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_7488
|
Caitlyn Marie Jenner is a retired Olympic gold medal-winning combined event in athletics that consists of ten track and field events, that were also played in the 1914 Summer Olympics, an international multi-sport event held in Stockholm, Sweden.
|
{
"evidence": [
"Caitlyn Jenner",
"Decathlon",
"1912 Summer Olympics"
],
"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:\nCaitlyn Marie Jenner is a retired Olympic gold medal-winning combined event in athletics that consists of ten track and field events, that were also played in the 1914 Summer Olympics, an international multi-sport event held in Stockholm, Sweden.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Caitlyn Jenner",
"Decathlon",
"1912 Summer Olympics"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5984,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
92376
|
Ivan Loiko was a World War I flying ace who served five years on Vaygach Island, an island in the Arctic Sea between the Pechora Sea and the Kara Sea.
|
{
"evidence": [
"Ivan Loiko_sentence_0",
"Ivan Loiko_sentence_7",
"Vaygach Island_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:\nIvan Loiko was a World War I flying ace who served five years on Vaygach Island, an island in the Arctic Sea between the Pechora Sea and the Kara Sea.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Ivan Loiko_sentence_0",
"Ivan Loiko_sentence_7",
"Vaygach Island_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5985,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.237",
"value": "?search="
},
{
"operation": "search",
"time": "11.156",
"value": "Ivan Loiko"
},
{
"operation": "Now on",
"time": "12.287",
"value": "Ivan Loiko"
},
{
"operation": "Highlighting",
"time": "17.315",
"value": "Ivan Loiko_sentence_0"
},
{
"operation": "Page search",
"time": "28.949",
"value": "Vaygach"
},
{
"operation": "Highlighting",
"time": "31.248",
"value": "Ivan Loiko_sentence_7"
},
{
"operation": "hyperlink",
"time": "32.777",
"value": "Vaygach Island"
},
{
"operation": "Now on",
"time": "33.471",
"value": "Vaygach Island"
},
{
"operation": "Highlighting",
"time": "37.376",
"value": "Vaygach Island_sentence_2"
},
{
"operation": "finish",
"time": "59.019",
"value": "finish"
}
]
}
|
exfever_7187
|
The Wars of the Roses were a series of civil wars fought over control of the English throne in the mid-to-late fifteenth century, fought between supporters of two rival cadet branches of the Royal house which originated from the lands of Anjou in France.
|
{
"evidence": [
"Wars of the Roses",
"House of Plantagenet"
],
"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 Wars of the Roses were a series of civil wars fought over control of the English throne in the mid-to-late fifteenth century, fought between supporters of two rival cadet branches of the Royal house which originated from the lands of Anjou in France.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Wars of the Roses",
"House of Plantagenet"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5986,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
43336
|
Dropshort Marsh is biological site in Bedfordshire is yet to be constructed.
|
{
"evidence": [
"Dropshort Marsh_sentence_0",
"Dropshort Marsh_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:\nDropshort Marsh is biological site in Bedfordshire is yet to be constructed.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Dropshort Marsh_sentence_0",
"Dropshort Marsh_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5987,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.861",
"value": "?search="
},
{
"operation": "search",
"time": "28.521",
"value": "Dropshort Marsh"
},
{
"operation": "Now on",
"time": "29.457",
"value": "Dropshort Marsh"
},
{
"operation": "Highlighting",
"time": "43.623",
"value": "Dropshort Marsh_sentence_0"
},
{
"operation": "Highlighting",
"time": "67.853",
"value": "Dropshort Marsh_sentence_1"
},
{
"operation": "finish",
"time": "71.814",
"value": "finish"
}
]
}
|
81760
|
In the 2008 United States House of Representatives elections in Michigan were held on November 4, 2008, to determine who would represent the state of Michigan in the United States House of Representatives; it had a make up of congressional delegation in 2008 which consisted of ten republicans; as a result, the delegation in 2009/10 was consisted of 20 Republicans.
|
{
"evidence": [
"2008 United States House of Representatives elections in Michigan_sentence_6",
"2008 United States House of Representatives elections in Michigan_sentence_7"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn the 2008 United States House of Representatives elections in Michigan were held on November 4, 2008, to determine who would represent the state of Michigan in the United States House of Representatives; it had a make up of congressional delegation in 2008 which consisted of ten republicans; as a result, the delegation in 2009/10 was consisted of 20 Republicans.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"2008 United States House of Representatives elections in Michigan_sentence_6",
"2008 United States House of Representatives elections in Michigan_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5988,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.889",
"value": "?search="
},
{
"operation": "search",
"time": "11.059",
"value": "2008 United States House of Representatives elections in Michigan"
},
{
"operation": "Now on",
"time": "13.589",
"value": "2008 United States House of Representatives elections in Michigan"
},
{
"operation": "Highlighting",
"time": "15.488",
"value": "2008 United States House of Representatives elections in Michigan_sentence_0"
},
{
"operation": "Highlighting",
"time": "27.917",
"value": "2008 United States House of Representatives elections in Michigan_sentence_6"
},
{
"operation": "Highlighting",
"time": "41.07",
"value": "2008 United States House of Representatives elections in Michigan_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "44.032",
"value": "2008 United States House of Representatives elections in Michigan_sentence_0"
},
{
"operation": "finish",
"time": "68.066",
"value": "finish"
}
]
}
|
53055
|
Bartłomiej Nowodworski was born 1552 in Krakow, in recognition of his contribution the school became known as Collegium Nowodworskiego and in 1898 the school relocated from the university area.
|
{
"evidence": [
"Bartłomiej Nowodworski High School_sentence_6",
"Bartłomiej Nowodworski High School_sentence_7",
"Bartłomiej Nowodworski High School_sentence_9"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBartłomiej Nowodworski was born 1552 in Krakow, in recognition of his contribution the school became known as Collegium Nowodworskiego and in 1898 the school relocated from the university area.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bartłomiej Nowodworski High School_sentence_6",
"Bartłomiej Nowodworski High School_sentence_7",
"Bartłomiej Nowodworski High School_sentence_9"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5989,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.729",
"value": "?search="
},
{
"operation": "search",
"time": "4.938",
"value": "Bartłomiej Nowodworski"
},
{
"operation": "Now on",
"time": "6.919",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "9.722",
"value": "Agata Kornhauser-Duda"
},
{
"operation": "Now on",
"time": "14.03",
"value": "Bartłomiej Nowodworski High School"
},
{
"operation": "Highlighting",
"time": "27.878",
"value": "Bartłomiej Nowodworski High School_sentence_6"
},
{
"operation": "Highlighting",
"time": "56.705",
"value": "Bartłomiej Nowodworski High School_sentence_7"
},
{
"operation": "Highlighting",
"time": "78.245",
"value": "Bartłomiej Nowodworski High School_sentence_9"
},
{
"operation": "finish",
"time": "130.17",
"value": "finish"
}
]
}
|
exfever_4087
|
Hell's Angels is a best-selling book by Hunter Stockton Thompson.
|
{
"evidence": [
"Hunter S. Thompson",
"Counterculture"
],
"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:\nHell's Angels is a best-selling book by Hunter Stockton Thompson.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hunter S. Thompson",
"Counterculture"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5990,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
52031
|
Luis Antonio Pavez Contreras, born on January 23, 1988 in Santiago, Chile, is currently playing for Unión Española (a professional Chilean football based in Independencia, Santiago de Chile).
|
{
"evidence": [
"Luis Pavez Contreras_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:\nLuis Antonio Pavez Contreras, born on January 23, 1988 in Santiago, Chile, is currently playing for Unión Española (a professional Chilean football based in Independencia, Santiago de Chile).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Luis Pavez Contreras_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5991,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.341",
"value": "?search="
},
{
"operation": "search",
"time": "11.457",
"value": "Contains...Luis Antonio Pavez Contreras"
},
{
"operation": "Now on",
"time": "13.089",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "15.646",
"value": "Luis Pavez Contreras"
},
{
"operation": "Now on",
"time": "16.338",
"value": "Luis Pavez Contreras"
},
{
"operation": "Highlighting",
"time": "28.554",
"value": "Luis Pavez Contreras_sentence_2"
},
{
"operation": "finish",
"time": "36.894",
"value": "finish"
}
]
}
|
94028
|
Texas SH 242 connects the Woodlands and 2.41 sq mile 89 ft above sea level, Patton Village has six major intersections including I-69 and I-45.
|
{
"evidence": [
"List of Farm to Market Roads in Texas (200–299)_sentence_222",
"List of Farm to Market Roads in Texas (200–299)_sentence_223"
],
"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:\nTexas SH 242 connects the Woodlands and 2.41 sq mile 89 ft above sea level, Patton Village has six major intersections including I-69 and I-45.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"List of Farm to Market Roads in Texas (200–299)_sentence_222",
"List of Farm to Market Roads in Texas (200–299)_sentence_223"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5992,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.64",
"value": "?search="
},
{
"operation": "search",
"time": "6.022",
"value": "Contains...Texas SH 242"
},
{
"operation": "Now on",
"time": "8.75",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "49.119",
"value": "List of Farm to Market Roads in Texas (200–299)"
},
{
"operation": "Now on",
"time": "55.712",
"value": "List of Farm to Market Roads in Texas (200–299)"
},
{
"operation": "hyperlink",
"time": "57.635",
"value": "43 FM 242 List of Farm to Market Roads in Texas (200–299)_section_49"
},
{
"operation": "Highlighting",
"time": "64.272",
"value": "List of Farm to Market Roads in Texas (200–299)_sentence_222"
},
{
"operation": "Highlighting",
"time": "67.118",
"value": "List of Farm to Market Roads in Texas (200–299)_sentence_223"
},
{
"operation": "finish",
"time": "391.292",
"value": "finish"
}
]
}
|
45557
|
Ian Dejardin (born 26 August 1955) is an art historian and Chief Executive of the McMichael Canadian Art Collection who succeeded Desmond Shawe-Taylor and became the director of the Dulwich Picture Gallery.
|
{
"evidence": [
"Ian Dejardin_sentence_0",
"Ian Dejardin_sentence_1",
"Ian Dejardin_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:\nIan Dejardin (born 26 August 1955) is an art historian and Chief Executive of the McMichael Canadian Art Collection who succeeded Desmond Shawe-Taylor and became the director of the Dulwich Picture Gallery.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Ian Dejardin_sentence_0",
"Ian Dejardin_sentence_1",
"Ian Dejardin_sentence_10"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5993,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.003",
"value": "undefined"
},
{
"operation": "search",
"time": "0.008",
"value": "ian Dejardin"
},
{
"operation": "Now on",
"time": "0.011",
"value": "Ian Dejardin"
},
{
"operation": "Highlighting",
"time": "0.047",
"value": "Ian Dejardin_sentence_0"
},
{
"operation": "Page search",
"time": "0.076",
"value": "art historian"
},
{
"operation": "Highlighting deleted",
"time": "0.079",
"value": "Ian Dejardin_sentence_0"
},
{
"operation": "Highlighting",
"time": "0.08",
"value": "Ian Dejardin_sentence_0"
},
{
"operation": "Highlighting",
"time": "0.086",
"value": "Ian Dejardin_sentence_1"
},
{
"operation": "Page search",
"time": "0.156",
"value": "mcmichael"
},
{
"operation": "Highlighting deleted",
"time": "0.159",
"value": "Ian Dejardin_sentence_1"
},
{
"operation": "Highlighting",
"time": "0.16",
"value": "Ian Dejardin_sentence_1"
},
{
"operation": "Highlighting",
"time": "0.171",
"value": "Ian Dejardin_sentence_10"
},
{
"operation": "finish",
"time": "0.279",
"value": "finish"
},
{
"operation": "Now on",
"time": "1620846084.917",
"value": "Ian Dejardin"
},
{
"operation": "finish",
"time": "1620846089.839",
"value": "finish"
}
]
}
|
exfever_3915
|
Elementary is a television series that presented a contemporary update of the character Sherlock Holmes, created by a British writer and physician.
|
{
"evidence": [
"Elementary (TV series)",
"Arthur Conan Doyle"
],
"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:\nElementary is a television series that presented a contemporary update of the character Sherlock Holmes, created by a British writer and physician.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Elementary (TV series)",
"Arthur Conan Doyle"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5994,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6320
|
Lindsay Dee Lohan is an American actress who starred in a 2004 American teen musical comedy film directed by Mel Gibson.
|
{
"evidence": [
"Lindsay Lohan",
"Confessions of a Teenage Drama Queen"
],
"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:\nLindsay Dee Lohan is an American actress who starred in a 2004 American teen musical comedy film directed by Mel Gibson.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lindsay Lohan",
"Confessions of a Teenage Drama Queen"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5995,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
53752
|
Brightling is a village in the England's Rother district and part in the High Weald Area of Outstanding Natural Beauty (one of England's finest landscape with 1461 square kilometers of land area).
|
{
"evidence": [
"Brightling_sentence_0",
"Brightling_sentence_2"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBrightling is a village in the England's Rother district and part in the High Weald Area of Outstanding Natural Beauty (one of England's finest landscape with 1461 square kilometers of land area).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Brightling_sentence_0",
"Brightling_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5996,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.039",
"value": "?search="
},
{
"operation": "search",
"time": "19.347",
"value": "Brightling"
},
{
"operation": "Now on",
"time": "21.219",
"value": "Brightling"
},
{
"operation": "Highlighting",
"time": "31.983",
"value": "Brightling_sentence_0"
},
{
"operation": "Highlighting",
"time": "61.035",
"value": "Brightling_sentence_2"
},
{
"operation": "search",
"time": "91.034",
"value": "Contains...High Weald Area of Outstanding Natural Beauty"
},
{
"operation": "Now on",
"time": "93.105",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "118.299",
"value": "High Weald Landscape Trail"
},
{
"operation": "Now on",
"time": "119.088",
"value": "High Weald Landscape Trail"
},
{
"operation": "search",
"time": "150.061",
"value": "AONB"
},
{
"operation": "Now on",
"time": "151.1",
"value": "Area of Outstanding Natural Beauty"
},
{
"operation": "hyperlink",
"time": "167.718",
"value": "7.1 England Area of Outstanding Natural Beauty_section_7"
},
{
"operation": "Page search",
"time": "208.508",
"value": "weald"
},
{
"operation": "hyperlink",
"time": "246.369",
"value": "National parks of England and Wales"
},
{
"operation": "Now on",
"time": "247.818",
"value": "National parks of the United Kingdom"
},
{
"operation": "search",
"time": "270.611",
"value": "aonbs of england"
},
{
"operation": "Now on",
"time": "272.716",
"value": "undefined"
},
{
"operation": "search",
"time": "294.042",
"value": "High Weald"
},
{
"operation": "Now on",
"time": "295.051",
"value": "Weald"
},
{
"operation": "hyperlink",
"time": "345.299",
"value": "Weald and Downland Open Air Museum"
},
{
"operation": "Now on",
"time": "350.532",
"value": "Weald and Downland Living Museum"
},
{
"operation": "search",
"time": "364.117",
"value": "high weald"
},
{
"operation": "Now on",
"time": "364.721",
"value": "High Weald"
},
{
"operation": "hyperlink",
"time": "403.781",
"value": "Recreational walks in Kent"
},
{
"operation": "Now on",
"time": "404.581",
"value": "Recreational walks in Kent"
},
{
"operation": "search",
"time": "427.292",
"value": "Contains...high weald"
},
{
"operation": "Now on",
"time": "428.886",
"value": "undefined"
},
{
"operation": "search",
"time": "473.067",
"value": "Contains...high weald aonb"
},
{
"operation": "Now on",
"time": "473.854",
"value": "undefined"
},
{
"operation": "search",
"time": "534.689",
"value": "Contains...one of England's finest landscape*"
},
{
"operation": "Now on",
"time": "537.844",
"value": "undefined"
},
{
"operation": "search",
"time": "568.614",
"value": "weald one of England's finest landscap*"
},
{
"operation": "Now on",
"time": "570.051",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "576.569",
"value": "A259 road"
},
{
"operation": "Now on",
"time": "600.627",
"value": "A259 road"
},
{
"operation": "back-button-clicked",
"time": "604.593",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "605.179",
"value": "undefined"
},
{
"operation": "finish",
"time": "641.382",
"value": "finish"
}
]
}
|
33776
|
The Center for Religious Freedom of Freedom House analyzed a set of Saudi Ministry of Education textbooks in Islamic studies courses for elementary and secondary school students, and they found statements promoting hatred of Christians and Jews, people originating from the Israelites and Hebrews.
|
{
"evidence": [
"Antisemitism in Saudi Arabia_sentence_10",
"Antisemitism in Saudi Arabia_sentence_11",
"Jews_sentence_4"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Center for Religious Freedom of Freedom House analyzed a set of Saudi Ministry of Education textbooks in Islamic studies courses for elementary and secondary school students, and they found statements promoting hatred of Christians and Jews, people originating from the Israelites and Hebrews.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Antisemitism in Saudi Arabia_sentence_10",
"Antisemitism in Saudi Arabia_sentence_11",
"Jews_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5997,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.782",
"value": "?search="
},
{
"operation": "search",
"time": "37.059",
"value": "center for religious freedom of freedom"
},
{
"operation": "Now on",
"time": "40.09",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "54.387",
"value": "Nina Shea"
},
{
"operation": "Now on",
"time": "55.153",
"value": "Nina Shea"
},
{
"operation": "Page search",
"time": "70.011",
"value": "freedom of freedom house"
},
{
"operation": "back-button-clicked",
"time": "79.34",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "80.229",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "85.701",
"value": "Nina Shea"
},
{
"operation": "Now on",
"time": "86.091",
"value": "Nina Shea"
},
{
"operation": "hyperlink",
"time": "93.195",
"value": "Freedom House"
},
{
"operation": "Now on",
"time": "94.075",
"value": "Freedom House"
},
{
"operation": "Page search",
"time": "116.588",
"value": "saudi ministry"
},
{
"operation": "search",
"time": "168.867",
"value": "they found statements promoting hatred of christians and jews"
},
{
"operation": "Now on",
"time": "170.939",
"value": "undefined"
},
{
"operation": "search",
"time": "202.576",
"value": "the center for religious freedom"
},
{
"operation": "Now on",
"time": "206.226",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "217.464",
"value": "Nina Shea"
},
{
"operation": "Now on",
"time": "217.879",
"value": "Nina Shea"
},
{
"operation": "Page search",
"time": "227.175",
"value": "center for religious freedom"
},
{
"operation": "hyperlink",
"time": "236.231",
"value": "Freedom House"
},
{
"operation": "Now on",
"time": "237.011",
"value": "Freedom House"
},
{
"operation": "Page search",
"time": "276.277",
"value": "analyzed a set of saudi ministry of education textbooks"
},
{
"operation": "search",
"time": "307.818",
"value": "analyzed a set of saudi ministry of education textbooks"
},
{
"operation": "Now on",
"time": "310.165",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "321.828",
"value": "Antisemitism in school textbooks Antisemitism in Saudi Arabia_section_1"
},
{
"operation": "Now on",
"time": "322.339",
"value": "Antisemitism in Saudi Arabia#Antisemitism in school textbooks Antisemitism in Saudi Arabia section 1"
},
{
"operation": "Highlighting",
"time": "334.048",
"value": "Antisemitism in Saudi Arabia_sentence_10"
},
{
"operation": "Highlighting",
"time": "354.834",
"value": "Antisemitism in Saudi Arabia_sentence_11"
},
{
"operation": "Page search",
"time": "378.54",
"value": "people originating from the Israelites and Hebrew"
},
{
"operation": "Highlighting",
"time": "398.748",
"value": "Antisemitism in Saudi Arabia_sentence_12"
},
{
"operation": "Highlighting deleted",
"time": "399.521",
"value": "Antisemitism in Saudi Arabia_sentence_11"
},
{
"operation": "Highlighting deleted",
"time": "400.184",
"value": "Antisemitism in Saudi Arabia_sentence_12"
},
{
"operation": "Highlighting",
"time": "400.896",
"value": "Antisemitism in Saudi Arabia_sentence_11"
},
{
"operation": "Highlighting deleted",
"time": "403.982",
"value": "Antisemitism in Saudi Arabia_sentence_11"
},
{
"operation": "Highlighting",
"time": "406.43",
"value": "Antisemitism in Saudi Arabia_sentence_11"
},
{
"operation": "search",
"time": "410.746",
"value": "polytheists"
},
{
"operation": "Now on",
"time": "411.964",
"value": "Polytheism"
},
{
"operation": "Page search",
"time": "522.508",
"value": "people originating from the Israelites and Hebrews."
},
{
"operation": "search",
"time": "534.242",
"value": "people originating from the Israelites and Hebrews."
},
{
"operation": "Now on",
"time": "536.313",
"value": "undefined"
},
{
"operation": "search",
"time": "550.017",
"value": "jew"
},
{
"operation": "Now on",
"time": "551.257",
"value": "Jew"
},
{
"operation": "Highlighting",
"time": "558.778",
"value": "Jews_sentence_4"
},
{
"operation": "finish",
"time": "566.094",
"value": "finish"
}
]
}
|
exfever_8535
|
The Blue Lagoon is a romance novel trilogy, which also includes a romance, and is the first sequel to his best-selling novel.
|
{
"evidence": [
"The Blue Lagoon (novel)",
"The Garden of God"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Blue Lagoon is a romance novel trilogy, which also includes a romance, and is the first sequel to his best-selling novel.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Blue Lagoon (novel)",
"The Garden of God"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5998,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
76348
|
Cork claimed the title beating Waterford on a score line of 3-16 to 3-12, though Waterford was leading by 1-9 to 1-4 at halftime.
|
{
"evidence": [
"2003 Munster Senior Hurling Championship Final_sentence_2",
"2003 Munster Senior Hurling Championship Final_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:\nCork claimed the title beating Waterford on a score line of 3-16 to 3-12, though Waterford was leading by 1-9 to 1-4 at halftime.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"2003 Munster Senior Hurling Championship Final_sentence_2",
"2003 Munster Senior Hurling Championship Final_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5999,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.199",
"value": "?search="
},
{
"operation": "search",
"time": "26.594",
"value": "cork waterford"
},
{
"operation": "Now on",
"time": "27.774",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "42.728",
"value": "Waterford GAA"
},
{
"operation": "Now on",
"time": "43.695",
"value": "Waterford GAA"
},
{
"operation": "Now on",
"time": "128.66",
"value": "Waterford GAA"
},
{
"operation": "Page search",
"time": "134.903",
"value": "halftime"
},
{
"operation": "search",
"time": "166.406",
"value": "Cork claimed the title beating Waterford"
},
{
"operation": "Now on",
"time": "167.801",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "183.893",
"value": "2003 Munster Senior Hurling Championship Final"
},
{
"operation": "Now on",
"time": "184.428",
"value": "2003 Munster Senior Hurling Championship Final"
},
{
"operation": "Highlighting",
"time": "195.319",
"value": "2003 Munster Senior Hurling Championship Final_sentence_2"
},
{
"operation": "Highlighting",
"time": "198.104",
"value": "2003 Munster Senior Hurling Championship Final_sentence_3"
},
{
"operation": "finish",
"time": "223.344",
"value": "finish"
}
]
}
|
13818
|
Tiger started his path towards his later career with his study on the decolonization of Africa and he was inspired by Weber's questions and new discoveries in Africa by Raymond Dart and Lewis Leaky.
|
{
"evidence": [
"Lionel Tiger_sentence_7",
"Lionel Tiger_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:\nTiger started his path towards his later career with his study on the decolonization of Africa and he was inspired by Weber's questions and new discoveries in Africa by Raymond Dart and Lewis Leaky.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Lionel Tiger_sentence_7",
"Lionel Tiger_sentence_10"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6000,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.913",
"value": "?search="
},
{
"operation": "search",
"time": "81.463",
"value": "lionel tiger"
},
{
"operation": "Now on",
"time": "82.321",
"value": "Lionel Tiger"
},
{
"operation": "Page search",
"time": "100.182",
"value": "decoloniz"
},
{
"operation": "Highlighting",
"time": "102.833",
"value": "Lionel Tiger_sentence_7"
},
{
"operation": "Highlighting",
"time": "116.791",
"value": "Lionel Tiger_sentence_10"
},
{
"operation": "finish",
"time": "129.275",
"value": "finish"
}
]
}
|
21059
|
Sluseholmen Canal District is residential development.
|
{
"evidence": [
"Sluseholmen Canal District_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:\nSluseholmen Canal District is residential development.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Sluseholmen Canal District_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6001,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.71",
"value": "?search="
},
{
"operation": "search",
"time": "17.539",
"value": "Sluseholmen Canal District"
},
{
"operation": "Now on",
"time": "18.382",
"value": "Sluseholmen Canal District"
},
{
"operation": "hyperlink",
"time": "22.244",
"value": "Denmark"
},
{
"operation": "Now on",
"time": "22.834",
"value": "Denmark"
},
{
"operation": "back-button-clicked",
"time": "24.65",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "24.725",
"value": "Sluseholmen Canal District"
},
{
"operation": "Highlighting",
"time": "26",
"value": "Sluseholmen Canal District_sentence_0"
},
{
"operation": "finish",
"time": "46.542",
"value": "finish"
}
]
}
|
exfever_1044
|
Jennifer Grey is a director who earned an accolade bestowed by the HFPA.
|
{
"evidence": [
"Jennifer Grey",
"Golden Globe Award",
"Hollywood Foreign Press Association"
],
"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:\nJennifer Grey is a director who earned an accolade bestowed by the HFPA.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jennifer Grey",
"Golden Globe Award",
"Hollywood Foreign Press Association"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6002,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3368
|
Grunge is an alternative rock genre that fuses elements of punk rock and a genre of rock music characterized by extended melodic passages, instrumental sections, or an entire piece of music that is pre-written to be played on an electric drum.
|
{
"evidence": [
"Grunge",
"Heavy metal music",
"Guitar solo"
],
"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:\nGrunge is an alternative rock genre that fuses elements of punk rock and a genre of rock music characterized by extended melodic passages, instrumental sections, or an entire piece of music that is pre-written to be played on an electric drum.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Grunge",
"Heavy metal music",
"Guitar solo"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6003,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
10898
|
Jenny Davidson (born 1971), an American historian and writer, had her dissertation Hypocrisy and the Politics of Politeness published by Cambridge University Press, and her second non-fiction publication Breeding: A Partial History of the Eighteenth Century continues the exploration of Age of Enlightenment authors.
|
{
"evidence": [
"Jenny Davidson_sentence_0",
"Jenny Davidson_sentence_5",
"Jenny Davidson_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:\nJenny Davidson (born 1971), an American historian and writer, had her dissertation Hypocrisy and the Politics of Politeness published by Cambridge University Press, and her second non-fiction publication Breeding: A Partial History of the Eighteenth Century continues the exploration of Age of Enlightenment authors.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Jenny Davidson_sentence_0",
"Jenny Davidson_sentence_5",
"Jenny Davidson_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6004,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.564",
"value": "?search="
},
{
"operation": "search",
"time": "5.666",
"value": "Jenny Davidson"
},
{
"operation": "Now on",
"time": "6.338",
"value": "Jenny Davidson"
},
{
"operation": "Highlighting",
"time": "7.787",
"value": "Jenny Davidson_sentence_0"
},
{
"operation": "Page search",
"time": "14.813",
"value": "Hypocrisy"
},
{
"operation": "Highlighting",
"time": "17.563",
"value": "Jenny Davidson_sentence_5"
},
{
"operation": "Page search",
"time": "31.023",
"value": "Breeding"
},
{
"operation": "Highlighting",
"time": "37.288",
"value": "Jenny Davidson_sentence_6"
},
{
"operation": "finish",
"time": "420.22",
"value": "finish"
}
]
}
|
exfever_2895
|
Nicholas King Nolte is an American actor whose other film appearances include a 1984 American satirical black comedy-drama series written by W. R. McKinney.
|
{
"evidence": [
"Nick Nolte",
"Teachers (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:\nNicholas King Nolte is an American actor whose other film appearances include a 1984 American satirical black comedy-drama series written by W. R. McKinney.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Nick Nolte",
"Teachers (film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6005,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
79148
|
Giorgio Mainerio sought to work as a chaplain in Udine, a city in north-eastern Italy.
|
{
"evidence": [
"Giorgio Mainerio_sentence_4",
"Udine_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:\nGiorgio Mainerio sought to work as a chaplain in Udine, a city in north-eastern Italy.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Giorgio Mainerio_sentence_4",
"Udine_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6006,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.141",
"value": "?search="
},
{
"operation": "search",
"time": "9.976",
"value": "Giorgio Mainerio"
},
{
"operation": "Now on",
"time": "12.698",
"value": "Giorgio Mainerio"
},
{
"operation": "Highlighting",
"time": "16.632",
"value": "Giorgio Mainerio_sentence_4"
},
{
"operation": "hyperlink",
"time": "21.693",
"value": "Udine"
},
{
"operation": "Now on",
"time": "22.985",
"value": "Udine"
},
{
"operation": "Highlighting",
"time": "30.623",
"value": "Udine_sentence_2"
},
{
"operation": "finish",
"time": "39.664",
"value": "finish"
}
]
}
|
exfever_8567
|
Assassin's Creed is an open-world action-adventure stealth video game created by an American video game writer who works as the lead writer of a 2009 action-adventure video game developed by Sony.
|
{
"evidence": [
"Assassin's Creed",
"Corey May",
"Assassin's Creed II"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAssassin's Creed is an open-world action-adventure stealth video game created by an American video game writer who works as the lead writer of a 2009 action-adventure video game developed by Sony.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Assassin's Creed",
"Corey May",
"Assassin's Creed II"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6007,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
60160
|
The debut single of The Brightlights, "Inspired By," was filmed in the band's hometown of Grimsby on the eastern coast of England while Jay Burridge was directing it, then later entered in the UK Indie Chart at No. 3 in 2007.
|
{
"evidence": [
"The Brightlights_sentence_16",
"The Brightlights_sentence_17"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe debut single of The Brightlights, \"Inspired By,\" was filmed in the band's hometown of Grimsby on the eastern coast of England while Jay Burridge was directing it, then later entered in the UK Indie Chart at No. 3 in 2007.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"The Brightlights_sentence_16",
"The Brightlights_sentence_17"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6008,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.309",
"value": "?search="
},
{
"operation": "search",
"time": "33.771",
"value": "The Brightlights"
},
{
"operation": "Now on",
"time": "34.827",
"value": "The Brightlights"
},
{
"operation": "Highlighting",
"time": "47.977",
"value": "The Brightlights_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "51.469",
"value": "The Brightlights_sentence_1"
},
{
"operation": "Highlighting",
"time": "60.309",
"value": "The Brightlights_sentence_12"
},
{
"operation": "Highlighting deleted",
"time": "64.757",
"value": "The Brightlights_sentence_12"
},
{
"operation": "Highlighting",
"time": "67.037",
"value": "The Brightlights_sentence_10"
},
{
"operation": "Highlighting",
"time": "70.693",
"value": "The Brightlights_sentence_12"
},
{
"operation": "Highlighting",
"time": "81.198",
"value": "The Brightlights_sentence_13"
},
{
"operation": "Highlighting",
"time": "96.318",
"value": "The Brightlights_sentence_16"
},
{
"operation": "Highlighting",
"time": "97.934",
"value": "The Brightlights_sentence_17"
},
{
"operation": "Highlighting deleted",
"time": "102.79",
"value": "The Brightlights_sentence_13"
},
{
"operation": "Highlighting deleted",
"time": "103.79",
"value": "The Brightlights_sentence_12"
},
{
"operation": "Highlighting deleted",
"time": "104.887",
"value": "The Brightlights_sentence_10"
},
{
"operation": "finish",
"time": "123.12",
"value": "finish"
}
]
}
|
exfever_726
|
The Semitic languages are a branch of the language family of about 300 languages that are spoken predominantly in Western Asia, North Africa, the Horn of Africa, and parts of the Sahara/Sahel.
|
{
"evidence": [
"Semitic languages",
"Afroasiatic languages"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Semitic languages are a branch of the language family of about 300 languages that are spoken predominantly in Western Asia, North Africa, the Horn of Africa, and parts of the Sahara/Sahel.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Semitic languages",
"Afroasiatic languages"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6009,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6376
|
Lightning Point, a television crime drama, was produced for Network Ten in association with ZDF and a British pay television network aimed at children aged under 18.
|
{
"evidence": [
"Lightning Point",
"Nickelodeon (UK and Ireland)"
],
"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:\nLightning Point, a television crime drama, was produced for Network Ten in association with ZDF and a British pay television network aimed at children aged under 18.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lightning Point",
"Nickelodeon (UK and Ireland)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6010,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
23703
|
The list shows the programs which are showing on an Australian television network but had previously premiered on another network, including those which premiered in free-to-air television.
|
{
"evidence": [
"List of programming changes on Australian television in 2008_sentence_3",
"List of programming changes on Australian television in 2008_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:\nThe list shows the programs which are showing on an Australian television network but had previously premiered on another network, including those which premiered in free-to-air television.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"List of programming changes on Australian television in 2008_sentence_3",
"List of programming changes on Australian television in 2008_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6011,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.006",
"value": "undefined"
},
{
"operation": "search",
"time": "15.824",
"value": "Contains...The list shows the programs which are showing on an Australian television network"
},
{
"operation": "Now on",
"time": "21.464",
"value": "undefined"
},
{
"operation": "search",
"time": "58.575",
"value": "Contains...premiered on another network, including those which premiered in free-to-air television"
},
{
"operation": "Now on",
"time": "61.541",
"value": "undefined"
},
{
"operation": "search",
"time": "84.159",
"value": "Contains...programs which are showing on an Australian television network"
},
{
"operation": "Now on",
"time": "85.036",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "99.438",
"value": "Australian Gold Coast"
},
{
"operation": "Now on",
"time": "101.374",
"value": "Australian Gold Coast"
},
{
"operation": "search",
"time": "117.511",
"value": "Contains...Subscription television in Australia"
},
{
"operation": "Now on",
"time": "119.215",
"value": "undefined"
},
{
"operation": "search",
"time": "147.456",
"value": "Contains...list of Television in Australia"
},
{
"operation": "Now on",
"time": "150.033",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "155.666",
"value": "List of digital television channels in Australia"
},
{
"operation": "Now on",
"time": "158.034",
"value": "List of digital television channels in Australia"
},
{
"operation": "hyperlink",
"time": "175.395",
"value": "Australian and New Zealand television frequencies"
},
{
"operation": "Now on",
"time": "178.326",
"value": "Australian and New Zealand television frequencies"
},
{
"operation": "search",
"time": "203.334",
"value": "List of Australian television news services"
},
{
"operation": "Now on",
"time": "204.412",
"value": "List of Australian television news services"
},
{
"operation": "hyperlink",
"time": "218.747",
"value": "List of Australian television series"
},
{
"operation": "Now on",
"time": "219.543",
"value": "List of Australian television series"
},
{
"operation": "hyperlink",
"time": "241.682",
"value": "List of longest-running Australian television series"
},
{
"operation": "Now on",
"time": "244.904",
"value": "List of longest-running Australian television series"
},
{
"operation": "search",
"time": "265.904",
"value": "Contains...Australian television related lists"
},
{
"operation": "Now on",
"time": "267.197",
"value": "undefined"
},
{
"operation": "search",
"time": "304.541",
"value": "Contains...programs which had previously premiered on another network, including those which premiered in free-to-air televisionashowing on an Australian television"
},
{
"operation": "Now on",
"time": "305.525",
"value": "undefined"
},
{
"operation": "search",
"time": "325.873",
"value": "Contains...Australian television network that had previously premiered on another Australian television network List of Australian television premieres"
},
{
"operation": "Now on",
"time": "327.521",
"value": "undefined"
},
{
"operation": "search",
"time": "348.217",
"value": "Contains...changes to Australian television network that had previously premiered on another Australian television network"
},
{
"operation": "Now on",
"time": "349.736",
"value": "undefined"
},
{
"operation": "search",
"time": "369.952",
"value": "Contains...Australian television in 2010"
},
{
"operation": "Now on",
"time": "371.236",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "379.16",
"value": "List of international specialty programme premieres on Australian television in 2010"
},
{
"operation": "Now on",
"time": "379.936",
"value": "List of international specialty programme premieres on Australian television in 2010"
},
{
"operation": "search",
"time": "443.691",
"value": "Contains...List of programming changes on Australian television network that had previously premiered on another Australian television network"
},
{
"operation": "Now on",
"time": "444.601",
"value": "undefined"
},
{
"operation": "search",
"time": "461.462",
"value": "List of programming changes on Australian television in 2008"
},
{
"operation": "Now on",
"time": "463.004",
"value": "List of programming changes on Australian television in 2008"
},
{
"operation": "Highlighting",
"time": "484.223",
"value": "List of programming changes on Australian television in 2008_sentence_3"
},
{
"operation": "Highlighting",
"time": "502.57",
"value": "List of programming changes on Australian television in 2008_sentence_5"
},
{
"operation": "finish",
"time": "566.005",
"value": "finish"
}
]
}
|
exfever_3134
|
Radioactive is an alternative rock song with elements of electronic dance music that originated in East London.
|
{
"evidence": [
"Radioactive (Imagine Dragons song)",
"Dubstep"
],
"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:\nRadioactive is an alternative rock song with elements of electronic dance music that originated in East London.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Radioactive (Imagine Dragons song)",
"Dubstep"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6012,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
74451
|
The United States required almost 19,000 tonnes of uranium between 2006 and 2008 while France required over 10,000 tonnes during the same time.
|
{
"evidence": [
"Peak uranium_sentence_46"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe United States required almost 19,000 tonnes of uranium between 2006 and 2008 while France required over 10,000 tonnes during the same time.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Peak uranium_sentence_46"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6013,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.314",
"value": "?search="
},
{
"operation": "search",
"time": "86.895",
"value": "Uranium mining"
},
{
"operation": "Now on",
"time": "88.761",
"value": "Uranium mining"
},
{
"operation": "Page search",
"time": "104.939",
"value": "france"
},
{
"operation": "hyperlink",
"time": "141.304",
"value": "Uranium mining by country"
},
{
"operation": "Now on",
"time": "144.171",
"value": "Uranium mining by country"
},
{
"operation": "Page search",
"time": "166.201",
"value": "france"
},
{
"operation": "hyperlink",
"time": "200.95",
"value": "List of countries by uranium production"
},
{
"operation": "Now on",
"time": "201.753",
"value": "List of countries by uranium production"
},
{
"operation": "back-button-clicked",
"time": "225.516",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "225.615",
"value": "Uranium mining by country"
},
{
"operation": "hyperlink",
"time": "228.406",
"value": "List of countries by uranium reserves"
},
{
"operation": "Now on",
"time": "229.286",
"value": "List of countries by uranium reserves"
},
{
"operation": "search",
"time": "282.198",
"value": "Contains...19,000 tonnes of uranium"
},
{
"operation": "Now on",
"time": "284.375",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "298.933",
"value": "Uranium mining"
},
{
"operation": "Now on",
"time": "299.187",
"value": "Uranium mining"
},
{
"operation": "Page search",
"time": "312.8",
"value": "19,000 tonnes"
},
{
"operation": "hyperlink",
"time": "340.444",
"value": "Uranium mining by country"
},
{
"operation": "Now on",
"time": "341.262",
"value": "Uranium mining by country"
},
{
"operation": "Page search",
"time": "345.825",
"value": "2006"
},
{
"operation": "search",
"time": "379",
"value": "Contains...uranium mining france"
},
{
"operation": "Now on",
"time": "381.298",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "390.762",
"value": "Uranium mining"
},
{
"operation": "Now on",
"time": "391.249",
"value": "Uranium mining"
},
{
"operation": "Page search",
"time": "395.65",
"value": "france"
},
{
"operation": "hyperlink",
"time": "460.522",
"value": "List of uranium mines"
},
{
"operation": "Now on",
"time": "476.577",
"value": "List of uranium projects"
},
{
"operation": "Page search",
"time": "484.475",
"value": "france"
},
{
"operation": "back-button-clicked",
"time": "508.99",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "509.092",
"value": "Uranium mining"
},
{
"operation": "hyperlink",
"time": "538.812",
"value": "Peak uranium"
},
{
"operation": "Now on",
"time": "540.784",
"value": "Peak uranium"
},
{
"operation": "Highlighting",
"time": "561.415",
"value": "Peak uranium_sentence_46"
},
{
"operation": "finish",
"time": "592.212",
"value": "finish"
}
]
}
|
88583
|
The Treasury staff prepared the Emergency Banking Act during Herbert Hoover's administration (31st president of the United States from 1929 to 1933), then passed on March 9, 1933, allowing the twelve Federal Reserve Banks to issue additional currency on good assets.
|
{
"evidence": [
"Emergency Banking Act_sentence_7",
"Emergency Banking Act_sentence_8",
"Herbert Hoover_sentence_2"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Treasury staff prepared the Emergency Banking Act during Herbert Hoover's administration (31st president of the United States from 1929 to 1933), then passed on March 9, 1933, allowing the twelve Federal Reserve Banks to issue additional currency on good assets.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Emergency Banking Act_sentence_7",
"Emergency Banking Act_sentence_8",
"Herbert Hoover_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6014,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.473",
"value": "?search="
},
{
"operation": "search",
"time": "15.151",
"value": "Contains...The Treasury staff prepared the Emergency Banking Act during Herbert Hoover's administration"
},
{
"operation": "Now on",
"time": "17.102",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "19.332",
"value": "Emergency Banking Act"
},
{
"operation": "Now on",
"time": "19.857",
"value": "Emergency Banking Act"
},
{
"operation": "Page search",
"time": "35.428",
"value": "Emergency Banking Act during Herbert Hoover's administration"
},
{
"operation": "Page search",
"time": "50.225",
"value": "Herbert Hoover"
},
{
"operation": "Highlighting",
"time": "55.276",
"value": "Emergency Banking Act_sentence_7"
},
{
"operation": "Highlighting",
"time": "61.9",
"value": "Emergency Banking Act_sentence_8"
},
{
"operation": "hyperlink",
"time": "66.3",
"value": "Herbert Hoover"
},
{
"operation": "Now on",
"time": "68.098",
"value": "Herbert Hoover"
},
{
"operation": "Highlighting",
"time": "69.45",
"value": "Herbert Hoover_sentence_2"
},
{
"operation": "finish",
"time": "78.91",
"value": "finish"
}
]
}
|
29342
|
Research on ABS methods involved over forty mathematicians from Hungary, UK, China, Iran and other countries, and has been the outcome of an international collaboration coordinated by Spedicato of University of Bergamo, located in the alpine Lombardy region of northern Italy.
|
{
"evidence": [
"ABS methods_sentence_4",
"ABS methods_sentence_3",
"Bergamo_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:\nResearch on ABS methods involved over forty mathematicians from Hungary, UK, China, Iran and other countries, and has been the outcome of an international collaboration coordinated by Spedicato of University of Bergamo, located in the alpine Lombardy region of northern Italy.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"ABS methods_sentence_4",
"ABS methods_sentence_3",
"Bergamo_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6015,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.331",
"value": "?search="
},
{
"operation": "search",
"time": "138.809",
"value": "ABS methods"
},
{
"operation": "Now on",
"time": "139.677",
"value": "ABS methods"
},
{
"operation": "Highlighting",
"time": "251.694",
"value": "ABS methods_sentence_4"
},
{
"operation": "Highlighting",
"time": "257.391",
"value": "ABS methods_sentence_3"
},
{
"operation": "hyperlink",
"time": "265.289",
"value": "Bergamo"
},
{
"operation": "Now on",
"time": "266.485",
"value": "Bergamo"
},
{
"operation": "Highlighting",
"time": "269.213",
"value": "Bergamo_sentence_3"
},
{
"operation": "finish",
"time": "280.96",
"value": "finish"
}
]
}
|
76144
|
Frederick Herzog, who was an impassioned champion of civil rights, taught Christian theology at Duke until his sudden death during a faculty meeting on October 9, 1995 and wrote the first North American article by a white theologian on liberation theology in the spring of1970.
|
{
"evidence": [
"Frederick Herzog_sentence_1",
"Frederick Herzog_sentence_6",
"Frederick Herzog_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:\nFrederick Herzog, who was an impassioned champion of civil rights, taught Christian theology at Duke until his sudden death during a faculty meeting on October 9, 1995 and wrote the first North American article by a white theologian on liberation theology in the spring of1970.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Frederick Herzog_sentence_1",
"Frederick Herzog_sentence_6",
"Frederick Herzog_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6016,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.644",
"value": "undefined"
},
{
"operation": "search",
"time": "15.478",
"value": "Frederick Herzog"
},
{
"operation": "Now on",
"time": "16.406",
"value": "Frederick Herzog"
},
{
"operation": "Highlighting",
"time": "23.633",
"value": "Frederick Herzog_sentence_1"
},
{
"operation": "Highlighting",
"time": "45.065",
"value": "Frederick Herzog_sentence_6"
},
{
"operation": "Highlighting",
"time": "64.654",
"value": "Frederick Herzog_sentence_7"
},
{
"operation": "finish",
"time": "80.738",
"value": "finish"
}
]
}
|
exfever_316
|
Jaime Lannister is a fictional character in the A Song of Ice and Fire series portrayed by the Danish actor who had his breakthrough role in Denmark with the 1994 thriller film directed and written by Ole Bornedal.
|
{
"evidence": [
"Jaime Lannister",
"Nikolaj Coster-Waldau",
"Nightwatch (1994 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:\nJaime Lannister is a fictional character in the A Song of Ice and Fire series portrayed by the Danish actor who had his breakthrough role in Denmark with the 1994 thriller film directed and written by Ole Bornedal.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jaime Lannister",
"Nikolaj Coster-Waldau",
"Nightwatch (1994 film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6017,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5224
|
Chris Weitz is an American film director, screenwriter, and producer who directed the film The Golden Compass a 2007 fantasy adventure film based on the 1995 novel Northern Lights starring an English actress who is known for the film Dustbin Baby.
|
{
"evidence": [
"Chris Weitz",
"The Golden Compass (film)",
"Dakota Blue Richards"
],
"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:\nChris Weitz is an American film director, screenwriter, and producer who directed the film The Golden Compass a 2007 fantasy adventure film based on the 1995 novel Northern Lights starring an English actress who is known for the film Dustbin Baby.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Chris Weitz",
"The Golden Compass (film)",
"Dakota Blue Richards"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6018,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
60333
|
The Northern Lights Local Exchange Point is a free Internet Exchange Point, which serves as the physical infrastructure which facilitates Internet traffic exchange between Internet Service Providers and content delivery networks.
|
{
"evidence": [
"Northern Lights Local Exchange Point_sentence_0",
"Internet exchange point_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Northern Lights Local Exchange Point is a free Internet Exchange Point, which serves as the physical infrastructure which facilitates Internet traffic exchange between Internet Service Providers and content delivery networks.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Northern Lights Local Exchange Point_sentence_0",
"Internet exchange point_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6019,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.678",
"value": "?search="
},
{
"operation": "search",
"time": "33.419",
"value": "Northern Lights Local Exchange Point"
},
{
"operation": "Now on",
"time": "34.517",
"value": "Northern Lights Local Exchange Point"
},
{
"operation": "Highlighting",
"time": "58.344",
"value": "Northern Lights Local Exchange Point_sentence_0"
},
{
"operation": "hyperlink",
"time": "69.216",
"value": "IXP"
},
{
"operation": "Now on",
"time": "69.717",
"value": "Internet exchange point"
},
{
"operation": "Highlighting",
"time": "77.464",
"value": "Internet exchange point_sentence_0"
},
{
"operation": "finish",
"time": "121.718",
"value": "finish"
}
]
}
|
37940
|
Delmer Berg was an American soldier who volunteered to serve with the XV International Brigade (nicknamed the Abraham Lincoln Brigade) during the Spanish Civil War that has four and who began serving as a union organizer in the 1950s, then became an official of the National Association for the Advancement of Colored People.
|
{
"evidence": [
"Delmer Berg_sentence_0",
"Delmer Berg_sentence_9",
"Delmer Berg_sentence_10"
],
"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:\nDelmer Berg was an American soldier who volunteered to serve with the XV International Brigade (nicknamed the Abraham Lincoln Brigade) during the Spanish Civil War that has four and who began serving as a union organizer in the 1950s, then became an official of the National Association for the Advancement of Colored People.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Delmer Berg_sentence_0",
"Delmer Berg_sentence_9",
"Delmer Berg_sentence_10"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6020,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.638",
"value": "?search="
},
{
"operation": "search",
"time": "15.402",
"value": "Delmer Berg"
},
{
"operation": "Now on",
"time": "16.29",
"value": "Delmer Berg"
},
{
"operation": "Highlighting",
"time": "26.14",
"value": "Delmer Berg_sentence_0"
},
{
"operation": "Highlighting",
"time": "51.653",
"value": "Delmer Berg_sentence_9"
},
{
"operation": "Highlighting",
"time": "62.929",
"value": "Delmer Berg_sentence_10"
},
{
"operation": "finish",
"time": "89.269",
"value": "finish"
}
]
}
|
exfever_6797
|
Eugenio Campari was the award-winning Italian sporting director.
|
{
"evidence": [
"Eugenio Canfari",
"Juventus F.C."
],
"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:\nEugenio Campari was the award-winning Italian sporting director.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Eugenio Canfari",
"Juventus F.C."
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6021,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_591
|
WALL-E is a 2008 American computer-animated science fiction film co-written by an American animator, storyboard artist, television writer, television director, and screenwriter.
|
{
"evidence": [
"WALL-E",
"Jim Reardon"
],
"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:\nWALL-E is a 2008 American computer-animated science fiction film co-written by an American animator, storyboard artist, television writer, television director, and screenwriter.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"WALL-E",
"Jim Reardon"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6022,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
31941
|
During the Jokela school shooting in Tuusula, Finland, head teacher Helena Kalmi immediately ordered all students and teachers to barricade themselves inside their classrooms after she was alerted to the shooting by the deputy head teacher.
|
{
"evidence": [
"Jokela school shooting_sentence_0",
"Jokela school shooting_sentence_13",
"Jokela school shooting_sentence_14"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDuring the Jokela school shooting in Tuusula, Finland, head teacher Helena Kalmi immediately ordered all students and teachers to barricade themselves inside their classrooms after she was alerted to the shooting by the deputy head teacher.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Jokela school shooting_sentence_0",
"Jokela school shooting_sentence_13",
"Jokela school shooting_sentence_14"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6023,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.409",
"value": "?search="
},
{
"operation": "search",
"time": "41.106",
"value": "Jokela school shooting in Tuusula, Finland,"
},
{
"operation": "Now on",
"time": "42.792",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "44.926",
"value": "Jokela school shooting"
},
{
"operation": "Now on",
"time": "46.253",
"value": "Jokela school shooting"
},
{
"operation": "Highlighting",
"time": "47.426",
"value": "Jokela school shooting_sentence_0"
},
{
"operation": "Page search",
"time": "66.248",
"value": "Helena Kalmi"
},
{
"operation": "Highlighting",
"time": "86.287",
"value": "Jokela school shooting_sentence_13"
},
{
"operation": "Highlighting",
"time": "89.194",
"value": "Jokela school shooting_sentence_14"
},
{
"operation": "finish",
"time": "120.364",
"value": "finish"
}
]
}
|
54099
|
Karim Olowu attended Saint Paul's Primary School, Breadfruit Lagos, then in 1936 he represented his school in the Empire Day Athletics, and he later attended CMS Grammar School, Lagos where he was the sports prefect.
|
{
"evidence": [
"Karim Olowu_sentence_8",
"Karim Olowu_sentence_9"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nKarim Olowu attended Saint Paul's Primary School, Breadfruit Lagos, then in 1936 he represented his school in the Empire Day Athletics, and he later attended CMS Grammar School, Lagos where he was the sports prefect.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Karim Olowu_sentence_8",
"Karim Olowu_sentence_9"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6024,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.422",
"value": "?search="
},
{
"operation": "search",
"time": "7.178",
"value": "Karim Olowu"
},
{
"operation": "Now on",
"time": "8.708",
"value": "Karim Olowu"
},
{
"operation": "Page search",
"time": "38.952",
"value": "primary"
},
{
"operation": "Page search",
"time": "39.126",
"value": "primary"
},
{
"operation": "Highlighting",
"time": "43.251",
"value": "Karim Olowu_sentence_8"
},
{
"operation": "Highlighting",
"time": "62.793",
"value": "Karim Olowu_sentence_9"
},
{
"operation": "finish",
"time": "99.971",
"value": "finish"
}
]
}
|
6461
|
Bongoyo was a Tropical low category cyclone with 45 km/h wind speed and 1005 hPa pressure, which affected the Cocos Islands (an Australian external territory in the Pacific Ocean) on Nov 24–30.
|
{
"evidence": [
"Cocos (Keeling) Islands_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:\nBongoyo was a Tropical low category cyclone with 45 km/h wind speed and 1005 hPa pressure, which affected the Cocos Islands (an Australian external territory in the Pacific Ocean) on Nov 24–30.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Cocos (Keeling) Islands_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6025,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.901",
"value": "?search="
},
{
"operation": "search",
"time": "7.595",
"value": "Bongoyo"
},
{
"operation": "Now on",
"time": "8.211",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "18.086",
"value": "Tropical cyclone naming"
},
{
"operation": "Now on",
"time": "19.178",
"value": "Tropical cyclone naming"
},
{
"operation": "Page search",
"time": "23.286",
"value": "Bongoyo"
},
{
"operation": "back-button-clicked",
"time": "61.981",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "62.546",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "69.213",
"value": "Tropical cyclones in 2020"
},
{
"operation": "Now on",
"time": "70.665",
"value": "Tropical cyclones in 2020"
},
{
"operation": "Page search",
"time": "74.641",
"value": "Bongoyo"
},
{
"operation": "hyperlink",
"time": "137.972",
"value": "2020–21 Australian region cyclone season"
},
{
"operation": "Now on",
"time": "138.725",
"value": "2020–21 Australian region cyclone season"
},
{
"operation": "Page search",
"time": "143.527",
"value": "Bongoyo"
},
{
"operation": "hyperlink",
"time": "147.168",
"value": "3.1 Tropical Low 01U (Bongoyo) 2020–21 Australian region cyclone season_section_3"
},
{
"operation": "Highlighting",
"time": "224.574",
"value": "2020–21 Australian region cyclone season_sentence_18"
},
{
"operation": "Highlighting",
"time": "224.579",
"value": "2020–21 Australian region cyclone season_title"
},
{
"operation": "Highlighting",
"time": "224.581",
"value": "2020–21 Australian region cyclone season_section_2"
},
{
"operation": "Highlighting",
"time": "224.585",
"value": "2020–21 Australian region cyclone season_section_3"
},
{
"operation": "Highlighting",
"time": "248.216",
"value": "2020–21 Australian region cyclone season_cell_1_0_1"
},
{
"operation": "Highlighting",
"time": "248.24",
"value": "2020–21 Australian region cyclone season_header_cell_1_0_0"
},
{
"operation": "Highlighting deleted",
"time": "249.301",
"value": "2020–21 Australian region cyclone season_cell_1_0_1"
},
{
"operation": "Highlighting deleted",
"time": "249.32",
"value": "2020–21 Australian region cyclone season_header_cell_1_0_0"
},
{
"operation": "Highlighting",
"time": "251.974",
"value": "2020–21 Australian region cyclone season_header_cell_1_0_0"
},
{
"operation": "Highlighting",
"time": "252.47",
"value": "2020–21 Australian region cyclone season_cell_1_0_1"
},
{
"operation": "Highlighting",
"time": "252.493",
"value": "2020–21 Australian region cyclone season_header_cell_1_0_0"
},
{
"operation": "hyperlink",
"time": "273.314",
"value": "Cocos Islands"
},
{
"operation": "Now on",
"time": "274.133",
"value": "Cocos (Keeling) Islands"
},
{
"operation": "back-button-clicked",
"time": "278.847",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "278.949",
"value": "2020–21 Australian region cyclone season#Tropical Low 01U .28Bongoyo.29 2020.E2.80.9321 Australian region cyclone season section 3"
},
{
"operation": "Highlighting",
"time": "305.008",
"value": "2020–21 Australian region cyclone season_cell_1_1_1"
},
{
"operation": "Highlighting",
"time": "305.035",
"value": "2020–21 Australian region cyclone season_header_cell_1_1_0"
},
{
"operation": "Highlighting deleted",
"time": "307.428",
"value": "2020–21 Australian region cyclone season_cell_1_1_1"
},
{
"operation": "Highlighting deleted",
"time": "307.452",
"value": "2020–21 Australian region cyclone season_header_cell_1_1_0"
},
{
"operation": "Highlighting",
"time": "309.662",
"value": "2020–21 Australian region cyclone season_cell_1_1_1"
},
{
"operation": "Highlighting",
"time": "309.686",
"value": "2020–21 Australian region cyclone season_header_cell_1_1_0"
},
{
"operation": "Highlighting deleted",
"time": "320.199",
"value": "2020–21 Australian region cyclone season_cell_1_0_1"
},
{
"operation": "Highlighting deleted",
"time": "320.223",
"value": "2020–21 Australian region cyclone season_header_cell_1_0_0"
},
{
"operation": "Highlighting deleted",
"time": "324.997",
"value": "2020–21 Australian region cyclone season_cell_1_1_1"
},
{
"operation": "Highlighting deleted",
"time": "325.017",
"value": "2020–21 Australian region cyclone season_header_cell_1_1_0"
},
{
"operation": "Highlighting",
"time": "329.124",
"value": "2020–21 Australian region cyclone season_header_cell_1_1_0"
},
{
"operation": "Highlighting",
"time": "330.197",
"value": "2020–21 Australian region cyclone season_cell_1_1_1"
},
{
"operation": "Highlighting",
"time": "330.221",
"value": "2020–21 Australian region cyclone season_header_cell_1_1_0"
},
{
"operation": "Highlighting deleted",
"time": "331.183",
"value": "2020–21 Australian region cyclone season_cell_1_1_1"
},
{
"operation": "Highlighting deleted",
"time": "331.202",
"value": "2020–21 Australian region cyclone season_header_cell_1_1_0"
},
{
"operation": "Highlighting",
"time": "331.777",
"value": "2020–21 Australian region cyclone season_cell_1_1_1"
},
{
"operation": "Highlighting",
"time": "331.795",
"value": "2020–21 Australian region cyclone season_header_cell_1_1_0"
},
{
"operation": "Highlighting",
"time": "389.453",
"value": "2020–21 Australian region cyclone season_sentence_30"
},
{
"operation": "Highlighting",
"time": "394.951",
"value": "2020–21 Australian region cyclone season_sentence_31"
},
{
"operation": "hyperlink",
"time": "407.689",
"value": "Cocos Islands"
},
{
"operation": "Now on",
"time": "408.331",
"value": "Cocos (Keeling) Islands"
},
{
"operation": "Highlighting",
"time": "429.357",
"value": "Cocos (Keeling) Islands_sentence_2"
},
{
"operation": "finish",
"time": "442.477",
"value": "finish"
}
]
}
|
exfever_2990
|
The David Geffen School of Drama at Yale University is a graduate professional school of Yale University, located in the coastal city of the U.S. state of Connecticut.
|
{
"evidence": [
"Yale School of Drama",
"New Haven, Connecticut"
],
"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 David Geffen School of Drama at Yale University is a graduate professional school of Yale University, located in the coastal city of the U.S. state of Connecticut.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Yale School of Drama",
"New Haven, Connecticut"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6026,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
6872
|
Dekker, who was not the only one to view the outbreak of plague as a divine consequence of Elizabeth's death, recalls the widespread lamentation at her death and includes in his work some of the epigrams written for her funeral at Whitehall.
|
{
"evidence": [
"The Wonderfull Yeare_sentence_10",
"The Wonderfull Yeare_sentence_9"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDekker, who was not the only one to view the outbreak of plague as a divine consequence of Elizabeth's death, recalls the widespread lamentation at her death and includes in his work some of the epigrams written for her funeral at Whitehall.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"The Wonderfull Yeare_sentence_10",
"The Wonderfull Yeare_sentence_9"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6027,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.872",
"value": "?search="
},
{
"operation": "search",
"time": "10.362",
"value": "Dekker"
},
{
"operation": "Now on",
"time": "11.236",
"value": "Dekker"
},
{
"operation": "search",
"time": "23.709",
"value": "The Wonderfull Yeare"
},
{
"operation": "Now on",
"time": "24.357",
"value": "The Wonderfull Yeare"
},
{
"operation": "Page search",
"time": "34.807",
"value": "dekker"
},
{
"operation": "Page search",
"time": "42.643",
"value": "plague"
},
{
"operation": "Highlighting",
"time": "105.01",
"value": "The Wonderfull Yeare_sentence_10"
},
{
"operation": "Page search",
"time": "117.534",
"value": "lamentation"
},
{
"operation": "Highlighting",
"time": "130.816",
"value": "The Wonderfull Yeare_sentence_9"
},
{
"operation": "Page search",
"time": "155.38",
"value": "whitehall"
},
{
"operation": "Highlighting deleted",
"time": "160.806",
"value": "The Wonderfull Yeare_sentence_9"
},
{
"operation": "Highlighting",
"time": "161.595",
"value": "The Wonderfull Yeare_sentence_9"
},
{
"operation": "finish",
"time": "183.068",
"value": "finish"
}
]
}
|
52059
|
Pamunuwa sees its visitor count increase during the Sinhalese New Year, a Sri Lankan holiday celebrating the traditional New Year of the Sinhalese People.
|
{
"evidence": [
"Pamunuwa_sentence_4",
"Sinhalese New Year_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:\nPamunuwa sees its visitor count increase during the Sinhalese New Year, a Sri Lankan holiday celebrating the traditional New Year of the Sinhalese People.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Pamunuwa_sentence_4",
"Sinhalese New Year_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6028,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.331",
"value": "?search="
},
{
"operation": "search",
"time": "13.589",
"value": "Pamunuwa"
},
{
"operation": "Now on",
"time": "14.786",
"value": "Pamunuwa"
},
{
"operation": "Highlighting",
"time": "31.209",
"value": "Pamunuwa_sentence_4"
},
{
"operation": "hyperlink",
"time": "41.521",
"value": "Sinhalese New Year"
},
{
"operation": "Now on",
"time": "42.487",
"value": "Sinhalese New Year"
},
{
"operation": "Highlighting",
"time": "45.288",
"value": "Sinhalese New Year_sentence_0"
},
{
"operation": "finish",
"time": "62.018",
"value": "finish"
}
]
}
|
32174
|
Broad money (defined differently by the European Central Bank and the Bank of England) is defined by OECD as all banknotes and coins, and has the typical OECD notation of M5.
|
{
"evidence": [
"Broad money_sentence_9"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBroad money (defined differently by the European Central Bank and the Bank of England) is defined by OECD as all banknotes and coins, and has the typical OECD notation of M5.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Broad money_sentence_9"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6029,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.957",
"value": "?search="
},
{
"operation": "search",
"time": "10.219",
"value": "Broad money"
},
{
"operation": "Now on",
"time": "11.104",
"value": "Broad money"
},
{
"operation": "Page search",
"time": "83.968",
"value": "European Central Bank"
},
{
"operation": "Page search",
"time": "102.047",
"value": "england"
},
{
"operation": "Highlighting",
"time": "104.387",
"value": "Broad money_sentence_1"
},
{
"operation": "Highlighting",
"time": "120.246",
"value": "Broad money_sentence_8"
},
{
"operation": "Highlighting",
"time": "128.705",
"value": "Broad money_sentence_9"
},
{
"operation": "Highlighting deleted",
"time": "138.324",
"value": "Broad money_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "140.895",
"value": "Broad money_sentence_1"
},
{
"operation": "finish",
"time": "146.747",
"value": "finish"
}
]
}
|
36058
|
The main building of Agios Andreas Hospital, which is the second largest hospital in Patras, was struck by an earthquake in 2008 and was restored with additional equipment.
|
{
"evidence": [
"Agios Andreas Hospital_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:\nThe main building of Agios Andreas Hospital, which is the second largest hospital in Patras, was struck by an earthquake in 2008 and was restored with additional equipment.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Agios Andreas Hospital_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6030,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.785",
"value": "?search="
},
{
"operation": "search",
"time": "12.671",
"value": "Agios Andreas Hospital"
},
{
"operation": "Now on",
"time": "13.519",
"value": "Agios Andreas Hospital"
},
{
"operation": "Highlighting",
"time": "63.462",
"value": "Agios Andreas Hospital_sentence_4"
},
{
"operation": "finish",
"time": "76.913",
"value": "finish"
}
]
}
|
exfever_8541
|
Home Alone is the most prominent live-action comedy of the 1990s.?
|
{
"evidence": [
"Home Alone",
"John Hughes (filmmaker)"
],
"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:\nHome Alone is the most prominent live-action comedy of the 1990s.?\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Home Alone",
"John Hughes (filmmaker)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6031,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3071
|
Abdul Rashid Salim Salman Khan is one of the most commercially successful actors in the film industry in India, which has more than 1800 movies produced annually, and its major centers of film production in the country include a major port city on the Malabar Coast of Russia, bordering the Laccadive Sea.
|
{
"evidence": [
"Salman Khan",
"Cinema of India",
"Kochi"
],
"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:\nAbdul Rashid Salim Salman Khan is one of the most commercially successful actors in the film industry in India, which has more than 1800 movies produced annually, and its major centers of film production in the country include a major port city on the Malabar Coast of Russia, bordering the Laccadive Sea.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Salman Khan",
"Cinema of India",
"Kochi"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6032,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8076
|
Carter was the youngest artist who released his self-titled debut album in 1997.
|
{
"evidence": [
"Aaron Carter",
"Preadolescence"
],
"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:\nCarter was the youngest artist who released his self-titled debut album in 1997.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Aaron Carter",
"Preadolescence"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6033,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4460
|
The 79th Academy Awards ceremony was hosted for the first time by a Mexican comedian who provided the voice of Dory in the film by an American computer animation studio based in Emeryville, California.
|
{
"evidence": [
"79th Academy Awards",
"Ellen DeGeneres",
"Pixar"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe 79th Academy Awards ceremony was hosted for the first time by a Mexican comedian who provided the voice of Dory in the film by an American computer animation studio based in Emeryville, California.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"79th Academy Awards",
"Ellen DeGeneres",
"Pixar"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6034,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
33982
|
KKVV is a radio station which transmits audio by radio waves which is not intended to reach a wide audience.
|
{
"evidence": [
"KKVV_sentence_0",
"Radio broadcasting_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:\nKKVV is a radio station which transmits audio by radio waves which is not intended to reach a wide audience.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"KKVV_sentence_0",
"Radio broadcasting_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6035,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.651",
"value": "?search="
},
{
"operation": "search",
"time": "20.04",
"value": "KKVV"
},
{
"operation": "Now on",
"time": "20.642",
"value": "KKVV"
},
{
"operation": "Highlighting",
"time": "37.297",
"value": "KKVV_sentence_0"
},
{
"operation": "hyperlink",
"time": "94.641",
"value": "radio station"
},
{
"operation": "Now on",
"time": "95.413",
"value": "Radio station"
},
{
"operation": "Highlighting",
"time": "100.154",
"value": "Radio broadcasting_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "126.082",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "126.137",
"value": "KKVV"
},
{
"operation": "Highlighting",
"time": "128.146",
"value": "KKVV_sentence_0"
},
{
"operation": "hyperlink",
"time": "129.042",
"value": "radio station"
},
{
"operation": "Now on",
"time": "129.405",
"value": "Radio station"
},
{
"operation": "Highlighting",
"time": "130.354",
"value": "Radio broadcasting_sentence_0"
},
{
"operation": "finish",
"time": "145.457",
"value": "finish"
}
]
}
|
exfever_3993
|
John Heard Jr. was an American actor who appeared in a number of successful films, including in a 1981 American neo-noir thriller directed by a Czech film director who is best known for his involvement in the Czechoslovak New Wave.
|
{
"evidence": [
"John Heard (actor)",
"Cutter's Way",
"Ivan Passer"
],
"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 Heard Jr. was an American actor who appeared in a number of successful films, including in a 1981 American neo-noir thriller directed by a Czech film director who is best known for his involvement in the Czechoslovak New Wave.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"John Heard (actor)",
"Cutter's Way",
"Ivan Passer"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6036,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4834
|
Mahendra Singh Dhoni is an Indian professional footballer known as a right-handed player on the fielding side who stands behind the wicket or stumps.
|
{
"evidence": [
"MS Dhoni",
"Wicket-keeper"
],
"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:\nMahendra Singh Dhoni is an Indian professional footballer known as a right-handed player on the fielding side who stands behind the wicket or stumps.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"MS Dhoni",
"Wicket-keeper"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6037,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2348
|
Cleopatra was Queen of the Ptolemaic Kingdom of Israel, a descendant of Ptolemy I Soter, who came from the ancient tribe that lived on the longest river that flows fully into Greece.
|
{
"evidence": [
"Cleopatra",
"Ancient Macedonians",
"Haliacmon"
],
"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:\nCleopatra was Queen of the Ptolemaic Kingdom of Israel, a descendant of Ptolemy I Soter, who came from the ancient tribe that lived on the longest river that flows fully into Greece.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Cleopatra",
"Ancient Macedonians",
"Haliacmon"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6038,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3658
|
In Dubious Battle is the first drama film directed and produced by James Franco.
|
{
"evidence": [
"In Dubious Battle (film)",
"Nat Wolff"
],
"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:\nIn Dubious Battle is the first drama film directed and produced by James Franco.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"In Dubious Battle (film)",
"Nat Wolff"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6039,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
93465
|
The song immediately fell to number 97 on the U.S. Billboard Hot Dance Club Play on the second day of its album release.
|
{
"evidence": [
"Dance Club Songs_sentence_2",
"Dance Club Songs_sentence_10"
],
"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 song immediately fell to number 97 on the U.S. Billboard Hot Dance Club Play on the second day of its album release.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Dance Club Songs_sentence_2",
"Dance Club Songs_sentence_10"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6040,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.434",
"value": "?search="
},
{
"operation": "search",
"time": "72.658",
"value": "Billboard Hot Dance Club Play"
},
{
"operation": "Now on",
"time": "74.268",
"value": "Dance Club Songs"
},
{
"operation": "hyperlink",
"time": "161.904",
"value": "Love Hangover 2020"
},
{
"operation": "Now on",
"time": "162.565",
"value": "Love Hangover 2020"
},
{
"operation": "back-button-clicked",
"time": "187.727",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "189.473",
"value": "Dance Club Songs"
},
{
"operation": "Highlighting",
"time": "201.178",
"value": "Dance Club Songs_sentence_2"
},
{
"operation": "Highlighting",
"time": "205.739",
"value": "Dance Club Songs_sentence_10"
},
{
"operation": "finish",
"time": "227.678",
"value": "finish"
}
]
}
|
exfever_466
|
Amy Winehouse's debut studio album, Frank, was released by a multinational music label owned by Universal Music Group.
|
{
"evidence": [
"Frank (Amy Winehouse album)",
"Island Records"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAmy Winehouse's debut studio album, Frank, was released by a multinational music label owned by Universal Music Group.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Frank (Amy Winehouse album)",
"Island Records"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6041,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8701
|
The Eighth Doctor is the fifteenth fictional character portrayed by Paul McGann.
|
{
"evidence": [
"Eighth Doctor",
"The Night of the Doctor",
"BBC iPlayer"
],
"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 Eighth Doctor is the fifteenth fictional character portrayed by Paul McGann.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Eighth Doctor",
"The Night of the Doctor",
"BBC iPlayer"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6042,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
77623
|
Changyon County is bordered to the north by Kwail and Songhwa and to the east by Samchon and Taetan and most of the land is flat plains with mountains forming its borders.
|
{
"evidence": [
"Changyon County_sentence_1",
"Changyon County_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:\nChangyon County is bordered to the north by Kwail and Songhwa and to the east by Samchon and Taetan and most of the land is flat plains with mountains forming its borders.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Changyon County_sentence_1",
"Changyon County_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6043,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.873",
"value": "?search="
},
{
"operation": "search",
"time": "6.431",
"value": "Changyon County"
},
{
"operation": "Now on",
"time": "7.741",
"value": "Changyon County"
},
{
"operation": "Highlighting",
"time": "12.025",
"value": "Changyon County_sentence_1"
},
{
"operation": "Highlighting",
"time": "17.988",
"value": "Changyon County_sentence_2"
},
{
"operation": "finish",
"time": "21.657",
"value": "finish"
}
]
}
|
33075
|
TSG101 (Tumor susceptibility gene 201) recognises short linear motif : P(T/S)AP via the UEV protein domain of the VPS23/TSG101 subunit and the structure is primarily composed of four long, parallel helical hairpins, each corresponding to a different subunit.
|
{
"evidence": [
"TSG101_sentence_10"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nTSG101 (Tumor susceptibility gene 201) recognises short linear motif : P(T/S)AP via the UEV protein domain of the VPS23/TSG101 subunit and the structure is primarily composed of four long, parallel helical hairpins, each corresponding to a different subunit.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"TSG101_sentence_10"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6044,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.427",
"value": "?search="
},
{
"operation": "search",
"time": "6.224",
"value": "TSG101"
},
{
"operation": "Now on",
"time": "7.479",
"value": "TSG101"
},
{
"operation": "Page search",
"time": "22.963",
"value": "linear"
},
{
"operation": "Highlighting",
"time": "27.181",
"value": "TSG101_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "37.54",
"value": "TSG101_sentence_8"
},
{
"operation": "Highlighting",
"time": "42.883",
"value": "TSG101_sentence_10"
},
{
"operation": "finish",
"time": "53.043",
"value": "finish"
}
]
}
|
12413
|
Marius Fourie(born 15 October 1990 in Cape Town, Canada) came in on the opening match and made his first start the following week.
|
{
"evidence": [
"Marius Fourie_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:\nMarius Fourie(born 15 October 1990 in Cape Town, Canada) came in on the opening match and made his first start the following week.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Marius Fourie_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6045,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.612",
"value": "?search="
},
{
"operation": "search",
"time": "6.314",
"value": "Marius Fourie"
},
{
"operation": "Now on",
"time": "7.451",
"value": "Marius Fourie"
},
{
"operation": "Highlighting",
"time": "11.174",
"value": "Marius Fourie_sentence_0"
},
{
"operation": "finish",
"time": "29.901",
"value": "finish"
}
]
}
|
exfever_7406
|
Stars Are Blind is a song recorded for the debut studio album, Paris by an American media personality model, singer, DJ, and actress who was born in New York City, and raised there and in Beverly Hills, California.
|
{
"evidence": [
"Stars Are Blind",
"Paris Hilton"
],
"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:\nStars Are Blind is a song recorded for the debut studio album, Paris by an American media personality model, singer, DJ, and actress who was born in New York City, and raised there and in Beverly Hills, California.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Stars Are Blind",
"Paris Hilton"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6046,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
34557
|
Love Prevails received five ratings one of which was from Christian Music Review.
|
{
"evidence": [
"Love Prevails_sentence_2",
"Love Prevails_sentence_3",
"Love Prevails_sentence_4",
"Love Prevails_sentence_5",
"Love Prevails_sentence_6",
"Love Prevails_sentence_7"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLove Prevails received five ratings one of which was from Christian Music Review.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Love Prevails_sentence_2",
"Love Prevails_sentence_3",
"Love Prevails_sentence_4",
"Love Prevails_sentence_5",
"Love Prevails_sentence_6",
"Love Prevails_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6047,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.786",
"value": "?search="
},
{
"operation": "search",
"time": "19.616",
"value": "Love Prevails"
},
{
"operation": "Now on",
"time": "20.131",
"value": "Love Prevails"
},
{
"operation": "Highlighting",
"time": "24.482",
"value": "Love Prevails_cell_0_2_0"
},
{
"operation": "Highlighting",
"time": "24.486",
"value": "Love Prevails_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "24.487",
"value": "Love Prevails_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "24.97",
"value": "Love Prevails_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "25.484",
"value": "Love Prevails_cell_0_4_0"
},
{
"operation": "Highlighting",
"time": "25.996",
"value": "Love Prevails_cell_0_5_0"
},
{
"operation": "Highlighting",
"time": "26.494",
"value": "Love Prevails_cell_0_6_0"
},
{
"operation": "Highlighting",
"time": "27.246",
"value": "Love Prevails_cell_0_4_1"
},
{
"operation": "Highlighting",
"time": "27.248",
"value": "Love Prevails_header_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "29.513",
"value": "Love Prevails_cell_0_4_1"
},
{
"operation": "Highlighting deleted",
"time": "29.517",
"value": "Love Prevails_header_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "168.73",
"value": "Love Prevails_cell_0_2_0"
},
{
"operation": "Highlighting deleted",
"time": "169.17",
"value": "Love Prevails_cell_0_3_0"
},
{
"operation": "Highlighting deleted",
"time": "169.61",
"value": "Love Prevails_cell_0_4_0"
},
{
"operation": "Highlighting deleted",
"time": "170.081",
"value": "Love Prevails_cell_0_5_0"
},
{
"operation": "Highlighting deleted",
"time": "170.554",
"value": "Love Prevails_cell_0_6_0"
},
{
"operation": "Highlighting deleted",
"time": "170.561",
"value": "Love Prevails_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "170.562",
"value": "Love Prevails_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "173.124",
"value": "Love Prevails_sentence_2"
},
{
"operation": "Highlighting",
"time": "173.877",
"value": "Love Prevails_sentence_3"
},
{
"operation": "Highlighting",
"time": "174.428",
"value": "Love Prevails_sentence_4"
},
{
"operation": "Highlighting",
"time": "175.957",
"value": "Love Prevails_sentence_5"
},
{
"operation": "Highlighting",
"time": "176.638",
"value": "Love Prevails_sentence_6"
},
{
"operation": "Highlighting",
"time": "177.254",
"value": "Love Prevails_sentence_7"
},
{
"operation": "finish",
"time": "218.718",
"value": "finish"
}
]
}
|
exfever_2345
|
The Formula is a 1980 American-West German mystery film directed by a Canadian film director who is best known for directing the film Rocky.
|
{
"evidence": [
"The Formula (1980 film)",
"John G. Avildsen"
],
"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 Formula is a 1980 American-West German mystery film directed by a Canadian film director who is best known for directing the film Rocky.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Formula (1980 film)",
"John G. Avildsen"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6048,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
25072
|
Netsanet Gudeta represented Ethiopia at the 2014 IAAF World Half Marathon Championships where she took sixth place and won the team silver and again at the IAAF World Cross Country Championships 2015 where she won bronze.
|
{
"evidence": [
"Netsanet Gudeta_sentence_9",
"Netsanet Gudeta_sentence_10",
"Netsanet Gudeta_sentence_15"
],
"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:\nNetsanet Gudeta represented Ethiopia at the 2014 IAAF World Half Marathon Championships where she took sixth place and won the team silver and again at the IAAF World Cross Country Championships 2015 where she won bronze.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Netsanet Gudeta_sentence_9",
"Netsanet Gudeta_sentence_10",
"Netsanet Gudeta_sentence_15"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6049,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.274",
"value": "?search="
},
{
"operation": "search",
"time": "13.474",
"value": "Netsanet Gudeta"
},
{
"operation": "Now on",
"time": "14.3",
"value": "Netsanet Gudeta"
},
{
"operation": "Highlighting",
"time": "33.904",
"value": "Netsanet Gudeta_header_cell_0_2_0"
},
{
"operation": "Highlighting deleted",
"time": "52.224",
"value": "Netsanet Gudeta_header_cell_0_2_0"
},
{
"operation": "Highlighting",
"time": "66.334",
"value": "Netsanet Gudeta_sentence_9"
},
{
"operation": "Highlighting",
"time": "77.481",
"value": "Netsanet Gudeta_sentence_10"
},
{
"operation": "Highlighting",
"time": "105.407",
"value": "Netsanet Gudeta_sentence_15"
},
{
"operation": "finish",
"time": "129.895",
"value": "finish"
}
]
}
|
90790
|
Potawatomi Islands are islands that transition from Green Bay to Lake Michigan, the largest of these is Washington Island and others include Detroit Island.
|
{
"evidence": [
"Potawatomi Islands_sentence_0",
"Potawatomi Islands_sentence_1",
"Potawatomi Islands_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:\nPotawatomi Islands are islands that transition from Green Bay to Lake Michigan, the largest of these is Washington Island and others include Detroit Island.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Potawatomi Islands_sentence_0",
"Potawatomi Islands_sentence_1",
"Potawatomi Islands_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6050,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.605",
"value": "?search="
},
{
"operation": "search",
"time": "13.534",
"value": "Potawatomi Islands"
},
{
"operation": "Now on",
"time": "14.327",
"value": "Potawatomi Islands"
},
{
"operation": "Highlighting",
"time": "17.482",
"value": "Potawatomi Islands_sentence_0"
},
{
"operation": "Highlighting",
"time": "19.981",
"value": "Potawatomi Islands_sentence_1"
},
{
"operation": "hyperlink",
"time": "21.477",
"value": "Plum Island"
},
{
"operation": "Now on",
"time": "22.049",
"value": "Plum Island (Wisconsin)"
},
{
"operation": "back-button-clicked",
"time": "22.822",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "23.005",
"value": "Potawatomi Islands"
},
{
"operation": "Highlighting",
"time": "23.743",
"value": "Potawatomi Islands_sentence_2"
},
{
"operation": "finish",
"time": "28.072",
"value": "finish"
}
]
}
|
exfever_7846
|
Shinsegae is a South Korean department store franchise with its flagship store and the world's largest department store in South Korea's fifth-most populous city with a population of over 3.4 million inhabitants.
|
{
"evidence": [
"Shinsegae",
"Busan"
],
"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:\nShinsegae is a South Korean department store franchise with its flagship store and the world's largest department store in South Korea's fifth-most populous city with a population of over 3.4 million inhabitants.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Shinsegae",
"Busan"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6051,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4783
|
Precious is Gabourey Sidibe's last film.
|
{
"evidence": [
"Precious (film)",
"Push (novel)"
],
"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:\nPrecious is Gabourey Sidibe's last film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Precious (film)",
"Push (novel)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6052,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
88376
|
Sir Harry was a good friend of the Prince of Wales (later King George IV), who stayed at Uppark, the same place he took the teenaged Emma Hamilton as a mistress to live with him.
|
{
"evidence": [
"Henry Fetherstonhaugh_sentence_4",
"Henry Fetherstonhaugh_sentence_5"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSir Harry was a good friend of the Prince of Wales (later King George IV), who stayed at Uppark, the same place he took the teenaged Emma Hamilton as a mistress to live with him.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Henry Fetherstonhaugh_sentence_4",
"Henry Fetherstonhaugh_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6053,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.593",
"value": "?search="
},
{
"operation": "search",
"time": "17.077",
"value": "Sir Harry"
},
{
"operation": "Now on",
"time": "17.938",
"value": "Sir Harry"
},
{
"operation": "search",
"time": "33.349",
"value": "King George IV"
},
{
"operation": "Now on",
"time": "35.203",
"value": "George IV"
},
{
"operation": "Page search",
"time": "39.977",
"value": "sir harry"
},
{
"operation": "Page search",
"time": "80.52",
"value": "emma"
},
{
"operation": "search",
"time": "154.991",
"value": "Emma Hilton"
},
{
"operation": "Now on",
"time": "156.765",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "180.968",
"value": "next 20"
},
{
"operation": "Now on",
"time": "181.948",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "209.544",
"value": "next 20"
},
{
"operation": "Now on",
"time": "210.341",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "235.473",
"value": "next 20"
},
{
"operation": "Now on",
"time": "236.612",
"value": "undefined"
},
{
"operation": "search",
"time": "257.14",
"value": "Sir Harry + Emma"
},
{
"operation": "Now on",
"time": "258.41",
"value": "undefined"
},
{
"operation": "search",
"time": "299.732",
"value": "King George IV"
},
{
"operation": "Now on",
"time": "300.842",
"value": "George IV"
},
{
"operation": "Highlighting",
"time": "329.603",
"value": "George IV_sentence_1"
},
{
"operation": "Highlighting",
"time": "329.605",
"value": "George IV_title"
},
{
"operation": "search",
"time": "332.928",
"value": "George Augustus Frederick"
},
{
"operation": "Now on",
"time": "334.005",
"value": "George Augustus Frederick"
},
{
"operation": "hyperlink",
"time": "337.095",
"value": "George IV of the United Kingdom"
},
{
"operation": "hyperlink",
"time": "337.207",
"value": "George IV of the United Kingdom"
},
{
"operation": "Now on",
"time": "338.511",
"value": "George IV"
},
{
"operation": "Highlighting deleted",
"time": "339.75",
"value": "George IV_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "339.752",
"value": "George IV_title"
},
{
"operation": "search",
"time": "363.164",
"value": "Sir Harry Prince of Wales"
},
{
"operation": "Now on",
"time": "365.615",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "386.718",
"value": "Emma, Lady Hamilton"
},
{
"operation": "Now on",
"time": "388.124",
"value": "Emma, Lady Hamilton"
},
{
"operation": "hyperlink",
"time": "397.505",
"value": "Lord Nelson"
},
{
"operation": "Now on",
"time": "399.805",
"value": "Horatio Nelson, 1st Viscount Nelson"
},
{
"operation": "search",
"time": "448.181",
"value": "Sir Harry was a good friend of the Prince of Wales"
},
{
"operation": "Now on",
"time": "450.165",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "453.649",
"value": "Henry Fetherstonhaugh"
},
{
"operation": "Now on",
"time": "454.325",
"value": "Henry Fetherstonhaugh"
},
{
"operation": "Highlighting",
"time": "457.523",
"value": "Henry Fetherstonhaugh_sentence_4"
},
{
"operation": "Highlighting",
"time": "465.146",
"value": "Henry Fetherstonhaugh_sentence_5"
},
{
"operation": "finish",
"time": "499.723",
"value": "finish"
}
]
}
|
exfever_1921
|
Serena Jameka Williams is an American professional tennis player who has been ranked singles world No. 1 by the principal organizing body of women's professional tennis.
|
{
"evidence": [
"Serena Williams",
"Women's Tennis Association"
],
"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:\nSerena Jameka Williams is an American professional tennis player who has been ranked singles world No. 1 by the principal organizing body of women's professional tennis.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Serena Williams",
"Women's Tennis Association"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6054,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
72384
|
The single "Don't Say You Love Me" from the album, This Road Goes Everywhere, did not chart in Canada.
|
{
"evidence": [
"This Road Goes Everywhere_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 single \"Don't Say You Love Me\" from the album, This Road Goes Everywhere, did not chart in Canada.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"This Road Goes Everywhere_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6055,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.279",
"value": "?search="
},
{
"operation": "search",
"time": "27.633",
"value": "Don't Say You Love Me"
},
{
"operation": "Now on",
"time": "28.991",
"value": "Don't Say You Love Me"
},
{
"operation": "search",
"time": "62.73",
"value": "This Road Goes Everywhere"
},
{
"operation": "Now on",
"time": "63.899",
"value": "This Road Goes Everywhere"
},
{
"operation": "Highlighting",
"time": "83.568",
"value": "This Road Goes Everywhere_sentence_2"
},
{
"operation": "finish",
"time": "111.406",
"value": "finish"
}
]
}
|
exfever_4692
|
The Leftovers is a supernatural drama television series that follows the life of police chief Kevin Garvey, portrayed by an American actor and filmmaker who was born on August 10, 1971.
|
{
"evidence": [
"The Leftovers (TV series)",
"Justin Theroux"
],
"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 Leftovers is a supernatural drama television series that follows the life of police chief Kevin Garvey, portrayed by an American actor and filmmaker who was born on August 10, 1971.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Leftovers (TV series)",
"Justin Theroux"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6056,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3341
|
The Lion King is a 1994 American animated musical drama film produced by an American animation studio that creates animated features, headquartered at the Roy E. Disney Animation Building in Burbank, California, the United States, which serves as the corporate headquarters for The Walt Disney Company media conglomerate.
|
{
"evidence": [
"The Lion King",
"Walt Disney Animation Studios",
"Walt Disney Studios (Burbank)"
],
"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 Lion King is a 1994 American animated musical drama film produced by an American animation studio that creates animated features, headquartered at the Roy E. Disney Animation Building in Burbank, California, the United States, which serves as the corporate headquarters for The Walt Disney Company media conglomerate.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Lion King",
"Walt Disney Animation Studios",
"Walt Disney Studios (Burbank)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6057,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
65242
|
Laura Ashley plc, family originally lived above the shop for five years before moving to Carno, Montgomeryshire and it was originally located in the social club in Carno.
|
{
"evidence": [
"Laura Ashley plc_sentence_24"
],
"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:\nLaura Ashley plc, family originally lived above the shop for five years before moving to Carno, Montgomeryshire and it was originally located in the social club in Carno.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Laura Ashley plc_sentence_24"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6058,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.78",
"value": "?search="
},
{
"operation": "search",
"time": "12.533",
"value": "Laura Ashley plc"
},
{
"operation": "Now on",
"time": "13.594",
"value": "Laura Ashley plc"
},
{
"operation": "Page search",
"time": "29.452",
"value": "family"
},
{
"operation": "Highlighting",
"time": "41.54",
"value": "Laura Ashley plc_sentence_24"
},
{
"operation": "finish",
"time": "51.273",
"value": "finish"
}
]
}
|
36217
|
The Australian Surf Life Saving Championships, held annually since 1915, is the largest surf lifesaving event in Australia and the largest even of its kind in the world.
|
{
"evidence": [
"Australian Surf Life Saving Championships_sentence_2",
"Australian Surf Life Saving Championships_sentence_1"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Australian Surf Life Saving Championships, held annually since 1915, is the largest surf lifesaving event in Australia and the largest even of its kind in the world.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Australian Surf Life Saving Championships_sentence_2",
"Australian Surf Life Saving Championships_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6059,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.652",
"value": "?search="
},
{
"operation": "search",
"time": "15.289",
"value": "Australian Surf Life Saving Championships"
},
{
"operation": "Now on",
"time": "16.243",
"value": "Australian Surf Life Saving Championships"
},
{
"operation": "Highlighting",
"time": "20.105",
"value": "Australian Surf Life Saving Championships_sentence_2"
},
{
"operation": "Highlighting",
"time": "28.843",
"value": "Australian Surf Life Saving Championships_sentence_1"
},
{
"operation": "finish",
"time": "47.647",
"value": "finish"
}
]
}
|
exfever_2648
|
Christopher Michael Pratt rose to prominence for his television roles, particularly as Andy Dwyer in an American political satire mockumentary sitcom television series created by Greg Daniels and Michael Schur.
|
{
"evidence": [
"Chris Pratt",
"Parks and Recreation"
],
"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:\nChristopher Michael Pratt rose to prominence for his television roles, particularly as Andy Dwyer in an American political satire mockumentary sitcom television series created by Greg Daniels and Michael Schur.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Chris Pratt",
"Parks and Recreation"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6060,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
75513
|
BRM P180 was a Formula One racing car, built by BRM and designed by Tony Southgate which raced in the 1972 Formula One season, and It was powered by a BRM 3.0-litre V12 engine. One of the main features of the P180 was that the radiators had moved to the rear of the car, allowing the nose of the car to be very wide and flat.
|
{
"evidence": [
"BRM P180_sentence_0",
"BRM P180_sentence_1",
"BRM P180_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:\nBRM P180 was a Formula One racing car, built by BRM and designed by Tony Southgate which raced in the 1972 Formula One season, and It was powered by a BRM 3.0-litre V12 engine. One of the main features of the P180 was that the radiators had moved to the rear of the car, allowing the nose of the car to be very wide and flat.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"BRM P180_sentence_0",
"BRM P180_sentence_1",
"BRM P180_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6061,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.882",
"value": "undefined"
},
{
"operation": "search",
"time": "9.629",
"value": "BRM P180"
},
{
"operation": "Now on",
"time": "10.958",
"value": "BRM P180"
},
{
"operation": "Highlighting",
"time": "15.618",
"value": "BRM P180_sentence_0"
},
{
"operation": "Highlighting",
"time": "20.794",
"value": "BRM P180_sentence_1"
},
{
"operation": "Highlighting",
"time": "29.984",
"value": "BRM P180_sentence_2"
},
{
"operation": "finish",
"time": "42.53",
"value": "finish"
}
]
}
|
67045
|
Meredith Gwynne Evans, a British physical chemist who made important theoretical contributions in the study of chemical reaction rates and reaction mechanisms, attended the elementary school at which his father was Headmaster and was the eldest son in a family of three sons and one daughter.
|
{
"evidence": [
"Meredith Gwynne Evans_sentence_0",
"Meredith Gwynne Evans_sentence_4",
"Meredith Gwynne Evans_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:\nMeredith Gwynne Evans, a British physical chemist who made important theoretical contributions in the study of chemical reaction rates and reaction mechanisms, attended the elementary school at which his father was Headmaster and was the eldest son in a family of three sons and one daughter.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Meredith Gwynne Evans_sentence_0",
"Meredith Gwynne Evans_sentence_4",
"Meredith Gwynne Evans_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6062,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.806",
"value": "?search="
},
{
"operation": "search",
"time": "9.242",
"value": "Meredith Gwynne Evans"
},
{
"operation": "Now on",
"time": "16.718",
"value": "Meredith Gwynne Evans"
},
{
"operation": "Highlighting",
"time": "52.585",
"value": "Meredith Gwynne Evans_sentence_0"
},
{
"operation": "Page search",
"time": "69.986",
"value": "elementary"
},
{
"operation": "Highlighting",
"time": "92.651",
"value": "Meredith Gwynne Evans_sentence_4"
},
{
"operation": "Highlighting",
"time": "100.201",
"value": "Meredith Gwynne Evans_sentence_3"
},
{
"operation": "finish",
"time": "449.413",
"value": "finish"
}
]
}
|
60005
|
Iara oil field is a large ultra-deepwater pre-salt oil field located 230 km (140 mi) off the coast of Portugal discovered in 2008, and is now partially owned and operated by Petrobras with an estimated 3,000−4,000 million barrels of oil once in production.
|
{
"evidence": [
"Iara Oil Field_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:\nIara oil field is a large ultra-deepwater pre-salt oil field located 230 km (140 mi) off the coast of Portugal discovered in 2008, and is now partially owned and operated by Petrobras with an estimated 3,000−4,000 million barrels of oil once in production.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Iara Oil Field_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6063,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.06",
"value": "?search="
},
{
"operation": "search",
"time": "9.412",
"value": "Iara Oil Field"
},
{
"operation": "Now on",
"time": "10.422",
"value": "Iara Oil Field"
},
{
"operation": "Highlighting",
"time": "17.793",
"value": "Iara Oil Field_sentence_0"
},
{
"operation": "finish",
"time": "44.157",
"value": "finish"
}
]
}
|
exfever_2820
|
George Dewey Cukor was a Canadian moviemaker and film producer who controlled a film's artistic and dramatic aspects and visualized the screenplay while guiding a group of people hired by a production company for the purpose of producing a film.
|
{
"evidence": [
"George Cukor",
"Film director",
"Film crew"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nGeorge Dewey Cukor was a Canadian moviemaker and film producer who controlled a film's artistic and dramatic aspects and visualized the screenplay while guiding a group of people hired by a production company for the purpose of producing a film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"George Cukor",
"Film director",
"Film crew"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6064,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
52187
|
Syracuse is a city in the county seat of Onodaga County, New York and Henninger High School is one of four high schools within the Syracuse School District.
|
{
"evidence": [
"Syracuse, New York_sentence_2",
"Henninger High School_sentence_1"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSyracuse is a city in the county seat of Onodaga County, New York and Henninger High School is one of four high schools within the Syracuse School District.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Syracuse, New York_sentence_2",
"Henninger High School_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6065,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.842",
"value": "?search="
},
{
"operation": "search",
"time": "7.932",
"value": "Syracuse"
},
{
"operation": "Now on",
"time": "8.663",
"value": "Syracuse"
},
{
"operation": "hyperlink",
"time": "27.74",
"value": "Syracuse, New York"
},
{
"operation": "Now on",
"time": "28.837",
"value": "Syracuse, New York"
},
{
"operation": "Highlighting",
"time": "40.788",
"value": "Syracuse, New York_sentence_2"
},
{
"operation": "search",
"time": "103.388",
"value": "Henninger High School"
},
{
"operation": "Now on",
"time": "104.366",
"value": "Henninger High School"
},
{
"operation": "Highlighting",
"time": "109.196",
"value": "Henninger High School_sentence_1"
},
{
"operation": "finish",
"time": "128.475",
"value": "finish"
}
]
}
|
88643
|
Henry Darwin was called to the Bar at Lincoln's Inn in 1963 and then worked as Director-General Legal Secretariat American Communities Brussels 1963-1986.
|
{
"evidence": [
"Henry Darwin_sentence_4",
"Henry Darwin_sentence_7"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHenry Darwin was called to the Bar at Lincoln's Inn in 1963 and then worked as Director-General Legal Secretariat American Communities Brussels 1963-1986.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Henry Darwin_sentence_4",
"Henry Darwin_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6066,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.2",
"value": "?search="
},
{
"operation": "search",
"time": "11.652",
"value": "Henry Darwin was called to the Bar at Lincoln's Inn"
},
{
"operation": "Now on",
"time": "14.039",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "22.386",
"value": "Biography Henry Darwin_section_0"
},
{
"operation": "Now on",
"time": "22.99",
"value": "Henry Darwin#Biography Henry Darwin section 0"
},
{
"operation": "Highlighting",
"time": "46.544",
"value": "Henry Darwin_sentence_4"
},
{
"operation": "Highlighting",
"time": "54.472",
"value": "Henry Darwin_sentence_7"
},
{
"operation": "finish",
"time": "58.853",
"value": "finish"
}
]
}
|
exfever_1527
|
Mel Columcille Gerard Gibson is an American actor who portrayed the fictional character Martin Riggs in the buddy cop film series Lethal Weapon.
|
{
"evidence": [
"Mel Gibson",
"Martin Riggs"
],
"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:\nMel Columcille Gerard Gibson is an American actor who portrayed the fictional character Martin Riggs in the buddy cop film series Lethal Weapon.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mel Gibson",
"Martin Riggs"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6067,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2434
|
Wilhelmina Vivian Slater is a fictional character in the American dramedy series Ugly Betty, who gained a supporting actress nomination at the Primetime Emmy Awards were held on Sunday, July 10, 2010, at the newly opened Nokia Theatre in Los Angeles, California.
|
{
"evidence": [
"Wilhelmina Slater",
"60th Primetime Emmy 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:\nWilhelmina Vivian Slater is a fictional character in the American dramedy series Ugly Betty, who gained a supporting actress nomination at the Primetime Emmy Awards were held on Sunday, July 10, 2010, at the newly opened Nokia Theatre in Los Angeles, California.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Wilhelmina Slater",
"60th Primetime Emmy Awards"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6068,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3219
|
RB Leipzig is based in Leipzig, in a landlocked state in Germany that uses the minority language spoken by Sorbs.
|
{
"evidence": [
"RB Leipzig",
"Saxony",
"Upper Sorbian language"
],
"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:\nRB Leipzig is based in Leipzig, in a landlocked state in Germany that uses the minority language spoken by Sorbs.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"RB Leipzig",
"Saxony",
"Upper Sorbian language"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6069,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
53583
|
Mount Eerie's ninth album, Now Only, has several songs that incorporate multiple tempo and instrument changes, and one of the songs, "Distortion" deals with the ideas of memory and mortality, and is a reflection on the death of Phil Elverum's wife, Genevieve Castree who died of pancreatic cancer in 2016. A studio album by Mount Eerie, Now Only, has several songs that incorporate multiple tempo and instrument changes, and one of the songs, "Distortion" deals with the ideas of memory and mortality.
|
{
"evidence": [
"Now Only_sentence_0",
"Now Only_sentence_35",
"Now Only_sentence_33",
"Now Only_sentence_1",
"Now Only_sentence_3",
"Now Only_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:\nMount Eerie's ninth album, Now Only, has several songs that incorporate multiple tempo and instrument changes, and one of the songs, \"Distortion\" deals with the ideas of memory and mortality, and is a reflection on the death of Phil Elverum's wife, Genevieve Castree who died of pancreatic cancer in 2016. A studio album by Mount Eerie, Now Only, has several songs that incorporate multiple tempo and instrument changes, and one of the songs, \"Distortion\" deals with the ideas of memory and mortality.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Now Only_sentence_0",
"Now Only_sentence_35",
"Now Only_sentence_33",
"Now Only_sentence_1",
"Now Only_sentence_3",
"Now Only_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6070,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.281",
"value": "?search="
},
{
"operation": "search",
"time": "36.68",
"value": "Mount Eerie (album)"
},
{
"operation": "Now on",
"time": "37.701",
"value": "Mount Eerie (album)"
},
{
"operation": "back-button-clicked",
"time": "43.253",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "43.336",
"value": "?search="
},
{
"operation": "search",
"time": "49.831",
"value": "Now Only"
},
{
"operation": "Now on",
"time": "51.697",
"value": "Now Only"
},
{
"operation": "Highlighting",
"time": "55.058",
"value": "Now Only_sentence_0"
},
{
"operation": "Highlighting",
"time": "101.791",
"value": "Now Only_sentence_35"
},
{
"operation": "Highlighting",
"time": "148.663",
"value": "Now Only_sentence_33"
},
{
"operation": "Highlighting",
"time": "301.425",
"value": "Now Only_sentence_1"
},
{
"operation": "Highlighting",
"time": "307.609",
"value": "Now Only_sentence_3"
},
{
"operation": "Highlighting",
"time": "308.927",
"value": "Now Only_sentence_4"
},
{
"operation": "finish",
"time": "355.085",
"value": "finish"
}
]
}
|
15654
|
After growing up in Guildford in Surrey, England, Kathryn Croft became a full-time writer.
|
{
"evidence": [
"Kathryn Croft_sentence_2",
"Guildford_sentence_3",
"Kathryn Croft_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:\nAfter growing up in Guildford in Surrey, England, Kathryn Croft became a full-time writer.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Kathryn Croft_sentence_2",
"Guildford_sentence_3",
"Kathryn Croft_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6071,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.468",
"value": "?search="
},
{
"operation": "search",
"time": "39.926",
"value": "Kathryn Croft"
},
{
"operation": "Now on",
"time": "40.591",
"value": "Kathryn Croft"
},
{
"operation": "Highlighting",
"time": "59.601",
"value": "Kathryn Croft_sentence_2"
},
{
"operation": "hyperlink",
"time": "63.542",
"value": "Guildford"
},
{
"operation": "Now on",
"time": "64.493",
"value": "Guildford"
},
{
"operation": "Highlighting",
"time": "66.386",
"value": "Guildford_sentence_3"
},
{
"operation": "back-button-clicked",
"time": "67.61",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "67.658",
"value": "Kathryn Croft"
},
{
"operation": "Highlighting",
"time": "74.821",
"value": "Kathryn Croft_sentence_4"
},
{
"operation": "finish",
"time": "87.681",
"value": "finish"
}
]
}
|
82390
|
In the Physick Book of Deliverance Dane, the debut novel by Katherine Howe which was a New York Times bestseller, Harvard graduate student Connie Goodwin tries to spend her summer doing research for her doctoral dissertation, but she has to take care of the sale of her grandmother's home, where she discovers an ancient key.
|
{
"evidence": [
"The Physick Book of Deliverance Dane_sentence_0",
"The Physick Book of Deliverance Dane_sentence_2",
"The Physick Book of Deliverance Dane_sentence_5",
"The Physick Book of Deliverance Dane_sentence_6",
"The Physick Book of Deliverance Dane_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:\nIn the Physick Book of Deliverance Dane, the debut novel by Katherine Howe which was a New York Times bestseller, Harvard graduate student Connie Goodwin tries to spend her summer doing research for her doctoral dissertation, but she has to take care of the sale of her grandmother's home, where she discovers an ancient key.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"The Physick Book of Deliverance Dane_sentence_0",
"The Physick Book of Deliverance Dane_sentence_2",
"The Physick Book of Deliverance Dane_sentence_5",
"The Physick Book of Deliverance Dane_sentence_6",
"The Physick Book of Deliverance Dane_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6072,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.756",
"value": "?search="
},
{
"operation": "search",
"time": "25.316",
"value": "Physick Book of Deliverance Dane"
},
{
"operation": "Now on",
"time": "26.603",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "35.663",
"value": "Physick"
},
{
"operation": "Now on",
"time": "36.126",
"value": "Physick"
},
{
"operation": "hyperlink",
"time": "39.529",
"value": "The Physick Book of Deliverance Dane"
},
{
"operation": "Now on",
"time": "39.963",
"value": "The Physick Book of Deliverance Dane"
},
{
"operation": "Highlighting",
"time": "42.355",
"value": "The Physick Book of Deliverance Dane_sentence_0"
},
{
"operation": "Highlighting",
"time": "50.016",
"value": "The Physick Book of Deliverance Dane_sentence_2"
},
{
"operation": "Highlighting",
"time": "57.971",
"value": "The Physick Book of Deliverance Dane_sentence_5"
},
{
"operation": "Highlighting",
"time": "66.687",
"value": "The Physick Book of Deliverance Dane_sentence_6"
},
{
"operation": "Highlighting",
"time": "68.766",
"value": "The Physick Book of Deliverance Dane_sentence_7"
},
{
"operation": "finish",
"time": "87.061",
"value": "finish"
}
]
}
|
exfever_5489
|
The Khmer Empire is the term that historians use to refer to Cambodia when the nation was a Hindu-Buddhist empire in the geographical southeastern subregion of the continent, which is bothered to the east by a geographic region that includes Australia, Melanesia, Micronesia, and Polynesia.
|
{
"evidence": [
"Khmer Empire",
"Southeast Asia",
"Oceania"
],
"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 Khmer Empire is the term that historians use to refer to Cambodia when the nation was a Hindu-Buddhist empire in the geographical southeastern subregion of the continent, which is bothered to the east by a geographic region that includes Australia, Melanesia, Micronesia, and Polynesia.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Khmer Empire",
"Southeast Asia",
"Oceania"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6073,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
82200
|
Bob Margarita was an American football halfback and coach, he played for the Chicago Bears and took the head coaching job at Georgetown University.
|
{
"evidence": [
"Bob Margarita_sentence_0",
"Bob Margarita_sentence_2",
"Bob Margarita_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:\nBob Margarita was an American football halfback and coach, he played for the Chicago Bears and took the head coaching job at Georgetown University.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bob Margarita_sentence_0",
"Bob Margarita_sentence_2",
"Bob Margarita_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6074,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.784",
"value": "?search="
},
{
"operation": "search",
"time": "14.382",
"value": "Bob Margarita"
},
{
"operation": "Now on",
"time": "15.282",
"value": "Bob Margarita"
},
{
"operation": "Highlighting",
"time": "19.503",
"value": "Bob Margarita_sentence_0"
},
{
"operation": "Highlighting",
"time": "24.885",
"value": "Bob Margarita_sentence_2"
},
{
"operation": "Highlighting",
"time": "30.923",
"value": "Bob Margarita_sentence_3"
},
{
"operation": "finish",
"time": "42.285",
"value": "finish"
}
]
}
|
exfever_8280
|
Franklin Delano Roosevelt was the most famous president of the United State.
|
{
"evidence": [
"Franklin D. Roosevelt",
"New Deal",
"United States Congress"
],
"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:\nFranklin Delano Roosevelt was the most famous president of the United State.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Franklin D. Roosevelt",
"New Deal",
"United States Congress"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6075,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7312
|
Carrie Fisher's film credits include a 1980 American musical comedy film directed by John Landis.
|
{
"evidence": [
"Carrie Fisher",
"The Blues Brothers (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:\nCarrie Fisher's film credits include a 1980 American musical comedy film directed by John Landis.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Carrie Fisher",
"The Blues Brothers (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6076,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3272
|
"California Gurls" is a song by American singer Taylor Swift that was written in response to a song performed by American rapper Jay-Z with vocals by American singer Alicia Keys from Jay-Z's 11th studio album, The Blueprint 3.
|
{
"evidence": [
"California Gurls",
"Empire State of Mind"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n\"California Gurls\" is a song by American singer Taylor Swift that was written in response to a song performed by American rapper Jay-Z with vocals by American singer Alicia Keys from Jay-Z's 11th studio album, The Blueprint 3.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"California Gurls",
"Empire State of Mind"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6077,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1815
|
Catching Fire was the most novel in America.
|
{
"evidence": [
"Catching Fire",
"Peeta Mellark",
"Suzanne Collins"
],
"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:\nCatching Fire was the most novel in America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Catching Fire",
"Peeta Mellark",
"Suzanne Collins"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6078,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
93611
|
Lee was elected to the Nevada Assembly in 1996 and served in the Assembly until 2000; then he was elected to the Nevada Senate in 2004, where he served as a member of the Legislative Operations and Elections, and the Energy, Infrastructure, and Transportation committees.
|
{
"evidence": [
"John Jay Lee_sentence_13",
"John Jay Lee_sentence_14"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLee was elected to the Nevada Assembly in 1996 and served in the Assembly until 2000; then he was elected to the Nevada Senate in 2004, where he served as a member of the Legislative Operations and Elections, and the Energy, Infrastructure, and Transportation committees.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"John Jay Lee_sentence_13",
"John Jay Lee_sentence_14"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6079,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.001",
"value": "?search="
},
{
"operation": "search",
"time": "0.013",
"value": "Nevada Assembly"
},
{
"operation": "Now on",
"time": "0.014",
"value": "Nevada Assembly"
},
{
"operation": "hyperlink",
"time": "0.08",
"value": "Nevada Legislature"
},
{
"operation": "Now on",
"time": "0.081",
"value": "Nevada Legislature"
},
{
"operation": "Highlighting",
"time": "0.082",
"value": "Nevada Legislature_item_0_5"
},
{
"operation": "Highlighting",
"time": "0.082",
"value": "Nevada Legislature_title"
},
{
"operation": "Highlighting",
"time": "0.082",
"value": "Nevada Legislature_section_3"
},
{
"operation": "Highlighting deleted",
"time": "0.085",
"value": "Nevada Legislature_item_0_5"
},
{
"operation": "Highlighting deleted",
"time": "0.085",
"value": "Nevada Legislature_title"
},
{
"operation": "Highlighting deleted",
"time": "0.085",
"value": "Nevada Legislature_section_3"
},
{
"operation": "Highlighting",
"time": "0.086",
"value": "Nevada Legislature_item_0_9"
},
{
"operation": "Highlighting",
"time": "0.086",
"value": "Nevada Legislature_title"
},
{
"operation": "Highlighting",
"time": "0.086",
"value": "Nevada Legislature_section_3"
},
{
"operation": "Highlighting deleted",
"time": "0.133",
"value": "Nevada Legislature_item_0_9"
},
{
"operation": "Highlighting deleted",
"time": "0.133",
"value": "Nevada Legislature_title"
},
{
"operation": "Highlighting deleted",
"time": "0.133",
"value": "Nevada Legislature_section_3"
},
{
"operation": "search",
"time": "0.166",
"value": "Lee was elected to the Nevada Assembly in 1996 and served in the Assembly until 2000"
},
{
"operation": "Now on",
"time": "0.17",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "0.172",
"value": "John Jay Lee"
},
{
"operation": "Now on",
"time": "0.173",
"value": "John Jay Lee"
},
{
"operation": "Highlighting",
"time": "0.177",
"value": "John Jay Lee_sentence_13"
},
{
"operation": "Highlighting",
"time": "0.18",
"value": "John Jay Lee_sentence_14"
},
{
"operation": "finish",
"time": "0.229",
"value": "finish"
},
{
"operation": "Now on",
"time": "1622841137.179",
"value": "John Jay Lee"
},
{
"operation": "finish",
"time": "1622841139.26",
"value": "finish"
}
]
}
|
69405
|
Hans Landa first appeared in Inglorious Bastards and was played by Brad Pitt.
|
{
"evidence": [
"Hans Landa_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:\nHans Landa first appeared in Inglorious Bastards and was played by Brad Pitt.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Hans Landa_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6080,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.992",
"value": "?search="
},
{
"operation": "search",
"time": "9.955",
"value": "Hans Landa"
},
{
"operation": "Now on",
"time": "11.275",
"value": "Hans Landa"
},
{
"operation": "Highlighting",
"time": "42.6",
"value": "Hans Landa_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "42.605",
"value": "Hans Landa_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "42.605",
"value": "Hans Landa_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "50.999",
"value": "Hans Landa_sentence_2"
},
{
"operation": "hyperlink",
"time": "61.456",
"value": "Inglourious Basterds"
},
{
"operation": "Now on",
"time": "62.564",
"value": "Inglourious Basterds"
},
{
"operation": "back-button-clicked",
"time": "67.173",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "67.294",
"value": "Hans Landa"
},
{
"operation": "Highlighting deleted",
"time": "68.617",
"value": "Hans Landa_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "68.622",
"value": "Hans Landa_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "68.624",
"value": "Hans Landa_header_cell_0_0_0"
},
{
"operation": "finish",
"time": "86.427",
"value": "finish"
}
]
}
|
37258
|
Murray Aynsley Hill, a suburb of Christchurch, New Zealand was named after early settler Hugh Murray-Aynsley but he never owned the land.
|
{
"evidence": [
"Murray Aynsley Hill_sentence_0",
"Murray Aynsley Hill_sentence_2"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMurray Aynsley Hill, a suburb of Christchurch, New Zealand was named after early settler Hugh Murray-Aynsley but he never owned the land.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Murray Aynsley Hill_sentence_0",
"Murray Aynsley Hill_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6081,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.748",
"value": "?search="
},
{
"operation": "search",
"time": "19.237",
"value": "Murray Aynsley Hill"
},
{
"operation": "Now on",
"time": "19.895",
"value": "Murray Aynsley Hill"
},
{
"operation": "Highlighting",
"time": "25.118",
"value": "Murray Aynsley Hill_sentence_0"
},
{
"operation": "Highlighting",
"time": "30.738",
"value": "Murray Aynsley Hill_sentence_2"
},
{
"operation": "finish",
"time": "90.524",
"value": "finish"
}
]
}
|
23350
|
Due to the Clausura 2018 Copa MX Final's regulations, the higher seed among both finalists during the group stage will host the final, thus Estadio Victoria, the venue that has been home to Necaxa since the club relocated from Mexico City in the Apertura 2003 season, hosted the final.
|
{
"evidence": [
"Clausura 2018 Copa MX Final_sentence_4",
"Clausura 2018 Copa MX Final_sentence_5"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDue to the Clausura 2018 Copa MX Final's regulations, the higher seed among both finalists during the group stage will host the final, thus Estadio Victoria, the venue that has been home to Necaxa since the club relocated from Mexico City in the Apertura 2003 season, hosted the final.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Clausura 2018 Copa MX Final_sentence_4",
"Clausura 2018 Copa MX Final_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6082,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "3.506",
"value": "?search="
},
{
"operation": "search",
"time": "16.05",
"value": "Clausura 2018 Copa MX Final"
},
{
"operation": "Now on",
"time": "17.373",
"value": "Clausura 2018 Copa MX Final"
},
{
"operation": "Highlighting",
"time": "31.866",
"value": "Clausura 2018 Copa MX Final_sentence_1"
},
{
"operation": "Highlighting",
"time": "45.701",
"value": "Clausura 2018 Copa MX Final_sentence_4"
},
{
"operation": "Highlighting",
"time": "82.82",
"value": "Clausura 2018 Copa MX Final_sentence_5"
},
{
"operation": "finish",
"time": "161.168",
"value": "finish"
}
]
}
|
87959
|
Five Talents was founded at the Lambeth Conference of Anglican Church leaders in 1998, and works with churches to provide financial literacy and business skills training to communities.
|
{
"evidence": [
"Five Talents_sentence_11",
"Five Talents_sentence_13"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nFive Talents was founded at the Lambeth Conference of Anglican Church leaders in 1998, and works with churches to provide financial literacy and business skills training to communities.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Five Talents_sentence_11",
"Five Talents_sentence_13"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6083,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.221",
"value": "?search="
},
{
"operation": "search",
"time": "9.389",
"value": "Five Talents"
},
{
"operation": "Now on",
"time": "10.477",
"value": "Five Talents"
},
{
"operation": "Page search",
"time": "28.195",
"value": "Conference"
},
{
"operation": "Highlighting",
"time": "32.883",
"value": "Five Talents_sentence_11"
},
{
"operation": "Highlighting",
"time": "47.417",
"value": "Five Talents_sentence_13"
},
{
"operation": "finish",
"time": "87.242",
"value": "finish"
}
]
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.