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_3567
|
The Times is a daily national newspaper published by Times Newspapers since 1981, as well as its sister paper, a British newspaper whose circulation makes it the largest in Britain's quality press market category.
|
{
"evidence": [
"The Times",
"The Sunday Times"
],
"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 Times is a daily national newspaper published by Times Newspapers since 1981, as well as its sister paper, a British newspaper whose circulation makes it the largest in Britain's quality press market category.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Times",
"The Sunday Times"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 88,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_6465
|
The Washington Naval Conference was a disarmament conference attended by nine nations including a historical nation-state and great power that existed from the Meiji Restoration in 1868 until the enactment of the post-World War II 1947 constitution.
|
{
"evidence": [
"Washington Naval Conference",
"Empire of Japan"
],
"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 Washington Naval Conference was a disarmament conference attended by nine nations including a historical nation-state and great power that existed from the Meiji Restoration in 1868 until the enactment of the post-World War II 1947 constitution.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Washington Naval Conference",
"Empire of Japan"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 89,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
74742
|
Catcott, who had been corresponding since 1733 with John Hutchinson, preached a sermon in St Mark's Church on May 16, 1735, which provoked "the first serious debate about Hutchinsonianism and the scientific evidence contained in the Old Testament".
|
{
"evidence": [
"Alexander Stopford Catcott_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:\nCatcott, who had been corresponding since 1733 with John Hutchinson, preached a sermon in St Mark's Church on May 16, 1735, which provoked \"the first serious debate about Hutchinsonianism and the scientific evidence contained in the Old Testament\".\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Alexander Stopford Catcott_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 90,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.976",
"value": "undefined"
},
{
"operation": "search",
"time": "13.815",
"value": "Contains...Catcott, who had been corresponding since 1733 with John Hutchinson"
},
{
"operation": "Now on",
"time": "15.373",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "18.933",
"value": "Alexander Stopford Catcott"
},
{
"operation": "Now on",
"time": "19.909",
"value": "Alexander Stopford Catcott"
},
{
"operation": "Highlighting",
"time": "24.742",
"value": "Alexander Stopford Catcott_sentence_6"
},
{
"operation": "Highlighting",
"time": "36.22",
"value": "Alexander Stopford Catcott_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "38.221",
"value": "Alexander Stopford Catcott_sentence_6"
},
{
"operation": "finish",
"time": "58.692",
"value": "finish"
}
]
}
|
exfever_4888
|
Victorious is Ariana Grande's first Nickelodeon television series.
|
{
"evidence": [
"Ariana Grande",
"Americans"
],
"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:\nVictorious is Ariana Grande's first Nickelodeon television series.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ariana Grande",
"Americans"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 91,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_3375
|
Her is a 2013 American science-fiction romantic drama film written, directed, and co-produced by an American actor, director, musician, and photographer who co-founded an American lifestyle magazine targeting young women.
|
{
"evidence": [
"Her (film)",
"Spike Jonze",
"Dirt (magazine)"
],
"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:\nHer is a 2013 American science-fiction romantic drama film written, directed, and co-produced by an American actor, director, musician, and photographer who co-founded an American lifestyle magazine targeting young women.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Her (film)",
"Spike Jonze",
"Dirt (magazine)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 92,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_2931
|
Machu Pichu is located in the Eastern Cordillera of southern Peru, built by an ancient empire that incorporated a large portion of South America, centered in the longest continental mountain range in Asia.
|
{
"evidence": [
"Machu Picchu",
"Inca Empire",
"Andes"
],
"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:\nMachu Pichu is located in the Eastern Cordillera of southern Peru, built by an ancient empire that incorporated a large portion of South America, centered in the longest continental mountain range in Asia.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Machu Picchu",
"Inca Empire",
"Andes"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 93,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_2899
|
Austria is a sovereign country comprising the island of an island state in Australia.
|
{
"evidence": [
"Australia",
"Tasmania"
],
"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:\nAustria is a sovereign country comprising the island of an island state in Australia.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Australia",
"Tasmania"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 94,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_1619
|
Hunter Stockton Thompson was an English journalist who founded a movement involving a style of reporting that is written without claims of objectivity.
|
{
"evidence": [
"Hunter S. Thompson",
"Gonzo journalism"
],
"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:\nHunter Stockton Thompson was an English journalist who founded a movement involving a style of reporting that is written without claims of objectivity.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hunter S. Thompson",
"Gonzo journalism"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 95,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_6593
|
Splash is a 1984 American fantasy romantic comedy film that involves a mysterious woman who is secretly an aquatic creature with the body of a female human and the tail of a fish.
|
{
"evidence": [
"Splash (film)",
"Mermaid"
],
"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:\nSplash is a 1984 American fantasy romantic comedy film that involves a mysterious woman who is secretly an aquatic creature with the body of a female human and the tail of a fish.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Splash (film)",
"Mermaid"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 96,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_6329
|
Jared Joseph Leto is the youngest member of Thirty Seconds to Mars.
|
{
"evidence": [
"Jared Leto",
"American Psycho (film)",
"Serial killer"
],
"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:\nJared Joseph Leto is the youngest member of Thirty Seconds to Mars.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jared Leto",
"American Psycho (film)",
"Serial killer"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 97,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_2482
|
Maggie Smith was also awarded the Shakespeare Prize by the Hamburg Alfred Toepfer Foundation.
|
{
"evidence": [
"Maggie Smith",
"Tony 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:\nMaggie Smith was also awarded the Shakespeare Prize by the Hamburg Alfred Toepfer Foundation.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Maggie Smith",
"Tony Award"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 98,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
25516
|
Neil Breslin was first elected to the State Senate in 1996, defeating one-term Republican incumbent Michael Hoblock, who graduated B.B.A from La Salle Institute in Troy, by just over eight points; however, he has never faced another contest even that close in what has historically been one of the few safe Democratic districts in Upstate.
|
{
"evidence": [
"Neil Breslin_sentence_9",
"Michael J. Hoblock Jr._sentence_2",
"Michael J. Hoblock Jr._sentence_3",
"Neil Breslin_sentence_10"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nNeil Breslin was first elected to the State Senate in 1996, defeating one-term Republican incumbent Michael Hoblock, who graduated B.B.A from La Salle Institute in Troy, by just over eight points; however, he has never faced another contest even that close in what has historically been one of the few safe Democratic districts in Upstate.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Neil Breslin_sentence_9",
"Michael J. Hoblock Jr._sentence_2",
"Michael J. Hoblock Jr._sentence_3",
"Neil Breslin_sentence_10"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 99,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.725",
"value": "?search="
},
{
"operation": "search",
"time": "25.721",
"value": "Neil Breslin"
},
{
"operation": "Now on",
"time": "42.011",
"value": "Neil Breslin"
},
{
"operation": "Highlighting",
"time": "99.733",
"value": "Neil Breslin_sentence_9"
},
{
"operation": "Page search",
"time": "120.626",
"value": "la salle"
},
{
"operation": "Page search",
"time": "137.584",
"value": "troy"
},
{
"operation": "Highlighting",
"time": "154.828",
"value": "Neil Breslin_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "166.834",
"value": "Neil Breslin_sentence_1"
},
{
"operation": "hyperlink",
"time": "171.62",
"value": "Michael J. Hoblock Jr."
},
{
"operation": "Now on",
"time": "172.137",
"value": "Michael J. Hoblock Jr."
},
{
"operation": "Highlighting",
"time": "190.87",
"value": "Michael J. Hoblock Jr._sentence_2"
},
{
"operation": "Highlighting",
"time": "191.984",
"value": "Michael J. Hoblock Jr._sentence_3"
},
{
"operation": "back-button-clicked",
"time": "329.724",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "329.839",
"value": "Neil Breslin"
},
{
"operation": "Highlighting",
"time": "368.994",
"value": "Neil Breslin_sentence_10"
},
{
"operation": "finish",
"time": "382.457",
"value": "finish"
}
]
}
|
89023
|
Cinderella was written by Nari Kusakawa, the pen name of a Japanese manga artist who debuted in 2000 with the short story My Grandfather, and published by Hakusensha.
|
{
"evidence": [
"Nari Kusakawa_sentence_0",
"Nari Kusakawa_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:\nCinderella was written by Nari Kusakawa, the pen name of a Japanese manga artist who debuted in 2000 with the short story My Grandfather, and published by Hakusensha.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Nari Kusakawa_sentence_0",
"Nari Kusakawa_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 100,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.087",
"value": "?search="
},
{
"operation": "search",
"time": "17.774",
"value": "Nari Kusakawa"
},
{
"operation": "Now on",
"time": "18.67",
"value": "Nari Kusakawa"
},
{
"operation": "Highlighting",
"time": "30.105",
"value": "Nari Kusakawa_sentence_0"
},
{
"operation": "Highlighting",
"time": "43.613",
"value": "Nari Kusakawa_sentence_2"
},
{
"operation": "search",
"time": "101.953",
"value": "Cinderella (2012 film)"
},
{
"operation": "Now on",
"time": "103.395",
"value": "Cinderella (2012 film)"
},
{
"operation": "finish",
"time": "120.536",
"value": "finish"
}
]
}
|
exfever_5223
|
Isaac Liev Schreiber is a singer who became known after appearing in several mainstream Hollywood films, including a 1996 American action thriller film starring an American actress and model born February 17, 1954.
|
{
"evidence": [
"Liev Schreiber",
"Ransom (1996 film)",
"Rene Russo"
],
"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:\nIsaac Liev Schreiber is a singer who became known after appearing in several mainstream Hollywood films, including a 1996 American action thriller film starring an American actress and model born February 17, 1954.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Liev Schreiber",
"Ransom (1996 film)",
"Rene Russo"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 101,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
49450
|
There is a white vest with the Saetbyoul logo and a gray jacket for both genders and there is also a cardigan, that can be worn at any time.
|
{
"evidence": [
"Saetbyoul Middle School_sentence_3",
"Saetbyoul Middle School_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:\nThere is a white vest with the Saetbyoul logo and a gray jacket for both genders and there is also a cardigan, that can be worn at any time.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Saetbyoul Middle School_sentence_3",
"Saetbyoul Middle School_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 102,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.177",
"value": "?search="
},
{
"operation": "search",
"time": "28.307",
"value": "Saetbyoul Middle School"
},
{
"operation": "Now on",
"time": "29.151",
"value": "Saetbyoul Middle School"
},
{
"operation": "Highlighting",
"time": "36.199",
"value": "Saetbyoul Middle School_sentence_3"
},
{
"operation": "Highlighting",
"time": "52.062",
"value": "Saetbyoul Middle School_sentence_5"
},
{
"operation": "finish",
"time": "211.658",
"value": "finish"
}
]
}
|
exfever_5072
|
STS-96 was a Space Shuttle mission flown by Space Shuttle Discovery to a modular space station in high Earth orbit.
|
{
"evidence": [
"STS-96",
"International Space Station"
],
"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:\nSTS-96 was a Space Shuttle mission flown by Space Shuttle Discovery to a modular space station in high Earth orbit.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"STS-96",
"International Space Station"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 103,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
14065
|
Bethany Kehdy competed in the Miss Lebanon pageant in 2002 and finished as 1st runner up.
|
{
"evidence": [
"Bethany Kehdy_sentence_8",
"Bethany Kehdy_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:\nBethany Kehdy competed in the Miss Lebanon pageant in 2002 and finished as 1st runner up.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Bethany Kehdy_sentence_8",
"Bethany Kehdy_sentence_9"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 104,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.652",
"value": "?search="
},
{
"operation": "search",
"time": "184.372",
"value": "bethany kehdy"
},
{
"operation": "Now on",
"time": "185.533",
"value": "Bethany Kehdy"
},
{
"operation": "Highlighting",
"time": "207.589",
"value": "Bethany Kehdy_sentence_8"
},
{
"operation": "Highlighting",
"time": "210.24",
"value": "Bethany Kehdy_sentence_9"
},
{
"operation": "finish",
"time": "229.006",
"value": "finish"
}
]
}
|
exfever_4821
|
Matthew Avery Modine rose to prominence through his role as U.S. Marine Private/Sergeant J.T. "Joker" Davis in a war drama film based on a screenplay by an American writer, best known as the author of Dispatches.
|
{
"evidence": [
"Matthew Modine",
"Full Metal Jacket",
"Michael Herr"
],
"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:\nMatthew Avery Modine rose to prominence through his role as U.S. Marine Private/Sergeant J.T. \"Joker\" Davis in a war drama film based on a screenplay by an American writer, best known as the author of Dispatches.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Matthew Modine",
"Full Metal Jacket",
"Michael Herr"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 105,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_8832
|
Philip Ng won an Academy Award for Best Actor for his role in the Birth of the Dragon.
|
{
"evidence": [
"Birth of the Dragon",
"Billy Magnussen"
],
"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:\nPhilip Ng won an Academy Award for Best Actor for his role in the Birth of the Dragon.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Birth of the Dragon",
"Billy Magnussen"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 106,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_7839
|
A mixtape in the 1980s describes a homemade compilation of music on a compact disc that is a flat circular disc that encodes binary data used in physical offline data distribution and long-term archival.
|
{
"evidence": [
"Mixtape",
"Compact disc",
"Optical disc"
],
"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:\nA mixtape in the 1980s describes a homemade compilation of music on a compact disc that is a flat circular disc that encodes binary data used in physical offline data distribution and long-term archival.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mixtape",
"Compact disc",
"Optical disc"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 107,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
46357
|
William Broadhead born in Whirlow, Sheffield was a saw grinder who was later on elected General Secretary of the Saw Grinders in 1850.
|
{
"evidence": [
"William Broadhead_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:\nWilliam Broadhead born in Whirlow, Sheffield was a saw grinder who was later on elected General Secretary of the Saw Grinders in 1850.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"William Broadhead_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 108,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.807",
"value": "?search="
},
{
"operation": "search",
"time": "5.334",
"value": "William Broadhead"
},
{
"operation": "Now on",
"time": "6.485",
"value": "William Broadhead"
},
{
"operation": "Highlighting",
"time": "9.483",
"value": "William Broadhead_sentence_1"
},
{
"operation": "Highlighting",
"time": "14.027",
"value": "William Broadhead_sentence_2"
},
{
"operation": "hyperlink",
"time": "17.279",
"value": "Owlerton"
},
{
"operation": "Now on",
"time": "18.305",
"value": "Owlerton"
},
{
"operation": "back-button-clicked",
"time": "18.565",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "18.669",
"value": "William Broadhead"
},
{
"operation": "Highlighting deleted",
"time": "19.423",
"value": "William Broadhead_sentence_1"
},
{
"operation": "finish",
"time": "23.431",
"value": "finish"
}
]
}
|
46855
|
The Ern Westmore Hollywood Glamour Show was produced by Kroger Babb that featured five more personalities.
|
{
"evidence": [
"The Ern Westmore Hollywood Glamour Show_sentence_2"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Ern Westmore Hollywood Glamour Show was produced by Kroger Babb that featured five more personalities.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"The Ern Westmore Hollywood Glamour Show_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 109,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.026",
"value": "?search="
},
{
"operation": "search",
"time": "21.474",
"value": "Ern Westmore Hollywood Glamour Show"
},
{
"operation": "Now on",
"time": "22.554",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "25.213",
"value": "The Ern Westmore Hollywood Glamour Show"
},
{
"operation": "Now on",
"time": "25.798",
"value": "The Ern Westmore Hollywood Glamour Show"
},
{
"operation": "Highlighting",
"time": "31.723",
"value": "The Ern Westmore Hollywood Glamour Show_sentence_2"
},
{
"operation": "Page search",
"time": "39.256",
"value": "five"
},
{
"operation": "finish",
"time": "64.484",
"value": "finish"
}
]
}
|
exfever_2385
|
John Heard was the first actor nominated for an Emmy Award in 1999.
|
{
"evidence": [
"John Heard (actor)",
"Prison Break",
"Serial (radio and television)"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJohn Heard was the first actor nominated for an Emmy Award in 1999.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"John Heard (actor)",
"Prison Break",
"Serial (radio and television)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 110,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_2547
|
The Nile is a major north-flowing river in northeastern Africa that flows into a sea connected to the second-largest of the world's five oceans.
|
{
"evidence": [
"Nile",
"Mediterranean Sea",
"Atlantic 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 Nile is a major north-flowing river in northeastern Africa that flows into a sea connected to the second-largest of the world's five oceans.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Nile",
"Mediterranean Sea",
"Atlantic Ocean"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 111,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
30135
|
Maurice Leitch's first novel was The Liberty Lad and second novel was Poor Lazarus he then joined BBC headquartered in Belfast.
|
{
"evidence": [
"Maurice Leitch_sentence_15",
"Maurice Leitch_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:\nMaurice Leitch's first novel was The Liberty Lad and second novel was Poor Lazarus he then joined BBC headquartered in Belfast.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Maurice Leitch_sentence_15",
"Maurice Leitch_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 112,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.877",
"value": "?search="
},
{
"operation": "search",
"time": "12.953",
"value": "Maurice Leitch"
},
{
"operation": "Now on",
"time": "14.601",
"value": "Maurice Leitch"
},
{
"operation": "Page search",
"time": "64.068",
"value": "first"
},
{
"operation": "Highlighting",
"time": "66.368",
"value": "Maurice Leitch_sentence_2"
},
{
"operation": "Highlighting",
"time": "70.437",
"value": "Maurice Leitch_sentence_3"
},
{
"operation": "Highlighting",
"time": "92.806",
"value": "Maurice Leitch_sentence_5"
},
{
"operation": "Page search",
"time": "116.96",
"value": "Belfast"
},
{
"operation": "Highlighting",
"time": "155.665",
"value": "Maurice Leitch_sentence_14"
},
{
"operation": "Highlighting",
"time": "159.256",
"value": "Maurice Leitch_sentence_15"
},
{
"operation": "Highlighting deleted",
"time": "181.753",
"value": "Maurice Leitch_sentence_15"
},
{
"operation": "Highlighting",
"time": "184.657",
"value": "Maurice Leitch_sentence_15"
},
{
"operation": "Highlighting deleted",
"time": "252.119",
"value": "Maurice Leitch_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "268.076",
"value": "Maurice Leitch_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "326.36",
"value": "Maurice Leitch_sentence_3"
},
{
"operation": "Highlighting",
"time": "328.87",
"value": "Maurice Leitch_sentence_3"
},
{
"operation": "finish",
"time": "393.472",
"value": "finish"
}
]
}
|
24802
|
The Dalai Lama Renaissance was directed and produced by Khashyar Darvich who graduated from Miami University in Oxford, Ohio.
|
{
"evidence": [
"Dalai Lama Renaissance_sentence_0",
"Khashyar Darvich_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 Dalai Lama Renaissance was directed and produced by Khashyar Darvich who graduated from Miami University in Oxford, Ohio.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Dalai Lama Renaissance_sentence_0",
"Khashyar Darvich_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 113,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.319",
"value": "?search="
},
{
"operation": "search",
"time": "9.941",
"value": "Dalai Lama Renaissance"
},
{
"operation": "Now on",
"time": "10.739",
"value": "Dalai Lama Renaissance"
},
{
"operation": "Highlighting",
"time": "23.342",
"value": "Dalai Lama Renaissance_sentence_0"
},
{
"operation": "hyperlink",
"time": "27.133",
"value": "Khashyar Darvich"
},
{
"operation": "Now on",
"time": "27.426",
"value": "Khashyar Darvich"
},
{
"operation": "Highlighting",
"time": "38.16",
"value": "Khashyar Darvich_sentence_6"
},
{
"operation": "finish",
"time": "55.444",
"value": "finish"
}
]
}
|
exfever_5123
|
Fire OS is also a mobile operating system.
|
{
"evidence": [
"Android (operating system)",
"Linux kernel"
],
"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:\nFire OS is also a mobile operating system.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Android (operating system)",
"Linux kernel"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 114,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_3817
|
New England's largest city is also the most populous city in the Commonwealth of Massachusetts, in a country primarily located in North America.
|
{
"evidence": [
"New England",
"Boston",
"United States"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nNew England's largest city is also the most populous city in the Commonwealth of Massachusetts, in a country primarily located in North America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"New England",
"Boston",
"United States"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 115,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_7277
|
Brokeback Mountain is the most nominated film at 63rd Golden Globe Awards.
|
{
"evidence": [
"Brokeback Mountain",
"Brokeback Mountain (short story)",
"The New Yorker"
],
"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:\nBrokeback Mountain is the most nominated film at 63rd Golden Globe Awards.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Brokeback Mountain",
"Brokeback Mountain (short story)",
"The New Yorker"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 116,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
73850
|
The United States has won more medals than any other country in the history of the 400 metres event at the Olympics, the most prestigious 400 m race at elite level.
|
{
"evidence": [
"400 metres at the Olympics_sentence_13",
"400 metres at the Olympics_sentence_2"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe United States has won more medals than any other country in the history of the 400 metres event at the Olympics, the most prestigious 400 m race at elite level.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"400 metres at the Olympics_sentence_13",
"400 metres at the Olympics_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 117,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.08",
"value": "?search="
},
{
"operation": "search",
"time": "38.196",
"value": "Contains...he United States has won more medals than any other country in the history of the 400 metres event at the Olympic"
},
{
"operation": "Now on",
"time": "42.459",
"value": "undefined"
},
{
"operation": "search",
"time": "73.829",
"value": "Contains...400 metres event at the Olympics"
},
{
"operation": "Now on",
"time": "77.287",
"value": "undefined"
},
{
"operation": "search",
"time": "99.581",
"value": "the United states in the 400 metres event at the Olympics"
},
{
"operation": "Now on",
"time": "100.678",
"value": "undefined"
},
{
"operation": "search",
"time": "134.245",
"value": "Contains...The United States has won more medals than any other country in the history of the 400 metres event at the Olympics, the most prestigious 400 m race at elite level.3"
},
{
"operation": "Now on",
"time": "136.481",
"value": "undefined"
},
{
"operation": "search",
"time": "183.382",
"value": "400 metre sprint at the olympics"
},
{
"operation": "Now on",
"time": "184.665",
"value": "undefined"
},
{
"operation": "search",
"time": "430.901",
"value": "400 metre olympics medal table"
},
{
"operation": "Now on",
"time": "433.855",
"value": "undefined"
},
{
"operation": "search",
"time": "455.01",
"value": "Olympics medal table"
},
{
"operation": "Now on",
"time": "455.885",
"value": "Olympic medal table"
},
{
"operation": "search",
"time": "477.962",
"value": "olympics medal count"
},
{
"operation": "Now on",
"time": "484.445",
"value": "Olympics medal count"
},
{
"operation": "hyperlink",
"time": "505.152",
"value": "United States"
},
{
"operation": "Now on",
"time": "506.538",
"value": "United States at the Olympics"
},
{
"operation": "Page search",
"time": "510.988",
"value": "400"
},
{
"operation": "hyperlink",
"time": "537.074",
"value": "United States at the Summer Olympics"
},
{
"operation": "Now on",
"time": "542.31",
"value": "United States at the Summer Olympics"
},
{
"operation": "Page search",
"time": "545.53",
"value": "400"
},
{
"operation": "hyperlink",
"time": "618.105",
"value": "400 metres"
},
{
"operation": "Now on",
"time": "620.077",
"value": "400 metres at the Olympics"
},
{
"operation": "Page search",
"time": "634.158",
"value": "united states"
},
{
"operation": "Highlighting",
"time": "640.622",
"value": "400 metres at the Olympics_sentence_13"
},
{
"operation": "Highlighting",
"time": "651.584",
"value": "400 metres at the Olympics_sentence_2"
},
{
"operation": "finish",
"time": "863.427",
"value": "finish"
}
]
}
|
exfever_2625
|
Foo Fighters is an American band was founded by former Nirvana drummer Dave Grohl a broad genre of popular music that originated as "rock and roll" in the United States in the late 1940s and early 1950s.
|
{
"evidence": [
"Foo Fighters",
"Rock music"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nFoo Fighters is an American band was founded by former Nirvana drummer Dave Grohl a broad genre of popular music that originated as \"rock and roll\" in the United States in the late 1940s and early 1950s.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Foo Fighters",
"Rock music"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 118,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
72569
|
José Antonio Romero Morilla in this spanish name, the first or paternal family name is Romero where in some cultures, a surname, family name, or last name is the portion of one's personal name that indicates their family, tribe or community.
|
{
"evidence": [
"José Antonio Romero Morilla_sentence_0",
"Surname_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:\nJosé Antonio Romero Morilla in this spanish name, the first or paternal family name is Romero where in some cultures, a surname, family name, or last name is the portion of one's personal name that indicates their family, tribe or community.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"José Antonio Romero Morilla_sentence_0",
"Surname_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 119,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.369",
"value": "?search="
},
{
"operation": "search",
"time": "76.765",
"value": "José Antonio Romero Morilla"
},
{
"operation": "Now on",
"time": "78.036",
"value": "José Antonio Romero Morilla"
},
{
"operation": "Highlighting",
"time": "83.734",
"value": "José Antonio Romero Morilla_sentence_0"
},
{
"operation": "hyperlink",
"time": "91.77",
"value": "Spanish name"
},
{
"operation": "Now on",
"time": "93.364",
"value": "Spanish naming customs"
},
{
"operation": "back-button-clicked",
"time": "116.178",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "116.282",
"value": "José Antonio Romero Morilla"
},
{
"operation": "hyperlink",
"time": "121.428",
"value": "family name"
},
{
"operation": "Now on",
"time": "123.44",
"value": "Surname"
},
{
"operation": "Highlighting",
"time": "146.571",
"value": "Surname_sentence_4"
},
{
"operation": "back-button-clicked",
"time": "149.538",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "149.644",
"value": "José Antonio Romero Morilla"
},
{
"operation": "finish",
"time": "395.72",
"value": "finish"
}
]
}
|
68407
|
The University of Cincinnati Bearcats have two NCAA Championships, while Xavier has three.
|
{
"evidence": [
"Xavier University_sentence_97"
],
"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 University of Cincinnati Bearcats have two NCAA Championships, while Xavier has three.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Xavier University_sentence_97"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 120,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.839",
"value": "?search="
},
{
"operation": "search",
"time": "39.561",
"value": "University of Cincinnati Bearcats"
},
{
"operation": "Now on",
"time": "40.925",
"value": "Cincinnati Bearcats"
},
{
"operation": "search",
"time": "121.436",
"value": "NCAA Championships"
},
{
"operation": "Now on",
"time": "147.631",
"value": "National Collegiate Athletic Association"
},
{
"operation": "Highlighting",
"time": "261.754",
"value": "National Collegiate Athletic Association_sentence_135"
},
{
"operation": "Highlighting",
"time": "261.755",
"value": "National Collegiate Athletic Association_title"
},
{
"operation": "Highlighting",
"time": "261.755",
"value": "National Collegiate Athletic Association_section_12"
},
{
"operation": "Highlighting",
"time": "261.756",
"value": "National Collegiate Athletic Association_section_13"
},
{
"operation": "Highlighting deleted",
"time": "263.406",
"value": "National Collegiate Athletic Association_sentence_135"
},
{
"operation": "Highlighting deleted",
"time": "263.408",
"value": "National Collegiate Athletic Association_title"
},
{
"operation": "Highlighting deleted",
"time": "263.409",
"value": "National Collegiate Athletic Association_section_12"
},
{
"operation": "Highlighting deleted",
"time": "263.411",
"value": "National Collegiate Athletic Association_section_13"
},
{
"operation": "back-button-clicked",
"time": "265.917",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "278.382",
"value": "Cincinnati Bearcats"
},
{
"operation": "Page search",
"time": "299.372",
"value": "bearcats"
},
{
"operation": "page-search-reset",
"time": "303.917",
"value": "page-search-reset"
},
{
"operation": "Page search",
"time": "316.407",
"value": "ncaa"
},
{
"operation": "hyperlink",
"time": "321.857",
"value": "4 Championships Cincinnati Bearcats_section_8"
},
{
"operation": "Highlighting",
"time": "324.835",
"value": "Cincinnati Bearcats_sentence_52"
},
{
"operation": "Highlighting",
"time": "324.836",
"value": "Cincinnati Bearcats_title"
},
{
"operation": "Highlighting",
"time": "324.836",
"value": "Cincinnati Bearcats_section_8"
},
{
"operation": "Highlighting",
"time": "324.836",
"value": "Cincinnati Bearcats_section_9"
},
{
"operation": "hyperlink",
"time": "348.542",
"value": "American Athletic Conference NCAA team championships"
},
{
"operation": "Now on",
"time": "390.307",
"value": "American Athletic Conference#NCAA team championships"
},
{
"operation": "Page search",
"time": "459.558",
"value": "xavier"
},
{
"operation": "back-button-clicked",
"time": "499.421",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "499.545",
"value": "Cincinnati Bearcats#Championships Cincinnati Bearcats section 8"
},
{
"operation": "hyperlink",
"time": "539.191",
"value": "Xavier University"
},
{
"operation": "Now on",
"time": "556.183",
"value": "Xavier Musketeers men's basketball"
},
{
"operation": "back-button-clicked",
"time": "588.67",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "588.793",
"value": "Cincinnati Bearcats#Championships Cincinnati Bearcats section 8"
},
{
"operation": "hyperlink",
"time": "693.091",
"value": "Xavier University"
},
{
"operation": "Now on",
"time": "694.073",
"value": "Xavier Musketeers men's basketball"
},
{
"operation": "Page search",
"time": "707.393",
"value": "ncaa"
},
{
"operation": "hyperlink",
"time": "714.443",
"value": "4 Postseason Xavier Musketeers men's basketball_section_13"
},
{
"operation": "hyperlink",
"time": "799.861",
"value": "Xavier University"
},
{
"operation": "Now on",
"time": "801.155",
"value": "Xavier University"
},
{
"operation": "hyperlink",
"time": "806.579",
"value": "4 Athletics Xavier University_section_11"
},
{
"operation": "Highlighting",
"time": "925.352",
"value": "Xavier University_sentence_95"
},
{
"operation": "Highlighting deleted",
"time": "927.101",
"value": "Xavier University_sentence_95"
},
{
"operation": "hyperlink",
"time": "932.154",
"value": "Xavier Musketeers men's basketball"
},
{
"operation": "Now on",
"time": "932.568",
"value": "Xavier Musketeers men's basketball"
},
{
"operation": "Page search",
"time": "953.811",
"value": "ncaa"
},
{
"operation": "hyperlink",
"time": "959.071",
"value": "4.1 NCAA tournament results Xavier Musketeers men's basketball_section_14"
},
{
"operation": "Highlighting",
"time": "959.786",
"value": "Xavier Musketeers men's basketball_header_cell_1_0_2"
},
{
"operation": "Highlighting",
"time": "959.786",
"value": "Xavier Musketeers men's basketball_title"
},
{
"operation": "Highlighting",
"time": "959.787",
"value": "Xavier Musketeers men's basketball_section_13"
},
{
"operation": "Highlighting",
"time": "959.788",
"value": "Xavier Musketeers men's basketball_section_14"
},
{
"operation": "Highlighting deleted",
"time": "961.069",
"value": "Xavier Musketeers men's basketball_header_cell_1_0_2"
},
{
"operation": "Highlighting deleted",
"time": "961.071",
"value": "Xavier Musketeers men's basketball_title"
},
{
"operation": "Highlighting deleted",
"time": "961.072",
"value": "Xavier Musketeers men's basketball_section_13"
},
{
"operation": "Highlighting deleted",
"time": "961.074",
"value": "Xavier Musketeers men's basketball_section_14"
},
{
"operation": "back-button-clicked",
"time": "978.06",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "988.94",
"value": "Xavier University#Athletics Xavier University section 11"
},
{
"operation": "Highlighting",
"time": "1004.072",
"value": "Xavier University_sentence_97"
},
{
"operation": "back-button-clicked",
"time": "1016.625",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "1017.243",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "1017.439",
"value": "Xavier Musketeers men's basketball#Postseason Xavier Musketeers men.27s basketball section 13"
},
{
"operation": "back-button-clicked",
"time": "1018.775",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "1021.912",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "1022.031",
"value": "Cincinnati Bearcats#Championships Cincinnati Bearcats section 8"
},
{
"operation": "finish",
"time": "1040.586",
"value": "finish"
}
]
}
|
exfever_6448
|
Paul von Hindenburg led the Imperial Roman Army during World War I and later became the German head of state under the Weimar constitution.
|
{
"evidence": [
"Paul von Hindenburg",
"President of Germany (1919–1945)"
],
"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:\nPaul von Hindenburg led the Imperial Roman Army during World War I and later became the German head of state under the Weimar constitution.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Paul von Hindenburg",
"President of Germany (1919–1945)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 121,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
31495
|
Beaver Stadium is located on the campus of Pennsylvania State University which was founded in 1855, its covered in natural grass and has a capacity of more than a hundred thousand.
|
{
"evidence": [
"Pennsylvania State University_sentence_2",
"Pennsylvania State University_sentence_131",
"Beaver Stadium_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:\nBeaver Stadium is located on the campus of Pennsylvania State University which was founded in 1855, its covered in natural grass and has a capacity of more than a hundred thousand.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Pennsylvania State University_sentence_2",
"Pennsylvania State University_sentence_131",
"Beaver Stadium_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 122,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.833",
"value": "?search="
},
{
"operation": "search",
"time": "14.911",
"value": "Pennsylvania State University"
},
{
"operation": "Now on",
"time": "15.767",
"value": "Pennsylvania State University"
},
{
"operation": "Highlighting",
"time": "38.455",
"value": "Pennsylvania State University_sentence_2"
},
{
"operation": "Highlighting",
"time": "40.52",
"value": "Pennsylvania State University_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "44.026",
"value": "Pennsylvania State University_sentence_1"
},
{
"operation": "Page search",
"time": "74.785",
"value": "beaver"
},
{
"operation": "Highlighting",
"time": "97.754",
"value": "Pennsylvania State University_sentence_131"
},
{
"operation": "search",
"time": "140.526",
"value": "Beaver Stadium"
},
{
"operation": "Now on",
"time": "141.627",
"value": "Beaver Stadium"
},
{
"operation": "Page search",
"time": "158.234",
"value": "grass"
},
{
"operation": "Highlighting",
"time": "162.55",
"value": "Beaver Stadium_sentence_5"
},
{
"operation": "finish",
"time": "173.053",
"value": "finish"
}
]
}
|
exfever_1547
|
Facebook is the most successful American online social media and social networking service.
|
{
"evidence": [
"Facebook",
"Dustin Moskovitz",
"Eduardo Saverin"
],
"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:\nFacebook is the most successful American online social media and social networking service.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Facebook",
"Dustin Moskovitz",
"Eduardo Saverin"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 123,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
59991
|
Hormone replacement therapy has resulted in numerous cases of breast cancer.
|
{
"evidence": [
"Hormone replacement therapy_sentence_58"
],
"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:\nHormone replacement therapy has resulted in numerous cases of breast cancer.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Hormone replacement therapy_sentence_58"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 124,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.709",
"value": "?search="
},
{
"operation": "search",
"time": "11.023",
"value": "Hormone replacement therapy"
},
{
"operation": "Now on",
"time": "14.655",
"value": "Hormone replacement therapy"
},
{
"operation": "Highlighting",
"time": "182.333",
"value": "Hormone replacement therapy_cell_1_2_1"
},
{
"operation": "Highlighting",
"time": "182.338",
"value": "Hormone replacement therapy_header_cell_1_1_0"
},
{
"operation": "Highlighting",
"time": "182.338",
"value": "Hormone replacement therapy_header_cell_1_0_1"
},
{
"operation": "Highlighting",
"time": "208.621",
"value": "Hormone replacement therapy_cell_1_13_1"
},
{
"operation": "Highlighting",
"time": "251.604",
"value": "Hormone replacement therapy_sentence_58"
},
{
"operation": "finish",
"time": "266.403",
"value": "finish"
}
]
}
|
exfever_2372
|
Trevor Noah is the first South African who became the host of an American satirical news program, The Daily Show.
|
{
"evidence": [
"Trevor Noah",
"DStv"
],
"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:\nTrevor Noah is the first South African who became the host of an American satirical news program, The Daily Show.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Trevor Noah",
"DStv"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 125,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_933
|
Pablo Picasso was born and raised in Spain.
|
{
"evidence": [
"Pablo Picasso",
"Bombing of Guernica"
],
"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:\nPablo Picasso was born and raised in Spain.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Pablo Picasso",
"Bombing of Guernica"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 126,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_788
|
Disney executives announced that the show Jake and the Never Land Pirates was canceled and would not receive the fifth season in light of Mickey and the Roadster Racers.
|
{
"evidence": [
"David Arquette",
"Slasher film"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDisney executives announced that the show Jake and the Never Land Pirates was canceled and would not receive the fifth season in light of Mickey and the Roadster Racers.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"David Arquette",
"Slasher film"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 127,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_3843
|
Ryder is also located in Puerto Rico.
|
{
"evidence": [
"Ryder",
"Miami",
"Florida"
],
"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:\nRyder is also located in Puerto Rico.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ryder",
"Miami",
"Florida"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 128,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
13666
|
The list of international goals scored by Raúl shows that he was able to contribute six out of the nine goals scored by the team during the Euro 2000 qualifying held in Spain on March 27, 1999.
|
{
"evidence": [
"Raúl (footballer)_sentence_119"
],
"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 list of international goals scored by Raúl shows that he was able to contribute six out of the nine goals scored by the team during the Euro 2000 qualifying held in Spain on March 27, 1999.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Raúl (footballer)_sentence_119"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 129,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.955",
"value": "?search="
},
{
"operation": "search",
"time": "13.361",
"value": "Euro 2000"
},
{
"operation": "Now on",
"time": "14.51",
"value": "UEFA Euro 2000"
},
{
"operation": "search",
"time": "65.987",
"value": "March 27, 1999"
},
{
"operation": "Now on",
"time": "67.535",
"value": "undefined"
},
{
"operation": "search",
"time": "92.436",
"value": "The list of international goals scored by Raúl"
},
{
"operation": "Now on",
"time": "95.166",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "101.283",
"value": "Raúl (footballer)"
},
{
"operation": "Now on",
"time": "102.403",
"value": "Raúl (footballer)"
},
{
"operation": "hyperlink",
"time": "132.013",
"value": "UEFA Euro 2000 qualifier"
},
{
"operation": "Now on",
"time": "132.801",
"value": "UEFA Euro 2000 qualifying Group 6"
},
{
"operation": "back-button-clicked",
"time": "142.18",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "142.436",
"value": "Raúl (footballer)"
},
{
"operation": "Highlighting",
"time": "177.192",
"value": "Raúl (footballer)_sentence_119"
},
{
"operation": "finish",
"time": "193.997",
"value": "finish"
}
]
}
|
87786
|
The 1914 Michigan State Normal Normalites football team played against Michigan freshman on October 20th at Ann Arbor, Michigan (founded in 1824 in Washtenaw county and home of University of Michigan) and faced a 0-7 loss.
|
{
"evidence": [
"Eastern Michigan Eagles football_sentence_9"
],
"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 1914 Michigan State Normal Normalites football team played against Michigan freshman on October 20th at Ann Arbor, Michigan (founded in 1824 in Washtenaw county and home of University of Michigan) and faced a 0-7 loss.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Eastern Michigan Eagles football_sentence_9"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 130,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.694",
"value": "?search="
},
{
"operation": "search",
"time": "35.106",
"value": "Michigan State Normal Normalites football"
},
{
"operation": "Now on",
"time": "37.491",
"value": "Eastern Michigan Eagles football"
},
{
"operation": "Page search",
"time": "85.796",
"value": "october 20"
},
{
"operation": "Page search",
"time": "113.626",
"value": "1914"
},
{
"operation": "hyperlink",
"time": "147.592",
"value": "Thomas Ransom"
},
{
"operation": "Now on",
"time": "148.459",
"value": "Thomas Ransom (coach)"
},
{
"operation": "hyperlink",
"time": "180.273",
"value": "Michigan State Normal"
},
{
"operation": "Now on",
"time": "184.863",
"value": "Eastern Michigan Eagles men's basketball"
},
{
"operation": "back-button-clicked",
"time": "187.231",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "187.752",
"value": "Thomas Ransom (coach)"
},
{
"operation": "back-button-clicked",
"time": "189.323",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "189.333",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "191.283",
"value": "Eastern Michigan Eagles football"
},
{
"operation": "Highlighting",
"time": "218.018",
"value": "Eastern Michigan Eagles football_sentence_9"
},
{
"operation": "hyperlink",
"time": "279.28",
"value": "Central Michigan–Eastern Michigan football rivalry"
},
{
"operation": "Now on",
"time": "280.241",
"value": "Central Michigan–Eastern Michigan football rivalry"
},
{
"operation": "back-button-clicked",
"time": "332.235",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "333.034",
"value": "Eastern Michigan Eagles football"
},
{
"operation": "Page search",
"time": "373.831",
"value": "michigan freshman"
},
{
"operation": "finish",
"time": "402.656",
"value": "finish"
}
]
}
|
exfever_2548
|
Naturi Cora Maria Naughton is best known as one-third of the R&B group 3LW and for her acting roles in the 2009 American musical drama film that follows talented high school students attending The High School of Performing Arts, a public magnet high school specializing in teaching visual arts and performing arts, located near Lincoln Center in Lincoln Square.
|
{
"evidence": [
"Naturi Naughton",
"Fame (2009 film)",
"Fiorello H. LaGuardia High School"
],
"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:\nNaturi Cora Maria Naughton is best known as one-third of the R&B group 3LW and for her acting roles in the 2009 American musical drama film that follows talented high school students attending The High School of Performing Arts, a public magnet high school specializing in teaching visual arts and performing arts, located near Lincoln Center in Lincoln Square.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Naturi Naughton",
"Fame (2009 film)",
"Fiorello H. LaGuardia High School"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 131,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_3949
|
The Fall is the fourth studio album of Gorillaz, with collaborators that include an American singer, songwriter, musician, and record producer.
|
{
"evidence": [
"The Fall (Gorillaz album)",
"Bobby Womack"
],
"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 Fall is the fourth studio album of Gorillaz, with collaborators that include an American singer, songwriter, musician, and record producer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Fall (Gorillaz album)",
"Bobby Womack"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 132,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_5155
|
True Detective is the most-watched American anthology crime drama series on HBO in 2014 and 2015.
|
{
"evidence": [
"True Detective",
"Matthew McConaughey"
],
"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:\nTrue Detective is the most-watched American anthology crime drama series on HBO in 2014 and 2015.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"True Detective",
"Matthew McConaughey"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 133,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
72318
|
There are various types of volcanic eruptions which are distinguished by geologists and are often named after famous volcanoes wherein one of which is the Strombolian eruption—driven by the bursting of gas bubbles within the magma which grow large enough to rise yet still weak enough to cause destruction and harm.
|
{
"evidence": [
"Types of volcanic eruptions_sentence_5",
"Volcanology_sentence_4",
"Types of volcanic eruptions_sentence_66",
"Types of volcanic eruptions_sentence_68"
],
"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:\nThere are various types of volcanic eruptions which are distinguished by geologists and are often named after famous volcanoes wherein one of which is the Strombolian eruption—driven by the bursting of gas bubbles within the magma which grow large enough to rise yet still weak enough to cause destruction and harm.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Types of volcanic eruptions_sentence_5",
"Volcanology_sentence_4",
"Types of volcanic eruptions_sentence_66",
"Types of volcanic eruptions_sentence_68"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 134,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.99",
"value": "?search="
},
{
"operation": "search",
"time": "12.303",
"value": "There are various types of volcanic eruptions which are distinguished by geologists"
},
{
"operation": "Now on",
"time": "14.672",
"value": "undefined"
},
{
"operation": "search",
"time": "33.859",
"value": "Strombolian eruption"
},
{
"operation": "Now on",
"time": "35.704",
"value": "Strombolian eruption"
},
{
"operation": "search",
"time": "50.223",
"value": "Types of volcanic eruptions"
},
{
"operation": "Now on",
"time": "54.05",
"value": "Types of volcanic eruptions"
},
{
"operation": "Page search",
"time": "58.956",
"value": "dist"
},
{
"operation": "Highlighting",
"time": "62.921",
"value": "Types of volcanic eruptions_sentence_5"
},
{
"operation": "hyperlink",
"time": "66.694",
"value": "volcanologists"
},
{
"operation": "Now on",
"time": "67.704",
"value": "Volcanology"
},
{
"operation": "Highlighting",
"time": "73.153",
"value": "Volcanology_sentence_4"
},
{
"operation": "back-button-clicked",
"time": "75.841",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "75.991",
"value": "Types of volcanic eruptions"
},
{
"operation": "Highlighting",
"time": "88.886",
"value": "Types of volcanic eruptions_sentence_66"
},
{
"operation": "Highlighting",
"time": "99.414",
"value": "Types of volcanic eruptions_sentence_68"
},
{
"operation": "Page search",
"time": "139.744",
"value": "destr"
},
{
"operation": "finish",
"time": "197.537",
"value": "finish"
}
]
}
|
45134
|
Found in Vietnam, Oreta shania (described by Watson in 1967) is a moth with forewings that has a length of about 15-20.5 mm and 19.5 mm for females, and with wings have two white cell-spots and black or dark brown spots on the costa and at the anal angle, while the hindwings are lightly speckled with black or dark brown.
|
{
"evidence": [
"Oreta shania_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:\nFound in Vietnam, Oreta shania (described by Watson in 1967) is a moth with forewings that has a length of about 15-20.5 mm and 19.5 mm for females, and with wings have two white cell-spots and black or dark brown spots on the costa and at the anal angle, while the hindwings are lightly speckled with black or dark brown.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Oreta shania_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 135,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.395",
"value": "?search="
},
{
"operation": "search",
"time": "12.565",
"value": "Oreta shania"
},
{
"operation": "Now on",
"time": "13.631",
"value": "Oreta shania"
},
{
"operation": "Highlighting",
"time": "15.849",
"value": "Oreta shania_sentence_2"
},
{
"operation": "finish",
"time": "62.717",
"value": "finish"
}
]
}
|
exfever_3644
|
Venice is a city in northeastern Italy and the capital of the region whose population is about five million, ranking fourth in Spain.
|
{
"evidence": [
"Venice",
"Veneto"
],
"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:\nVenice is a city in northeastern Italy and the capital of the region whose population is about five million, ranking fourth in Spain.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Venice",
"Veneto"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 136,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
2629
|
Alexa Micek studied at Burnham Park High School and later went to New York City College.
|
{
"evidence": [
"Alexa Micek_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:\nAlexa Micek studied at Burnham Park High School and later went to New York City College.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Alexa Micek_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 137,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "5.739",
"value": "alexa micek"
},
{
"operation": "Highlighting",
"time": "14.402",
"value": "Alexa Micek_sentence_3"
},
{
"operation": "Page search",
"time": "27.618",
"value": "burnham"
},
{
"operation": "finish",
"time": "31.069",
"value": "finish"
}
]
}
|
91030
|
There are a total of 20 Central Bureau of Investigation (CBI) cases and several other vigilance department investigations against Chief Minister Bhupinder Singh Hooda underway which included several scams, mostly related to illegal land grab, that took place during his rule in Haryana.
|
{
"evidence": [
"Bhupinder Singh Hooda_sentence_28"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThere are a total of 20 Central Bureau of Investigation (CBI) cases and several other vigilance department investigations against Chief Minister Bhupinder Singh Hooda underway which included several scams, mostly related to illegal land grab, that took place during his rule in Haryana.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bhupinder Singh Hooda_sentence_28"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 138,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.17",
"value": "?search="
},
{
"operation": "search",
"time": "30.27",
"value": "Bhupinder Singh Hooda"
},
{
"operation": "Now on",
"time": "31.769",
"value": "Bhupinder Singh Hooda"
},
{
"operation": "Highlighting",
"time": "50.952",
"value": "Bhupinder Singh Hooda_sentence_28"
},
{
"operation": "finish",
"time": "68.736",
"value": "finish"
}
]
}
|
39444
|
Platydemus is a genus of four-legged insects in the tribe Rhynchodemini, in the family Geoplanidae.
|
{
"evidence": [
"Platydemus_sentence_0",
"Geoplanidae_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:\nPlatydemus is a genus of four-legged insects in the tribe Rhynchodemini, in the family Geoplanidae.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Platydemus_sentence_0",
"Geoplanidae_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 139,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.886",
"value": "?search="
},
{
"operation": "search",
"time": "7.601",
"value": "Platydemus"
},
{
"operation": "Now on",
"time": "9.161",
"value": "Platydemus"
},
{
"operation": "Highlighting",
"time": "26.994",
"value": "Platydemus_cell_0_8_0"
},
{
"operation": "Highlighting",
"time": "27.007",
"value": "Platydemus_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "27.008",
"value": "Platydemus_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "29.943",
"value": "Platydemus_cell_0_8_1"
},
{
"operation": "Highlighting",
"time": "32.651",
"value": "Platydemus_cell_0_6_0"
},
{
"operation": "Highlighting",
"time": "33.945",
"value": "Platydemus_cell_0_6_1"
},
{
"operation": "Highlighting",
"time": "37.782",
"value": "Platydemus_sentence_1"
},
{
"operation": "Highlighting",
"time": "37.783",
"value": "Platydemus_section_0"
},
{
"operation": "Highlighting",
"time": "42.033",
"value": "Platydemus_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "47.4",
"value": "Platydemus_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "47.406",
"value": "Platydemus_section_0"
},
{
"operation": "hyperlink",
"time": "50.149",
"value": "land planarians"
},
{
"operation": "Now on",
"time": "51.497",
"value": "Land planarians"
},
{
"operation": "Highlighting",
"time": "86.206",
"value": "Geoplanidae_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "88.689",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "88.826",
"value": "Platydemus"
},
{
"operation": "Highlighting deleted",
"time": "93.229",
"value": "Platydemus_cell_0_6_0"
},
{
"operation": "Highlighting deleted",
"time": "94.151",
"value": "Platydemus_cell_0_6_1"
},
{
"operation": "Highlighting deleted",
"time": "95.272",
"value": "Platydemus_cell_0_8_1"
},
{
"operation": "Highlighting deleted",
"time": "96.174",
"value": "Platydemus_cell_0_8_0"
},
{
"operation": "Highlighting deleted",
"time": "96.182",
"value": "Platydemus_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "96.187",
"value": "Platydemus_header_cell_0_0_0"
},
{
"operation": "hyperlink",
"time": "98.022",
"value": "land planarians"
},
{
"operation": "Now on",
"time": "99.006",
"value": "Land planarians"
},
{
"operation": "finish",
"time": "130.1",
"value": "finish"
}
]
}
|
exfever_8350
|
Liev Schreiber is an American actor known for appearing in mainstream Hollywood films, such as the 2004 political thriller film directed by Jonathan Demme, based on Richard Condon's novel.
|
{
"evidence": [
"Liev Schreiber",
"The Manchurian Candidate (2004 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:\nLiev Schreiber is an American actor known for appearing in mainstream Hollywood films, such as the 2004 political thriller film directed by Jonathan Demme, based on Richard Condon's novel.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Liev Schreiber",
"The Manchurian Candidate (2004 film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 140,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
77150
|
Alistair Bray is an Australian professional footballer who plays as a goalkeeper for South Melbourne FC in National Premier Leagues Victoria.
|
{
"evidence": [
"South Melbourne FC_sentence_3",
"South Melbourne FC_sentence_151"
],
"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:\nAlistair Bray is an Australian professional footballer who plays as a goalkeeper for South Melbourne FC in National Premier Leagues Victoria.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"South Melbourne FC_sentence_3",
"South Melbourne FC_sentence_151"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 141,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.956",
"value": "?search="
},
{
"operation": "search",
"time": "28.995",
"value": "South Melbourne FC"
},
{
"operation": "Now on",
"time": "30.141",
"value": "South Melbourne FC"
},
{
"operation": "Page search",
"time": "112.412",
"value": "bray"
},
{
"operation": "Highlighting",
"time": "121.93",
"value": "South Melbourne FC_sentence_3"
},
{
"operation": "Highlighting",
"time": "147.983",
"value": "South Melbourne FC_sentence_151"
},
{
"operation": "Highlighting",
"time": "149.245",
"value": "South Melbourne FC_sentence_150"
},
{
"operation": "Highlighting deleted",
"time": "151.365",
"value": "South Melbourne FC_sentence_150"
},
{
"operation": "search",
"time": "173.907",
"value": "alistair bray"
},
{
"operation": "Now on",
"time": "175.026",
"value": "undefined"
},
{
"operation": "finish",
"time": "197.058",
"value": "finish"
}
]
}
|
exfever_4242
|
The Great British Sewing Bee was the first presented reality show by Claudia Winkleman.
|
{
"evidence": [
"The Great British Sewing Bee",
"BBC One"
],
"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 Great British Sewing Bee was the first presented reality show by Claudia Winkleman.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Great British Sewing Bee",
"BBC One"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 142,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_6266
|
Karl Malone is a Ukrainian former professional player of a team sport in which two teams, most commonly of five players each, opposing one another on a rectangular court, compete with the primary objective of shooting a spherical ball usually ranging in size from very small promotional items that are only a few inches in diameter to extra large balls nearly 2 feet in diameter used in training exercises.
|
{
"evidence": [
"Karl Malone",
"Basketball",
"Basketball (ball)"
],
"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:\nKarl Malone is a Ukrainian former professional player of a team sport in which two teams, most commonly of five players each, opposing one another on a rectangular court, compete with the primary objective of shooting a spherical ball usually ranging in size from very small promotional items that are only a few inches in diameter to extra large balls nearly 2 feet in diameter used in training exercises.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Karl Malone",
"Basketball",
"Basketball (ball)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 143,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
40486
|
In a cricket match in India on 25 December during the 2012-2013 season, when the Pakistan cricket team toured India from 25 December 2012 to 6 January 2013, Pakistan won by 5 wickets, but on 28 December, India won by 11 runs.
|
{
"evidence": [
"India–Pakistan cricket rivalry_sentence_38",
"Pakistani cricket team in India in 2012–13_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 a cricket match in India on 25 December during the 2012-2013 season, when the Pakistan cricket team toured India from 25 December 2012 to 6 January 2013, Pakistan won by 5 wickets, but on 28 December, India won by 11 runs.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"India–Pakistan cricket rivalry_sentence_38",
"Pakistani cricket team in India in 2012–13_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 144,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.682",
"value": "?search="
},
{
"operation": "search",
"time": "52.784",
"value": "Contains...2012-2013 pakistan cricket team"
},
{
"operation": "Now on",
"time": "54.633",
"value": "undefined"
},
{
"operation": "search",
"time": "105.625",
"value": "2012-13 pakistan cricket"
},
{
"operation": "Now on",
"time": "107.099",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "127.219",
"value": "India–Pakistan cricket rivalry"
},
{
"operation": "Now on",
"time": "128.331",
"value": "India–Pakistan cricket rivalry"
},
{
"operation": "Highlighting",
"time": "141.496",
"value": "India–Pakistan cricket rivalry_sentence_38"
},
{
"operation": "hyperlink",
"time": "227.554",
"value": "Pakistan's tour of India"
},
{
"operation": "Now on",
"time": "228.088",
"value": "Pakistani cricket team in India in 2012–13"
},
{
"operation": "Highlighting",
"time": "245.947",
"value": "Pakistani cricket team in India in 2012–13_sentence_0"
},
{
"operation": "hyperlink",
"time": "276.862",
"value": "One Day International"
},
{
"operation": "Now on",
"time": "277.537",
"value": "One Day International"
},
{
"operation": "back-button-clicked",
"time": "325.976",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "326.094",
"value": "Pakistani cricket team in India in 2012–13"
},
{
"operation": "back-button-clicked",
"time": "331.714",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "331.719",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "332.045",
"value": "India–Pakistan cricket rivalry"
},
{
"operation": "hyperlink",
"time": "335.501",
"value": "2012–13"
},
{
"operation": "Now on",
"time": "335.929",
"value": "Pakistani cricket team in India in 2012–13"
},
{
"operation": "back-button-clicked",
"time": "339.49",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "339.816",
"value": "India–Pakistan cricket rivalry"
},
{
"operation": "hyperlink",
"time": "442.007",
"value": "5 List of ODI series India–Pakistan cricket rivalry_section_6"
},
{
"operation": "search",
"time": "482.004",
"value": "pakistan india cricket match december 25"
},
{
"operation": "Now on",
"time": "484.343",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "493.331",
"value": "India national cricket team"
},
{
"operation": "Now on",
"time": "494.456",
"value": "India national cricket team"
},
{
"operation": "back-button-clicked",
"time": "534.874",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "535.649",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "560.943",
"value": "next 20"
},
{
"operation": "Now on",
"time": "561.683",
"value": "undefined"
},
{
"operation": "search",
"time": "601.255",
"value": "pakistan cricket team performance in 2012"
},
{
"operation": "Now on",
"time": "602.62",
"value": "undefined"
},
{
"operation": "search",
"time": "624.244",
"value": "2012-2013 cricket"
},
{
"operation": "Now on",
"time": "625.049",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "643.849",
"value": "India national cricket team"
},
{
"operation": "Now on",
"time": "644.693",
"value": "India national cricket team"
},
{
"operation": "hyperlink",
"time": "658.372",
"value": "11.2 One-Day Internationals India national cricket team_section_22"
},
{
"operation": "Page search",
"time": "676.517",
"value": "2012"
},
{
"operation": "search",
"time": "739.471",
"value": "In a cricket match in India on 25 December during the 2012-2013 season, when the Pakistan cricket team toured India from 25 December 2012 to 6 January 2013, Pakistan won by 5 wickets, but on 28 December, India won by 11 runs"
},
{
"operation": "Now on",
"time": "744.777",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "770.257",
"value": "Pakistan national cricket team"
},
{
"operation": "Now on",
"time": "771.811",
"value": "Pakistan national cricket team"
},
{
"operation": "hyperlink",
"time": "802.44",
"value": "1.7 2012 ICC World T20 Pakistan national cricket team_section_7"
},
{
"operation": "hyperlink",
"time": "850.149",
"value": "5 List of international grounds Pakistan national cricket team_section_29"
},
{
"operation": "hyperlink",
"time": "878.359",
"value": "India versus Pakistan cricket rivalry"
},
{
"operation": "Now on",
"time": "879.383",
"value": "India versus Pakistan cricket rivalry"
},
{
"operation": "hyperlink",
"time": "939.684",
"value": "2012–13"
},
{
"operation": "Now on",
"time": "940.094",
"value": "Pakistani cricket team in India in 2012–13"
},
{
"operation": "finish",
"time": "970.014",
"value": "finish"
}
]
}
|
exfever_8441
|
Hairspray: Soundtrack to the Motion Picture is the soundtrack album for the 2007 musical film Hairspray which features performances from the film's cast, including an American actor who has appeared in more than 100 films and television programs, including a 1992 American superhero film directed by Tim Burton.
|
{
"evidence": [
"Hairspray (2007 soundtrack)",
"Christopher Walken",
"Batman Returns"
],
"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:\nHairspray: Soundtrack to the Motion Picture is the soundtrack album for the 2007 musical film Hairspray which features performances from the film's cast, including an American actor who has appeared in more than 100 films and television programs, including a 1992 American superhero film directed by Tim Burton.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hairspray (2007 soundtrack)",
"Christopher Walken",
"Batman Returns"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 145,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_583
|
Stephanie Daley is a 2006 film about teen pregnancy that earned Tamblyn a nomination for best-supporting female at the Friends of Independents Awards, an award dedicated to independent filmmakers and produced by Film Independent.
|
{
"evidence": [
"Stephanie Daley",
"Independent Spirit Awards"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nStephanie Daley is a 2006 film about teen pregnancy that earned Tamblyn a nomination for best-supporting female at the Friends of Independents Awards, an award dedicated to independent filmmakers and produced by Film Independent.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Stephanie Daley",
"Independent Spirit Awards"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 146,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_8026
|
Juliette Binoche is a French actress who appeared in a 1989 French film by Jacques Doillon.
|
{
"evidence": [
"Juliette Binoche",
"Family Life (1985 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:\nJuliette Binoche is a French actress who appeared in a 1989 French film by Jacques Doillon.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Juliette Binoche",
"Family Life (1985 film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 147,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_3106
|
The Celtic are a Scottish professional football club based in Glasgow who had a friendly match against the Scottish professional football club based in the Govan district of Birmingham.
|
{
"evidence": [
"Celtic F.C.",
"Rangers F.C."
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Celtic are a Scottish professional football club based in Glasgow who had a friendly match against the Scottish professional football club based in the Govan district of Birmingham.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Celtic F.C.",
"Rangers F.C."
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 148,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_942
|
Friendship is a relationship of mutual affection between people, and theories have been proposed, including a sociological and psychological theory that studies the social behavior in the interaction of two parties that implements a cost-benefit analysis to determine risks and benefits.
|
{
"evidence": [
"Friendship",
"Social exchange theory"
],
"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:\nFriendship is a relationship of mutual affection between people, and theories have been proposed, including a sociological and psychological theory that studies the social behavior in the interaction of two parties that implements a cost-benefit analysis to determine risks and benefits.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Friendship",
"Social exchange theory"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 149,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_7746
|
Take Me High is a 1970 British film directed by David Askey, starring an English singer who is the third-top-selling artist in UK Singles Chart history behind the Beatles and the King of Rock and Roll.
|
{
"evidence": [
"Take Me High",
"Cliff Richard",
"Elvis Presley"
],
"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:\nTake Me High is a 1970 British film directed by David Askey, starring an English singer who is the third-top-selling artist in UK Singles Chart history behind the Beatles and the King of Rock and Roll.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Take Me High",
"Cliff Richard",
"Elvis Presley"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 150,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_2694
|
Omar Sharif was an Egyptian actor who played opposite Peter O'Toole as Sherif Ali in a 1962 Belgian epic historical drama film based on the life of T. E. Lawrence and his 1926 book Seven Pillars of Wisdom.
|
{
"evidence": [
"Omar Sharif",
"Lawrence of Arabia (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:\nOmar Sharif was an Egyptian actor who played opposite Peter O'Toole as Sherif Ali in a 1962 Belgian epic historical drama film based on the life of T. E. Lawrence and his 1926 book Seven Pillars of Wisdom.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Omar Sharif",
"Lawrence of Arabia (film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 151,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
81136
|
Hexage developed the RPG Reaper: Tale of a Pale Swordsman in 2013 and was the first title by the company that is available on Steam (a standalone software client to provide automatic updates for their games, and expanded to include games from third-party publishers).
|
{
"evidence": [
"Hexage_sentence_3",
"Hexage_sentence_6",
"Hexage_sentence_5"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHexage developed the RPG Reaper: Tale of a Pale Swordsman in 2013 and was the first title by the company that is available on Steam (a standalone software client to provide automatic updates for their games, and expanded to include games from third-party publishers).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Hexage_sentence_3",
"Hexage_sentence_6",
"Hexage_sentence_5"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 152,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.868",
"value": "?search="
},
{
"operation": "search",
"time": "82.797",
"value": "Hexage"
},
{
"operation": "Now on",
"time": "83.607",
"value": "Hexage"
},
{
"operation": "Highlighting",
"time": "112.633",
"value": "Hexage_sentence_3"
},
{
"operation": "Highlighting",
"time": "134.367",
"value": "Hexage_sentence_6"
},
{
"operation": "hyperlink",
"time": "193.679",
"value": "Steam"
},
{
"operation": "Now on",
"time": "195.172",
"value": "Steam (software)"
},
{
"operation": "Highlighting",
"time": "203.463",
"value": "Steam (service)_sentence_1"
},
{
"operation": "Highlighting",
"time": "203.463",
"value": "Steam (service)_title"
},
{
"operation": "back-button-clicked",
"time": "226.609",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "226.857",
"value": "Hexage"
},
{
"operation": "search",
"time": "401.209",
"value": "Contains...Reaper"
},
{
"operation": "Now on",
"time": "403.788",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "421.311",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "421.502",
"value": "Hexage"
},
{
"operation": "search",
"time": "494.88",
"value": "Reaper"
},
{
"operation": "Now on",
"time": "495.995",
"value": "Reaper"
},
{
"operation": "back-button-clicked",
"time": "506.287",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "506.503",
"value": "Hexage"
},
{
"operation": "search",
"time": "685.544",
"value": "Contains...RPG Reaper"
},
{
"operation": "Now on",
"time": "686.361",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "694.015",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "694.272",
"value": "Hexage"
},
{
"operation": "Highlighting",
"time": "726.015",
"value": "Hexage_sentence_5"
},
{
"operation": "finish",
"time": "961.796",
"value": "finish"
}
]
}
|
exfever_914
|
Gerald Rudolph Ford Jr. previously served as the leader of the Republican Party in the lower house of the United States Congress.
|
{
"evidence": [
"Gerald Ford",
"United States House of Representatives"
],
"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:\nGerald Rudolph Ford Jr. previously served as the leader of the Republican Party in the lower house of the United States Congress.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Gerald Ford",
"United States House of Representatives"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 153,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_8597
|
Red is a 2010 British action comedy film loosely inspired by the comic book publishing imprint, a subdivision of Wildstorm.
|
{
"evidence": [
"Red (2010 film)",
"Homage 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:\nRed is a 2010 British action comedy film loosely inspired by the comic book publishing imprint, a subdivision of Wildstorm.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Red (2010 film)",
"Homage Comics"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 154,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_6056
|
Céline Marie Claudette Dion is a Chinese singer with a music style incorporated into the genre of popular music, which originated in its modern form during the mid-1950s in the United States and the United Kingdom.
|
{
"evidence": [
"Celine Dion",
"Pop music"
],
"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:\nCéline Marie Claudette Dion is a Chinese singer with a music style incorporated into the genre of popular music, which originated in its modern form during the mid-1950s in the United States and the United Kingdom.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Celine Dion",
"Pop music"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 155,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
17246
|
The Nazi umbrella group National Movement of Switzerland has its roots in the foundation of the Bund by Swiss politician and supporter of National Socialism Rolf Henne.
|
{
"evidence": [
"National Movement of Switzerland_sentence_0",
"National Movement of Switzerland_sentence_1",
"Rolf Henne_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 Nazi umbrella group National Movement of Switzerland has its roots in the foundation of the Bund by Swiss politician and supporter of National Socialism Rolf Henne.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"National Movement of Switzerland_sentence_0",
"National Movement of Switzerland_sentence_1",
"Rolf Henne_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 156,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.283",
"value": "?search="
},
{
"operation": "search",
"time": "19.753",
"value": "National Movement of Switzerland"
},
{
"operation": "Now on",
"time": "21.271",
"value": "National Movement of Switzerland"
},
{
"operation": "Highlighting",
"time": "24.199",
"value": "National Movement of Switzerland_sentence_0"
},
{
"operation": "Highlighting",
"time": "28.612",
"value": "National Movement of Switzerland_sentence_1"
},
{
"operation": "hyperlink",
"time": "46.047",
"value": "Rolf Henne"
},
{
"operation": "Now on",
"time": "46.907",
"value": "Rolf Henne"
},
{
"operation": "Highlighting",
"time": "49.947",
"value": "Rolf Henne_sentence_0"
},
{
"operation": "finish",
"time": "79.749",
"value": "finish"
}
]
}
|
71881
|
György Almásy was a Hungarian Asiologist, traveler, and zoologist which is the study of the animal kingdom.
|
{
"evidence": [
"György Almásy_sentence_2",
"Zoology_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:\nGyörgy Almásy was a Hungarian Asiologist, traveler, and zoologist which is the study of the animal kingdom.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"György Almásy_sentence_2",
"Zoology_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 157,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.466",
"value": "?search="
},
{
"operation": "search",
"time": "13.561",
"value": "György Almásy"
},
{
"operation": "Now on",
"time": "14.727",
"value": "György Almásy"
},
{
"operation": "Highlighting",
"time": "32.999",
"value": "György Almásy_sentence_2"
},
{
"operation": "hyperlink",
"time": "48.228",
"value": "zoologist"
},
{
"operation": "Now on",
"time": "49.291",
"value": "Zoology"
},
{
"operation": "Highlighting",
"time": "67.792",
"value": "Zoology_sentence_5"
},
{
"operation": "finish",
"time": "100.79",
"value": "finish"
}
]
}
|
exfever_3804
|
Psychological horror is a subgenre of horror and psychological fiction that frequently overlaps with the related subgenre of psychological thriller and often uses mystery elements and characters with disturbed psychological states to enhance the instinct or thought process that is believed to be heavily influenced by anxiety or fear.
|
{
"evidence": [
"Psychological horror",
"Paranoia"
],
"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:\nPsychological horror is a subgenre of horror and psychological fiction that frequently overlaps with the related subgenre of psychological thriller and often uses mystery elements and characters with disturbed psychological states to enhance the instinct or thought process that is believed to be heavily influenced by anxiety or fear.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Psychological horror",
"Paranoia"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 158,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_8345
|
The Wachowskis are American film and television directors, best known for writing and directing The Matrix and its sequel, the 2003 American science-fiction action film released by Warner Bros.
|
{
"evidence": [
"The Wachowskis",
"The Matrix Reloaded"
],
"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 Wachowskis are American film and television directors, best known for writing and directing The Matrix and its sequel, the 2003 American science-fiction action film released by Warner Bros.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Wachowskis",
"The Matrix Reloaded"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 159,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
37866
|
The Ameriprise Financial Center, a 498-foot-tall skyscraper located at 707 2nd Avenue South in Minneapolis, Minnesota, finished its construction in 2020.
|
{
"evidence": [
"Ameriprise Financial Center_sentence_1"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Ameriprise Financial Center, a 498-foot-tall skyscraper located at 707 2nd Avenue South in Minneapolis, Minnesota, finished its construction in 2020.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ameriprise Financial Center_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 160,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.691",
"value": "?search="
},
{
"operation": "search",
"time": "12.483",
"value": "Ameriprise Financial Center"
},
{
"operation": "Now on",
"time": "13.146",
"value": "Ameriprise Financial Center"
},
{
"operation": "Highlighting",
"time": "29.482",
"value": "Ameriprise Financial Center_sentence_0"
},
{
"operation": "Highlighting",
"time": "38.242",
"value": "Ameriprise Financial Center_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "40.104",
"value": "Ameriprise Financial Center_sentence_0"
},
{
"operation": "finish",
"time": "48.652",
"value": "finish"
}
]
}
|
60531
|
As a young man, Louis never served as a Lieutenant in France during the revolution; and was Colonel commandant of the 2nd Regiment of Swiss Guards in the French Royal Service.
|
{
"evidence": [
"Louis Gugy_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:\nAs a young man, Louis never served as a Lieutenant in France during the revolution; and was Colonel commandant of the 2nd Regiment of Swiss Guards in the French Royal Service.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Louis Gugy_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 161,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.218",
"value": "?search="
},
{
"operation": "search",
"time": "25.513",
"value": "louis was Colonel commandant of the 2nd Regiment of Swiss Guards in the French Royal Service"
},
{
"operation": "Now on",
"time": "28.96",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "32.028",
"value": "Louis Gugy"
},
{
"operation": "Now on",
"time": "32.601",
"value": "Louis Gugy"
},
{
"operation": "Highlighting",
"time": "40.965",
"value": "Louis Gugy_sentence_8"
},
{
"operation": "finish",
"time": "54.983",
"value": "finish"
}
]
}
|
68843
|
The film "Sunrise at Campobello" begins at the Roosevelt family's summer home on Campobello Island, New Brunswick, Canada (the largest and only inhabited island in Campobello, a civil parish in southwestern New Brunswick, near the border with Maine), in the summer of 1911, where Franklin D. Roosevelt is depicted in early scenes as vigorously athletic, enjoying games with his children and sailing his boat.
|
{
"evidence": [
"Sunrise at Campobello_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 film \"Sunrise at Campobello\" begins at the Roosevelt family's summer home on Campobello Island, New Brunswick, Canada (the largest and only inhabited island in Campobello, a civil parish in southwestern New Brunswick, near the border with Maine), in the summer of 1911, where Franklin D. Roosevelt is depicted in early scenes as vigorously athletic, enjoying games with his children and sailing his boat.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sunrise at Campobello_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 162,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "4.633",
"value": "?search="
},
{
"operation": "search",
"time": "71.793",
"value": "Sunrise at Campobello"
},
{
"operation": "Now on",
"time": "74.028",
"value": "Sunrise at Campobello"
},
{
"operation": "Highlighting",
"time": "302.215",
"value": "Sunrise at Campobello_sentence_6"
},
{
"operation": "finish",
"time": "309.824",
"value": "finish"
}
]
}
|
exfever_8207
|
Babe Ruth was an American professional player in a bat-and-ball game played between two opposing teams.
|
{
"evidence": [
"Babe Ruth",
"Baseball"
],
"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:\nBabe Ruth was an American professional player in a bat-and-ball game played between two opposing teams.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Babe Ruth",
"Baseball"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 163,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
35754
|
Among Square Enix's Final Fantasy VII metaseries, Advent Children got the highest metacritic rating.
|
{
"evidence": [
"Kazushige Nojima_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:\nAmong Square Enix's Final Fantasy VII metaseries, Advent Children got the highest metacritic rating.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Kazushige Nojima_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 164,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.626",
"value": "undefined"
},
{
"operation": "search",
"time": "20.365",
"value": "Square Enix's Final Fantasy"
},
{
"operation": "Now on",
"time": "22.134",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "42.782",
"value": "Kazushige Nojima"
},
{
"operation": "Now on",
"time": "43.571",
"value": "Kazushige Nojima"
},
{
"operation": "Highlighting",
"time": "84.974",
"value": "Kazushige Nojima_sentence_1"
},
{
"operation": "Page search",
"time": "168.066",
"value": "advent"
},
{
"operation": "Highlighting deleted",
"time": "176.609",
"value": "Kazushige Nojima_sentence_1"
},
{
"operation": "Highlighting",
"time": "180.654",
"value": "Kazushige Nojima_sentence_1"
},
{
"operation": "hyperlink",
"time": "184.313",
"value": "Final Fantasy VII"
},
{
"operation": "Now on",
"time": "185.192",
"value": "Final Fantasy VII"
},
{
"operation": "Page search",
"time": "361.92",
"value": "advent"
},
{
"operation": "Highlighting",
"time": "380.893",
"value": "Final Fantasy VII_sentence_366"
},
{
"operation": "Highlighting",
"time": "380.894",
"value": "Final Fantasy VII_title"
},
{
"operation": "Highlighting",
"time": "380.898",
"value": "Final Fantasy VII_section_13"
},
{
"operation": "Highlighting",
"time": "380.898",
"value": "Final Fantasy VII_section_15"
},
{
"operation": "Highlighting deleted",
"time": "386.729",
"value": "Final Fantasy VII_sentence_366"
},
{
"operation": "Highlighting deleted",
"time": "386.731",
"value": "Final Fantasy VII_title"
},
{
"operation": "Highlighting deleted",
"time": "386.737",
"value": "Final Fantasy VII_section_13"
},
{
"operation": "Highlighting deleted",
"time": "386.739",
"value": "Final Fantasy VII_section_15"
},
{
"operation": "Page search",
"time": "509.425",
"value": "metric"
},
{
"operation": "finish",
"time": "525.604",
"value": "finish"
}
]
}
|
43138
|
The Unity of Command II introduces the concept of headquarters, whose functionality was expanded to include repairing or constructing bridges on the map.
|
{
"evidence": [
"Unity of Command II_sentence_4",
"Unity of Command II_sentence_7"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Unity of Command II introduces the concept of headquarters, whose functionality was expanded to include repairing or constructing bridges on the map.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Unity of Command II_sentence_4",
"Unity of Command II_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 165,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.056",
"value": "?search="
},
{
"operation": "search",
"time": "15.557",
"value": "Unity of Command II"
},
{
"operation": "Now on",
"time": "16.828",
"value": "Unity of Command II"
},
{
"operation": "Highlighting",
"time": "27.614",
"value": "Unity of Command II_sentence_4"
},
{
"operation": "Highlighting",
"time": "40.414",
"value": "Unity of Command II_sentence_7"
},
{
"operation": "finish",
"time": "69.039",
"value": "finish"
}
]
}
|
exfever_8931
|
New England is a region comprising six states in the Northeastern United States including the most populous city of the Commonwealth of Massachusetts which is officially the city that physically encompasses the government's offices and meeting places.
|
{
"evidence": [
"New England",
"Boston",
"Capital city"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nNew England is a region comprising six states in the Northeastern United States including the most populous city of the Commonwealth of Massachusetts which is officially the city that physically encompasses the government's offices and meeting places.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"New England",
"Boston",
"Capital city"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 166,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
11324
|
Two of the functions of Peptidylprolyl isomerase A, or PPIA, an 18 kDa, 165-amino acid long cytosolic protein, are a PPIA-mediated activation of the ERK, JNK, and p38 kinase pathways contributing to angiogenesis, and helping to induce an inflammatory response which mitigates tissue injury.
|
{
"evidence": [
"Peptidylprolyl isomerase A_sentence_3",
"Peptidylprolyl isomerase A_sentence_21",
"Peptidylprolyl isomerase A_sentence_18"
],
"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:\nTwo of the functions of Peptidylprolyl isomerase A, or PPIA, an 18 kDa, 165-amino acid long cytosolic protein, are a PPIA-mediated activation of the ERK, JNK, and p38 kinase pathways contributing to angiogenesis, and helping to induce an inflammatory response which mitigates tissue injury.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Peptidylprolyl isomerase A_sentence_3",
"Peptidylprolyl isomerase A_sentence_21",
"Peptidylprolyl isomerase A_sentence_18"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 167,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.208",
"value": "?search="
},
{
"operation": "search",
"time": "19.722",
"value": "PPIA"
},
{
"operation": "Now on",
"time": "20.673",
"value": "PPIA"
},
{
"operation": "Highlighting",
"time": "93.579",
"value": "Peptidylprolyl isomerase A_sentence_3"
},
{
"operation": "Highlighting",
"time": "146.713",
"value": "Peptidylprolyl isomerase A_sentence_21"
},
{
"operation": "Highlighting",
"time": "155.305",
"value": "Peptidylprolyl isomerase A_sentence_18"
},
{
"operation": "finish",
"time": "203.386",
"value": "finish"
}
]
}
|
exfever_8382
|
The Grey Album is a mashup album by an American musician who came to prominence in 2004 when he released The Grey Album.
|
{
"evidence": [
"The Grey Album",
"Danger Mouse (musician)"
],
"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 Grey Album is a mashup album by an American musician who came to prominence in 2004 when he released The Grey Album.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Grey Album",
"Danger Mouse (musician)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 168,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
77673
|
The mechanisms of customary law in Africa (a large and populous continent) are becoming more widely implemented outside of small communities for civil and criminal matters ; however, the employment of customary law differs greatly between nations and this is based on conditions for human rights and conflicts with the western laws employed.
|
{
"evidence": [
"Customary law in South Africa_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 mechanisms of customary law in Africa (a large and populous continent) are becoming more widely implemented outside of small communities for civil and criminal matters ; however, the employment of customary law differs greatly between nations and this is based on conditions for human rights and conflicts with the western laws employed.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Customary law in South Africa_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 169,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.877",
"value": "?search="
},
{
"operation": "search",
"time": "7.164",
"value": "The mechanisms of customary law in Africa"
},
{
"operation": "Now on",
"time": "10.176",
"value": "undefined"
},
{
"operation": "search",
"time": "27.547",
"value": "customary law in africa"
},
{
"operation": "Now on",
"time": "29.078",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "30.538",
"value": "Customary law in South Africa"
},
{
"operation": "Now on",
"time": "31.411",
"value": "Customary law in South Africa"
},
{
"operation": "Page search",
"time": "34.744",
"value": "mechan"
},
{
"operation": "Page search",
"time": "46.594",
"value": "widely"
},
{
"operation": "Page search",
"time": "91.072",
"value": "employment of customary law differs"
},
{
"operation": "Highlighting",
"time": "339.456",
"value": "Customary law in South Africa_sentence_0"
},
{
"operation": "finish",
"time": "344.661",
"value": "finish"
}
]
}
|
exfever_414
|
Martina Hingis was born and raised in Slovakia.
|
{
"evidence": [
"Martina Hingis",
"Grand Slam (tennis)",
"Tennis"
],
"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:\nMartina Hingis was born and raised in Slovakia.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Martina Hingis",
"Grand Slam (tennis)",
"Tennis"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 170,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_731
|
The Bostonians was the third film directed by James Ivory.
|
{
"evidence": [
"The Bostonians (film)",
"Drama (film and television)"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Bostonians was the third film directed by James Ivory.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Bostonians (film)",
"Drama (film and television)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 171,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
67957
|
Disappointment Island has some interesting history, including a shipwreck which stranded sixteen people on it in 1907, and is part of the Auckland Island group Important Bird Area (IBA), which are locations determined by an internationally agreed set of criteria.
|
{
"evidence": [
"Disappointment Island_sentence_5",
"Disappointment Island_sentence_6",
"Disappointment Island_sentence_11",
"Important Bird Area_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:\nDisappointment Island has some interesting history, including a shipwreck which stranded sixteen people on it in 1907, and is part of the Auckland Island group Important Bird Area (IBA), which are locations determined by an internationally agreed set of criteria.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Disappointment Island_sentence_5",
"Disappointment Island_sentence_6",
"Disappointment Island_sentence_11",
"Important Bird Area_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 172,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.821",
"value": "?search="
},
{
"operation": "search",
"time": "12.782",
"value": "Disappointment Island"
},
{
"operation": "Now on",
"time": "13.76",
"value": "Disappointment Island"
},
{
"operation": "Highlighting",
"time": "25.757",
"value": "Disappointment Island_sentence_5"
},
{
"operation": "Highlighting",
"time": "35.548",
"value": "Disappointment Island_sentence_6"
},
{
"operation": "Highlighting",
"time": "58.468",
"value": "Disappointment Island_sentence_11"
},
{
"operation": "hyperlink",
"time": "68.973",
"value": "Important Bird Area"
},
{
"operation": "Now on",
"time": "70.18",
"value": "Important Bird Area"
},
{
"operation": "Highlighting",
"time": "85.54",
"value": "Important Bird Area_sentence_0"
},
{
"operation": "finish",
"time": "93.334",
"value": "finish"
}
]
}
|
exfever_974
|
One-Punch Man was created by an artist in a Japanese graphic novel originating from Japan.?
|
{
"evidence": [
"One-Punch Man",
"One (manga artist)",
"Manga"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nOne-Punch Man was created by an artist in a Japanese graphic novel originating from Japan.?\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"One-Punch Man",
"One (manga artist)",
"Manga"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 173,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
47180
|
The United States required almost 19,000 tonnes of uranium over three years while France required about half of that total.
|
{
"evidence": [
"Uranium mining_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:\nThe United States required almost 19,000 tonnes of uranium over three years while France required about half of that total.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Uranium mining_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 174,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.874",
"value": "?search="
},
{
"operation": "search",
"time": "23.634",
"value": "tonnes of uranium"
},
{
"operation": "Now on",
"time": "25.19",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "28.697",
"value": "List of countries by uranium reserves"
},
{
"operation": "Now on",
"time": "29.36",
"value": "List of countries by uranium reserves"
},
{
"operation": "Highlighting",
"time": "35.443",
"value": "List of countries by uranium reserves_cell_0_53_0"
},
{
"operation": "Highlighting",
"time": "35.444",
"value": "List of countries by uranium reserves_title"
},
{
"operation": "Highlighting",
"time": "35.463",
"value": "List of countries by uranium reserves_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "44.112",
"value": "List of countries by uranium reserves_cell_0_53_0"
},
{
"operation": "Highlighting deleted",
"time": "44.113",
"value": "List of countries by uranium reserves_title"
},
{
"operation": "Highlighting deleted",
"time": "44.132",
"value": "List of countries by uranium reserves_header_cell_0_0_0"
},
{
"operation": "hyperlink",
"time": "62.8",
"value": "List of countries by uranium production"
},
{
"operation": "Now on",
"time": "64.049",
"value": "List of countries by uranium production"
},
{
"operation": "hyperlink",
"time": "79.48",
"value": "Uranium mining by country"
},
{
"operation": "Now on",
"time": "80.776",
"value": "Uranium mining by country"
},
{
"operation": "hyperlink",
"time": "117.136",
"value": "uranium"
},
{
"operation": "Now on",
"time": "118.808",
"value": "Uranium"
},
{
"operation": "search",
"time": "136.431",
"value": "united states uranium"
},
{
"operation": "Now on",
"time": "137.425",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "154.503",
"value": "Depleted uranium"
},
{
"operation": "Now on",
"time": "156.35",
"value": "Depleted uranium"
},
{
"operation": "Page search",
"time": "204.128",
"value": "tonnes of uranium"
},
{
"operation": "Page search",
"time": "205.041",
"value": "tonnes of uranium"
},
{
"operation": "search",
"time": "207.125",
"value": "tonnes of uranium\\"
},
{
"operation": "Now on",
"time": "207.701",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "333.321",
"value": "Mine operations Rössing uranium mine_section_0"
},
{
"operation": "Now on",
"time": "334.134",
"value": "Rössing uranium mine#Mine operations R.C3.B6ssing uranium mine section 0"
},
{
"operation": "hyperlink",
"time": "337.027",
"value": "Uranium market"
},
{
"operation": "Now on",
"time": "337.459",
"value": "Uranium market"
},
{
"operation": "hyperlink",
"time": "357.826",
"value": "List of countries by uranium production"
},
{
"operation": "Now on",
"time": "358.577",
"value": "List of countries by uranium production"
},
{
"operation": "hyperlink",
"time": "404.553",
"value": "List of countries by uranium reserves"
},
{
"operation": "Now on",
"time": "404.863",
"value": "List of countries by uranium reserves"
},
{
"operation": "search",
"time": "439.222",
"value": "United States required almost 19,000 tonnes of uranium"
},
{
"operation": "Now on",
"time": "440.972",
"value": "undefined"
},
{
"operation": "search",
"time": "474.523",
"value": "19,000 tonnes of uranium"
},
{
"operation": "Now on",
"time": "475.208",
"value": "undefined"
},
{
"operation": "search",
"time": "507.44",
"value": "uranium import"
},
{
"operation": "Now on",
"time": "508.658",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "512.155",
"value": "Uranium mining"
},
{
"operation": "Now on",
"time": "513.513",
"value": "Uranium mining"
},
{
"operation": "hyperlink",
"time": "522.703",
"value": "List of uranium mines"
},
{
"operation": "Now on",
"time": "525.992",
"value": "List of uranium projects"
},
{
"operation": "hyperlink",
"time": "529.057",
"value": "Uranium production"
},
{
"operation": "Now on",
"time": "529.4",
"value": "Uranium mining"
},
{
"operation": "Page search",
"time": "546.562",
"value": "united"
},
{
"operation": "hyperlink",
"time": "575.214",
"value": "Uranium mining by country"
},
{
"operation": "Now on",
"time": "575.547",
"value": "Uranium mining by country"
},
{
"operation": "back-button-clicked",
"time": "593.556",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "593.647",
"value": "Uranium mining"
},
{
"operation": "Highlighting",
"time": "609.208",
"value": "Uranium mining_sentence_1"
},
{
"operation": "hyperlink",
"time": "617.002",
"value": "Nuclear fuel cycle"
},
{
"operation": "Now on",
"time": "619.256",
"value": "Nuclear fuel cycle"
},
{
"operation": "back-button-clicked",
"time": "621.033",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "621.138",
"value": "Uranium mining"
},
{
"operation": "back-button-clicked",
"time": "621.241",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "621.244",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "621.25",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "622.254",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "622.366",
"value": "Uranium mining"
},
{
"operation": "back-button-clicked",
"time": "622.784",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "622.788",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "623.402",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "623.484",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "628.183",
"value": "Uranium mining"
},
{
"operation": "Now on",
"time": "628.524",
"value": "Uranium mining"
},
{
"operation": "hyperlink",
"time": "633.267",
"value": "Uranium mining by country"
},
{
"operation": "Now on",
"time": "634.11",
"value": "Uranium mining by country"
},
{
"operation": "hyperlink",
"time": "639.448",
"value": "List of countries by uranium production"
},
{
"operation": "Now on",
"time": "639.787",
"value": "List of countries by uranium production"
},
{
"operation": "hyperlink",
"time": "670.895",
"value": "World Nuclear Association"
},
{
"operation": "Now on",
"time": "671.695",
"value": "World Nuclear Association"
},
{
"operation": "back-button-clicked",
"time": "675.322",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "675.389",
"value": "List of countries by uranium production"
},
{
"operation": "search",
"time": "693.267",
"value": "uranium use"
},
{
"operation": "Now on",
"time": "694.377",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "704.275",
"value": "Depleted uranium"
},
{
"operation": "Now on",
"time": "704.658",
"value": "Depleted uranium"
},
{
"operation": "Page search",
"time": "746.813",
"value": "united states"
},
{
"operation": "finish",
"time": "792.5",
"value": "finish"
}
]
}
|
48853
|
2016–17 UEFA Champions League knockout phase had six teams, one of which was France Monaco.
|
{
"evidence": [
"2016–17 UEFA Champions League knockout phase_sentence_1"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n2016–17 UEFA Champions League knockout phase had six teams, one of which was France Monaco.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"2016–17 UEFA Champions League knockout phase_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 175,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.092",
"value": "?search="
},
{
"operation": "search",
"time": "16.073",
"value": "2016–17 UEFA Champions League knockout phase"
},
{
"operation": "Now on",
"time": "17.903",
"value": "2016–17 UEFA Champions League knockout phase"
},
{
"operation": "Highlighting",
"time": "28.672",
"value": "2016–17 UEFA Champions League knockout phase_sentence_1"
},
{
"operation": "finish",
"time": "47.483",
"value": "finish"
}
]
}
|
exfever_2863
|
The stratosphere is the second layer of the atmosphere of the Earth whose temperatures also vary within the stratosphere as the seasons change, reaching particularly low temperatures in a phenomenon where the nighttime lasts for more than 24 hours that only occurs only inside the polar circles where it is a day in the Arctic Circle and it is night in the most southerly of the five major circles of latitude.
|
{
"evidence": [
"Stratosphere",
"Polar night",
"Antarctic Circle"
],
"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 stratosphere is the second layer of the atmosphere of the Earth whose temperatures also vary within the stratosphere as the seasons change, reaching particularly low temperatures in a phenomenon where the nighttime lasts for more than 24 hours that only occurs only inside the polar circles where it is a day in the Arctic Circle and it is night in the most southerly of the five major circles of latitude.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Stratosphere",
"Polar night",
"Antarctic Circle"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 176,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
19871
|
During the Modern pentathlon at the 1952 Summer Olympics held in Helsinki, Finland did not win any medal.
|
{
"evidence": [
"Finland at the 1952 Summer Olympics_sentence_20"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDuring the Modern pentathlon at the 1952 Summer Olympics held in Helsinki, Finland did not win any medal.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Finland at the 1952 Summer Olympics_sentence_20"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 177,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.796",
"value": "?search="
},
{
"operation": "search",
"time": "14.371",
"value": "1952 Summer Olympics"
},
{
"operation": "Now on",
"time": "15.418",
"value": "1952 Summer Olympics"
},
{
"operation": "Highlighting",
"time": "23.885",
"value": "1952 Summer Olympics_sentence_0"
},
{
"operation": "Highlighting",
"time": "23.886",
"value": "1952 Summer Olympics_title"
},
{
"operation": "Page search",
"time": "37.084",
"value": "pentathlon"
},
{
"operation": "hyperlink",
"time": "59.787",
"value": "1952 Summer Olympics medal table"
},
{
"operation": "Now on",
"time": "60.854",
"value": "1952 Summer Olympics medal table"
},
{
"operation": "back-button-clicked",
"time": "79.424",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "79.584",
"value": "1952 Summer Olympics"
},
{
"operation": "Highlighting",
"time": "94.392",
"value": "1952 Summer Olympics_item_1_2"
},
{
"operation": "Highlighting",
"time": "94.395",
"value": "1952 Summer Olympics_section_4"
},
{
"operation": "Highlighting deleted",
"time": "97.652",
"value": "1952 Summer Olympics_item_1_2"
},
{
"operation": "Highlighting deleted",
"time": "97.655",
"value": "1952 Summer Olympics_section_4"
},
{
"operation": "hyperlink",
"time": "117.96",
"value": "1952 Summer Olympics medal table"
},
{
"operation": "Now on",
"time": "118.535",
"value": "1952 Summer Olympics medal table"
},
{
"operation": "hyperlink",
"time": "123.088",
"value": "Finland"
},
{
"operation": "Now on",
"time": "123.942",
"value": "Finland at the 1952 Summer Olympics"
},
{
"operation": "Highlighting",
"time": "141.4",
"value": "Finland at the 1952 Summer Olympics_sentence_20"
},
{
"operation": "finish",
"time": "174.609",
"value": "finish"
}
]
}
|
94161
|
KAP1 acts as a transcriptional corepressor of the viral genome, where the protein separates from histones of the viral chromatin and then recruits Mi2α and SETB1.
|
{
"evidence": [
"TRIM28_sentence_22"
],
"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:\nKAP1 acts as a transcriptional corepressor of the viral genome, where the protein separates from histones of the viral chromatin and then recruits Mi2α and SETB1.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"TRIM28_sentence_22"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 178,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.914",
"value": "?search="
},
{
"operation": "search",
"time": "11.122",
"value": "KAP1"
},
{
"operation": "Now on",
"time": "11.824",
"value": "TRIM28"
},
{
"operation": "Page search",
"time": "47.812",
"value": "corepressor"
},
{
"operation": "Highlighting",
"time": "53.118",
"value": "TRIM28_sentence_21"
},
{
"operation": "Highlighting",
"time": "65.585",
"value": "TRIM28_sentence_22"
},
{
"operation": "Highlighting deleted",
"time": "71.772",
"value": "TRIM28_sentence_21"
},
{
"operation": "finish",
"time": "81.323",
"value": "finish"
}
]
}
|
exfever_8964
|
Felicity Huffman is an American actress who has won numerous awards, including a nomination for the Oscars, an award for artistic and technical merit in the film industry, first hosted by an American actor, screenwriter, director, and producer.
|
{
"evidence": [
"Felicity Huffman",
"Academy Awards",
"Douglas Fairbanks"
],
"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:\nFelicity Huffman is an American actress who has won numerous awards, including a nomination for the Oscars, an award for artistic and technical merit in the film industry, first hosted by an American actor, screenwriter, director, and producer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Felicity Huffman",
"Academy Awards",
"Douglas Fairbanks"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 179,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_4667
|
Iceland is a Nordic island country in the South Atlantic Ocean where North Germanic ethnic groups and nations who are native to the island country of Iceland speak the North Germanic language.
|
{
"evidence": [
"Iceland",
"Icelanders"
],
"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:\nIceland is a Nordic island country in the South Atlantic Ocean where North Germanic ethnic groups and nations who are native to the island country of Iceland speak the North Germanic language.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Iceland",
"Icelanders"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 180,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
15641
|
Born Nobermber 13, 1931 in Seattle, Washington, Robert Landsburg was. a photographer who never photographed the 1980 eruption of Mount St. Helens.
|
{
"evidence": [
"Robert Landsburg_sentence_0",
"Robert Landsburg_sentence_4",
"Robert Landsburg_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:\nBorn Nobermber 13, 1931 in Seattle, Washington, Robert Landsburg was. a photographer who never photographed the 1980 eruption of Mount St. Helens.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Robert Landsburg_sentence_0",
"Robert Landsburg_sentence_4",
"Robert Landsburg_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 181,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.748",
"value": "?search="
},
{
"operation": "search",
"time": "8.726",
"value": "Robert Landsburg"
},
{
"operation": "Now on",
"time": "10.13",
"value": "Robert Landsburg"
},
{
"operation": "Highlighting",
"time": "12.342",
"value": "Robert Landsburg_sentence_0"
},
{
"operation": "Highlighting",
"time": "29.509",
"value": "Robert Landsburg_sentence_4"
},
{
"operation": "Highlighting",
"time": "36.959",
"value": "Robert Landsburg_sentence_7"
},
{
"operation": "finish",
"time": "56.671",
"value": "finish"
}
]
}
|
16275
|
Index Herbariorum is the only publisher that has ever produced a document of its kind that is of a global directory of staff.
|
{
"evidence": [
"Index Herbariorum_sentence_0",
"Index Herbariorum_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:\nIndex Herbariorum is the only publisher that has ever produced a document of its kind that is of a global directory of staff.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Index Herbariorum_sentence_0",
"Index Herbariorum_sentence_6"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 182,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.697",
"value": "?search="
},
{
"operation": "search",
"time": "40.633",
"value": "Index Herbariorum"
},
{
"operation": "Now on",
"time": "41.457",
"value": "Index Herbariorum"
},
{
"operation": "Highlighting",
"time": "50.116",
"value": "Index Herbariorum_sentence_0"
},
{
"operation": "Highlighting",
"time": "55.249",
"value": "Index Herbariorum_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "56.575",
"value": "Index Herbariorum_sentence_3"
},
{
"operation": "Highlighting",
"time": "64.876",
"value": "Index Herbariorum_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "91.684",
"value": "Index Herbariorum_sentence_6"
},
{
"operation": "Highlighting",
"time": "92.792",
"value": "Index Herbariorum_sentence_6"
},
{
"operation": "finish",
"time": "119.946",
"value": "finish"
}
]
}
|
46263
|
Adrian Ungur, a Romanian tennis player who was once coached by Fabrizio Fanucci, ranked 122 in the singles main draw entrant for the 2014 Open BNP Paribas Banque de Bretagne.
|
{
"evidence": [
"Adrian Ungur_sentence_1",
"Adrian Ungur_sentence_2"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAdrian Ungur, a Romanian tennis player who was once coached by Fabrizio Fanucci, ranked 122 in the singles main draw entrant for the 2014 Open BNP Paribas Banque de Bretagne.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Adrian Ungur_sentence_1",
"Adrian Ungur_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 183,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.15",
"value": "?search="
},
{
"operation": "search",
"time": "65.132",
"value": "Adrian Ungur"
},
{
"operation": "Now on",
"time": "66.16",
"value": "Adrian Ungur"
},
{
"operation": "Highlighting",
"time": "68.648",
"value": "Adrian Ungur_sentence_1"
},
{
"operation": "Highlighting",
"time": "73.304",
"value": "Adrian Ungur_sentence_2"
},
{
"operation": "Page search",
"time": "90.537",
"value": "breta"
},
{
"operation": "search",
"time": "113.898",
"value": "2014 Open BNP Paribas Banque de Bretagne – Singles"
},
{
"operation": "Now on",
"time": "114.741",
"value": "2014 Open BNP Paribas Banque de Bretagne – Singles"
},
{
"operation": "hyperlink",
"time": "121.708",
"value": "2014 Open BNP Paribas Banque de Bretagne"
},
{
"operation": "Now on",
"time": "122.249",
"value": "2014 Open BNP Paribas Banque de Bretagne"
},
{
"operation": "hyperlink",
"time": "131.068",
"value": "2014 Open BNP Paribas Banque de Bretagne – Singles"
},
{
"operation": "Now on",
"time": "131.426",
"value": "2014 Open BNP Paribas Banque de Bretagne – Singles"
},
{
"operation": "back-button-clicked",
"time": "198.303",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "198.335",
"value": "2014 Open BNP Paribas Banque de Bretagne"
},
{
"operation": "back-button-clicked",
"time": "199.84",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "199.842",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "199.867",
"value": "2014 Open BNP Paribas Banque de Bretagne – Singles"
},
{
"operation": "back-button-clicked",
"time": "200.893",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "200.895",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "200.955",
"value": "Adrian Ungur"
},
{
"operation": "Page search",
"time": "234.13",
"value": "rank"
},
{
"operation": "finish",
"time": "260.612",
"value": "finish"
}
]
}
|
44819
|
Hollowell-Dhar was introduced at Ultra Music Festival 2015, his song "Secrets" released on March 16 was a huge EDM hit and had a number of other Beatport hits.
|
{
"evidence": [
"Kshmr_sentence_24",
"Kshmr_sentence_25",
"Secrets (Tiësto and KSHMR song)_sentence_1",
"Kshmr_sentence_26"
],
"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:\nHollowell-Dhar was introduced at Ultra Music Festival 2015, his song \"Secrets\" released on March 16 was a huge EDM hit and had a number of other Beatport hits.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Kshmr_sentence_24",
"Kshmr_sentence_25",
"Secrets (Tiësto and KSHMR song)_sentence_1",
"Kshmr_sentence_26"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 184,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.86",
"value": "?search="
},
{
"operation": "search",
"time": "8.752",
"value": "Hollowell-Dhar"
},
{
"operation": "Now on",
"time": "9.98",
"value": "undefined"
},
{
"operation": "search",
"time": "13.6",
"value": "Ultra Music Festival 2015"
},
{
"operation": "Now on",
"time": "15.542",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "35.145",
"value": "Ultra Music Festival"
},
{
"operation": "Now on",
"time": "36.856",
"value": "Ultra Music Festival"
},
{
"operation": "search",
"time": "106.334",
"value": "kshmr"
},
{
"operation": "Now on",
"time": "107.38",
"value": "Kshmr"
},
{
"operation": "Highlighting",
"time": "120.502",
"value": "Kshmr_sentence_24"
},
{
"operation": "Highlighting",
"time": "123.186",
"value": "Kshmr_sentence_25"
},
{
"operation": "hyperlink",
"time": "126.007",
"value": "Secrets"
},
{
"operation": "Now on",
"time": "126.89",
"value": "Secrets (Tiësto and KSHMR song)"
},
{
"operation": "Highlighting",
"time": "129.311",
"value": "Secrets (Tiësto and KSHMR song)_sentence_1"
},
{
"operation": "back-button-clicked",
"time": "147.354",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "147.495",
"value": "Kshmr"
},
{
"operation": "Highlighting",
"time": "151.291",
"value": "Kshmr_sentence_26"
},
{
"operation": "finish",
"time": "157.149",
"value": "finish"
}
]
}
|
11804
|
Paul Montel (29 April 1876 – 22 January 1975) was a French mathematician from Nice, who taught Henri Cartan, Jean Dieudonné and Miron Nicolescu and contributed to mathematics through the introduction and systematic development of the notion of normal family.
|
{
"evidence": [
"Paul Montel_sentence_0",
"Paul Montel_sentence_4",
"Paul Montel_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 Montel (29 April 1876 – 22 January 1975) was a French mathematician from Nice, who taught Henri Cartan, Jean Dieudonné and Miron Nicolescu and contributed to mathematics through the introduction and systematic development of the notion of normal family.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Paul Montel_sentence_0",
"Paul Montel_sentence_4",
"Paul Montel_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 185,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.842",
"value": "undefined"
},
{
"operation": "search",
"time": "16.663",
"value": "Paul Montel"
},
{
"operation": "Now on",
"time": "17.807",
"value": "Paul Montel"
},
{
"operation": "Highlighting",
"time": "22.351",
"value": "Paul Montel_sentence_0"
},
{
"operation": "Highlighting",
"time": "32.899",
"value": "Paul Montel_sentence_3"
},
{
"operation": "Highlighting",
"time": "44.932",
"value": "Paul Montel_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "48.516",
"value": "Paul Montel_sentence_3"
},
{
"operation": "Highlighting",
"time": "68.916",
"value": "Paul Montel_sentence_5"
},
{
"operation": "finish",
"time": "110.382",
"value": "finish"
}
]
}
|
78301
|
Juraqan is in the Northern District of Hamadan in Iran and has a total population of 15,000 according to the 2016 Census.
|
{
"evidence": [
"Juraqan_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:\nJuraqan is in the Northern District of Hamadan in Iran and has a total population of 15,000 according to the 2016 Census.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Juraqan_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 186,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.238",
"value": "?search="
},
{
"operation": "search",
"time": "31.16",
"value": "Juraqan"
},
{
"operation": "Now on",
"time": "32.098",
"value": "Juraqan"
},
{
"operation": "Highlighting",
"time": "37.232",
"value": "Juraqan_sentence_0"
},
{
"operation": "finish",
"time": "50.495",
"value": "finish"
}
]
}
|
exfever_594
|
Brian Helgeland is an American screenwriter known for writing the screenplays for the film that based on James Ellroy's 2000 novel of the same name.
|
{
"evidence": [
"Brian Helgeland",
"L.A. Confidential (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:\nBrian Helgeland is an American screenwriter known for writing the screenplays for the film that based on James Ellroy's 2000 novel of the same name.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Brian Helgeland",
"L.A. Confidential (film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 187,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.