angelina Jolie",
"Lookin' to Get Out"
],
"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:\nAngelina Jolie is an American actress, filmmaker, and humanitarian who made her screen debut as a child in a 1982 American comedy film directed by Jon Voight.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Angelina Jolie",
"Lookin' to Get Out"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5444,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8537
|
Breaking Bad is an American neo-Western crime drama television series created and produced by an American writer, producer, and director, also known as the writer and producer of an American science fiction drama television series.
|
{
"evidence": [
"Breaking Bad",
"Vince Gilligan",
"The X-Files"
],
"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:\nBreaking Bad is an American neo-Western crime drama television series created and produced by an American writer, producer, and director, also known as the writer and producer of an American science fiction drama television series.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Breaking Bad",
"Vince Gilligan",
"The X-Files"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5445,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
45507
|
Polonia Triangle functioned as the capital of the American Polonia with the headquarters for almost every major Polish organization in the United States clustered within its vicinity and is bound by Division, Ashland and Milwaukee Avenue.
|
{
"evidence": [
"Polonia Triangle_sentence_4",
"Polonia Triangle_sentence_5"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPolonia Triangle functioned as the capital of the American Polonia with the headquarters for almost every major Polish organization in the United States clustered within its vicinity and is bound by Division, Ashland and Milwaukee Avenue.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Polonia Triangle_sentence_4",
"Polonia Triangle_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5446,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.729",
"value": "?search="
},
{
"operation": "search",
"time": "7.481",
"value": "Polonia Triange"
},
{
"operation": "Now on",
"time": "8.989",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "11.299",
"value": "polonia triangle"
},
{
"operation": "Now on",
"time": "11.793",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "16.032",
"value": "Polonia Triangle"
},
{
"operation": "Now on",
"time": "16.77",
"value": "Polonia Triangle"
},
{
"operation": "Highlighting",
"time": "41.77",
"value": "Polonia Triangle_sentence_4"
},
{
"operation": "Highlighting",
"time": "53.518",
"value": "Polonia Triangle_sentence_5"
},
{
"operation": "finish",
"time": "76.096",
"value": "finish"
}
]
}
|
exfever_7924
|
Troy Baker is known for his video game role as Booker DeWitt in a first-person shooter video game developed by Irrational Games and published by 2K Games.
|
{
"evidence": [
"Troy Baker",
"BioShock Infinite"
],
"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:\nTroy Baker is known for his video game role as Booker DeWitt in a first-person shooter video game developed by Irrational Games and published by 2K Games.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Troy Baker",
"BioShock Infinite"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5447,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
2
|
Luis Salgado appeared in various musicals in 2007.
|
{
"evidence": [
"Luis Salgado_sentence_13"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLuis Salgado appeared in various musicals in 2007.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Luis Salgado_sentence_13"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5448,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "55.666",
"value": "Luis Salgado"
},
{
"operation": "Highlighting",
"time": "71.567",
"value": "Luis Salgado_sentence_8"
},
{
"operation": "hyperlink",
"time": "94.804",
"value": "Fame"
},
{
"operation": "Highlighting",
"time": "96.998",
"value": "Fame (musical)_sentence_0"
},
{
"operation": "Highlighting",
"time": "97.013",
"value": "Fame (musical)_title"
},
{
"operation": "back-button-clicked",
"time": "99.946",
"value": "back-button-clicked"
},
{
"operation": "Highlighting",
"time": "965.343",
"value": "Luis Salgado_sentence_10"
},
{
"operation": "Highlighting",
"time": "989.781",
"value": "Luis Salgado_sentence_13"
},
{
"operation": "Highlighting deleted",
"time": "1002.844",
"value": "Luis Salgado_sentence_13"
},
{
"operation": "Highlighting",
"time": "1018.17",
"value": "Luis Salgado_sentence_13"
},
{
"operation": "Highlighting",
"time": "1045.323",
"value": "Luis Salgado_sentence_13"
},
{
"operation": "hyperlink",
"time": "1068.557",
"value": "Off-Broadway"
},
{
"operation": "back-button-clicked",
"time": "1077.035",
"value": "back-button-clicked"
},
{
"operation": "finish",
"time": "1447.64",
"value": "finish"
}
]
}
|
exfever_1491
|
Ellen Bursty is an American actress who earned a nomination for an Academy Award for her appearance in a 1973 American supernatural horror film directed by James Cameron.
|
{
"evidence": [
"Ellen Burstyn",
"The Exorcist (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:\nEllen Bursty is an American actress who earned a nomination for an Academy Award for her appearance in a 1973 American supernatural horror film directed by James Cameron.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ellen Burstyn",
"The Exorcist (film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5449,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
25930
|
Erotic Review was founded in 1995 by Jamie Maclean and later sold to Rowan Pelling for only £1 plus liabilities.
|
{
"evidence": [
"Erotic Review_sentence_11",
"Erotic Review_sentence_12",
"Erotic Review_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:\nErotic Review was founded in 1995 by Jamie Maclean and later sold to Rowan Pelling for only £1 plus liabilities.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Erotic Review_sentence_11",
"Erotic Review_sentence_12",
"Erotic Review_sentence_14"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5450,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.561",
"value": "?search="
},
{
"operation": "search",
"time": "9.031",
"value": "Erotic Review"
},
{
"operation": "Now on",
"time": "9.644",
"value": "Erotic Review"
},
{
"operation": "Highlighting",
"time": "17.14",
"value": "Erotic Review_sentence_2"
},
{
"operation": "Highlighting",
"time": "32.362",
"value": "Erotic Review_sentence_11"
},
{
"operation": "Highlighting deleted",
"time": "34.162",
"value": "Erotic Review_sentence_2"
},
{
"operation": "Highlighting",
"time": "38.389",
"value": "Erotic Review_sentence_12"
},
{
"operation": "Highlighting",
"time": "54.082",
"value": "Erotic Review_sentence_14"
},
{
"operation": "finish",
"time": "77.515",
"value": "finish"
}
]
}
|
exfever_7623
|
Sammy Cahn was an Indian lyricist who worked at an American record label that was founded by an American lyricist as well as a record label executive.
|
{
"evidence": [
"Sammy Cahn",
"Capitol Records",
"Johnny Mercer"
],
"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:\nSammy Cahn was an Indian lyricist who worked at an American record label that was founded by an American lyricist as well as a record label executive.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sammy Cahn",
"Capitol Records",
"Johnny Mercer"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5451,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_478
|
George R. R. Martin is the author of the series of epic fantasy novels, A Song of Ice and Fire, which was given an award by the Academy of Television Arts and Sciences for its award-winning HBO series, Game of Chess.
|
{
"evidence": [
"George R. R. Martin",
"Primetime Emmy Award"
],
"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:\nGeorge R. R. Martin is the author of the series of epic fantasy novels, A Song of Ice and Fire, which was given an award by the Academy of Television Arts and Sciences for its award-winning HBO series, Game of Chess.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"George R. R. Martin",
"Primetime Emmy Award"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5452,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2004
|
Marlon Brando is the highest-paid actor in America.
|
{
"evidence": [
"Marlon Brando",
"British Academy Film Awards"
],
"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:\nMarlon Brando is the highest-paid actor in America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Marlon Brando",
"British Academy Film Awards"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5453,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2888
|
Deadpool is a 2016 American superhero film starring a Canadian-American actor who also performed in dramatic roles in a biographical drama film directed by Simon Curtis and written by Alexi Kaye Campbell.
|
{
"evidence": [
"Deadpool (film)",
"Ryan Reynolds",
"Woman in Gold (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:\nDeadpool is a 2016 American superhero film starring a Canadian-American actor who also performed in dramatic roles in a biographical drama film directed by Simon Curtis and written by Alexi Kaye Campbell.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Deadpool (film)",
"Ryan Reynolds",
"Woman in Gold (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5454,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
92970
|
The Cook Islands included Sam Mataora and Tokelau replaced New Caledonia who pulled out a week before the tournament.
|
{
"evidence": [
"2006 Pacific Cup_sentence_7",
"2006 Pacific Cup_sentence_6"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Cook Islands included Sam Mataora and Tokelau replaced New Caledonia who pulled out a week before the tournament.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"2006 Pacific Cup_sentence_7",
"2006 Pacific Cup_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5455,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.001",
"value": "?search="
},
{
"operation": "search",
"time": "0.013",
"value": "The Cook Islands"
},
{
"operation": "Now on",
"time": "0.016",
"value": "Cook Islands"
},
{
"operation": "search",
"time": "0.025",
"value": "Sam Mataora"
},
{
"operation": "Now on",
"time": "0.027",
"value": "Sam Mataora"
},
{
"operation": "Highlighting",
"time": "0.04",
"value": "Sam Mataora_sentence_6"
},
{
"operation": "Highlighting",
"time": "0.04",
"value": "Sam Mataora_title"
},
{
"operation": "Highlighting",
"time": "0.04",
"value": "Sam Mataora_section_1"
},
{
"operation": "Highlighting",
"time": "0.04",
"value": "Sam Mataora_section_2"
},
{
"operation": "hyperlink",
"time": "0.052",
"value": "2009 Pacific Cup"
},
{
"operation": "Now on",
"time": "0.053",
"value": "2009 Pacific Cup"
},
{
"operation": "Page search",
"time": "0.056",
"value": "Sam Mataora"
},
{
"operation": "back-button-clicked",
"time": "0.076",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "0.076",
"value": "Sam Mataora"
},
{
"operation": "Highlighting deleted",
"time": "0.079",
"value": "Sam Mataora_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "0.079",
"value": "Sam Mataora_title"
},
{
"operation": "Highlighting deleted",
"time": "0.079",
"value": "Sam Mataora_section_1"
},
{
"operation": "Highlighting deleted",
"time": "0.079",
"value": "Sam Mataora_section_2"
},
{
"operation": "hyperlink",
"time": "0.085",
"value": "2013 Rugby League World Cup"
},
{
"operation": "Now on",
"time": "0.087",
"value": "2013 Rugby League World Cup"
},
{
"operation": "Page search",
"time": "0.095",
"value": "tokelau"
},
{
"operation": "search",
"time": "0.115",
"value": "\"Tokelau replaced New Caledonia\""
},
{
"operation": "Now on",
"time": "0.116",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "0.118",
"value": "2006 Pacific Cup"
},
{
"operation": "Now on",
"time": "0.119",
"value": "2006 Pacific Cup"
},
{
"operation": "Page search",
"time": "0.123",
"value": "Tokelau replaced New Caledonia"
},
{
"operation": "Highlighting",
"time": "0.126",
"value": "2006 Pacific Cup_sentence_7"
},
{
"operation": "Highlighting",
"time": "0.127",
"value": "2006 Pacific Cup_sentence_6"
},
{
"operation": "finish",
"time": "0.138",
"value": "finish"
},
{
"operation": "Now on",
"time": "1622829018.828",
"value": "2006 Pacific Cup"
},
{
"operation": "finish",
"time": "1622829024.696",
"value": "finish"
}
]
}
|
14468
|
Jollification is the third album by The Lightning Seeds and there were 6 songs released from the album.
|
{
"evidence": [
"Jollification_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:\nJollification is the third album by The Lightning Seeds and there were 6 songs released from the album.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Jollification_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5456,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.228",
"value": "?search="
},
{
"operation": "search",
"time": "20.572",
"value": "Jollification"
},
{
"operation": "Now on",
"time": "22.045",
"value": "Jollification"
},
{
"operation": "Highlighting",
"time": "23.528",
"value": "Jollification_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "30.386",
"value": "Jollification_sentence_0"
},
{
"operation": "Highlighting",
"time": "31.541",
"value": "Jollification_sentence_2"
},
{
"operation": "Page search",
"time": "54.359",
"value": "release"
},
{
"operation": "finish",
"time": "92.325",
"value": "finish"
}
]
}
|
exfever_2589
|
William Bradley Pitt is the only actor who received two Golden Globe Awards for acting.
|
{
"evidence": [
"Brad Pitt",
"Plan B Entertainment",
"The Departed"
],
"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:\nWilliam Bradley Pitt is the only actor who received two Golden Globe Awards for acting.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Brad Pitt",
"Plan B Entertainment",
"The Departed"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5457,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2432
|
The Tesla Model S is a battery-powered liftback car produced by an American automotive and clean energy company based in Orlando, Florida.
|
{
"evidence": [
"Tesla Model S",
"Tesla, Inc."
],
"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 Tesla Model S is a battery-powered liftback car produced by an American automotive and clean energy company based in Orlando, Florida.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tesla Model S",
"Tesla, Inc."
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5458,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
40932
|
Arnold Alfred Schmidt, born in 1930 in Plainsfield, New Jersey, was a Commercial Art teacher at New Jersey's Kearny High School who died of cancer in 1993.
|
{
"evidence": [
"Arnold Alfred Schmidt_sentence_0",
"Arnold Alfred Schmidt_sentence_2",
"Arnold Alfred Schmidt_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:\nArnold Alfred Schmidt, born in 1930 in Plainsfield, New Jersey, was a Commercial Art teacher at New Jersey's Kearny High School who died of cancer in 1993.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Arnold Alfred Schmidt_sentence_0",
"Arnold Alfred Schmidt_sentence_2",
"Arnold Alfred Schmidt_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5459,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.095",
"value": "?search="
},
{
"operation": "search",
"time": "9.556",
"value": "Arnold Alfred Schmidt"
},
{
"operation": "Now on",
"time": "13.257",
"value": "Arnold Alfred Schmidt"
},
{
"operation": "Highlighting",
"time": "15.199",
"value": "Arnold Alfred Schmidt_sentence_0"
},
{
"operation": "Highlighting",
"time": "19.757",
"value": "Arnold Alfred Schmidt_sentence_2"
},
{
"operation": "Highlighting",
"time": "24.099",
"value": "Arnold Alfred Schmidt_sentence_3"
},
{
"operation": "finish",
"time": "44.272",
"value": "finish"
}
]
}
|
72388
|
Stockman is the son of Jayne Anne Phillips, he graduated from Concord Academy and was was an active member of the theatre program performing as Nick Bottom.
|
{
"evidence": [
"Theo Stockman_sentence_2",
"Theo Stockman_sentence_3",
"Theo Stockman_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:\nStockman is the son of Jayne Anne Phillips, he graduated from Concord Academy and was was an active member of the theatre program performing as Nick Bottom.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Theo Stockman_sentence_2",
"Theo Stockman_sentence_3",
"Theo Stockman_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5460,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.708",
"value": "?search="
},
{
"operation": "search",
"time": "17.908",
"value": "Jayne Anne Phillips"
},
{
"operation": "Now on",
"time": "18.654",
"value": "Jayne Anne Phillips"
},
{
"operation": "Page search",
"time": "26.161",
"value": "stockman"
},
{
"operation": "hyperlink",
"time": "33.974",
"value": "Theo Stockman"
},
{
"operation": "Now on",
"time": "34.531",
"value": "Theo Stockman"
},
{
"operation": "Highlighting",
"time": "43.784",
"value": "Theo Stockman_sentence_3"
},
{
"operation": "Highlighting",
"time": "45.572",
"value": "Theo Stockman_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "46.453",
"value": "Theo Stockman_sentence_3"
},
{
"operation": "Highlighting",
"time": "46.668",
"value": "Theo Stockman_sentence_3"
},
{
"operation": "Highlighting",
"time": "53.98",
"value": "Theo Stockman_sentence_4"
},
{
"operation": "finish",
"time": "79.067",
"value": "finish"
}
]
}
|
29458
|
Basketball player, Dragan Apic was loaded (sic) out to FMP, a Serbian club and successor of FMP Zeleznik, which merged into Crvena zvezda in 2011) for the rest of the 2017-2018 season before signing in 2019 with Lokomotiv Kuban and shirtly being traded to San Pablo Burgos in 2020.
|
{
"evidence": [
"Dragan Apić_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:\nBasketball player, Dragan Apic was loaded (sic) out to FMP, a Serbian club and successor of FMP Zeleznik, which merged into Crvena zvezda in 2011) for the rest of the 2017-2018 season before signing in 2019 with Lokomotiv Kuban and shirtly being traded to San Pablo Burgos in 2020.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Dragan Apić_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5461,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.799",
"value": "?search="
},
{
"operation": "search",
"time": "17.593",
"value": "Dragan Apić"
},
{
"operation": "Now on",
"time": "18.464",
"value": "Dragan Apić"
},
{
"operation": "Highlighting",
"time": "20.909",
"value": "Dragan Apić_sentence_0"
},
{
"operation": "Highlighting",
"time": "26.856",
"value": "Dragan Apić_sentence_1"
},
{
"operation": "hyperlink",
"time": "36.847",
"value": "FMP"
},
{
"operation": "Now on",
"time": "37.843",
"value": "KK FMP"
},
{
"operation": "Highlighting",
"time": "47.756",
"value": "KK FMP_sentence_4"
},
{
"operation": "Highlighting",
"time": "47.757",
"value": "KK FMP_title"
},
{
"operation": "Highlighting",
"time": "69.257",
"value": "KK FMP_sentence_7"
},
{
"operation": "back-button-clicked",
"time": "79.586",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "79.775",
"value": "Dragan Apić"
},
{
"operation": "Highlighting",
"time": "94.968",
"value": "Dragan Apić_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "101.956",
"value": "Dragan Apić_sentence_1"
},
{
"operation": "Highlighting",
"time": "117.923",
"value": "Dragan Apić_sentence_5"
},
{
"operation": "Highlighting",
"time": "141.324",
"value": "Dragan Apić_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "149.896",
"value": "Dragan Apić_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "150.737",
"value": "Dragan Apić_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "151.641",
"value": "Dragan Apić_sentence_0"
},
{
"operation": "finish",
"time": "171.088",
"value": "finish"
}
]
}
|
76353
|
The New Zealand Arthropod Collection, which does not include terrestrial invertebrates, has over 7 million specimens.
|
{
"evidence": [
"New Zealand Arthropod Collection_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 New Zealand Arthropod Collection, which does not include terrestrial invertebrates, has over 7 million specimens.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"New Zealand Arthropod Collection_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5462,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.852",
"value": "?search="
},
{
"operation": "search",
"time": "4.911",
"value": "New Zealand Arthropod Collection"
},
{
"operation": "Now on",
"time": "5.812",
"value": "New Zealand Arthropod Collection"
},
{
"operation": "Highlighting",
"time": "23.196",
"value": "New Zealand Arthropod Collection_sentence_0"
},
{
"operation": "finish",
"time": "32.955",
"value": "finish"
}
]
}
|
exfever_4190
|
Pineapple Express is a 2008 American stoner comedy film directed by an American filmmaker who directed the drama, Goodfellas.
|
{
"evidence": [
"Pineapple Express (film)",
"David Gordon Green"
],
"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:\nPineapple Express is a 2008 American stoner comedy film directed by an American filmmaker who directed the drama, Goodfellas.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Pineapple Express (film)",
"David Gordon Green"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5463,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
78295
|
Symphyllocarpus used to be classified into Kingdom Plantae and family Asteraceae but now it is classified into Kingdom Eubacteria.
|
{
"evidence": [
"Symphyllocarpus_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:\nSymphyllocarpus used to be classified into Kingdom Plantae and family Asteraceae but now it is classified into Kingdom Eubacteria.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Symphyllocarpus_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5464,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.754",
"value": "?search="
},
{
"operation": "search",
"time": "14.246",
"value": "Symphyllocarpus"
},
{
"operation": "Now on",
"time": "14.81",
"value": "Symphyllocarpus"
},
{
"operation": "Highlighting",
"time": "44.926",
"value": "Symphyllocarpus_cell_0_2_1"
},
{
"operation": "Highlighting",
"time": "44.928",
"value": "Symphyllocarpus_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "44.928",
"value": "Symphyllocarpus_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "52.605",
"value": "Symphyllocarpus_cell_0_2_1"
},
{
"operation": "Highlighting deleted",
"time": "52.607",
"value": "Symphyllocarpus_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "52.608",
"value": "Symphyllocarpus_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "65.013",
"value": "Symphyllocarpus_cell_0_2_0"
},
{
"operation": "Highlighting",
"time": "65.014",
"value": "Symphyllocarpus_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "65.014",
"value": "Symphyllocarpus_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "65.645",
"value": "Symphyllocarpus_cell_0_2_1"
},
{
"operation": "Highlighting",
"time": "78.748",
"value": "Symphyllocarpus_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "80.98",
"value": "Symphyllocarpus_cell_0_2_1"
},
{
"operation": "Highlighting deleted",
"time": "82.508",
"value": "Symphyllocarpus_cell_0_2_0"
},
{
"operation": "Highlighting deleted",
"time": "82.51",
"value": "Symphyllocarpus_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "82.511",
"value": "Symphyllocarpus_header_cell_0_0_0"
},
{
"operation": "hyperlink",
"time": "84.204",
"value": "Symphyllocarpus exilis"
},
{
"operation": "Now on",
"time": "84.843",
"value": "Symphyllocarpus exilis"
},
{
"operation": "search",
"time": "103.078",
"value": "Eubacteria"
},
{
"operation": "Now on",
"time": "104.35",
"value": "Bacteria"
},
{
"operation": "finish",
"time": "159.077",
"value": "finish"
}
]
}
|
exfever_1265
|
Treasure is the first music video directory of Cameron Duddy.
|
{
"evidence": [
"Treasure (Bruno Mars song)",
"Bruno Mars"
],
"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:\nTreasure is the first music video directory of Cameron Duddy.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Treasure (Bruno Mars song)",
"Bruno Mars"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5465,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
82455
|
On June 23, 2010, was the first season premiere of the twelve weekly episodes of Black Panther in iTunes .
|
{
"evidence": [
"Black panther (disambiguation)_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:\nOn June 23, 2010, was the first season premiere of the twelve weekly episodes of Black Panther in iTunes .\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Black panther (disambiguation)_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5466,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.072",
"value": "?search="
},
{
"operation": "search",
"time": "14.593",
"value": "Contains...weekly episodes of Black Panther in iTunes"
},
{
"operation": "Now on",
"time": "16.687",
"value": "undefined"
},
{
"operation": "search",
"time": "33.641",
"value": "Contains...Black Panther in iTunes"
},
{
"operation": "Now on",
"time": "34.919",
"value": "undefined"
},
{
"operation": "search",
"time": "65.101",
"value": "Black Panther (soundtrack)"
},
{
"operation": "Now on",
"time": "67.056",
"value": "Black Panther (soundtrack)"
},
{
"operation": "Page search",
"time": "127.709",
"value": "episodes"
},
{
"operation": "search",
"time": "148.101",
"value": "Contains...Panther in iTunes"
},
{
"operation": "Now on",
"time": "149.034",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "167.005",
"value": "Steel Panther"
},
{
"operation": "Now on",
"time": "167.864",
"value": "Steel Panther"
},
{
"operation": "Page search",
"time": "183.432",
"value": "episodes"
},
{
"operation": "search",
"time": "228.857",
"value": "Contains...weekly episodes in iTunes"
},
{
"operation": "Now on",
"time": "230.471",
"value": "undefined"
},
{
"operation": "search",
"time": "263.663",
"value": "Contains...season premiere of the twelve weekly episodes of"
},
{
"operation": "Now on",
"time": "265.389",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "284.328",
"value": "Lost (TV series)"
},
{
"operation": "Now on",
"time": "286.047",
"value": "Lost (TV series)"
},
{
"operation": "search",
"time": "333.494",
"value": "Contains...weekly episodes of Black Panther"
},
{
"operation": "Now on",
"time": "334.762",
"value": "undefined"
},
{
"operation": "search",
"time": "362.775",
"value": "Black Panther (film)"
},
{
"operation": "Now on",
"time": "364.998",
"value": "Black Panther (film)"
},
{
"operation": "search",
"time": "402.303",
"value": "Contains... weekly episodes of Black Panther in iTunes"
},
{
"operation": "Now on",
"time": "403.164",
"value": "undefined"
},
{
"operation": "search",
"time": "453.513",
"value": "Contains...June 23, 2010 Panther"
},
{
"operation": "Now on",
"time": "455.66",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "476.11",
"value": "Marvel Comics"
},
{
"operation": "Now on",
"time": "477.527",
"value": "Marvel Comics"
},
{
"operation": "search",
"time": "496.914",
"value": "Black panther (disambiguation)"
},
{
"operation": "Now on",
"time": "497.837",
"value": "Black panther (disambiguation)"
},
{
"operation": "hyperlink",
"time": "507.571",
"value": "2.3 Other uses in arts and entertainment Black panther (disambiguation)_section_4"
},
{
"operation": "hyperlink",
"time": "545.992",
"value": "Black Panther (band)"
},
{
"operation": "Now on",
"time": "547.225",
"value": "Black Panther (band)"
},
{
"operation": "Page search",
"time": "555.142",
"value": "iTunes"
},
{
"operation": "Now on",
"time": "592.18",
"value": "Black Panther (band)"
},
{
"operation": "search",
"time": "608.944",
"value": "Black Panther (TV series)"
},
{
"operation": "Now on",
"time": "610.256",
"value": "Black Panther (TV series)"
},
{
"operation": "back-button-clicked",
"time": "636.101",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "636.426",
"value": "Black Panther (band)"
},
{
"operation": "search",
"time": "643.786",
"value": "Black panther (disambiguation)"
},
{
"operation": "Now on",
"time": "644.768",
"value": "Black panther (disambiguation)"
},
{
"operation": "Highlighting",
"time": "650.391",
"value": "Black panther (disambiguation)_sentence_1"
},
{
"operation": "finish",
"time": "656.627",
"value": "finish"
}
]
}
|
90850
|
India and China have a strained relationship especially with political and border issues which has made treaty adjustments difficult to finalize.
|
{
"evidence": [
"India-China border infrastructure_sentence_1",
"China–India relations_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:\nIndia and China have a strained relationship especially with political and border issues which has made treaty adjustments difficult to finalize.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"India-China border infrastructure_sentence_1",
"China–India relations_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5467,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.8",
"value": "?search="
},
{
"operation": "search",
"time": "35.45",
"value": "foreign relations with china"
},
{
"operation": "Now on",
"time": "37.618",
"value": "undefined"
},
{
"operation": "search",
"time": "50.139",
"value": "China-India Relations"
},
{
"operation": "Now on",
"time": "51.437",
"value": "China–India relations"
},
{
"operation": "Page search",
"time": "55.979",
"value": "strain"
},
{
"operation": "search",
"time": "87.144",
"value": "Foreign relations of China"
},
{
"operation": "Now on",
"time": "88.666",
"value": "Foreign relations of China"
},
{
"operation": "Page search",
"time": "91.597",
"value": "strained"
},
{
"operation": "Page search",
"time": "129.288",
"value": "india"
},
{
"operation": "hyperlink",
"time": "146.92",
"value": "7 Relations by region and country Foreign relations of China_section_9"
},
{
"operation": "search",
"time": "214.677",
"value": "India-China border"
},
{
"operation": "Now on",
"time": "215.8",
"value": "India-China border"
},
{
"operation": "Page search",
"time": "220.45",
"value": "strain"
},
{
"operation": "search",
"time": "231.882",
"value": "India-China relations"
},
{
"operation": "Now on",
"time": "233.253",
"value": "China–India relations"
},
{
"operation": "search",
"time": "241.116",
"value": "India-China border infrastructure"
},
{
"operation": "Now on",
"time": "242.125",
"value": "India-China border infrastructure"
},
{
"operation": "Highlighting",
"time": "258.612",
"value": "India-China border infrastructure_sentence_1"
},
{
"operation": "search",
"time": "272.789",
"value": "India-China border"
},
{
"operation": "Now on",
"time": "273.237",
"value": "India-China border"
},
{
"operation": "Page search",
"time": "285.993",
"value": "treaty"
},
{
"operation": "search",
"time": "297.684",
"value": "India-China relations"
},
{
"operation": "Now on",
"time": "298.34",
"value": "India-China relations"
},
{
"operation": "Page search",
"time": "300.531",
"value": "treaty"
},
{
"operation": "Highlighting",
"time": "360.73",
"value": "China–India relations_sentence_1"
},
{
"operation": "Now on",
"time": "376.98",
"value": "India-China relations"
},
{
"operation": "finish",
"time": "383.998",
"value": "finish"
}
]
}
|
8567
|
Anthony Cherrington was acquitted of assault charges against his ex even though the judge claimed he thought he was special and untouchable.
|
{
"evidence": [
"Anthony Cherrington_sentence_6",
"Anthony Cherrington_sentence_7",
"Anthony Cherrington_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:\nAnthony Cherrington was acquitted of assault charges against his ex even though the judge claimed he thought he was special and untouchable.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Anthony Cherrington_sentence_6",
"Anthony Cherrington_sentence_7",
"Anthony Cherrington_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5468,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.714",
"value": "?search="
},
{
"operation": "search",
"time": "60.91",
"value": "Anthony Cherrington"
},
{
"operation": "Now on",
"time": "61.956",
"value": "Anthony Cherrington"
},
{
"operation": "Page search",
"time": "539.202",
"value": "assault"
},
{
"operation": "Page search",
"time": "552.751",
"value": "acquitted"
},
{
"operation": "Highlighting",
"time": "562.539",
"value": "Anthony Cherrington_sentence_6"
},
{
"operation": "Highlighting",
"time": "564.956",
"value": "Anthony Cherrington_sentence_7"
},
{
"operation": "Highlighting",
"time": "567.001",
"value": "Anthony Cherrington_sentence_8"
},
{
"operation": "finish",
"time": "583.914",
"value": "finish"
}
]
}
|
exfever_5084
|
The Adjustment Bureau is a 2011 American science fiction written and directed by George Nolfi that has narrative that involves elements of the romance and thriller genres.
|
{
"evidence": [
"The Adjustment Bureau",
"Romantic thriller"
],
"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 Adjustment Bureau is a 2011 American science fiction written and directed by George Nolfi that has narrative that involves elements of the romance and thriller genres.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Adjustment Bureau",
"Romantic thriller"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5469,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
16938
|
The Veissella, which is a genus of African jumping monkeys that were first described by F. R. Wanless in 1984 and the genus name is an arbitrary combination of letters; the gender is considered feminine and as of August 2019 it contains two species, found in Africa: V.
|
{
"evidence": [
"Veissella_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 Veissella, which is a genus of African jumping monkeys that were first described by F. R. Wanless in 1984 and the genus name is an arbitrary combination of letters; the gender is considered feminine and as of August 2019 it contains two species, found in Africa: V.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Veissella_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5470,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.894",
"value": "?search="
},
{
"operation": "search",
"time": "14.497",
"value": "Veissella"
},
{
"operation": "Now on",
"time": "15.289",
"value": "Veissella"
},
{
"operation": "Highlighting",
"time": "67.666",
"value": "Veissella_sentence_0"
},
{
"operation": "finish",
"time": "78.729",
"value": "finish"
}
]
}
|
exfever_8286
|
The stratosphere is the fourth layer of the atmosphere of the Earth whose temperatures also vary within the stratosphere as the seasons change, reaching particularly low temperatures in the phenomenon where the nighttime lasts for more than 24 hours and occurs in the northernmost and southernmost regions of the planet that surround its geographical poles.
|
{
"evidence": [
"Stratosphere",
"Polar night",
"Polar regions of Earth"
],
"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 stratosphere is the fourth layer of the atmosphere of the Earth whose temperatures also vary within the stratosphere as the seasons change, reaching particularly low temperatures in the phenomenon where the nighttime lasts for more than 24 hours and occurs in the northernmost and southernmost regions of the planet that surround its geographical poles.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Stratosphere",
"Polar night",
"Polar regions of Earth"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5471,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2836
|
The Beverly Hillbillies has been adapted into a comic book series.
|
{
"evidence": [
"The Beverly Hillbillies",
"Filmways"
],
"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 Beverly Hillbillies has been adapted into a comic book series.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Beverly Hillbillies",
"Filmways"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5472,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
68094
|
In the 2017-2018 season, Albania Erblira Bici was number one as an opposite.
|
{
"evidence": [
"Erblira Bici_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn the 2017-2018 season, Albania Erblira Bici was number one as an opposite.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Erblira Bici_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5473,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "4.452",
"value": "?search="
},
{
"operation": "search",
"time": "59.266",
"value": "Albania Erblira Bici"
},
{
"operation": "Now on",
"time": "60.002",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "62.55",
"value": "Erblira Bici"
},
{
"operation": "Now on",
"time": "63.076",
"value": "Erblira Bici"
},
{
"operation": "Highlighting",
"time": "164.408",
"value": "Erblira Bici_sentence_0"
},
{
"operation": "finish",
"time": "288.053",
"value": "finish"
}
]
}
|
exfever_8739
|
The orca or killer whale is a toothed whale belonging to the oceanic dolphin family, of which it is the largest member identifiable by its black-and-white patterned body found in the smallest and shallowest of the world's major oceans.
|
{
"evidence": [
"Killer whale",
"Arctic Ocean"
],
"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 orca or killer whale is a toothed whale belonging to the oceanic dolphin family, of which it is the largest member identifiable by its black-and-white patterned body found in the smallest and shallowest of the world's major oceans.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Killer whale",
"Arctic Ocean"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5474,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2678
|
Woody Harrelson is a British actor known for his role in the NBC sitcom created by American screenwriters best known for the movie Taxi.
|
{
"evidence": [
"Woody Harrelson",
"Cheers",
"Glen and Les Charles"
],
"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:\nWoody Harrelson is a British actor known for his role in the NBC sitcom created by American screenwriters best known for the movie Taxi.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Woody Harrelson",
"Cheers",
"Glen and Les Charles"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5475,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
86338
|
Ernie Caddel was an American football running back who played college football for Glenn "Pop" Warner at Stanford University.
|
{
"evidence": [
"Ernie Caddel_sentence_0",
"Ernie Caddel_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:\nErnie Caddel was an American football running back who played college football for Glenn \"Pop\" Warner at Stanford University.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ernie Caddel_sentence_0",
"Ernie Caddel_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5476,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.903",
"value": "?search="
},
{
"operation": "search",
"time": "6.404",
"value": "Ernie Caddel"
},
{
"operation": "Now on",
"time": "7.222",
"value": "Ernie Caddel"
},
{
"operation": "Highlighting",
"time": "16.127",
"value": "Ernie Caddel_sentence_0"
},
{
"operation": "Highlighting",
"time": "22.865",
"value": "Ernie Caddel_sentence_1"
},
{
"operation": "finish",
"time": "48.482",
"value": "finish"
}
]
}
|
78744
|
Gerhard Häge was a West German rower who was discovered by rowing trainer Ludwig Marquardt, and for the 1954 rowing season, Marquardt looked for another rower to form a double scull boat with Häge and found Schneider.
|
{
"evidence": [
"Gerhard Häge_sentence_0",
"Gerhard Häge_sentence_3",
"Gerhard Häge_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:\nGerhard Häge was a West German rower who was discovered by rowing trainer Ludwig Marquardt, and for the 1954 rowing season, Marquardt looked for another rower to form a double scull boat with Häge and found Schneider.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Gerhard Häge_sentence_0",
"Gerhard Häge_sentence_3",
"Gerhard Häge_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5477,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.147",
"value": "?search="
},
{
"operation": "search",
"time": "20.927",
"value": "Gerhard Häge"
},
{
"operation": "Now on",
"time": "21.708",
"value": "Gerhard Häge"
},
{
"operation": "Highlighting",
"time": "43.136",
"value": "Gerhard Häge_sentence_0"
},
{
"operation": "Highlighting",
"time": "58.586",
"value": "Gerhard Häge_sentence_3"
},
{
"operation": "Highlighting",
"time": "67.787",
"value": "Gerhard Häge_sentence_4"
},
{
"operation": "finish",
"time": "134.183",
"value": "finish"
}
]
}
|
63162
|
North Korean athlete Om Yun-chol currently owns three records in the 56kg men's weightlifting events, which have been contested three times as optional sport, at the Summer Universiade
|
{
"evidence": [
"Om Yun-chol_sentence_1",
"Om Yun-chol_sentence_4",
"Om Yun-chol_sentence_6"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nNorth Korean athlete Om Yun-chol currently owns three records in the 56kg men's weightlifting events, which have been contested three times as optional sport, at the Summer Universiade\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Om Yun-chol_sentence_1",
"Om Yun-chol_sentence_4",
"Om Yun-chol_sentence_6"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5478,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.986",
"value": "?search="
},
{
"operation": "search",
"time": "14.167",
"value": "Om Yun-chol"
},
{
"operation": "Now on",
"time": "15.201",
"value": "Om Yun-chol"
},
{
"operation": "Highlighting",
"time": "20.009",
"value": "Om Yun-chol_sentence_1"
},
{
"operation": "Highlighting",
"time": "100.734",
"value": "Om Yun-chol_sentence_4"
},
{
"operation": "hyperlink",
"time": "132.757",
"value": "men's 56 kg event"
},
{
"operation": "Now on",
"time": "133.747",
"value": "Weightlifting at the 2012 Summer Olympics – Men's 56 kg"
},
{
"operation": "back-button-clicked",
"time": "146.67",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "146.789",
"value": "Om Yun-chol"
},
{
"operation": "search",
"time": "194.638",
"value": "Summer Universiade"
},
{
"operation": "Now on",
"time": "197.312",
"value": "Universiade"
},
{
"operation": "back-button-clicked",
"time": "210.148",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "210.264",
"value": "Om Yun-chol"
},
{
"operation": "Highlighting",
"time": "230.489",
"value": "Om Yun-chol_sentence_6"
},
{
"operation": "finish",
"time": "236.271",
"value": "finish"
}
]
}
|
exfever_2552
|
Aishwarya Rai is an actress who has received numerous accolades, including two awards that honor artistic and technical excellence in the Hindi-language film industry of a country in South Asia, which is the most populous democracy in the world.
|
{
"evidence": [
"Aishwarya Rai",
"Filmfare Awards",
"India"
],
"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:\nAishwarya Rai is an actress who has received numerous accolades, including two awards that honor artistic and technical excellence in the Hindi-language film industry of a country in South Asia, which is the most populous democracy in the world.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Aishwarya Rai",
"Filmfare Awards",
"India"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5479,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3342
|
Winona Ryder gained attention with her performance in Tim Burton's 1988 fantasy comedy film starring an Australian actor.
|
{
"evidence": [
"Winona Ryder",
"Beetlejuice",
"Michael Keaton"
],
"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:\nWinona Ryder gained attention with her performance in Tim Burton's 1988 fantasy comedy film starring an Australian actor.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Winona Ryder",
"Beetlejuice",
"Michael Keaton"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5480,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5037
|
"California Gurls" is a 2010 song recorded by Swedish singer Katy Perry with additional production from an American record producer who is the recipient of the 2013 Hal David Starlight Award from the Songwriters Hall of Fame.
|
{
"evidence": [
"California Gurls",
"Benny Blanco"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n\"California Gurls\" is a 2010 song recorded by Swedish singer Katy Perry with additional production from an American record producer who is the recipient of the 2013 Hal David Starlight Award from the Songwriters Hall of Fame.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"California Gurls",
"Benny Blanco"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5481,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1925
|
The Woman in Red is a 1984 Spanish film that won one of the awards given annually to people working in the motion picture industry by AMPAS.
|
{
"evidence": [
"The Woman in Red (1984 film)",
"Academy Award for Best Original Song"
],
"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 Woman in Red is a 1984 Spanish film that won one of the awards given annually to people working in the motion picture industry by AMPAS.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Woman in Red (1984 film)",
"Academy Award for Best Original Song"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5482,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
69747
|
Oemar Said Tjokroaminoto was the first to submit to the Netherlands and he shared his political ideas on many ideologies of Indonesia at that time to his students Semaoen, Alimin, Musso, Sukarno, Kartosuwiryo, and Tan Malaka whom he taught in his own household.
|
{
"evidence": [
"Oemar Said Tjokroaminoto_sentence_14"
],
"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:\nOemar Said Tjokroaminoto was the first to submit to the Netherlands and he shared his political ideas on many ideologies of Indonesia at that time to his students Semaoen, Alimin, Musso, Sukarno, Kartosuwiryo, and Tan Malaka whom he taught in his own household.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Oemar Said Tjokroaminoto_sentence_14"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5483,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.279",
"value": "?search="
},
{
"operation": "Now on",
"time": "7.376",
"value": "?search="
},
{
"operation": "search",
"time": "252.937",
"value": "Oemar Said Tjokroaminoto"
},
{
"operation": "Now on",
"time": "253.849",
"value": "Oemar Said Tjokroaminoto"
},
{
"operation": "Highlighting",
"time": "448.725",
"value": "Oemar Said Tjokroaminoto_sentence_14"
},
{
"operation": "finish",
"time": "463.147",
"value": "finish"
}
]
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.