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_3259
|
Pearl Jam was also named Planet Defenders by Rock The Earth.
|
{
"evidence": [
"Pearl Jam",
"Seattle"
],
"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:\nPearl Jam was also named Planet Defenders by Rock The Earth.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Pearl Jam",
"Seattle"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4784,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5386
|
Spider-Man 3 was the fifth directed film by Raimi.
|
{
"evidence": [
"Spider-Man 3",
"Spider-Man 2"
],
"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:\nSpider-Man 3 was the fifth directed film by Raimi.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Spider-Man 3",
"Spider-Man 2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4785,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6845
|
Wayne Mark Rooney is an English former player who is the record goal scorer for a professional football club that has won 12 Football Association Challenge Cups, an annual knockout football competition in men's domestic English football.
|
{
"evidence": [
"Wayne Rooney",
"Manchester United F.C.",
"FA Cup"
],
"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:\nWayne Mark Rooney is an English former player who is the record goal scorer for a professional football club that has won 12 Football Association Challenge Cups, an annual knockout football competition in men's domestic English football.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Wayne Rooney",
"Manchester United F.C.",
"FA Cup"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4786,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
89231
|
"HAM" is a song from a collaborative album released in January 2011 by American rappers Jay-Z and Kanye West (who was born June 8, 1977 and grew up in Chicago), with the help of producer Lexus "Lex Luger" Lewis,
|
{
"evidence": [
"H•A•M_sentence_2",
"H•A•M_sentence_1",
"H•A•M_sentence_3",
"Kanye West_sentence_4",
"Kanye West_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:\n\"HAM\" is a song from a collaborative album released in January 2011 by American rappers Jay-Z and Kanye West (who was born June 8, 1977 and grew up in Chicago), with the help of producer Lexus \"Lex Luger\" Lewis,\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"H•A•M_sentence_2",
"H•A•M_sentence_1",
"H•A•M_sentence_3",
"Kanye West_sentence_4",
"Kanye West_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4787,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.921",
"value": "?search="
},
{
"operation": "search",
"time": "194.188",
"value": "HAM"
},
{
"operation": "Now on",
"time": "195.12",
"value": "HAM"
},
{
"operation": "hyperlink",
"time": "196.963",
"value": "HAM"
},
{
"operation": "Now on",
"time": "197.347",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "198.579",
"value": "Ham (disambiguation)"
},
{
"operation": "Now on",
"time": "199.053",
"value": "Ham (disambiguation)"
},
{
"operation": "Highlighting",
"time": "207.661",
"value": "Ham (disambiguation)_item_5_24"
},
{
"operation": "Highlighting",
"time": "207.661",
"value": "Ham (disambiguation)_title"
},
{
"operation": "Highlighting",
"time": "207.662",
"value": "Ham (disambiguation)_section_6"
},
{
"operation": "Highlighting deleted",
"time": "208.499",
"value": "Ham (disambiguation)_item_5_24"
},
{
"operation": "Highlighting deleted",
"time": "208.5",
"value": "Ham (disambiguation)_title"
},
{
"operation": "Highlighting deleted",
"time": "208.501",
"value": "Ham (disambiguation)_section_6"
},
{
"operation": "hyperlink",
"time": "209.043",
"value": "H A M"
},
{
"operation": "Now on",
"time": "209.718",
"value": "H•A•M"
},
{
"operation": "Highlighting",
"time": "222.796",
"value": "H•A•M_sentence_2"
},
{
"operation": "Highlighting",
"time": "223.419",
"value": "H•A•M_sentence_1"
},
{
"operation": "Highlighting",
"time": "227.844",
"value": "H•A•M_sentence_3"
},
{
"operation": "hyperlink",
"time": "234.363",
"value": "Kanye West"
},
{
"operation": "Now on",
"time": "237.042",
"value": "Kanye West"
},
{
"operation": "Highlighting",
"time": "241.307",
"value": "Kanye West_sentence_4"
},
{
"operation": "Highlighting",
"time": "246.787",
"value": "Kanye West_sentence_2"
},
{
"operation": "finish",
"time": "250.739",
"value": "finish"
}
]
}
|
23492
|
Most Somali men (residents of Australia who are of culturally and linguistically Arabized people native to Greater Somalia ancestry) have found jobs in the transport and production industries, and a rising number of women in the community do clerical work; while communal support networks have also expanded.
|
{
"evidence": [
"Somali Australians_sentence_11"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMost Somali men (residents of Australia who are of culturally and linguistically Arabized people native to Greater Somalia ancestry) have found jobs in the transport and production industries, and a rising number of women in the community do clerical work; while communal support networks have also expanded.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Somali Australians_sentence_11"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4788,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.27",
"value": "?search="
},
{
"operation": "search",
"time": "29.858",
"value": "Contains...Somali men"
},
{
"operation": "Now on",
"time": "31.292",
"value": "undefined"
},
{
"operation": "search",
"time": "51.417",
"value": "Contains...have found jobs in the transport and production industries, and a rising number of women in the community do clerical work"
},
{
"operation": "Now on",
"time": "54.523",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "59.415",
"value": "Somali Australians"
},
{
"operation": "Now on",
"time": "60.555",
"value": "Somali Australians"
},
{
"operation": "Page search",
"time": "86.262",
"value": "nativ"
},
{
"operation": "search",
"time": "161.515",
"value": "Contains...residents of Australia who are of culturally and linguistically Arabized people native to Greater Somalia ancestry"
},
{
"operation": "Now on",
"time": "163.81",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "177.898",
"value": "Autosomal ancestry Somalis_section_27"
},
{
"operation": "Now on",
"time": "180.389",
"value": "Somalis#Autosomal ancestry Somalis section 27"
},
{
"operation": "search",
"time": "282.641",
"value": "Somali Australians"
},
{
"operation": "Now on",
"time": "283.513",
"value": "Somali Australians"
},
{
"operation": "Page search",
"time": "296.534",
"value": "transport"
},
{
"operation": "Highlighting",
"time": "305.2",
"value": "Somali Australians_sentence_11"
},
{
"operation": "Page search",
"time": "330.831",
"value": "cultural"
},
{
"operation": "Page search",
"time": "353.364",
"value": "Arabized"
},
{
"operation": "hyperlink",
"time": "360.249",
"value": "Somali diaspora"
},
{
"operation": "Now on",
"time": "361.191",
"value": "Somali diaspora"
},
{
"operation": "Page search",
"time": "366.319",
"value": "Arabized"
},
{
"operation": "Page search",
"time": "368.961",
"value": "Arabized"
},
{
"operation": "search",
"time": "387.18",
"value": "Contains...residents of Australia who are of culturally and linguistically Arabized"
},
{
"operation": "Now on",
"time": "388.912",
"value": "undefined"
},
{
"operation": "search",
"time": "436.791",
"value": "Contains...people native to Greater Somalia ancestry"
},
{
"operation": "Now on",
"time": "438.255",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "445.355",
"value": "Autosomal ancestry Somalis_section_27"
},
{
"operation": "Now on",
"time": "446.667",
"value": "Somalis#Autosomal ancestry Somalis section 27"
},
{
"operation": "Page search",
"time": "465.614",
"value": "people native to Greater Somalia ancestry"
},
{
"operation": "Page search",
"time": "518.035",
"value": "native to Greater Somalia ancestry"
},
{
"operation": "search",
"time": "530.033",
"value": "Contains...culturally and linguistically Arabized"
},
{
"operation": "Now on",
"time": "531.246",
"value": "undefined"
},
{
"operation": "search",
"time": "589.712",
"value": "somali arabs"
},
{
"operation": "Now on",
"time": "590.512",
"value": "undefined"
},
{
"operation": "search",
"time": "655.119",
"value": "Contains... have found jobs in the transport and production industries"
},
{
"operation": "Now on",
"time": "658.068",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "667.468",
"value": "Somali Australians"
},
{
"operation": "Now on",
"time": "668.289",
"value": "Somali Australians"
},
{
"operation": "Page search",
"time": "690.17",
"value": "clerical"
},
{
"operation": "Highlighting deleted",
"time": "695.333",
"value": "Somali Australians_sentence_11"
},
{
"operation": "Highlighting",
"time": "696.755",
"value": "Somali Australians_sentence_11"
},
{
"operation": "finish",
"time": "750.653",
"value": "finish"
}
]
}
|
exfever_8874
|
"The Long Run" is a song recorded by the Eagles which was written by an American musician, singer, songwriter, and record producer who was the drummer and co-lead vocalist for the Eagles from 1971 until the band broke up in 1980.
|
{
"evidence": [
"The Long Run (song)",
"Don Henley"
],
"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:\n\"The Long Run\" is a song recorded by the Eagles which was written by an American musician, singer, songwriter, and record producer who was the drummer and co-lead vocalist for the Eagles from 1971 until the band broke up in 1980.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Long Run (song)",
"Don Henley"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4789,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
40569
|
Within a few weeks of having arrived, his father died in Rockhampton; so, Parnell started work at a station in the St Lawrence area of Queensland and later managed a pharmacy in Peak Downs.
|
{
"evidence": [
"Arthur Horatio Parnell_sentence_5",
"Arthur Horatio Parnell_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:\nWithin a few weeks of having arrived, his father died in Rockhampton; so, Parnell started work at a station in the St Lawrence area of Queensland and later managed a pharmacy in Peak Downs.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Arthur Horatio Parnell_sentence_5",
"Arthur Horatio Parnell_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4790,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.655",
"value": "?search="
},
{
"operation": "search",
"time": "14.666",
"value": "Parnell Peak Downs"
},
{
"operation": "Now on",
"time": "16.3",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "17.738",
"value": "Arthur Horatio Parnell"
},
{
"operation": "Now on",
"time": "18.114",
"value": "Arthur Horatio Parnell"
},
{
"operation": "Highlighting",
"time": "22.658",
"value": "Arthur Horatio Parnell_sentence_5"
},
{
"operation": "Highlighting",
"time": "25.77",
"value": "Arthur Horatio Parnell_sentence_6"
},
{
"operation": "finish",
"time": "54.986",
"value": "finish"
}
]
}
|
exfever_4031
|
The Houston Rockets are an American professional basketball team based in Houston that was established in a city in the U.S. state of California on the coast of the Pacific Ocean in 1968.
|
{
"evidence": [
"Houston Rockets",
"San Diego"
],
"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 Houston Rockets are an American professional basketball team based in Houston that was established in a city in the U.S. state of California on the coast of the Pacific Ocean in 1968.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Houston Rockets",
"San Diego"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4791,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_607
|
The first song Fernando Garibay ever wrote is titled Stars Are Blind.
|
{
"evidence": [
"Stars Are Blind",
"Paris Hilton"
],
"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 first song Fernando Garibay ever wrote is titled Stars Are Blind.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Stars Are Blind",
"Paris Hilton"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4792,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
43062
|
In 1910, the Michigan State Normal Normalites football team lost against Alma and Central Michigan, headed by captain Earl McCarty.
|
{
"evidence": [
"1910 Michigan State Normal Normalites football team_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:\nIn 1910, the Michigan State Normal Normalites football team lost against Alma and Central Michigan, headed by captain Earl McCarty.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"1910 Michigan State Normal Normalites football team_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4793,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.692",
"value": "?search="
},
{
"operation": "search",
"time": "25.034",
"value": "1910 Michigan State Normal Normalites football team"
},
{
"operation": "Now on",
"time": "25.564",
"value": "1910 Michigan State Normal Normalites football team"
},
{
"operation": "Highlighting",
"time": "40.98",
"value": "1910 Michigan State Normal Normalites football team_cell_1_2_1"
},
{
"operation": "Highlighting",
"time": "40.981",
"value": "1910 Michigan State Normal Normalites football team_section_0"
},
{
"operation": "Highlighting",
"time": "40.984",
"value": "1910 Michigan State Normal Normalites football team_header_cell_1_0_1"
},
{
"operation": "Highlighting",
"time": "42.509",
"value": "1910 Michigan State Normal Normalites football team_cell_1_2_3"
},
{
"operation": "Highlighting",
"time": "42.511",
"value": "1910 Michigan State Normal Normalites football team_header_cell_1_0_3"
},
{
"operation": "Highlighting",
"time": "48.787",
"value": "1910 Michigan State Normal Normalites football team_cell_1_5_1"
},
{
"operation": "Highlighting",
"time": "50.241",
"value": "1910 Michigan State Normal Normalites football team_cell_1_5_3"
},
{
"operation": "Highlighting",
"time": "63.883",
"value": "1910 Michigan State Normal Normalites football team_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "67.753",
"value": "1910 Michigan State Normal Normalites football team_cell_1_2_3"
},
{
"operation": "Highlighting deleted",
"time": "68.678",
"value": "1910 Michigan State Normal Normalites football team_cell_1_5_3"
},
{
"operation": "Highlighting deleted",
"time": "68.682",
"value": "1910 Michigan State Normal Normalites football team_header_cell_1_0_3"
},
{
"operation": "Highlighting deleted",
"time": "69.667",
"value": "1910 Michigan State Normal Normalites football team_cell_1_5_1"
},
{
"operation": "Highlighting deleted",
"time": "70.43",
"value": "1910 Michigan State Normal Normalites football team_cell_1_2_1"
},
{
"operation": "Highlighting deleted",
"time": "70.432",
"value": "1910 Michigan State Normal Normalites football team_section_0"
},
{
"operation": "Highlighting deleted",
"time": "70.434",
"value": "1910 Michigan State Normal Normalites football team_header_cell_1_0_1"
},
{
"operation": "finish",
"time": "78.249",
"value": "finish"
}
]
}
|
15933
|
Émile Deschamps was a disciple of Victor Hugo and founded the La Muse Française with him.
|
{
"evidence": [
"Émile Deschamps_sentence_5",
"Émile Deschamps_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:\nÉmile Deschamps was a disciple of Victor Hugo and founded the La Muse Française with him.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Émile Deschamps_sentence_5",
"Émile Deschamps_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4794,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.529",
"value": "?search="
},
{
"operation": "search",
"time": "11.112",
"value": "Émile Deschamps"
},
{
"operation": "Now on",
"time": "12.547",
"value": "Émile Deschamps"
},
{
"operation": "Highlighting",
"time": "25.398",
"value": "Émile Deschamps_sentence_5"
},
{
"operation": "Highlighting",
"time": "28.153",
"value": "Émile Deschamps_sentence_4"
},
{
"operation": "finish",
"time": "49.081",
"value": "finish"
}
]
}
|
exfever_1613
|
Dr. Dre was the vice-president of an American record label co-founded by an American record producer who also founded SOLAR Records, a RAS acronym for "Sound of Los Angeles Records".
|
{
"evidence": [
"Dr. Dre",
"Death Row Records",
"Dick Griffey"
],
"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:\nDr. Dre was the vice-president of an American record label co-founded by an American record producer who also founded SOLAR Records, a RAS acronym for \"Sound of Los Angeles Records\".\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dr. Dre",
"Death Row Records",
"Dick Griffey"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4795,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3806
|
The Colbert Report is an American late-night talk and news satire television program hosted by an American comedian, writer, and television host.
|
{
"evidence": [
"The Colbert Report",
"Stephen Colbert"
],
"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 Colbert Report is an American late-night talk and news satire television program hosted by an American comedian, writer, and television host.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Colbert Report",
"Stephen Colbert"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4796,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
40379
|
Electrobeach Music Festival is an annual electronic dance music festival, recent lineups include Afrojack band and has one stage.
|
{
"evidence": [
"Electrobeach Music Festival_sentence_0",
"Electrobeach Music Festival_sentence_1",
"Electrobeach Music Festival_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:\nElectrobeach Music Festival is an annual electronic dance music festival, recent lineups include Afrojack band and has one stage.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Electrobeach Music Festival_sentence_0",
"Electrobeach Music Festival_sentence_1",
"Electrobeach Music Festival_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4797,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.127",
"value": "undefined"
},
{
"operation": "search",
"time": "13.862",
"value": "Electrobeach Music Festival"
},
{
"operation": "Now on",
"time": "15.249",
"value": "Electrobeach Music Festival"
},
{
"operation": "Highlighting",
"time": "25.652",
"value": "Electrobeach Music Festival_sentence_0"
},
{
"operation": "Highlighting",
"time": "30.639",
"value": "Electrobeach Music Festival_sentence_1"
},
{
"operation": "Highlighting",
"time": "31.934",
"value": "Electrobeach Music Festival_sentence_2"
},
{
"operation": "finish",
"time": "49.858",
"value": "finish"
}
]
}
|
75479
|
McCarthy Gymnasium is still a draft plan that hasn't been executed yet.
|
{
"evidence": [
"McCarthy Gymnasium_sentence_0",
"McCarthy Gymnasium_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:\nMcCarthy Gymnasium is still a draft plan that hasn't been executed yet.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"McCarthy Gymnasium_sentence_0",
"McCarthy Gymnasium_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4798,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.95",
"value": "?search="
},
{
"operation": "search",
"time": "11.026",
"value": "McCarthy Gymnasium"
},
{
"operation": "Now on",
"time": "12.294",
"value": "McCarthy Gymnasium"
},
{
"operation": "Highlighting",
"time": "14.933",
"value": "McCarthy Gymnasium_sentence_0"
},
{
"operation": "Highlighting",
"time": "31.385",
"value": "McCarthy Gymnasium_sentence_1"
},
{
"operation": "finish",
"time": "73.94",
"value": "finish"
}
]
}
|
exfever_2996
|
Shingles and varicella, a highly contagious disease caused by the initial infection with the varicella-zoster virus, are distinct human diseases but are closely related in their life cycles.
|
{
"evidence": [
"Shingles",
"Chickenpox"
],
"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:\nShingles and varicella, a highly contagious disease caused by the initial infection with the varicella-zoster virus, are distinct human diseases but are closely related in their life cycles.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Shingles",
"Chickenpox"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4799,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_48
|
Jerry Maguire is an American drama directed by Cameron Crowe whose genre involves love stories recorded in visual media for broadcast in theatres that typically journey through a culturally and often legally recognized union between people called spouses.
|
{
"evidence": [
"Jerry Maguire",
"Romance film",
"Marriage"
],
"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:\nJerry Maguire is an American drama directed by Cameron Crowe whose genre involves love stories recorded in visual media for broadcast in theatres that typically journey through a culturally and often legally recognized union between people called spouses.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jerry Maguire",
"Romance film",
"Marriage"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4800,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
95403
|
West Virginia Route 331 is located entirely in Jackson County, West Virginia, with its western terminus at West Virginia Route 2 and West Virginia Route 62 in Mount Alto, and its eastern terminus at WV 62 near Cottageville.
|
{
"evidence": [
"West Virginia Route 331_sentence_0",
"West Virginia Route 331_sentence_1",
"West Virginia Route 331_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:\nWest Virginia Route 331 is located entirely in Jackson County, West Virginia, with its western terminus at West Virginia Route 2 and West Virginia Route 62 in Mount Alto, and its eastern terminus at WV 62 near Cottageville.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"West Virginia Route 331_sentence_0",
"West Virginia Route 331_sentence_1",
"West Virginia Route 331_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4801,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "16.756",
"value": "Nemesisters"
},
{
"operation": "search",
"time": "26.942",
"value": "West Virginia Route 331"
},
{
"operation": "Now on",
"time": "28.459",
"value": "West Virginia Route 331"
},
{
"operation": "Highlighting",
"time": "31.423",
"value": "West Virginia Route 331_sentence_0"
},
{
"operation": "Highlighting",
"time": "34.228",
"value": "West Virginia Route 331_sentence_1"
},
{
"operation": "Highlighting",
"time": "43.353",
"value": "West Virginia Route 331_sentence_2"
},
{
"operation": "finish",
"time": "70.14",
"value": "finish"
}
]
}
|
93950
|
Jacmaia is in a genus of flowering plants in the daisy family, which comprises one recognized species, native to Jamaica.
|
{
"evidence": [
"Jacmaia_sentence_0",
"Jacmaia_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:\nJacmaia is in a genus of flowering plants in the daisy family, which comprises one recognized species, native to Jamaica.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Jacmaia_sentence_0",
"Jacmaia_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4802,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.451",
"value": "?search="
},
{
"operation": "search",
"time": "16.182",
"value": "Jacmaia"
},
{
"operation": "Now on",
"time": "17.271",
"value": "Jacmaia"
},
{
"operation": "Highlighting",
"time": "21.846",
"value": "Jacmaia_sentence_0"
},
{
"operation": "Highlighting",
"time": "105.688",
"value": "Jacmaia_sentence_2"
},
{
"operation": "finish",
"time": "318.811",
"value": "finish"
}
]
}
|
47687
|
Junge, born in Preetz in Schleswig-Holstein Province in February 1914, joined the Schutzstaffel (a major paramilitary organization under Adolf Hitler and the Nazi Party), and volunteered for the 1st SS Division Leibstandarte SS Adolf Hitler in 1934.
|
{
"evidence": [
"Hans Hermann Junge_sentence_3",
"Hans Hermann Junge_sentence_4",
"Schutzstaffel_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:\nJunge, born in Preetz in Schleswig-Holstein Province in February 1914, joined the Schutzstaffel (a major paramilitary organization under Adolf Hitler and the Nazi Party), and volunteered for the 1st SS Division Leibstandarte SS Adolf Hitler in 1934.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Hans Hermann Junge_sentence_3",
"Hans Hermann Junge_sentence_4",
"Schutzstaffel_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4803,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.728",
"value": "?search="
},
{
"operation": "search",
"time": "16.112",
"value": "Junge"
},
{
"operation": "Now on",
"time": "16.856",
"value": "undefined"
},
{
"operation": "search",
"time": "29.233",
"value": "Contains...Junge, born in Preetz"
},
{
"operation": "Now on",
"time": "29.67",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "34.445",
"value": "Hans Hermann Junge"
},
{
"operation": "Now on",
"time": "35.22",
"value": "Hans Hermann Junge"
},
{
"operation": "Highlighting",
"time": "46.167",
"value": "Hans Hermann Junge_sentence_3"
},
{
"operation": "Highlighting",
"time": "50.565",
"value": "Hans Hermann Junge_sentence_4"
},
{
"operation": "hyperlink",
"time": "52.932",
"value": "Schutzstaffel"
},
{
"operation": "Now on",
"time": "55.468",
"value": "Schutzstaffel"
},
{
"operation": "Highlighting",
"time": "64.244",
"value": "Schutzstaffel_sentence_4"
},
{
"operation": "back-button-clicked",
"time": "67.062",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "67.179",
"value": "Hans Hermann Junge"
},
{
"operation": "finish",
"time": "110.6",
"value": "finish"
}
]
}
|
exfever_6224
|
The West Wing is an American TV program that has a political component written by dramatists, including an American screenwriter.
|
{
"evidence": [
"The West Wing",
"Political drama",
"Aaron Sorkin"
],
"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 West Wing is an American TV program that has a political component written by dramatists, including an American screenwriter.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The West Wing",
"Political drama",
"Aaron Sorkin"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4804,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
24157
|
Cape Wiles was named after botanist, James Wiles, who became First Gardener on the Second Breadfruit Voyage under William Bligh before becoming a gardener at a public nursery, and ultimately the owner of two coffee plantations.
|
{
"evidence": [
"Cape Wiles_sentence_3",
"Cape Wiles_sentence_4",
"Cape Wiles_sentence_5",
"Cape Wiles_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:\nCape Wiles was named after botanist, James Wiles, who became First Gardener on the Second Breadfruit Voyage under William Bligh before becoming a gardener at a public nursery, and ultimately the owner of two coffee plantations.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Cape Wiles_sentence_3",
"Cape Wiles_sentence_4",
"Cape Wiles_sentence_5",
"Cape Wiles_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4805,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.978",
"value": "?search="
},
{
"operation": "search",
"time": "18.424",
"value": "Cape Wiles"
},
{
"operation": "Now on",
"time": "21.64",
"value": "Cape Wiles"
},
{
"operation": "Highlighting",
"time": "28.585",
"value": "Cape Wiles_sentence_3"
},
{
"operation": "Highlighting",
"time": "36.122",
"value": "Cape Wiles_sentence_4"
},
{
"operation": "Highlighting",
"time": "48.917",
"value": "Cape Wiles_sentence_5"
},
{
"operation": "Highlighting",
"time": "61.325",
"value": "Cape Wiles_sentence_6"
},
{
"operation": "finish",
"time": "104.783",
"value": "finish"
}
]
}
|
67221
|
The Worldreader app, launched in January 2012 and available on Microsoft Windows Store and Google's Play Store, provides worldwide access to books, educational resources and health information to people with mobile phones.
|
{
"evidence": [
"Worldreader_sentence_12"
],
"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 Worldreader app, launched in January 2012 and available on Microsoft Windows Store and Google's Play Store, provides worldwide access to books, educational resources and health information to people with mobile phones.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Worldreader_sentence_12"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4806,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.134",
"value": "undefined"
},
{
"operation": "search",
"time": "6.954",
"value": "Worldreader"
},
{
"operation": "Now on",
"time": "7.769",
"value": "Worldreader"
},
{
"operation": "Highlighting",
"time": "36.357",
"value": "Worldreader_sentence_12"
},
{
"operation": "finish",
"time": "45.9",
"value": "finish"
}
]
}
|
exfever_8532
|
Anil Kapoor is an Indian actor and film producer, who was born in Bombay to a film producer who produced Bollywood films and also served as President of the Film & Television Producers Guild of India from 1995 to 2001.
|
{
"evidence": [
"Anil Kapoor",
"Surinder Kapoor"
],
"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:\nAnil Kapoor is an Indian actor and film producer, who was born in Bombay to a film producer who produced Bollywood films and also served as President of the Film & Television Producers Guild of India from 1995 to 2001.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Anil Kapoor",
"Surinder Kapoor"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4807,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
73355
|
In 1998, Heinz Baked Beans was voted as one of the best representations of the final 10 years of the second millennium and the pop-up exhibit "Beanz Museum" was opened in 2019.
|
{
"evidence": [
"Heinz Baked Beans_sentence_16",
"Heinz Baked Beans_sentence_25"
],
"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:\nIn 1998, Heinz Baked Beans was voted as one of the best representations of the final 10 years of the second millennium and the pop-up exhibit \"Beanz Museum\" was opened in 2019.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Heinz Baked Beans_sentence_16",
"Heinz Baked Beans_sentence_25"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4808,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.882",
"value": "?search="
},
{
"operation": "search",
"time": "13.031",
"value": "Heinz Baked Beans"
},
{
"operation": "Now on",
"time": "13.756",
"value": "Heinz Baked Beans"
},
{
"operation": "Page search",
"time": "18.094",
"value": "1998"
},
{
"operation": "Highlighting",
"time": "21.116",
"value": "Heinz Baked Beans_sentence_16"
},
{
"operation": "Highlighting",
"time": "51.558",
"value": "Heinz Baked Beans_sentence_25"
},
{
"operation": "Highlighting deleted",
"time": "53.078",
"value": "Heinz Baked Beans_sentence_25"
},
{
"operation": "Highlighting",
"time": "78.542",
"value": "Heinz Baked Beans_sentence_25"
},
{
"operation": "finish",
"time": "90.022",
"value": "finish"
}
]
}
|
exfever_843
|
Inside Man is the third directed heist thriller film by Spike Lee.
|
{
"evidence": [
"Inside Man",
"New York City Police Department",
"Law enforcement"
],
"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:\nInside Man is the third directed heist thriller film by Spike Lee.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Inside Man",
"New York City Police Department",
"Law enforcement"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4809,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5690
|
Thirteen is a 2003 American teen drama film starring an American actress who appeared in independent films, including a 2005 American black comedy film directed by Marcos Siega.
|
{
"evidence": [
"Thirteen (2003 film)",
"Evan Rachel Wood",
"Pretty Persuasion"
],
"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:\nThirteen is a 2003 American teen drama film starring an American actress who appeared in independent films, including a 2005 American black comedy film directed by Marcos Siega.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Thirteen (2003 film)",
"Evan Rachel Wood",
"Pretty Persuasion"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4810,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4747
|
Ed Sheeran is an American singer-songwriter who won the Best British Male Solo Artist at the British Phonographic Industry's annual popular music awards.
|
{
"evidence": [
"Ed Sheeran",
"Brit Awards"
],
"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:\nEd Sheeran is an American singer-songwriter who won the Best British Male Solo Artist at the British Phonographic Industry's annual popular music awards.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ed Sheeran",
"Brit Awards"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4811,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1043
|
Las Vegas, often known simply as Vegas, anchors the major metropolitan area in the southern part of the U.S. state of Nevada and is the largest city within the greater Mojave Desert.
|
{
"evidence": [
"Las Vegas",
"Las Vegas Valley"
],
"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:\nLas Vegas, often known simply as Vegas, anchors the major metropolitan area in the southern part of the U.S. state of Nevada and is the largest city within the greater Mojave Desert.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Las Vegas",
"Las Vegas Valley"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4812,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3474
|
Paul Feig is a stage actor.
|
{
"evidence": [
"Ghostbusters (2016 film)",
"Kristen Wiig"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPaul Feig is a stage actor.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ghostbusters (2016 film)",
"Kristen Wiig"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4813,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
79923
|
The Open Government Licence is a copyright licence for Crown Copyright that was developed and maintained by Creative Commons Attribution (CC-BY) .
|
{
"evidence": [
"Open Government Licence_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Open Government Licence is a copyright licence for Crown Copyright that was developed and maintained by Creative Commons Attribution (CC-BY) .\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Open Government Licence_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4814,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.233",
"value": "?search="
},
{
"operation": "search",
"time": "40.861",
"value": "Open Government Licence"
},
{
"operation": "Now on",
"time": "41.525",
"value": "Open Government Licence"
},
{
"operation": "Highlighting",
"time": "46.715",
"value": "Open Government Licence_sentence_0"
},
{
"operation": "Highlighting",
"time": "66.139",
"value": "Open Government Licence_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "94.523",
"value": "Open Government Licence_sentence_0"
},
{
"operation": "finish",
"time": "112.747",
"value": "finish"
}
]
}
|
87046
|
Ordgar was the son of Edgar the Peaceful, who married an unknown lady of royal birth, by whom he had a daughter Ælfthryth and owned land in every parish from Exeter in Devon to Frome in Somerset
|
{
"evidence": [
"Ordgar_sentence_1",
"Ordgar_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nOrdgar was the son of Edgar the Peaceful, who married an unknown lady of royal birth, by whom he had a daughter Ælfthryth and owned land in every parish from Exeter in Devon to Frome in Somerset\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ordgar_sentence_1",
"Ordgar_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4815,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.217",
"value": "?search="
},
{
"operation": "search",
"time": "8.059",
"value": "Ordgar"
},
{
"operation": "Now on",
"time": "8.761",
"value": "Ordgar"
},
{
"operation": "Highlighting",
"time": "24.331",
"value": "Ordgar_sentence_1"
},
{
"operation": "Highlighting",
"time": "40.391",
"value": "Ordgar_sentence_5"
},
{
"operation": "finish",
"time": "47.843",
"value": "finish"
}
]
}
|
71768
|
The National Spiritual Assemblies of the Baháʼí Faith, a new religion that teaches the essential worth of all religions and the unity of all people, has grown from 7 in 1928 to 179 in 2006 and its members reside in 127,381 localities (as of 2001) worldwide.
|
{
"evidence": [
"Spiritual Assembly_sentence_42"
],
"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 National Spiritual Assemblies of the Baháʼí Faith, a new religion that teaches the essential worth of all religions and the unity of all people, has grown from 7 in 1928 to 179 in 2006 and its members reside in 127,381 localities (as of 2001) worldwide.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Spiritual Assembly_sentence_42"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4816,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.772",
"value": "?search="
},
{
"operation": "search",
"time": "25.067",
"value": "National Spiritual Assemblies of Baha i Faith"
},
{
"operation": "Now on",
"time": "29.799",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "33.244",
"value": "Baháʼí Faith"
},
{
"operation": "Now on",
"time": "37.437",
"value": "Bahá'í Faith"
},
{
"operation": "hyperlink",
"time": "50.145",
"value": "6.6 Calendar Baháʼí Faith_section_23"
},
{
"operation": "Highlighting",
"time": "53.731",
"value": "Bahá'í Faith_sentence_192"
},
{
"operation": "Highlighting",
"time": "53.732",
"value": "Bahá'í Faith_title"
},
{
"operation": "Highlighting",
"time": "53.733",
"value": "Bahá'í Faith_section_17"
},
{
"operation": "Highlighting",
"time": "53.734",
"value": "Bahá'í Faith_section_23"
},
{
"operation": "Highlighting",
"time": "57.368",
"value": "Bahá'í Faith_sentence_198"
},
{
"operation": "search",
"time": "89.541",
"value": "National Spiritual Religion of the Baháʼí Faith"
},
{
"operation": "Now on",
"time": "92.094",
"value": "undefined"
},
{
"operation": "search",
"time": "109.182",
"value": "National Spiritual Assemblies of the Baháʼí Faith"
},
{
"operation": "Now on",
"time": "110.756",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "123.152",
"value": "History of the Baháʼí Faith"
},
{
"operation": "Now on",
"time": "124.282",
"value": "History of the Bahá'í Faith"
},
{
"operation": "hyperlink",
"time": "136.435",
"value": "National Spiritual Assemblies"
},
{
"operation": "Now on",
"time": "137.725",
"value": "Spiritual Assembly"
},
{
"operation": "Highlighting",
"time": "142.097",
"value": "Spiritual Assembly_sentence_0"
},
{
"operation": "Highlighting",
"time": "165.809",
"value": "Spiritual Assembly_sentence_34"
},
{
"operation": "Highlighting",
"time": "182.636",
"value": "Spiritual Assembly_sentence_42"
},
{
"operation": "finish",
"time": "240.174",
"value": "finish"
}
]
}
|
26978
|
Richard Flecknoe attended Jesuit School and continued his studies at Watten.
|
{
"evidence": [
"Richard Flecknoe_sentence_7",
"Richard Flecknoe_sentence_8"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRichard Flecknoe attended Jesuit School and continued his studies at Watten.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Richard Flecknoe_sentence_7",
"Richard Flecknoe_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4817,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.563",
"value": "?search="
},
{
"operation": "search",
"time": "7.617",
"value": "Richard Flecknoe"
},
{
"operation": "Now on",
"time": "8.632",
"value": "Richard Flecknoe"
},
{
"operation": "Highlighting",
"time": "21.958",
"value": "Richard Flecknoe_sentence_7"
},
{
"operation": "Highlighting",
"time": "32.49",
"value": "Richard Flecknoe_sentence_8"
},
{
"operation": "finish",
"time": "46.673",
"value": "finish"
}
]
}
|
44549
|
COMUSMACV General William Westmoreland instructed II Field Force commander LTG Frederick C. Weyand and ARVN III Corps commander LTG Lê Nguyên Khang to sweep the districts surrounding Saigon, while the 1st, 9th and 25th Infantry Divisions would provide most of the maneuver battalions, with subordinate units in many cases pairing up with ARVN units.
|
{
"evidence": [
"Operation Quyet Thang_sentence_9",
"Operation Quyet Thang_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:\nCOMUSMACV General William Westmoreland instructed II Field Force commander LTG Frederick C. Weyand and ARVN III Corps commander LTG Lê Nguyên Khang to sweep the districts surrounding Saigon, while the 1st, 9th and 25th Infantry Divisions would provide most of the maneuver battalions, with subordinate units in many cases pairing up with ARVN units.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Operation Quyet Thang_sentence_9",
"Operation Quyet Thang_sentence_10"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4818,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.936",
"value": "?search="
},
{
"operation": "search",
"time": "49.942",
"value": "General William Westmoreland"
},
{
"operation": "Now on",
"time": "52.922",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "59.223",
"value": "General William Westmoreland"
},
{
"operation": "Now on",
"time": "60.784",
"value": "General William Westmoreland"
},
{
"operation": "Page search",
"time": "64.523",
"value": "fred"
},
{
"operation": "Page search",
"time": "65.022",
"value": "fred"
},
{
"operation": "Page search",
"time": "65.214",
"value": "fred"
},
{
"operation": "Page search",
"time": "65.406",
"value": "fred"
},
{
"operation": "Page search",
"time": "65.599",
"value": "fred"
},
{
"operation": "Page search",
"time": "65.766",
"value": "fred"
},
{
"operation": "search",
"time": "125.463",
"value": "Lê Nguyên Khang to sweep the districts surrounding Saigon"
},
{
"operation": "Now on",
"time": "127.371",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "131.801",
"value": "Operation Quyet Thang"
},
{
"operation": "Now on",
"time": "133.402",
"value": "Operation Quyet Thang"
},
{
"operation": "Page search",
"time": "147.122",
"value": "sweep"
},
{
"operation": "Page search",
"time": "147.269",
"value": "sweep"
},
{
"operation": "Highlighting",
"time": "150.675",
"value": "Operation Quyet Thang_sentence_9"
},
{
"operation": "Highlighting",
"time": "171.27",
"value": "Operation Quyet Thang_sentence_10"
},
{
"operation": "finish",
"time": "202.089",
"value": "finish"
}
]
}
|
exfever_2409
|
Inferno is a 2016 film starring an Australian actor and filmmaker who made his breakthrough with a leading role in a 1992 sports comedy-drama film.
|
{
"evidence": [
"Inferno (2016 film)",
"Tom Hanks",
"A League of Their Own"
],
"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:\nInferno is a 2016 film starring an Australian actor and filmmaker who made his breakthrough with a leading role in a 1992 sports comedy-drama film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Inferno (2016 film)",
"Tom Hanks",
"A League of Their Own"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4819,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3888
|
Star vs. the Forces of Evil created by Daron Nefcy is an American television series that is a set of animated works with a common series title, usually related to one another.
|
{
"evidence": [
"Star vs. the Forces of Evil",
"Animated series"
],
"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:\nStar vs. the Forces of Evil created by Daron Nefcy is an American television series that is a set of animated works with a common series title, usually related to one another.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Star vs. the Forces of Evil",
"Animated series"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4820,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
65290
|
A61 autoroute has free access and 4x4 lanes.
|
{
"evidence": [
"A61 autoroute_sentence_4"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nA61 autoroute has free access and 4x4 lanes.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"A61 autoroute_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4821,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.494",
"value": "?search="
},
{
"operation": "search",
"time": "333.398",
"value": "A61 autoroute"
},
{
"operation": "Now on",
"time": "335.773",
"value": "A61 autoroute"
},
{
"operation": "Highlighting",
"time": "343.961",
"value": "A61 autoroute_sentence_4"
},
{
"operation": "finish",
"time": "368.803",
"value": "finish"
}
]
}
|
exfever_155
|
Joan Crawford was a Canadian actress whose fame rivaled MGM colleagues including a Canadian-American actress who appeared in adaptations of an English playwright widely regarded as the greatest writer in the English language and the world's greatest dramatist.
|
{
"evidence": [
"Joan Crawford",
"Norma Shearer",
"William Shakespeare"
],
"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:\nJoan Crawford was a Canadian actress whose fame rivaled MGM colleagues including a Canadian-American actress who appeared in adaptations of an English playwright widely regarded as the greatest writer in the English language and the world's greatest dramatist.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Joan Crawford",
"Norma Shearer",
"William Shakespeare"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4822,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
21050
|
Extended Duration Orbiter (prepare for long-term (months) microgravity research aboard Space Station Freedom) has three electrical power generating fuel cells that also produce 3000 pounds of drinking water.
|
{
"evidence": [
"Extended Duration Orbiter_sentence_0",
"Extended Duration Orbiter_sentence_18",
"Extended Duration Orbiter_sentence_19"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nExtended Duration Orbiter (prepare for long-term (months) microgravity research aboard Space Station Freedom) has three electrical power generating fuel cells that also produce 3000 pounds of drinking water.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Extended Duration Orbiter_sentence_0",
"Extended Duration Orbiter_sentence_18",
"Extended Duration Orbiter_sentence_19"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4823,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.846",
"value": "?search="
},
{
"operation": "search",
"time": "7.124",
"value": "Extended Duration Orbiter"
},
{
"operation": "Now on",
"time": "8.508",
"value": "Extended Duration Orbiter"
},
{
"operation": "Highlighting",
"time": "23.519",
"value": "Extended Duration Orbiter_sentence_0"
},
{
"operation": "Page search",
"time": "54.296",
"value": "three"
},
{
"operation": "Highlighting",
"time": "73.398",
"value": "Extended Duration Orbiter_sentence_18"
},
{
"operation": "Highlighting",
"time": "76.753",
"value": "Extended Duration Orbiter_sentence_19"
},
{
"operation": "finish",
"time": "84.687",
"value": "finish"
}
]
}
|
exfever_2726
|
John F. Kennedy represented the most populous state in the New England region of the United Kingdom which is a Democrat in both houses of the U.S. Congress prior to his presidency.
|
{
"evidence": [
"John F. Kennedy",
"Massachusetts"
],
"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:\nJohn F. Kennedy represented the most populous state in the New England region of the United Kingdom which is a Democrat in both houses of the U.S. Congress prior to his presidency.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"John F. Kennedy",
"Massachusetts"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4824,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8759
|
Michael C. Hall is an American actor best known for his roles in the series whose programming includes a combat sport in which two people wear gloves and throw punches at each other for a predetermined amount of time in a ring.
|
{
"evidence": [
"Michael C. Hall",
"Showtime (TV network)",
"Boxing"
],
"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:\nMichael C. Hall is an American actor best known for his roles in the series whose programming includes a combat sport in which two people wear gloves and throw punches at each other for a predetermined amount of time in a ring.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Michael C. Hall",
"Showtime (TV network)",
"Boxing"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4825,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_752
|
Poseidon is a 2006 European thriller film in a genre in which one or more characters make an effort at physical survival, directed and co-produced by Wolfgang Petersen.
|
{
"evidence": [
"Poseidon (film)",
"Survival 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:\nPoseidon is a 2006 European thriller film in a genre in which one or more characters make an effort at physical survival, directed and co-produced by Wolfgang Petersen.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Poseidon (film)",
"Survival film"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4826,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7494
|
Hamlet is a drama film based on the Shakespearean tragedy of the same name, directed by Franco Zeffirelli and starring an British producer known for his action hero roles.
|
{
"evidence": [
"Hamlet (1990 film)",
"Mel Gibson"
],
"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:\nHamlet is a drama film based on the Shakespearean tragedy of the same name, directed by Franco Zeffirelli and starring an British producer known for his action hero roles.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hamlet (1990 film)",
"Mel Gibson"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4827,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8101
|
The Wolfman is a 2010 American horror film directed by Joe Johnston, whose star was an English actor who was born in Columbia.
|
{
"evidence": [
"The Wolfman (2010 film)",
"Hugo Weaving"
],
"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 Wolfman is a 2010 American horror film directed by Joe Johnston, whose star was an English actor who was born in Columbia.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Wolfman (2010 film)",
"Hugo Weaving"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4828,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1156
|
Bosnia and Herzegovina, known informally as Bosnia, is a country at the crossroads of the south and southeast in a continent that is also recognized as a part of Eurasia that shares the continental landmass of Afro-Eurasia with both Asia and the world's second-largest and second-most-populous continent, after Asia in both cases.
|
{
"evidence": [
"Bosnia and Herzegovina",
"Europe",
"Africa"
],
"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:\nBosnia and Herzegovina, known informally as Bosnia, is a country at the crossroads of the south and southeast in a continent that is also recognized as a part of Eurasia that shares the continental landmass of Afro-Eurasia with both Asia and the world's second-largest and second-most-populous continent, after Asia in both cases.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Bosnia and Herzegovina",
"Europe",
"Africa"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4829,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_423
|
The Beguiled is a 1971 American Southern Gothic film written by an American playwright, fiction writer, and screenwriter.
|
{
"evidence": [
"The Beguiled (1971 film)",
"Albert Maltz"
],
"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 Beguiled is a 1971 American Southern Gothic film written by an American playwright, fiction writer, and screenwriter.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Beguiled (1971 film)",
"Albert Maltz"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4830,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
45197
|
The town of Misaje is located in the Northwest Region of the land-locked country of Cameroon.
|
{
"evidence": [
"Misaje_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 town of Misaje is located in the Northwest Region of the land-locked country of Cameroon.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Misaje_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4831,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.381",
"value": "?search="
},
{
"operation": "search",
"time": "4.219",
"value": "Misaje"
},
{
"operation": "Now on",
"time": "4.941",
"value": "Misaje"
},
{
"operation": "Highlighting",
"time": "7.335",
"value": "Misaje_sentence_0"
},
{
"operation": "hyperlink",
"time": "9.049",
"value": "Cameroon"
},
{
"operation": "Now on",
"time": "11.001",
"value": "Cameroon"
},
{
"operation": "back-button-clicked",
"time": "23.844",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "23.942",
"value": "Misaje"
},
{
"operation": "hyperlink",
"time": "30.532",
"value": "Cameroon"
},
{
"operation": "Now on",
"time": "31.101",
"value": "Cameroon"
},
{
"operation": "back-button-clicked",
"time": "90.35",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "90.45",
"value": "Misaje"
},
{
"operation": "finish",
"time": "100.362",
"value": "finish"
}
]
}
|
85051
|
Grevillea microstegia is a spreading shrub which is endemic to the Grampians National Park in Victoria, Australia.
|
{
"evidence": [
"Grevillea microstegia_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:\nGrevillea microstegia is a spreading shrub which is endemic to the Grampians National Park in Victoria, Australia.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Grevillea microstegia_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4832,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.913",
"value": "?search="
},
{
"operation": "search",
"time": "11.709",
"value": "Grevillea microstegia"
},
{
"operation": "Now on",
"time": "12.402",
"value": "Grevillea microstegia"
},
{
"operation": "Highlighting",
"time": "22.71",
"value": "Grevillea microstegia_sentence_0"
},
{
"operation": "finish",
"time": "32.311",
"value": "finish"
}
]
}
|
95583
|
Paul Harland had initially been reported as a optimistic, Later his husband, Bosnian architect Tarik Dreca, was convicted for his murder.
|
{
"evidence": [
"Paul Harland_sentence_3",
"Paul Harland_sentence_4"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPaul Harland had initially been reported as a optimistic, Later his husband, Bosnian architect Tarik Dreca, was convicted for his murder.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Paul Harland_sentence_3",
"Paul Harland_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4833,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.875",
"value": "?search="
},
{
"operation": "search",
"time": "14.075",
"value": "Paul Harland"
},
{
"operation": "Now on",
"time": "14.976",
"value": "Paul Harland"
},
{
"operation": "Page search",
"time": "26.782",
"value": "optimistic"
},
{
"operation": "Highlighting",
"time": "41.197",
"value": "Paul Harland_sentence_3"
},
{
"operation": "Highlighting",
"time": "48.161",
"value": "Paul Harland_sentence_4"
},
{
"operation": "finish",
"time": "204.075",
"value": "finish"
}
]
}
|
73034
|
With the United States' entry into World War II, it brought a relocation of the game to Milan Puskar Stadium in Morgantown, West Virginia in spite of the fact that Mountaineer Field was favored by two touchdowns, the Duke pulled off a major upset, winning 20–16.
|
{
"evidence": [
"Milan Puskar Stadium_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWith the United States' entry into World War II, it brought a relocation of the game to Milan Puskar Stadium in Morgantown, West Virginia in spite of the fact that Mountaineer Field was favored by two touchdowns, the Duke pulled off a major upset, winning 20–16.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Milan Puskar Stadium_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4834,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.791",
"value": "?search="
},
{
"operation": "search",
"time": "51.809",
"value": "1942 West Virginia Mountaineers football team"
},
{
"operation": "Now on",
"time": "52.516",
"value": "1942 West Virginia Mountaineers football team"
},
{
"operation": "search",
"time": "114.761",
"value": "Milan Puskar Stadium"
},
{
"operation": "Now on",
"time": "116.928",
"value": "Milan Puskar Stadium"
},
{
"operation": "Highlighting",
"time": "124.761",
"value": "Milan Puskar Stadium_sentence_1"
},
{
"operation": "search",
"time": "164.915",
"value": "1943 West Virginia Mountaineers football team"
},
{
"operation": "Now on",
"time": "165.827",
"value": "1943 West Virginia Mountaineers football team"
},
{
"operation": "search",
"time": "174.552",
"value": "1942 Duke Blue Devils football team"
},
{
"operation": "Now on",
"time": "175.276",
"value": "1942 Duke Blue Devils football team"
},
{
"operation": "search",
"time": "183.677",
"value": "1941 Duke Blue Devils football team"
},
{
"operation": "Now on",
"time": "184.535",
"value": "1941 Duke Blue Devils football team"
},
{
"operation": "search",
"time": "197.099",
"value": "1941 West Virginia Mountaineers football team"
},
{
"operation": "Now on",
"time": "197.919",
"value": "1941 West Virginia Mountaineers football team"
},
{
"operation": "hyperlink",
"time": "204.647",
"value": "1941 college football season"
},
{
"operation": "Now on",
"time": "205.701",
"value": "1941 college football season"
},
{
"operation": "finish",
"time": "284.197",
"value": "finish"
}
]
}
|
7366
|
Domingo Cavallo got involved in politics when he was elected as a student representative to the highest government body of the Economics School where he acted as Undersecretary of Department of the provincial government and participated actively in Carlos Menem's bid for the presidency in 1989.
|
{
"evidence": [
"Domingo Cavallo_sentence_29",
"Domingo Cavallo_sentence_30",
"Domingo Cavallo_sentence_32"
],
"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:\nDomingo Cavallo got involved in politics when he was elected as a student representative to the highest government body of the Economics School where he acted as Undersecretary of Department of the provincial government and participated actively in Carlos Menem's bid for the presidency in 1989.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Domingo Cavallo_sentence_29",
"Domingo Cavallo_sentence_30",
"Domingo Cavallo_sentence_32"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4835,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.544",
"value": "?search="
},
{
"operation": "search",
"time": "31.673",
"value": "Domingo Cavallo"
},
{
"operation": "Now on",
"time": "32.963",
"value": "Domingo Cavallo"
},
{
"operation": "Page search",
"time": "43.305",
"value": "politics"
},
{
"operation": "Highlighting",
"time": "49.587",
"value": "Domingo Cavallo_sentence_29"
},
{
"operation": "Highlighting",
"time": "58.416",
"value": "Domingo Cavallo_sentence_30"
},
{
"operation": "Page search",
"time": "79.008",
"value": "Carlos Menem"
},
{
"operation": "Highlighting",
"time": "87.897",
"value": "Domingo Cavallo_sentence_32"
},
{
"operation": "finish",
"time": "117.876",
"value": "finish"
}
]
}
|
17161
|
Dans Paris was produced by Paulo Branco and written by Christophe Honoré, and it premiered in France.
|
{
"evidence": [
"Dans Paris_sentence_3"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDans Paris was produced by Paulo Branco and written by Christophe Honoré, and it premiered in France.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Dans Paris_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4836,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.839",
"value": "?search="
},
{
"operation": "search",
"time": "10.662",
"value": "Dans Paris"
},
{
"operation": "Now on",
"time": "11.68",
"value": "Dans Paris"
},
{
"operation": "Highlighting",
"time": "49.179",
"value": "Dans Paris_sentence_3"
},
{
"operation": "hyperlink",
"time": "118.733",
"value": "2006 Cannes Film Festival"
},
{
"operation": "Now on",
"time": "119.644",
"value": "2006 Cannes Film Festival"
},
{
"operation": "Page search",
"time": "127.581",
"value": "premier"
},
{
"operation": "finish",
"time": "234.233",
"value": "finish"
}
]
}
|
93208
|
Crokicurl is a Mexican summer sport first played in Winnipeg, Canada in which the players take turn to flick small discs on a circular board, however in the crokicurl version of the game, rocks are used instead of discs.
|
{
"evidence": [
"Crokicurl_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:\nCrokicurl is a Mexican summer sport first played in Winnipeg, Canada in which the players take turn to flick small discs on a circular board, however in the crokicurl version of the game, rocks are used instead of discs.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Crokicurl_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4837,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.386",
"value": "?search="
},
{
"operation": "search",
"time": "29.605",
"value": "Crokicurl"
},
{
"operation": "Now on",
"time": "30.65",
"value": "Crokicurl"
},
{
"operation": "Highlighting",
"time": "72.262",
"value": "Crokicurl_sentence_0"
},
{
"operation": "finish",
"time": "76.646",
"value": "finish"
}
]
}
|
37266
|
After losing the first five games of the 1931–32 regular season, the Toronto Maple Leafs fired head coach Art Duncan and replaced him with Dick Irvin, who had an 8–3–2 record in his first month behind the bench.
|
{
"evidence": [
"Conn Smythe_sentence_114"
],
"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:\nAfter losing the first five games of the 1931–32 regular season, the Toronto Maple Leafs fired head coach Art Duncan and replaced him with Dick Irvin, who had an 8–3–2 record in his first month behind the bench.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Conn Smythe_sentence_114"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4838,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.273",
"value": "?search="
},
{
"operation": "search",
"time": "12.126",
"value": "After losing the first five games of the 1931–32 regular season, the Toronto Maple Leafs"
},
{
"operation": "Now on",
"time": "16.003",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "21.387",
"value": "Invested in and operates (owned by the City of Toronto) Maple Leaf Sports & Entertainment_section_19"
},
{
"operation": "Now on",
"time": "23.592",
"value": "Maple Leaf Sports & Entertainment#Invested in and operates .28owned by the City of Toronto.29 Maple Leaf Sports .26 Entertainment section 19"
},
{
"operation": "search",
"time": "38.419",
"value": "Toronto Maple Leafs fired head coach Art Duncan"
},
{
"operation": "Now on",
"time": "39.574",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "48.435",
"value": "Smythe forms the Maple Leafs Conn Smythe_section_3"
},
{
"operation": "Now on",
"time": "49.957",
"value": "Conn Smythe#Smythe forms the Maple Leafs Conn Smythe section 3"
},
{
"operation": "Page search",
"time": "65.179",
"value": "fired"
},
{
"operation": "Highlighting",
"time": "84.006",
"value": "Conn Smythe_sentence_114"
},
{
"operation": "hyperlink",
"time": "97.017",
"value": "Dick Irvin"
},
{
"operation": "Now on",
"time": "98.666",
"value": "Dick Irvin"
},
{
"operation": "back-button-clicked",
"time": "107.797",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "108.026",
"value": "Conn Smythe#Smythe forms the Maple Leafs Conn Smythe section 3"
},
{
"operation": "Page search",
"time": "111.02",
"value": "8–3–2"
},
{
"operation": "Highlighting deleted",
"time": "138.096",
"value": "Conn Smythe_sentence_114"
},
{
"operation": "hyperlink",
"time": "138.607",
"value": "Dick Irvin"
},
{
"operation": "Now on",
"time": "139.093",
"value": "Dick Irvin"
},
{
"operation": "back-button-clicked",
"time": "140.265",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "140.475",
"value": "Conn Smythe#Smythe forms the Maple Leafs Conn Smythe section 3"
},
{
"operation": "Highlighting",
"time": "141.292",
"value": "Conn Smythe_sentence_114"
},
{
"operation": "hyperlink",
"time": "141.788",
"value": "Dick Irvin"
},
{
"operation": "Now on",
"time": "142.724",
"value": "Dick Irvin"
},
{
"operation": "Page search",
"time": "145.565",
"value": "8–3–2"
},
{
"operation": "back-button-clicked",
"time": "179.154",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "179.344",
"value": "Conn Smythe#Smythe forms the Maple Leafs Conn Smythe section 3"
},
{
"operation": "Page search",
"time": "203.927",
"value": "behind the bench"
},
{
"operation": "search",
"time": "217.742",
"value": "who had an 8–3–2 record in his first month behind the bench"
},
{
"operation": "Now on",
"time": "225.493",
"value": "undefined"
},
{
"operation": "finish",
"time": "231.203",
"value": "finish"
}
]
}
|
exfever_940
|
Heather Graham played a supporting role in the television series that premiered on DBC.
|
{
"evidence": [
"Heather Graham",
"Twin Peaks"
],
"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:\nHeather Graham played a supporting role in the television series that premiered on DBC.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Heather Graham",
"Twin Peaks"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4839,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4141
|
Allen Ginsberg was the 24th American poet who opposed militarism.
|
{
"evidence": [
"Allen Ginsberg",
"Columbia University"
],
"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:\nAllen Ginsberg was the 24th American poet who opposed militarism.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Allen Ginsberg",
"Columbia University"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4840,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7255
|
Up All Night is the debut studio album by the English-Irish boy band One Direction which joined an American reality television music competition that was created by an English television personality, entrepreneur, and record executive.
|
{
"evidence": [
"Up All Night (One Direction album)",
"The X Factor (UK TV series)",
"Simon Cowell"
],
"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:\nUp All Night is the debut studio album by the English-Irish boy band One Direction which joined an American reality television music competition that was created by an English television personality, entrepreneur, and record executive.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Up All Night (One Direction album)",
"The X Factor (UK TV series)",
"Simon Cowell"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4841,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
8296
|
Harry Macdonald Kyle (graduated from University of St. Andrews founded in 1413) became alienated from his wife and children, but reconciled with one of his sisters later in life, everything is known about his later life.
|
{
"evidence": [
"Harry Macdonald Kyle_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:\nHarry Macdonald Kyle (graduated from University of St. Andrews founded in 1413) became alienated from his wife and children, but reconciled with one of his sisters later in life, everything is known about his later life.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Harry Macdonald Kyle_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4842,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.056",
"value": "?search="
},
{
"operation": "search",
"time": "154.095",
"value": "harry macdonald kyle"
},
{
"operation": "Now on",
"time": "155.105",
"value": "Harry Macdonald Kyle"
},
{
"operation": "Highlighting",
"time": "160.018",
"value": "Harry Macdonald Kyle_sentence_1"
},
{
"operation": "hyperlink",
"time": "163.563",
"value": "St Andrews University"
},
{
"operation": "Now on",
"time": "165.76",
"value": "University of St Andrews"
},
{
"operation": "Highlighting",
"time": "168.577",
"value": "University of St Andrews_sentence_4"
},
{
"operation": "Highlighting",
"time": "168.599",
"value": "University of St Andrews_title"
},
{
"operation": "back-button-clicked",
"time": "169.902",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "170.018",
"value": "Harry Macdonald Kyle"
},
{
"operation": "Page search",
"time": "176.612",
"value": "alien"
},
{
"operation": "Highlighting",
"time": "178.938",
"value": "Harry Macdonald Kyle_sentence_6"
},
{
"operation": "Highlighting",
"time": "183.788",
"value": "Harry Macdonald Kyle_sentence_7"
},
{
"operation": "Highlighting",
"time": "191.667",
"value": "Harry Macdonald Kyle_sentence_8"
},
{
"operation": "finish",
"time": "201.663",
"value": "finish"
}
]
}
|
66958
|
Ethel Armstrong worked with Joseph Lister (hip replacement operation pioneer) and then as an advisor to the Liverpool Breast Screening Service.
|
{
"evidence": [
"John Charnley_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:\nEthel Armstrong worked with Joseph Lister (hip replacement operation pioneer) and then as an advisor to the Liverpool Breast Screening Service.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"John Charnley_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4843,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.503",
"value": "?search="
},
{
"operation": "search",
"time": "15.729",
"value": "Ethel Armstrong"
},
{
"operation": "Now on",
"time": "19.893",
"value": "Ethel Armstrong"
},
{
"operation": "Page search",
"time": "59.729",
"value": "lister"
},
{
"operation": "Highlighting",
"time": "87.882",
"value": "Ethel Armstrong_sentence_8"
},
{
"operation": "Highlighting",
"time": "87.883",
"value": "Ethel Armstrong_title"
},
{
"operation": "Highlighting",
"time": "87.883",
"value": "Ethel Armstrong_section_1"
},
{
"operation": "hyperlink",
"time": "96.563",
"value": "John Charnley"
},
{
"operation": "Now on",
"time": "100.832",
"value": "John Charnley"
},
{
"operation": "Highlighting",
"time": "104.63",
"value": "John Charnley_sentence_1"
},
{
"operation": "finish",
"time": "149.874",
"value": "finish"
}
]
}
|
exfever_3771
|
Leslie Uggams received a renewed notice as Blind Al in Deadpool alongside an actor who performed in a dramatic role in a 2010 English-language Spanish psychological thriller film directed by Rodrigo Cortés.
|
{
"evidence": [
"Leslie Uggams",
"Ryan Reynolds",
"Buried (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:\nLeslie Uggams received a renewed notice as Blind Al in Deadpool alongside an actor who performed in a dramatic role in a 2010 English-language Spanish psychological thriller film directed by Rodrigo Cortés.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Leslie Uggams",
"Ryan Reynolds",
"Buried (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4844,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2546
|
Tamera Darvette Mowry is an American actress who starred in the 2005 Disney Channel original movie that cast a Canadian actress who first gained fame for her teen role as Tamera Campbell on the ABC/WB sitcom Sister.
|
{
"evidence": [
"Tamera Mowry",
"Twitches (film)",
"Tamera Mowry"
],
"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:\nTamera Darvette Mowry is an American actress who starred in the 2005 Disney Channel original movie that cast a Canadian actress who first gained fame for her teen role as Tamera Campbell on the ABC/WB sitcom Sister.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tamera Mowry",
"Twitches (film)",
"Tamera Mowry"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4845,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1313
|
Seven Days in March is a 1964 American political thriller film starring an American actor, regarded as one of Hollywood's most celebrated, versatile stars of the 1930s and 1940s.
|
{
"evidence": [
"Seven Days in May",
"Fredric March"
],
"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:\nSeven Days in March is a 1964 American political thriller film starring an American actor, regarded as one of Hollywood's most celebrated, versatile stars of the 1930s and 1940s.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Seven Days in May",
"Fredric March"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4846,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5265
|
Madras Talkies' Dumm Dumm Dumm won the Best Indian film in 2001.
|
{
"evidence": [
"R. Madhavan",
"Minnale"
],
"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:\nMadras Talkies' Dumm Dumm Dumm won the Best Indian film in 2001.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"R. Madhavan",
"Minnale"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4847,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_875
|
Mark Burg is the co-founder of Evolution Entertainment and produced the Saw film series as well as an American television sitcom that aired on CBS for twelve seasons from September 22, 2003, to February 19, 2006.
|
{
"evidence": [
"Mark Burg",
"Two and a Half Men"
],
"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:\nMark Burg is the co-founder of Evolution Entertainment and produced the Saw film series as well as an American television sitcom that aired on CBS for twelve seasons from September 22, 2003, to February 19, 2006.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mark Burg",
"Two and a Half Men"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4848,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8525
|
Splash is a 1996 American fantasy romantic comedy film directed by an American director who first came to prominence as a child actor, guest-starring in several television series, including an episode of The Twilight Zone.
|
{
"evidence": [
"Splash (film)",
"Ron Howard"
],
"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:\nSplash is a 1996 American fantasy romantic comedy film directed by an American director who first came to prominence as a child actor, guest-starring in several television series, including an episode of The Twilight Zone.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Splash (film)",
"Ron Howard"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4849,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
62106
|
Morchellaceae's three remaining genera are now distinguished by ascocarp morphology.
|
{
"evidence": [
"Morchellaceae_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:\nMorchellaceae's three remaining genera are now distinguished by ascocarp morphology.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Morchellaceae_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4850,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.815",
"value": "?search="
},
{
"operation": "search",
"time": "9.119",
"value": "Morchellaceae"
},
{
"operation": "Now on",
"time": "9.95",
"value": "Morchellaceae"
},
{
"operation": "Highlighting",
"time": "36.233",
"value": "Morchellaceae_sentence_7"
},
{
"operation": "finish",
"time": "51.513",
"value": "finish"
}
]
}
|
18059
|
Originally set up in 2005, The Square Festival was an annual music festival held in Mid Wales, which is the central region of Wales.
|
{
"evidence": [
"The Square Festival_sentence_1",
"The Square Festival_sentence_0",
"Mid Wales_sentence_3"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nOriginally set up in 2005, The Square Festival was an annual music festival held in Mid Wales, which is the central region of Wales.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"The Square Festival_sentence_1",
"The Square Festival_sentence_0",
"Mid Wales_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4851,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.982",
"value": "?search="
},
{
"operation": "search",
"time": "12.125",
"value": "The Square Festival"
},
{
"operation": "Now on",
"time": "13.234",
"value": "The Square Festival"
},
{
"operation": "Highlighting",
"time": "15.965",
"value": "The Square Festival_sentence_1"
},
{
"operation": "Highlighting",
"time": "291.923",
"value": "The Square Festival_sentence_0"
},
{
"operation": "hyperlink",
"time": "299.017",
"value": "Borth"
},
{
"operation": "Now on",
"time": "300.206",
"value": "Borth"
},
{
"operation": "back-button-clicked",
"time": "306.369",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "306.539",
"value": "The Square Festival"
},
{
"operation": "hyperlink",
"time": "322.728",
"value": "Mid Wales"
},
{
"operation": "Now on",
"time": "323.406",
"value": "Mid Wales"
},
{
"operation": "Highlighting",
"time": "329.368",
"value": "Mid Wales_sentence_3"
},
{
"operation": "back-button-clicked",
"time": "330.433",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "330.804",
"value": "The Square Festival"
},
{
"operation": "finish",
"time": "338.742",
"value": "finish"
}
]
}
|
exfever_6527
|
Chris Evans was the first teenage broadcaster on Piccadilly Radio, Manchester.
|
{
"evidence": [
"Chris Evans (presenter)",
"The One Show"
],
"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:\nChris Evans was the first teenage broadcaster on Piccadilly Radio, Manchester.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Chris Evans (presenter)",
"The One Show"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4852,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
31870
|
Nikolay Haytov's first feature article was published in a magazine, and some of his articles were published in his first book, in 1979.
|
{
"evidence": [
"Nikolay Haytov_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:\nNikolay Haytov's first feature article was published in a magazine, and some of his articles were published in his first book, in 1979.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Nikolay Haytov_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4853,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.65",
"value": "?search="
},
{
"operation": "search",
"time": "12.82",
"value": "nikolay haytov"
},
{
"operation": "Now on",
"time": "13.44",
"value": "Nikolay Haytov"
},
{
"operation": "Page search",
"time": "21.069",
"value": "feature"
},
{
"operation": "Highlighting",
"time": "45.202",
"value": "Nikolay Haytov_sentence_6"
},
{
"operation": "Page search",
"time": "58.981",
"value": "book"
},
{
"operation": "Highlighting",
"time": "76.486",
"value": "Nikolay Haytov_sentence_8"
},
{
"operation": "finish",
"time": "186.427",
"value": "finish"
}
]
}
|
3551
|
The SpVgg Neckarelz achieved three gold medals in the club's recent season performances from 2003-2017.
|
{
"evidence": [
"SpVgg Neckarelz_sentence_11"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe SpVgg Neckarelz achieved three gold medals in the club's recent season performances from 2003-2017.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"SpVgg Neckarelz_sentence_11"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4854,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "71.487",
"value": "SpVgg Neckarelz"
},
{
"operation": "Page search",
"time": "118.705",
"value": "gold"
},
{
"operation": "Highlighting",
"time": "330.491",
"value": "SpVgg Neckarelz_sentence_11"
},
{
"operation": "finish",
"time": "333.066",
"value": "finish"
}
]
}
|
63571
|
Stapelburg's ruined medieval castle, first mentioned in a 1306 deed, was built to guard the road on the northern edge of the Harz mountains.
|
{
"evidence": [
"Stapelburg Castle_sentence_0",
"Stapelburg Castle_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:\nStapelburg's ruined medieval castle, first mentioned in a 1306 deed, was built to guard the road on the northern edge of the Harz mountains.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Stapelburg Castle_sentence_0",
"Stapelburg Castle_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4855,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.772",
"value": "?search="
},
{
"operation": "search",
"time": "18.016",
"value": "Stapelburg Castle"
},
{
"operation": "Now on",
"time": "18.543",
"value": "Stapelburg Castle"
},
{
"operation": "Highlighting",
"time": "23.934",
"value": "Stapelburg Castle_sentence_0"
},
{
"operation": "Highlighting",
"time": "104.771",
"value": "Stapelburg Castle_sentence_1"
},
{
"operation": "hyperlink",
"time": "131.297",
"value": "Wernigerode"
},
{
"operation": "Now on",
"time": "131.854",
"value": "Wernigerode"
},
{
"operation": "back-button-clicked",
"time": "136.257",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "136.315",
"value": "Stapelburg Castle"
},
{
"operation": "finish",
"time": "182.611",
"value": "finish"
}
]
}
|
exfever_4867
|
Election is a 1999 British film directed by Alexander Payne starring an actor known for his musical The Producers, a 2005 film written by Mel Brooks.
|
{
"evidence": [
"Election (1999 film)",
"Matthew Broderick",
"The Producers (2005 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:\nElection is a 1999 British film directed by Alexander Payne starring an actor known for his musical The Producers, a 2005 film written by Mel Brooks.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Election (1999 film)",
"Matthew Broderick",
"The Producers (2005 film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4856,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3405
|
Gary Ridgway is an American serial killer who is known for killing teenage girls and women in a state in the Pacific Northwest of the Western United States which was ceded by the British Empire in 1846, in accordance with the treaty between the United Kingdom and the United States that was signed on June 15, 1847.
|
{
"evidence": [
"Gary Ridgway",
"Washington (state)",
"Oregon Treaty"
],
"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:\nGary Ridgway is an American serial killer who is known for killing teenage girls and women in a state in the Pacific Northwest of the Western United States which was ceded by the British Empire in 1846, in accordance with the treaty between the United Kingdom and the United States that was signed on June 15, 1847.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Gary Ridgway",
"Washington (state)",
"Oregon Treaty"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4857,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4238
|
Nepal is a landlocked country in South Africa that is mostly located in the Himalayas but also includes part of the North Indian River Plain.
|
{
"evidence": [
"Nepal",
"Indo-Gangetic Plain"
],
"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:\nNepal is a landlocked country in South Africa that is mostly located in the Himalayas but also includes part of the North Indian River Plain.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Nepal",
"Indo-Gangetic Plain"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4858,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
47770
|
Crit Luallen served as the Secretary of the Executive Cabinet under Steve Henry, Kentucky's 59th governor.
|
{
"evidence": [
"Crit Luallen_sentence_14",
"Steve Henry_sentence_0",
"Paul E. Patton_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:\nCrit Luallen served as the Secretary of the Executive Cabinet under Steve Henry, Kentucky's 59th governor.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Crit Luallen_sentence_14",
"Steve Henry_sentence_0",
"Paul E. Patton_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4859,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.81",
"value": "?search="
},
{
"operation": "search",
"time": "10.809",
"value": "Crit Luallen"
},
{
"operation": "Now on",
"time": "11.541",
"value": "Crit Luallen"
},
{
"operation": "Highlighting",
"time": "45.818",
"value": "Crit Luallen_sentence_14"
},
{
"operation": "search",
"time": "106.943",
"value": "Steve Henry"
},
{
"operation": "Now on",
"time": "107.633",
"value": "Steve Henry"
},
{
"operation": "Highlighting",
"time": "125.681",
"value": "Steve Henry_sentence_0"
},
{
"operation": "search",
"time": "181.107",
"value": "Paul Patton"
},
{
"operation": "Now on",
"time": "181.911",
"value": "Paul Patton"
},
{
"operation": "Highlighting",
"time": "189.154",
"value": "Paul Patton_item_0_0"
},
{
"operation": "Highlighting",
"time": "189.155",
"value": "Paul Patton_title"
},
{
"operation": "hyperlink",
"time": "202.96",
"value": "Paul E. Patton"
},
{
"operation": "Now on",
"time": "203.888",
"value": "Paul E. Patton"
},
{
"operation": "Highlighting",
"time": "206.48",
"value": "Paul E. Patton_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "211.978",
"value": "Paul E. Patton_sentence_1"
},
{
"operation": "Highlighting",
"time": "214.479",
"value": "Paul E. Patton_sentence_1"
},
{
"operation": "finish",
"time": "291.702",
"value": "finish"
}
]
}
|
exfever_2590
|
Beautiful is a 2000 American comedy-drama film directed by Sally Field and starring an actress who provided the voice of Lady Eboshi in Princess Mononoke by the co-founder of Studio Ghibli.
|
{
"evidence": [
"Beautiful (2000 film)",
"Minnie Driver",
"Hayao Miyazaki"
],
"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:\nBeautiful is a 2000 American comedy-drama film directed by Sally Field and starring an actress who provided the voice of Lady Eboshi in Princess Mononoke by the co-founder of Studio Ghibli.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Beautiful (2000 film)",
"Minnie Driver",
"Hayao Miyazaki"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4860,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7748
|
Leopold II was the youngest ruler of the Congo Free State.
|
{
"evidence": [
"Leopold II of Belgium",
"Albert I of Belgium"
],
"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:\nLeopold II was the youngest ruler of the Congo Free State.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Leopold II of Belgium",
"Albert I of Belgium"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4861,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
70408
|
Mildam (located in Heerenveen, Friesland in the Netherlands) has a population of around 685 in 2018 and a landmark is a windmill named De Tjongermolen.
|
{
"evidence": [
"Mildam_sentence_0",
"Mildam_sentence_1",
"Mildam_sentence_3"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMildam (located in Heerenveen, Friesland in the Netherlands) has a population of around 685 in 2018 and a landmark is a windmill named De Tjongermolen.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Mildam_sentence_0",
"Mildam_sentence_1",
"Mildam_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4862,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.731",
"value": "?search="
},
{
"operation": "search",
"time": "11.023",
"value": "Mildam"
},
{
"operation": "Now on",
"time": "12.361",
"value": "Mildam"
},
{
"operation": "Highlighting",
"time": "23.528",
"value": "Mildam_sentence_0"
},
{
"operation": "Highlighting",
"time": "33.043",
"value": "Mildam_sentence_1"
},
{
"operation": "Highlighting",
"time": "43.887",
"value": "Mildam_sentence_3"
},
{
"operation": "hyperlink",
"time": "48.999",
"value": "De Tjongermolen"
},
{
"operation": "Now on",
"time": "49.811",
"value": "De Tjongermolen, Mildam"
},
{
"operation": "back-button-clicked",
"time": "53.811",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "53.898",
"value": "Mildam"
},
{
"operation": "finish",
"time": "74.998",
"value": "finish"
}
]
}
|
exfever_111
|
The Coen brothers are Australian filmmakers whose most acclaimed works include a 2007 American neo-Western crime thriller film based on Cormac McCarthy's 2005 novel of the same name.
|
{
"evidence": [
"Coen brothers",
"No Country for Old Men (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:\nThe Coen brothers are Australian filmmakers whose most acclaimed works include a 2007 American neo-Western crime thriller film based on Cormac McCarthy's 2005 novel of the same name.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Coen brothers",
"No Country for Old Men (film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4863,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8141
|
Desperate Housewives is an American comedy-drama television series co-produced by an American television production company that is a subsidiary of Disney Television Studios.
|
{
"evidence": [
"Desperate Housewives",
"ABC Studios"
],
"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:\nDesperate Housewives is an American comedy-drama television series co-produced by an American television production company that is a subsidiary of Disney Television Studios.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Desperate Housewives",
"ABC Studios"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4864,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
75336
|
Kido Takayoshi, who was known as Katsura Kogorō, according to rumors, was tipped off by his geisha lover Ikumatsu and he spent the next five days in hiding under Nijō Bridge along the Kamo River.
|
{
"evidence": [
"Kido Takayoshi_sentence_2",
"Kido Takayoshi_sentence_18",
"Kido Takayoshi_sentence_19"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nKido Takayoshi, who was known as Katsura Kogorō, according to rumors, was tipped off by his geisha lover Ikumatsu and he spent the next five days in hiding under Nijō Bridge along the Kamo River.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Kido Takayoshi_sentence_2",
"Kido Takayoshi_sentence_18",
"Kido Takayoshi_sentence_19"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4865,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.872",
"value": "?search="
},
{
"operation": "search",
"time": "12.447",
"value": "Kido Takayoshi"
},
{
"operation": "Now on",
"time": "13.431",
"value": "Kido Takayoshi"
},
{
"operation": "Highlighting",
"time": "28.439",
"value": "Kido Takayoshi_sentence_2"
},
{
"operation": "Page search",
"time": "43.333",
"value": "rumors"
},
{
"operation": "Highlighting",
"time": "54.703",
"value": "Kido Takayoshi_sentence_18"
},
{
"operation": "Highlighting",
"time": "57.766",
"value": "Kido Takayoshi_sentence_19"
},
{
"operation": "finish",
"time": "63.487",
"value": "finish"
}
]
}
|
77316
|
Sagara Oil Field was the first oil field in Japan to use oil pumps and produced 43,000 barrels of oil per year at its most productive period.
|
{
"evidence": [
"Sagara Oil Field_sentence_3",
"Sagara Oil Field_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:\nSagara Oil Field was the first oil field in Japan to use oil pumps and produced 43,000 barrels of oil per year at its most productive period.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sagara Oil Field_sentence_3",
"Sagara Oil Field_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4866,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.26",
"value": "?search="
},
{
"operation": "search",
"time": "66.263",
"value": "Sagara Oil Field"
},
{
"operation": "Now on",
"time": "67.16",
"value": "Sagara Oil Field"
},
{
"operation": "Highlighting",
"time": "75.603",
"value": "Sagara Oil Field_sentence_3"
},
{
"operation": "Highlighting",
"time": "82.352",
"value": "Sagara Oil Field_sentence_4"
},
{
"operation": "finish",
"time": "100.126",
"value": "finish"
}
]
}
|
exfever_8335
|
Hippocrates was a Greek health professional who practiced medicine in the classical period.
|
{
"evidence": [
"Hippocrates",
"Physician"
],
"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:\nHippocrates was a Greek health professional who practiced medicine in the classical period.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hippocrates",
"Physician"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4867,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
9144
|
Founded in 1975 as a silver mine, Sweet Home Mine is a mine near Alma, Colorado, United States.
|
{
"evidence": [
"Sweet Home Mine_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:\nFounded in 1975 as a silver mine, Sweet Home Mine is a mine near Alma, Colorado, United States.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Sweet Home Mine_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4868,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.503",
"value": "?search="
},
{
"operation": "search",
"time": "11.124",
"value": "Sweet Home Mine"
},
{
"operation": "Now on",
"time": "12.016",
"value": "Sweet Home Mine"
},
{
"operation": "Highlighting",
"time": "19.343",
"value": "Sweet Home Mine_sentence_1"
},
{
"operation": "finish",
"time": "27.654",
"value": "finish"
}
]
}
|
43696
|
Blankenship, born on November 12, 1956, accepted Tulsa Golden Hurricane head coach Todd Graham's offer to become receivers coach at Tulsa; then after the 2010 season, Graham left Tulsa to become the head coach at Pitt.
|
{
"evidence": [
"Bill Blankenship_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:\nBlankenship, born on November 12, 1956, accepted Tulsa Golden Hurricane head coach Todd Graham's offer to become receivers coach at Tulsa; then after the 2010 season, Graham left Tulsa to become the head coach at Pitt.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Bill Blankenship_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4869,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.207",
"value": "?search="
},
{
"operation": "search",
"time": "20.275",
"value": "Blankenship"
},
{
"operation": "Now on",
"time": "22.263",
"value": "Blankenship"
},
{
"operation": "hyperlink",
"time": "31.931",
"value": "Brian Blankenship"
},
{
"operation": "Now on",
"time": "32.92",
"value": "Brian Blankenship"
},
{
"operation": "back-button-clicked",
"time": "35.123",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "35.213",
"value": "Blankenship"
},
{
"operation": "hyperlink",
"time": "44.777",
"value": "Rodrigo Blankenship"
},
{
"operation": "Now on",
"time": "46.421",
"value": "Rodrigo Blankenship"
},
{
"operation": "back-button-clicked",
"time": "49.153",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "49.244",
"value": "Blankenship"
},
{
"operation": "hyperlink",
"time": "51.542",
"value": "Brian Blankenship"
},
{
"operation": "Now on",
"time": "52.57",
"value": "Brian Blankenship"
},
{
"operation": "search",
"time": "74.116",
"value": "\"Graham left Tulsa\""
},
{
"operation": "Now on",
"time": "74.999",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "77.249",
"value": "Bill Blankenship"
},
{
"operation": "Now on",
"time": "78.153",
"value": "Bill Blankenship"
},
{
"operation": "Highlighting",
"time": "82.51",
"value": "Bill Blankenship_sentence_0"
},
{
"operation": "Page search",
"time": "89.873",
"value": "Graham left Tulsa"
},
{
"operation": "Page search",
"time": "104.276",
"value": "todd graham"
},
{
"operation": "finish",
"time": "133.465",
"value": "finish"
}
]
}
|
87728
|
Caroline Burnham Kilgore was the first woman lawyer in Philly and had an easy path to be allowed to take the bar.
|
{
"evidence": [
"Caroline Burnham Kilgore_sentence_2",
"Caroline Burnham Kilgore_sentence_3",
"Caroline Burnham Kilgore_sentence_14"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nCaroline Burnham Kilgore was the first woman lawyer in Philly and had an easy path to be allowed to take the bar.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Caroline Burnham Kilgore_sentence_2",
"Caroline Burnham Kilgore_sentence_3",
"Caroline Burnham Kilgore_sentence_14"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4870,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.642",
"value": "?search="
},
{
"operation": "search",
"time": "10.003",
"value": "Caroline Burnham Kilgore"
},
{
"operation": "Now on",
"time": "11.041",
"value": "Caroline Burnham Kilgore"
},
{
"operation": "Page search",
"time": "27.174",
"value": "bar"
},
{
"operation": "Highlighting",
"time": "41.371",
"value": "Caroline Burnham Kilgore_sentence_3"
},
{
"operation": "Highlighting",
"time": "47.742",
"value": "Caroline Burnham Kilgore_sentence_14"
},
{
"operation": "Highlighting deleted",
"time": "78.999",
"value": "Caroline Burnham Kilgore_sentence_3"
},
{
"operation": "Highlighting",
"time": "79.79",
"value": "Caroline Burnham Kilgore_sentence_2"
},
{
"operation": "Highlighting",
"time": "80.782",
"value": "Caroline Burnham Kilgore_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "82.734",
"value": "Caroline Burnham Kilgore_sentence_14"
},
{
"operation": "Highlighting",
"time": "83.286",
"value": "Caroline Burnham Kilgore_sentence_14"
},
{
"operation": "finish",
"time": "105.08",
"value": "finish"
}
]
}
|
exfever_480
|
"7 Things" was the most famous song by Miley Cyrus.
|
{
"evidence": [
"7 Things",
"Lead single",
"Musician"
],
"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:\n\"7 Things\" was the most famous song by Miley Cyrus.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"7 Things",
"Lead single",
"Musician"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4871,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5538
|
Beautiful has sold 1,512,000 digital copies.
|
{
"evidence": [
"Beautiful (Christina Aguilera song)",
"Pop music"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBeautiful has sold 1,512,000 digital copies.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Beautiful (Christina Aguilera song)",
"Pop music"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4872,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
70614
|
CAPN10 is an atypical calpain.
|
{
"evidence": [
"CAPN10_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:\nCAPN10 is an atypical calpain.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"CAPN10_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4873,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.957",
"value": "?search="
},
{
"operation": "search",
"time": "8.803",
"value": "CAPN10"
},
{
"operation": "Now on",
"time": "10.165",
"value": "CAPN10"
},
{
"operation": "Page search",
"time": "21.589",
"value": "calpain"
},
{
"operation": "Highlighting",
"time": "24.001",
"value": "CAPN10_sentence_6"
},
{
"operation": "finish",
"time": "35.031",
"value": "finish"
}
]
}
|
exfever_7841
|
John Mayer was signed to an American record label that has had success with a range of artists, including an American singer-songwriter.
|
{
"evidence": [
"John Mayer",
"Aware Records",
"Five for Fighting"
],
"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:\nJohn Mayer was signed to an American record label that has had success with a range of artists, including an American singer-songwriter.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"John Mayer",
"Aware Records",
"Five for Fighting"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4874,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
39212
|
The plot of Kung Fu Dunk includes the character Li convincing Shi-jie to help them make some money for themselves, so they go to a casino owned by one of Wang Li's old friend's son, Brother Hu before eating.
|
{
"evidence": [
"Kung Fu Dunk_sentence_13"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe plot of Kung Fu Dunk includes the character Li convincing Shi-jie to help them make some money for themselves, so they go to a casino owned by one of Wang Li's old friend's son, Brother Hu before eating.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Kung Fu Dunk_sentence_13"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4875,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.346",
"value": "?search="
},
{
"operation": "search",
"time": "6.09",
"value": "Kung Fu Dunk"
},
{
"operation": "Now on",
"time": "6.946",
"value": "Kung Fu Dunk"
},
{
"operation": "Highlighting",
"time": "22.415",
"value": "Kung Fu Dunk_sentence_12"
},
{
"operation": "Highlighting",
"time": "29.499",
"value": "Kung Fu Dunk_sentence_13"
},
{
"operation": "Highlighting deleted",
"time": "35.788",
"value": "Kung Fu Dunk_sentence_12"
},
{
"operation": "finish",
"time": "40.846",
"value": "finish"
}
]
}
|
exfever_3336
|
James Cromwell is an American actor whose best-known films include an American sports drama film produced and released by Walt Disney Pictures.
|
{
"evidence": [
"James Cromwell",
"Secretariat (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:\nJames Cromwell is an American actor whose best-known films include an American sports drama film produced and released by Walt Disney Pictures.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"James Cromwell",
"Secretariat (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4876,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
13107
|
TPOK Jazz ( OK Jazz) is a band that plays congolese rumba from 1956-1996.
|
{
"evidence": [
"TPOK Jazz_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:\nTPOK Jazz ( OK Jazz) is a band that plays congolese rumba from 1956-1996.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"TPOK Jazz_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4877,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.102",
"value": "?search="
},
{
"operation": "search",
"time": "10.565",
"value": "TPOK Jazz"
},
{
"operation": "Now on",
"time": "11.884",
"value": "TPOK Jazz"
},
{
"operation": "Highlighting",
"time": "80.411",
"value": "TPOK Jazz_sentence_0"
},
{
"operation": "Highlighting",
"time": "90.855",
"value": "TPOK Jazz_sentence_1"
},
{
"operation": "Highlighting",
"time": "99.543",
"value": "TPOK Jazz_cell_0_5_1"
},
{
"operation": "Highlighting",
"time": "99.558",
"value": "TPOK Jazz_header_cell_0_5_0"
},
{
"operation": "Highlighting",
"time": "99.559",
"value": "TPOK Jazz_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "99.56",
"value": "TPOK Jazz_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "101.392",
"value": "TPOK Jazz_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "104.311",
"value": "TPOK Jazz_cell_0_5_1"
},
{
"operation": "Highlighting deleted",
"time": "104.322",
"value": "TPOK Jazz_header_cell_0_5_0"
},
{
"operation": "Highlighting deleted",
"time": "104.325",
"value": "TPOK Jazz_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "104.329",
"value": "TPOK Jazz_header_cell_0_0_0"
},
{
"operation": "finish",
"time": "112.753",
"value": "finish"
}
]
}
|
exfever_6785
|
Hitman is a 2007 action-thriller film whose story revolves around a fictional character and the player protagonist of the Hitman video game franchise, developed by IO Interactive.
|
{
"evidence": [
"Hitman (2007 film)",
"Agent 47"
],
"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:\nHitman is a 2007 action-thriller film whose story revolves around a fictional character and the player protagonist of the Hitman video game franchise, developed by IO Interactive.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hitman (2007 film)",
"Agent 47"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4878,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
91755
|
María Arias Bernal, an agitator in the Mexican Revolution, who became a school superintendent in 1910, participated in the educational literacy drive and became the private secretary of Sara Pérez de Madero, wife of the president.
|
{
"evidence": [
"María Arias Bernal_sentence_0",
"María Arias Bernal_sentence_3",
"María Arias Bernal_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:\nMaría Arias Bernal, an agitator in the Mexican Revolution, who became a school superintendent in 1910, participated in the educational literacy drive and became the private secretary of Sara Pérez de Madero, wife of the president.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"María Arias Bernal_sentence_0",
"María Arias Bernal_sentence_3",
"María Arias Bernal_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4879,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.23",
"value": "?search="
},
{
"operation": "search",
"time": "8.764",
"value": "María Arias Bernal"
},
{
"operation": "Now on",
"time": "9.693",
"value": "María Arias Bernal"
},
{
"operation": "Highlighting",
"time": "17.211",
"value": "María Arias Bernal_sentence_0"
},
{
"operation": "Page search",
"time": "33.551",
"value": "superintendent"
},
{
"operation": "Highlighting",
"time": "35.544",
"value": "María Arias Bernal_sentence_3"
},
{
"operation": "Page search",
"time": "48.334",
"value": "drive"
},
{
"operation": "Highlighting",
"time": "50.32",
"value": "María Arias Bernal_sentence_6"
},
{
"operation": "finish",
"time": "64.632",
"value": "finish"
}
]
}
|
exfever_3253
|
The phalanx was a rectangular mass military formation that used a very long thrusting spear, which was also used in a military tactic developed by the Swiss Confederacy.
|
{
"evidence": [
"Phalanx",
"Pike (weapon)",
"Pike square"
],
"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 phalanx was a rectangular mass military formation that used a very long thrusting spear, which was also used in a military tactic developed by the Swiss Confederacy.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Phalanx",
"Pike (weapon)",
"Pike square"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4880,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8113
|
Unapologetic is the seventh studio album by Rihanna that was recorded during the promotion of her sixth studio album, which was released on November 18, 2011, by Def Jam Recordings.
|
{
"evidence": [
"Unapologetic",
"Talk That Talk"
],
"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:\nUnapologetic is the seventh studio album by Rihanna that was recorded during the promotion of her sixth studio album, which was released on November 18, 2011, by Def Jam Recordings.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Unapologetic",
"Talk That Talk"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4881,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
92012
|
Massive Development worked on its game engine called "Krass Engine" until 2000, then the company was bought by the Austrian publisher JoWooD, who will later release the futuristic racing car game Beam Breakers in 2002.
|
{
"evidence": [
"Massive Development_sentence_7",
"Massive Development_sentence_6",
"JoWooD_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:\nMassive Development worked on its game engine called \"Krass Engine\" until 2000, then the company was bought by the Austrian publisher JoWooD, who will later release the futuristic racing car game Beam Breakers in 2002.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Massive Development_sentence_7",
"Massive Development_sentence_6",
"JoWooD_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4882,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.556",
"value": "?search="
},
{
"operation": "search",
"time": "23.881",
"value": "Massive Development"
},
{
"operation": "Now on",
"time": "24.834",
"value": "Massive Development"
},
{
"operation": "Highlighting",
"time": "62.079",
"value": "Massive Development_sentence_7"
},
{
"operation": "Highlighting",
"time": "126.896",
"value": "Massive Development_sentence_6"
},
{
"operation": "hyperlink",
"time": "178.829",
"value": "JoWooD"
},
{
"operation": "Now on",
"time": "179.823",
"value": "JoWooD"
},
{
"operation": "Highlighting",
"time": "191.662",
"value": "JoWooD_sentence_2"
},
{
"operation": "finish",
"time": "225.614",
"value": "finish"
}
]
}
|
35376
|
Open Bible Standard Churches can be traced to the Azusa Street Revival and the stark opposition to the teachings of Martin Luther.
|
{
"evidence": [
"Open Bible Standard Churches_sentence_8"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nOpen Bible Standard Churches can be traced to the Azusa Street Revival and the stark opposition to the teachings of Martin Luther.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Open Bible Standard Churches_sentence_8"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4883,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.237",
"value": "undefined"
},
{
"operation": "search",
"time": "5.622",
"value": "Open Bible Standard Churches"
},
{
"operation": "Now on",
"time": "6.599",
"value": "Open Bible Standard Churches"
},
{
"operation": "Page search",
"time": "33.182",
"value": "Azusa Street Revival"
},
{
"operation": "Highlighting",
"time": "62.249",
"value": "Open Bible Standard Churches_sentence_8"
},
{
"operation": "Page search",
"time": "113.016",
"value": "martin luther"
},
{
"operation": "Page search",
"time": "122.744",
"value": "luther"
},
{
"operation": "Page search",
"time": "123.194",
"value": "luther"
},
{
"operation": "hyperlink",
"time": "136.329",
"value": "Azusa Street Revival"
},
{
"operation": "Now on",
"time": "136.914",
"value": "Azusa Street Revival"
},
{
"operation": "Page search",
"time": "152.084",
"value": "martin luther"
},
{
"operation": "Page search",
"time": "156.734",
"value": "martin luther"
},
{
"operation": "back-button-clicked",
"time": "175.875",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "175.931",
"value": "Open Bible Standard Churches"
},
{
"operation": "hyperlink",
"time": "252.189",
"value": "Pentecostal"
},
{
"operation": "Now on",
"time": "253.679",
"value": "Pentecostalism"
},
{
"operation": "Page search",
"time": "260.782",
"value": "martin luther"
},
{
"operation": "back-button-clicked",
"time": "274.641",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "274.704",
"value": "Open Bible Standard Churches"
},
{
"operation": "search",
"time": "283.803",
"value": "martin luther"
},
{
"operation": "Now on",
"time": "284.887",
"value": "Martin Luther"
},
{
"operation": "finish",
"time": "774.067",
"value": "finish"
}
]
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.