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 |
|---|---|---|---|---|---|---|---|---|
44538
|
Charlton Island (Sivukutaitiarruvik) is a308 km (119 sq mi). uninhabited island located in James Bay, Qikiqtaaluk Region, Nunavut, Australia that was named after Prince Charles.
|
{
"evidence": [
"Charlton Island_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:\nCharlton Island (Sivukutaitiarruvik) is a308 km (119 sq mi). uninhabited island located in James Bay, Qikiqtaaluk Region, Nunavut, Australia that was named after Prince Charles.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Charlton Island_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5484,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.11",
"value": "?search="
},
{
"operation": "search",
"time": "8.889",
"value": "Charlton Island"
},
{
"operation": "Now on",
"time": "10.05",
"value": "Charlton Island"
},
{
"operation": "Highlighting",
"time": "15.102",
"value": "Charlton Island_sentence_0"
},
{
"operation": "Highlighting",
"time": "22.864",
"value": "Charlton Island_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "22.869",
"value": "Charlton Island_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "22.869",
"value": "Charlton Island_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "25.863",
"value": "Charlton Island_cell_0_4_1"
},
{
"operation": "Highlighting",
"time": "25.868",
"value": "Charlton Island_header_cell_0_4_0"
},
{
"operation": "Highlighting",
"time": "28.674",
"value": "Charlton Island_cell_0_9_1"
},
{
"operation": "Highlighting",
"time": "28.679",
"value": "Charlton Island_header_cell_0_9_0"
},
{
"operation": "Highlighting",
"time": "28.68",
"value": "Charlton Island_header_cell_0_8_0"
},
{
"operation": "Highlighting",
"time": "35.282",
"value": "Charlton Island_cell_0_7_1"
},
{
"operation": "Highlighting",
"time": "35.287",
"value": "Charlton Island_header_cell_0_7_0"
},
{
"operation": "Highlighting",
"time": "35.287",
"value": "Charlton Island_header_cell_0_5_0"
},
{
"operation": "Highlighting",
"time": "37.29",
"value": "Charlton Island_cell_0_6_1"
},
{
"operation": "Highlighting",
"time": "37.296",
"value": "Charlton Island_header_cell_0_6_0"
},
{
"operation": "Highlighting deleted",
"time": "52.697",
"value": "Charlton Island_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "52.702",
"value": "Charlton Island_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "53.355",
"value": "Charlton Island_cell_0_4_1"
},
{
"operation": "Highlighting deleted",
"time": "53.361",
"value": "Charlton Island_header_cell_0_4_0"
},
{
"operation": "Highlighting deleted",
"time": "53.362",
"value": "Charlton Island_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "53.958",
"value": "Charlton Island_cell_0_6_1"
},
{
"operation": "Highlighting deleted",
"time": "53.964",
"value": "Charlton Island_header_cell_0_6_0"
},
{
"operation": "Highlighting deleted",
"time": "54.357",
"value": "Charlton Island_cell_0_7_1"
},
{
"operation": "Highlighting deleted",
"time": "54.363",
"value": "Charlton Island_header_cell_0_7_0"
},
{
"operation": "Highlighting deleted",
"time": "54.364",
"value": "Charlton Island_header_cell_0_5_0"
},
{
"operation": "Highlighting deleted",
"time": "55.044",
"value": "Charlton Island_cell_0_9_1"
},
{
"operation": "Highlighting deleted",
"time": "55.051",
"value": "Charlton Island_header_cell_0_9_0"
},
{
"operation": "Highlighting deleted",
"time": "55.052",
"value": "Charlton Island_header_cell_0_8_0"
},
{
"operation": "Highlighting",
"time": "61.135",
"value": "Charlton Island_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "62.719",
"value": "Charlton Island_sentence_2"
},
{
"operation": "finish",
"time": "75.615",
"value": "finish"
}
]
}
|
10872
|
Ben Viljoen wrote his autobiography while in prison and later on returned to South Africa to take on British citizenship.
|
{
"evidence": [
"Ben Viljoen_sentence_21"
],
"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:\nBen Viljoen wrote his autobiography while in prison and later on returned to South Africa to take on British citizenship.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ben Viljoen_sentence_21"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5485,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.787",
"value": "?search="
},
{
"operation": "search",
"time": "17.219",
"value": "Ben Viljoen"
},
{
"operation": "Now on",
"time": "19.048",
"value": "Ben Viljoen"
},
{
"operation": "Highlighting",
"time": "49.793",
"value": "Ben Viljoen_sentence_16Ben Viljoen_sentence_17"
},
{
"operation": "Highlighting",
"time": "50.276",
"value": "Ben Viljoen_sentence_18"
},
{
"operation": "Highlighting",
"time": "60.976",
"value": "Ben Viljoen_sentence_21"
},
{
"operation": "Highlighting deleted",
"time": "61.808",
"value": "Ben Viljoen_sentence_18"
},
{
"operation": "Highlighting deleted",
"time": "62.202",
"value": "Ben Viljoen_sentence_16Ben Viljoen_sentence_17"
},
{
"operation": "finish",
"time": "113.85",
"value": "finish"
}
]
}
|
40507
|
European Closing Bell was presented from London, and covered the last half-hour of European equity market trades.
|
{
"evidence": [
"European Closing Bell_sentence_1",
"European Closing Bell_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:\nEuropean Closing Bell was presented from London, and covered the last half-hour of European equity market trades.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"European Closing Bell_sentence_1",
"European Closing Bell_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5486,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.681",
"value": "?search="
},
{
"operation": "search",
"time": "24.055",
"value": "European Closing Bell"
},
{
"operation": "Now on",
"time": "24.93",
"value": "European Closing Bell"
},
{
"operation": "Highlighting",
"time": "43.7",
"value": "European Closing Bell_cell_0_3_1"
},
{
"operation": "Highlighting",
"time": "43.705",
"value": "European Closing Bell_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "43.705",
"value": "European Closing Bell_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "46.828",
"value": "European Closing Bell_cell_0_3_1"
},
{
"operation": "Highlighting deleted",
"time": "46.835",
"value": "European Closing Bell_header_cell_0_3_0"
},
{
"operation": "Highlighting deleted",
"time": "46.837",
"value": "European Closing Bell_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "48.748",
"value": "European Closing Bell_sentence_1"
},
{
"operation": "Highlighting",
"time": "52.611",
"value": "European Closing Bell_sentence_2"
},
{
"operation": "finish",
"time": "67.858",
"value": "finish"
}
]
}
|
exfever_1261
|
Baby Driver is a 2017 drama film with a supporting role like an American actor and producer.
|
{
"evidence": [
"Baby Driver",
"Jon Hamm"
],
"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:\nBaby Driver is a 2017 drama film with a supporting role like an American actor and producer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Baby Driver",
"Jon Hamm"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5487,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5569
|
Get Out is Daniel Kaluuya's first film.
|
{
"evidence": [
"Get Out (film)",
"Stephen Root"
],
"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:\nGet Out is Daniel Kaluuya's first film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Get Out (film)",
"Stephen Root"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5488,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1047
|
Macaulay Culkin is a British actor who starred in George Balanchine's 1993 American Christmas film, in which the film genre in which songs by the characters are interwoven into the narrative.
|
{
"evidence": [
"Macaulay Culkin",
"The Nutcracker (1993 film)",
"Musical 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:\nMacaulay Culkin is a British actor who starred in George Balanchine's 1993 American Christmas film, in which the film genre in which songs by the characters are interwoven into the narrative.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Macaulay Culkin",
"The Nutcracker (1993 film)",
"Musical film"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5489,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8471
|
Hannibal was a Carthaginian general and statesman whose father, Hamilcar Barca, was a leading Carthaginian general during the first of three wars fought between Rome and Carthage.
|
{
"evidence": [
"Hannibal",
"First Punic War"
],
"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:\nHannibal was a Carthaginian general and statesman whose father, Hamilcar Barca, was a leading Carthaginian general during the first of three wars fought between Rome and Carthage.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hannibal",
"First Punic War"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5490,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2930
|
The American Broadcasting Company, a multinational commercial broadcast radio network, is directly across the street from the corporate headquarters of The Walt Disney Company.
|
{
"evidence": [
"American Broadcasting Company",
"Walt Disney Studios (Burbank)"
],
"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 American Broadcasting Company, a multinational commercial broadcast radio network, is directly across the street from the corporate headquarters of The Walt Disney Company.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"American Broadcasting Company",
"Walt Disney Studios (Burbank)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5491,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
54928
|
1302 Werra is an asteroid that was discovered by K. Reinmuth in 1921 and named after Werra, a river in central Germany.
|
{
"evidence": [
"1302 Werra_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:\n1302 Werra is an asteroid that was discovered by K. Reinmuth in 1921 and named after Werra, a river in central Germany.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"1302 Werra_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5492,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.083",
"value": "?search="
},
{
"operation": "search",
"time": "12.888",
"value": "1302 Werra"
},
{
"operation": "Now on",
"time": "13.818",
"value": "1302 Werra"
},
{
"operation": "Highlighting",
"time": "19.709",
"value": "1302 Werra_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "19.736",
"value": "1302 Werra_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "19.736",
"value": "1302 Werra_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "33.459",
"value": "1302 Werra_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "33.49",
"value": "1302 Werra_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "33.492",
"value": "1302 Werra_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "37.523",
"value": "1302 Werra_sentence_1"
},
{
"operation": "finish",
"time": "52.371",
"value": "finish"
}
]
}
|
84702
|
In the Runnymede Borough Council elections, Howard Butterfield resigned his seat in March 2017 and the Conservatives held the seat at the by-election held on May 4, 2017.
|
{
"evidence": [
"Runnymede Borough Council elections_sentence_4",
"Runnymede Borough Council elections_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:\nIn the Runnymede Borough Council elections, Howard Butterfield resigned his seat in March 2017 and the Conservatives held the seat at the by-election held on May 4, 2017.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Runnymede Borough Council elections_sentence_4",
"Runnymede Borough Council elections_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5493,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.913",
"value": "?search="
},
{
"operation": "search",
"time": "10.963",
"value": "Runnymede Borough Council elections"
},
{
"operation": "Now on",
"time": "12.222",
"value": "Runnymede Borough Council elections"
},
{
"operation": "Page search",
"time": "81.178",
"value": "Howard Butterfield"
},
{
"operation": "Page search",
"time": "81.436",
"value": "Howard Butterfield"
},
{
"operation": "Page search",
"time": "81.72",
"value": "Howard Butterfield"
},
{
"operation": "Highlighting",
"time": "90.609",
"value": "Runnymede Borough Council elections_sentence_4"
},
{
"operation": "Highlighting",
"time": "94.053",
"value": "Runnymede Borough Council elections_sentence_5"
},
{
"operation": "finish",
"time": "104.737",
"value": "finish"
}
]
}
|
exfever_519
|
Heinrich Luitpold Himmler was a leading member of a political party that emerged from the birth of romantic nationalism during World War I.
|
{
"evidence": [
"Heinrich Himmler",
"Nazi Party",
"German nationalism"
],
"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:\nHeinrich Luitpold Himmler was a leading member of a political party that emerged from the birth of romantic nationalism during World War I.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Heinrich Himmler",
"Nazi Party",
"German nationalism"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5494,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2980
|
Kirsten Dunst is the highest-paid actress in the United States.
|
{
"evidence": [
"Bring It On (film)",
"Peyton Reed"
],
"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:\nKirsten Dunst is the highest-paid actress in the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Bring It On (film)",
"Peyton Reed"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5495,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8394
|
The Bridge of Spies is a 2015 film starring one of the most popular film stars worldwide.
|
{
"evidence": [
"Bridge of Spies (film)",
"Tom Hanks"
],
"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 Bridge of Spies is a 2015 film starring one of the most popular film stars worldwide.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Bridge of Spies (film)",
"Tom Hanks"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5496,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5603
|
Oliver Reed was an English actor known for portraying a fictional character in the novels The Three Musketeers and a novel by Alexandre Dumas, first serialized from January to August 1845.
|
{
"evidence": [
"Oliver Reed",
"Athos (character)",
"Twenty Years After"
],
"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:\nOliver Reed was an English actor known for portraying a fictional character in the novels The Three Musketeers and a novel by Alexandre Dumas, first serialized from January to August 1845.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Oliver Reed",
"Athos (character)",
"Twenty Years After"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5497,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7556
|
Rogue is a fictional character appearing in American comic books published by the brand name and primary imprint of Marvel Worldwide Inc.,?which was started in 1939 during the era called the Silver Age of comic books.
|
{
"evidence": [
"Rogue (comics)",
"Marvel Comics",
"Timely Comics"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRogue is a fictional character appearing in American comic books published by the brand name and primary imprint of Marvel Worldwide Inc.,?which was started in 1939 during the era called the Silver Age of comic books.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Rogue (comics)",
"Marvel Comics",
"Timely Comics"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5498,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
2645
|
List of airports in the Melbourne area consists of seven airports one of which was Avalon Airports.
|
{
"evidence": [
"Melbourne_sentence_339"
],
"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:\nList of airports in the Melbourne area consists of seven airports one of which was Avalon Airports.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Melbourne_sentence_339"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5499,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "10.649",
"value": "list of airports in melbourne"
},
{
"operation": "search",
"time": "18.358",
"value": "airports in melbourne"
},
{
"operation": "hyperlink",
"time": "169.333",
"value": "Melbourne Airport"
},
{
"operation": "search",
"time": "178.985",
"value": "melbourne"
},
{
"operation": "Page search",
"time": "185.646",
"value": "airport"
},
{
"operation": "Highlighting",
"time": "222.646",
"value": "Melbourne_sentence_339"
},
{
"operation": "finish",
"time": "231.076",
"value": "finish"
}
]
}
|
46112
|
Água de Pau is a civil parish which is the third-level administrative subdivision of Portugal, as defined by the 1976 Constitution.
|
{
"evidence": [
"Água de Pau_sentence_2",
"Freguesia_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:\nÁgua de Pau is a civil parish which is the third-level administrative subdivision of Portugal, as defined by the 1976 Constitution.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Água de Pau_sentence_2",
"Freguesia_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5500,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.672",
"value": "undefined"
},
{
"operation": "search",
"time": "22.684",
"value": "Água de Pau"
},
{
"operation": "Now on",
"time": "24.049",
"value": "Água de Pau"
},
{
"operation": "Highlighting",
"time": "27.768",
"value": "Água de Pau_sentence_2"
},
{
"operation": "hyperlink",
"time": "48.922",
"value": "civil parish"
},
{
"operation": "Now on",
"time": "49.796",
"value": "Freguesia"
},
{
"operation": "Highlighting",
"time": "55.051",
"value": "Freguesia_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "57.48",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "57.61",
"value": "Água de Pau"
},
{
"operation": "finish",
"time": "123.074",
"value": "finish"
}
]
}
|
exfever_3676
|
Lost Highway is a 1997 neo-noir film directed by David Lynch, financed by a French film production and distribution company founded in 1993 by Francis Bouygues.
|
{
"evidence": [
"Lost Highway (film)",
"Ciby 2000"
],
"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:\nLost Highway is a 1997 neo-noir film directed by David Lynch, financed by a French film production and distribution company founded in 1993 by Francis Bouygues.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lost Highway (film)",
"Ciby 2000"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5501,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2266
|
The Wolfman in 2010 is a remake of a horror film written by a German-American novelist and screenwriter.
|
{
"evidence": [
"The Wolfman (2010 film)",
"The Wolf Man (1941 film)",
"Curt Siodmak"
],
"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 Wolfman in 2010 is a remake of a horror film written by a German-American novelist and screenwriter.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Wolfman (2010 film)",
"The Wolf Man (1941 film)",
"Curt Siodmak"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5502,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
20171
|
Des Dowling started his hobby in journalism in 1997, and is known for his sketch comedy writing, having written and produced for the AFL Footy Show (a show that was dedicated to the Australian Football League) and the Logie award-nominated mockumentary series, Mark Loves Sharon.
|
{
"evidence": [
"Des Dowling_sentence_2",
"Des Dowling_sentence_5",
"The Footy Show (AFL)_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:\nDes Dowling started his hobby in journalism in 1997, and is known for his sketch comedy writing, having written and produced for the AFL Footy Show (a show that was dedicated to the Australian Football League) and the Logie award-nominated mockumentary series, Mark Loves Sharon.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Des Dowling_sentence_2",
"Des Dowling_sentence_5",
"The Footy Show (AFL)_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5503,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.579",
"value": "?search="
},
{
"operation": "search",
"time": "17.959",
"value": "Des Dowling"
},
{
"operation": "Now on",
"time": "18.617",
"value": "Des Dowling"
},
{
"operation": "Highlighting",
"time": "28.895",
"value": "Des Dowling_sentence_2"
},
{
"operation": "Highlighting",
"time": "46.311",
"value": "Des Dowling_sentence_5"
},
{
"operation": "hyperlink",
"time": "55.19",
"value": "AFL Footy Show"
},
{
"operation": "Now on",
"time": "55.865",
"value": "The Footy Show (AFL)"
},
{
"operation": "Highlighting",
"time": "62.112",
"value": "The Footy Show (AFL)_sentence_1"
},
{
"operation": "finish",
"time": "76.054",
"value": "finish"
}
]
}
|
exfever_1201
|
Logan is a 2017 American superhero film that is part of a trilogy and the one following the sixth installment in the X-Men film series.
|
{
"evidence": [
"Logan (film)",
"The Wolverine (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:\nLogan is a 2017 American superhero film that is part of a trilogy and the one following the sixth installment in the X-Men film series.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Logan (film)",
"The Wolverine (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5504,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8963
|
The Godfather Part III is a 1990 American crime film starring talented actors, including an American actor who won Tony Awards for Best Featured Actor in his role in a play by David Mamet in 1984.
|
{
"evidence": [
"The Godfather Part III",
"Glengarry Glen Ross"
],
"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 Godfather Part III is a 1990 American crime film starring talented actors, including an American actor who won Tony Awards for Best Featured Actor in his role in a play by David Mamet in 1984.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Godfather Part III",
"Glengarry Glen Ross"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5505,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
40707
|
Following the American Civil War, the Army and Navy were unsupportive of each other at either the planning or operational level; thus the Joint Board, acting as an "advisory committee", was created to plan joint operations and resolve problems of common rivalry between the two services.
|
{
"evidence": [
"American Civil War_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nFollowing the American Civil War, the Army and Navy were unsupportive of each other at either the planning or operational level; thus the Joint Board, acting as an \"advisory committee\", was created to plan joint operations and resolve problems of common rivalry between the two services.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"American Civil War_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5506,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.309",
"value": "?search="
},
{
"operation": "search",
"time": "30.862",
"value": "American Civil War"
},
{
"operation": "Now on",
"time": "33.077",
"value": "American Civil War"
},
{
"operation": "Page search",
"time": "45.273",
"value": "supportive"
},
{
"operation": "Page search",
"time": "45.771",
"value": "supportive"
},
{
"operation": "Page search",
"time": "45.993",
"value": "supportive"
},
{
"operation": "Page search",
"time": "46.161",
"value": "supportive"
},
{
"operation": "Page search",
"time": "81.464",
"value": "navy"
},
{
"operation": "Page search",
"time": "81.657",
"value": "navy"
},
{
"operation": "search",
"time": "198.445",
"value": "Following the American Civil War, the Army and Navy were unsupportive"
},
{
"operation": "Now on",
"time": "200.939",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "217.407",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "218.534",
"value": "American Civil War"
},
{
"operation": "Page search",
"time": "232.426",
"value": "Army and Navy"
},
{
"operation": "Page search",
"time": "232.589",
"value": "Army and Navy"
},
{
"operation": "Page search",
"time": "232.756",
"value": "Army and Navy"
},
{
"operation": "Page search",
"time": "232.927",
"value": "Army and Navy"
},
{
"operation": "search",
"time": "666.26",
"value": "thus the Joint Board, acting as an \"advisory committee\""
},
{
"operation": "Now on",
"time": "668.708",
"value": "undefined"
},
{
"operation": "search",
"time": "683.973",
"value": "was created to plan joint operations and resolve problems of common rivalry between the two services."
},
{
"operation": "Now on",
"time": "687.054",
"value": "undefined"
},
{
"operation": "search",
"time": "704.365",
"value": "Following the American Civil War"
},
{
"operation": "Now on",
"time": "706.649",
"value": "undefined"
},
{
"operation": "search",
"time": "750.319",
"value": "\"joint board\" \" American Civil War\""
},
{
"operation": "Now on",
"time": "751.255",
"value": "undefined"
},
{
"operation": "search",
"time": "778.639",
"value": "American Civil War"
},
{
"operation": "Now on",
"time": "780.71",
"value": "American Civil War"
},
{
"operation": "Highlighting",
"time": "784.128",
"value": "American Civil War_sentence_1"
},
{
"operation": "finish",
"time": "789.151",
"value": "finish"
}
]
}
|
exfever_5351
|
Practical Magic is sixth film directed by Griffin Dunne.
|
{
"evidence": [
"Practical Magic",
"Nicole Kidman"
],
"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:\nPractical Magic is sixth film directed by Griffin Dunne.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Practical Magic",
"Nicole Kidman"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5507,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8806
|
Ronin is a 1998 American action thriller film written by a British filmmaker who won the Pulitzer Prize.
|
{
"evidence": [
"Ronin (film)",
"David Mamet"
],
"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:\nRonin is a 1998 American action thriller film written by a British filmmaker who won the Pulitzer Prize.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ronin (film)",
"David Mamet"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5508,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2776
|
Hogwarts School of Witchcraft and Wizardry is a fictional American boarding school of magic that is the primary setting for a series of seven fantasy novels written by British author J. K. Rowling.
|
{
"evidence": [
"Hogwarts",
"Harry Potter"
],
"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:\nHogwarts School of Witchcraft and Wizardry is a fictional American boarding school of magic that is the primary setting for a series of seven fantasy novels written by British author J. K. Rowling.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hogwarts",
"Harry Potter"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5509,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3103
|
Steven Knight wrote the screenplays for the British-American political thriller comedy film directed by John Crowley.
|
{
"evidence": [
"Steven Knight",
"Closed Circuit (2013 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:\nSteven Knight wrote the screenplays for the British-American political thriller comedy film directed by John Crowley.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Steven Knight",
"Closed Circuit (2013 film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5510,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5439
|
Observer Media is the first American online media company based in Lower Manhattan, New York City.
|
{
"evidence": [
"Observer Media",
"Jared Kushner"
],
"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:\nObserver Media is the first American online media company based in Lower Manhattan, New York City.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Observer Media",
"Jared Kushner"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5511,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
30579
|
The Nantou railway station is a flat, not elevated station in Japan and the first station in Zhongshan for the rail going to Zhuhai.
|
{
"evidence": [
"Nantou railway station_sentence_0"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Nantou railway station is a flat, not elevated station in Japan and the first station in Zhongshan for the rail going to Zhuhai.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Nantou railway station_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5512,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.354",
"value": "?search="
},
{
"operation": "search",
"time": "8.093",
"value": "Nantou railway station"
},
{
"operation": "Now on",
"time": "8.866",
"value": "Nantou railway station"
},
{
"operation": "Highlighting",
"time": "164.446",
"value": "Nantou railway station_sentence_0"
},
{
"operation": "finish",
"time": "176.857",
"value": "finish"
}
]
}
|
72643
|
Clarence Bolton became enthralled with the local Catskill Mountains landscape and began painting landscape paintings under John F. Carlson.
|
{
"evidence": [
"Clarence Bolton_sentence_8",
"Clarence Bolton_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:\nClarence Bolton became enthralled with the local Catskill Mountains landscape and began painting landscape paintings under John F. Carlson.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Clarence Bolton_sentence_8",
"Clarence Bolton_sentence_9"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5513,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.409",
"value": "?search="
},
{
"operation": "search",
"time": "17.116",
"value": "Clarence Bolton"
},
{
"operation": "Now on",
"time": "18.213",
"value": "Clarence Bolton"
},
{
"operation": "Highlighting",
"time": "61.742",
"value": "Clarence Bolton_sentence_8"
},
{
"operation": "Highlighting",
"time": "92.243",
"value": "Clarence Bolton_sentence_9"
},
{
"operation": "finish",
"time": "125.772",
"value": "finish"
}
]
}
|
exfever_8479
|
The Entire History of You is the third and final episode of the first series of the British science fiction anthology television series Black Mirror, the first episode not written by an English television presenter, author, screenwriter, producer, satirist and social critic.
|
{
"evidence": [
"The Entire History of You",
"Charlie Brooker"
],
"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 Entire History of You is the third and final episode of the first series of the British science fiction anthology television series Black Mirror, the first episode not written by an English television presenter, author, screenwriter, producer, satirist and social critic.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Entire History of You",
"Charlie Brooker"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5514,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
33523
|
Trimingham was awarded a medal for his community work in the field of drugs and alcohol, then has brought about International Family Drug Support Day which commenced on February 24, 2016.
|
{
"evidence": [
"Anthony Trimingham_sentence_8",
"Anthony Trimingham_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:\nTrimingham was awarded a medal for his community work in the field of drugs and alcohol, then has brought about International Family Drug Support Day which commenced on February 24, 2016.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Anthony Trimingham_sentence_8",
"Anthony Trimingham_sentence_9"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5515,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.091",
"value": "?search="
},
{
"operation": "search",
"time": "4.67",
"value": "Trimingham"
},
{
"operation": "Now on",
"time": "5.5",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "15.95",
"value": "Anthony Trimingham"
},
{
"operation": "Now on",
"time": "16.532",
"value": "Anthony Trimingham"
},
{
"operation": "Page search",
"time": "28.995",
"value": "medal"
},
{
"operation": "Highlighting",
"time": "31.849",
"value": "Anthony Trimingham_sentence_8"
},
{
"operation": "Highlighting",
"time": "43.847",
"value": "Anthony Trimingham_sentence_9"
},
{
"operation": "finish",
"time": "78.472",
"value": "finish"
}
]
}
|
79561
|
In the 1978 Seattle Seahawks season, Steve Largent made his first Pro Bowl with 71 receptions and eight touchdowns, Jim Zorn earned an All-Pro honor, and David Sims led the AFC in touchdowns.
|
{
"evidence": [
"1978 Seattle Seahawks season_sentence_5",
"1978 Seattle Seahawks season_sentence_6",
"1978 Seattle Seahawks season_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:\nIn the 1978 Seattle Seahawks season, Steve Largent made his first Pro Bowl with 71 receptions and eight touchdowns, Jim Zorn earned an All-Pro honor, and David Sims led the AFC in touchdowns.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"1978 Seattle Seahawks season_sentence_5",
"1978 Seattle Seahawks season_sentence_6",
"1978 Seattle Seahawks season_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5516,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.405",
"value": "?search="
},
{
"operation": "search",
"time": "11.665",
"value": "1978 Seattle Seahawks season"
},
{
"operation": "Now on",
"time": "13.492",
"value": "1978 Seattle Seahawks season"
},
{
"operation": "Highlighting",
"time": "20.329",
"value": "1978 Seattle Seahawks season_sentence_5"
},
{
"operation": "Highlighting",
"time": "25.568",
"value": "1978 Seattle Seahawks season_sentence_6"
},
{
"operation": "Page search",
"time": "56.592",
"value": "sims"
},
{
"operation": "Highlighting",
"time": "58.867",
"value": "1978 Seattle Seahawks season_sentence_4"
},
{
"operation": "finish",
"time": "77.501",
"value": "finish"
}
]
}
|
75120
|
The Tunisian Pirate Party is a small political party formed in 2010 and legalized on March 12, 2012, becoming one of the first outgrowth of the Pirate Party, a political incarnation of the freedom of expression movement to achieve their goals using the established political system rather than through activism.
|
{
"evidence": [
"Tunisian Pirate Party_sentence_1",
"Tunisian Pirate Party_sentence_2",
"Pirate Parties International_sentence_3"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Tunisian Pirate Party is a small political party formed in 2010 and legalized on March 12, 2012, becoming one of the first outgrowth of the Pirate Party, a political incarnation of the freedom of expression movement to achieve their goals using the established political system rather than through activism.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Tunisian Pirate Party_sentence_1",
"Tunisian Pirate Party_sentence_2",
"Pirate Parties International_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5517,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.235",
"value": "?search="
},
{
"operation": "search",
"time": "8.151",
"value": "The Tunisian Pirate Party"
},
{
"operation": "Now on",
"time": "9.845",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "158.069",
"value": "Tunisian Pirate Party"
},
{
"operation": "Now on",
"time": "158.86",
"value": "Tunisian Pirate Party"
},
{
"operation": "Highlighting",
"time": "367.98",
"value": "Tunisian Pirate Party_sentence_1"
},
{
"operation": "Highlighting",
"time": "368.612",
"value": "Tunisian Pirate Party_sentence_2"
},
{
"operation": "hyperlink",
"time": "383.261",
"value": "Pirate Party movement"
},
{
"operation": "Now on",
"time": "384.172",
"value": "Pirate Parties International"
},
{
"operation": "Highlighting",
"time": "395.428",
"value": "Pirate Parties International_sentence_3"
},
{
"operation": "finish",
"time": "402.54",
"value": "finish"
}
]
}
|
exfever_714
|
Steffi Graf became the first tennis player to win all four major championships,?overseen by the International Tennis Federation rather than the governing body of the men's professional tennis circuit.
|
{
"evidence": [
"Steffi Graf",
"Grand Slam (tennis)",
"Association of Tennis Professionals"
],
"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:\nSteffi Graf became the first tennis player to win all four major championships,?overseen by the International Tennis Federation rather than the governing body of the men's professional tennis circuit.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Steffi Graf",
"Grand Slam (tennis)",
"Association of Tennis Professionals"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5518,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
74470
|
Tōkyū Tōyoko Line (a major railway line connecting Tokyo (Shibuya) to Yokohama) Express does not stop at the Motosumiyoshi Station, as well as Tōkyū Meguro Line (MG12) Express.
|
{
"evidence": [
"Tōkyū Tōyoko Line_sentence_0",
"Tōkyū Tōyoko Line_sentence_4",
"Tōkyū Tōyoko Line_sentence_47"
],
"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:\nTōkyū Tōyoko Line (a major railway line connecting Tokyo (Shibuya) to Yokohama) Express does not stop at the Motosumiyoshi Station, as well as Tōkyū Meguro Line (MG12) Express.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Tōkyū Tōyoko Line_sentence_0",
"Tōkyū Tōyoko Line_sentence_4",
"Tōkyū Tōyoko Line_sentence_47"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5519,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.724",
"value": "?search="
},
{
"operation": "search",
"time": "24.088",
"value": "Tōkyū Tōyoko Line"
},
{
"operation": "Now on",
"time": "25.935",
"value": "Tōkyū Tōyoko Line"
},
{
"operation": "Highlighting",
"time": "29.614",
"value": "Tōkyū Tōyoko Line_sentence_0"
},
{
"operation": "Page search",
"time": "53.843",
"value": "stop"
},
{
"operation": "Page search",
"time": "86.491",
"value": "motosum"
},
{
"operation": "Page search",
"time": "137.323",
"value": "meguro"
},
{
"operation": "Highlighting",
"time": "146.07",
"value": "Tōkyū Tōyoko Line_sentence_4"
},
{
"operation": "Page search",
"time": "153.539",
"value": "motosum"
},
{
"operation": "Highlighting",
"time": "158.057",
"value": "Tōkyū Tōyoko Line_sentence_47"
},
{
"operation": "finish",
"time": "164.162",
"value": "finish"
}
]
}
|
exfever_5079
|
John Weldon "J. J." Cale was a Korean guitarist where influence as a musical artist has been widely acknowledged by figures such as an American singer who started playing guitar at the age of eight and performed at age 14 on KVOW radio, after which he formed his first band, The Texas Longhorns.
|
{
"evidence": [
"J. J. Cale",
"Waylon Jennings"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJohn Weldon \"J. J.\" Cale was a Korean guitarist where influence as a musical artist has been widely acknowledged by figures such as an American singer who started playing guitar at the age of eight and performed at age 14 on KVOW radio, after which he formed his first band, The Texas Longhorns.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"J. J. Cale",
"Waylon Jennings"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5520,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7923
|
Joni Mitchell is an Indian-American singer-songwriter who began singing in small nightclubs in western Canada, before moving on to the nightclubs of the capital city of the Canadian province of Ontario which is the most populous city in Canada and the fourth most populous city in North America.
|
{
"evidence": [
"Joni Mitchell",
"Toronto"
],
"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:\nJoni Mitchell is an Indian-American singer-songwriter who began singing in small nightclubs in western Canada, before moving on to the nightclubs of the capital city of the Canadian province of Ontario which is the most populous city in Canada and the fourth most populous city in North America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Joni Mitchell",
"Toronto"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5521,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3911
|
The Godfather Part III is a 1990 American crime film starring an American actor who named as one of the few performers to have achieved the Triple Crown of Acting.
|
{
"evidence": [
"The Godfather Part III",
"Al Pacino"
],
"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 Godfather Part III is a 1990 American crime film starring an American actor who named as one of the few performers to have achieved the Triple Crown of Acting.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Godfather Part III",
"Al Pacino"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5522,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_331
|
iO Theater was an improv theater and training center in Rockford that had notable alumni, including an American comedian who is also the co-founder of the improvisational-comedy troupe Upright Citizen Brigade.
|
{
"evidence": [
"IO Theater",
"Amy Poehler"
],
"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:\niO Theater was an improv theater and training center in Rockford that had notable alumni, including an American comedian who is also the co-founder of the improvisational-comedy troupe Upright Citizen Brigade.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"IO Theater",
"Amy Poehler"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5523,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
54281
|
The show business phrase "The show must go on", originating in the 19th century, means that the exhibition must be performed, regardless of what else happens.
|
{
"evidence": [
"The show must go on_sentence_2",
"The show must go on_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 show business phrase \"The show must go on\", originating in the 19th century, means that the exhibition must be performed, regardless of what else happens.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"The show must go on_sentence_2",
"The show must go on_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5524,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.977",
"value": "?search="
},
{
"operation": "search",
"time": "27.965",
"value": "The show business phrase \"The show must go on\""
},
{
"operation": "Now on",
"time": "30.284",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "34.541",
"value": "The show must go on"
},
{
"operation": "Now on",
"time": "34.92",
"value": "The show must go on"
},
{
"operation": "Highlighting",
"time": "37.389",
"value": "The show must go on_sentence_2"
},
{
"operation": "Highlighting",
"time": "54.676",
"value": "The show must go on_sentence_4"
},
{
"operation": "Highlighting",
"time": "85.155",
"value": "The show must go on_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "91.691",
"value": "The show must go on_sentence_7"
},
{
"operation": "finish",
"time": "95.603",
"value": "finish"
}
]
}
|
exfever_5543
|
Andrew Geoffrey Kaufman was an American actor and performance artist who had performed in comedy clubs and theaters as an absurdly foul-mouthed and domineering lounge singer claiming to hail from Las Vegas.
|
{
"evidence": [
"Andy Kaufman",
"Tony Clifton"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAndrew Geoffrey Kaufman was an American actor and performance artist who had performed in comedy clubs and theaters as an absurdly foul-mouthed and domineering lounge singer claiming to hail from Las Vegas.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Andy Kaufman",
"Tony Clifton"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5525,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
77796
|
Sadam Ali is an American professional boxer whose bouts are fought for a purse that is divided between the boxers as determined by contract.
|
{
"evidence": [
"Sadam Ali_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSadam Ali is an American professional boxer whose bouts are fought for a purse that is divided between the boxers as determined by contract.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sadam Ali_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5526,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.869",
"value": "?search="
},
{
"operation": "search",
"time": "5.836",
"value": "Sadam Ali"
},
{
"operation": "Now on",
"time": "7.584",
"value": "Sadam Ali"
},
{
"operation": "Highlighting",
"time": "10.661",
"value": "Sadam Ali_sentence_0"
},
{
"operation": "Page search",
"time": "38.594",
"value": "bout"
},
{
"operation": "Page search",
"time": "69.014",
"value": "purse"
},
{
"operation": "Page search",
"time": "79.438",
"value": "contract"
},
{
"operation": "finish",
"time": "127.112",
"value": "finish"
}
]
}
|
exfever_392
|
Jim Henson's last creation is The Muppets and Fraggle Rock.
|
{
"evidence": [
"Jim Henson",
"University Park, Maryland"
],
"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:\nJim Henson's last creation is The Muppets and Fraggle Rock.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jim Henson",
"University Park, Maryland"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5527,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8201
|
The cell came from the dialect spoken in present-day Rome, then known as a region in central western Italy.
|
{
"evidence": [
"Cell (biology)",
"Latin",
"Latium"
],
"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 cell came from the dialect spoken in present-day Rome, then known as a region in central western Italy.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Cell (biology)",
"Latin",
"Latium"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5528,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
32092
|
Sarah Meriwether Nutter was one of the original twenty founders of Alpha Kappa Alpha Sorority who worked in the profession considered most detrimental to the advancement of African-American citizens.
|
{
"evidence": [
"Sarah Meriwether Nutter_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:\nSarah Meriwether Nutter was one of the original twenty founders of Alpha Kappa Alpha Sorority who worked in the profession considered most detrimental to the advancement of African-American citizens.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sarah Meriwether Nutter_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5529,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.419",
"value": "?search="
},
{
"operation": "search",
"time": "15.564",
"value": "Sarah Meriwether Nutter"
},
{
"operation": "Now on",
"time": "16.729",
"value": "Sarah Meriwether Nutter"
},
{
"operation": "Highlighting",
"time": "28.221",
"value": "Sarah Meriwether Nutter_sentence_0"
},
{
"operation": "Highlighting",
"time": "50.771",
"value": "Sarah Meriwether Nutter_sentence_1"
},
{
"operation": "Highlighting",
"time": "78.191",
"value": "Sarah Meriwether Nutter_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "130.259",
"value": "Sarah Meriwether Nutter_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "131.552",
"value": "Sarah Meriwether Nutter_sentence_0"
},
{
"operation": "finish",
"time": "191.913",
"value": "finish"
}
]
}
|
exfever_380
|
Android is a mobile operating system based on a modified version of a mostly free and open-source, monolithic, modular, multitasking, Unix-like operating system kernel.
|
{
"evidence": [
"Android (operating system)",
"Linux kernel"
],
"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:\nAndroid is a mobile operating system based on a modified version of a mostly free and open-source, monolithic, modular, multitasking, Unix-like operating system kernel.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Android (operating system)",
"Linux kernel"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5530,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
44833
|
Mark Solonin, born on April 29, 1958 at Kuybyshev, Russian SFSR, Soviet Union, is a Russian historian focused on Soviet history who studied at Kuibyshev Aviation Institute—one of Russia’s leading engineering and technical institutions that is currently known as Samara National Research University.
|
{
"evidence": [
"Mark Solonin_sentence_0",
"Mark Solonin_sentence_3"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMark Solonin, born on April 29, 1958 at Kuybyshev, Russian SFSR, Soviet Union, is a Russian historian focused on Soviet history who studied at Kuibyshev Aviation Institute—one of Russia’s leading engineering and technical institutions that is currently known as Samara National Research University.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Mark Solonin_sentence_0",
"Mark Solonin_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5531,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.605",
"value": "?search="
},
{
"operation": "search",
"time": "20.413",
"value": "Mark Solonin"
},
{
"operation": "Now on",
"time": "21.152",
"value": "Mark Solonin"
},
{
"operation": "Highlighting",
"time": "23.413",
"value": "Mark Solonin_sentence_0"
},
{
"operation": "Highlighting",
"time": "69.629",
"value": "Mark Solonin_sentence_3"
},
{
"operation": "finish",
"time": "108.74",
"value": "finish"
}
]
}
|
90389
|
The INEC card reader uses cryptographic technology and was first used in Nigeria's presidential election in 2011.
|
{
"evidence": [
"INEC card reader_sentence_6"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe INEC card reader uses cryptographic technology and was first used in Nigeria's presidential election in 2011.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"INEC card reader_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5532,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.283",
"value": "?search="
},
{
"operation": "search",
"time": "17.459",
"value": "INEC card reader"
},
{
"operation": "Now on",
"time": "18.224",
"value": "INEC card reader"
},
{
"operation": "Highlighting",
"time": "23.569",
"value": "INEC card reader_sentence_3"
},
{
"operation": "Highlighting",
"time": "36.56",
"value": "INEC card reader_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "43.768",
"value": "INEC card reader_sentence_3"
},
{
"operation": "Highlighting",
"time": "116.985",
"value": "INEC card reader_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "118.274",
"value": "INEC card reader_sentence_0"
},
{
"operation": "Highlighting",
"time": "121.448",
"value": "INEC card reader_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "121.968",
"value": "INEC card reader_sentence_1"
},
{
"operation": "finish",
"time": "219.176",
"value": "finish"
}
]
}
|
48389
|
The Grand Avenue Water Tower has a diameter of five feet inside but was never used.
|
{
"evidence": [
"Grand Avenue Water Tower_sentence_6"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Grand Avenue Water Tower has a diameter of five feet inside but was never used.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Grand Avenue Water Tower_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5533,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.685",
"value": "?search="
},
{
"operation": "search",
"time": "7.71",
"value": "Grand Avenue Water Tower"
},
{
"operation": "Now on",
"time": "8.647",
"value": "Grand Avenue Water Tower"
},
{
"operation": "Highlighting",
"time": "22.974",
"value": "Grand Avenue Water Tower_sentence_5"
},
{
"operation": "Highlighting",
"time": "219.768",
"value": "Grand Avenue Water Tower_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "248.187",
"value": "Grand Avenue Water Tower_sentence_5"
},
{
"operation": "finish",
"time": "251.275",
"value": "finish"
}
]
}
|
28585
|
Among the wireless communications service providers in the United States, Verizon has continued service on their non-LTE networks and still support VoLTE.
|
{
"evidence": [
"List of United States wireless communications service providers_sentence_15",
"List of United States wireless communications service providers_sentence_16"
],
"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:\nAmong the wireless communications service providers in the United States, Verizon has continued service on their non-LTE networks and still support VoLTE.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"List of United States wireless communications service providers_sentence_15",
"List of United States wireless communications service providers_sentence_16"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5534,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.512",
"value": "?search="
},
{
"operation": "search",
"time": "24.326",
"value": "Verizon\\"
},
{
"operation": "Now on",
"time": "24.909",
"value": "undefined"
},
{
"operation": "search",
"time": "49.581",
"value": "non-LTE network support"
},
{
"operation": "Now on",
"time": "51.196",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "66.213",
"value": "Verizon Wireless"
},
{
"operation": "Now on",
"time": "66.754",
"value": "Verizon Wireless"
},
{
"operation": "Page search",
"time": "81.157",
"value": "non-LTE"
},
{
"operation": "Highlighting",
"time": "109.214",
"value": "Verizon Wireless_sentence_60"
},
{
"operation": "Highlighting",
"time": "109.214",
"value": "Verizon Wireless_title"
},
{
"operation": "Highlighting",
"time": "109.214",
"value": "Verizon Wireless_section_1"
},
{
"operation": "hyperlink",
"time": "168.132",
"value": "CDMA2000"
},
{
"operation": "Now on",
"time": "168.886",
"value": "CDMA2000"
},
{
"operation": "search",
"time": "204.837",
"value": "\"among the wireless communication service providers\""
},
{
"operation": "Now on",
"time": "206.105",
"value": "undefined"
},
{
"operation": "search",
"time": "221.676",
"value": "wireless communications service providers"
},
{
"operation": "Now on",
"time": "222.779",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "230.324",
"value": "List of United States wireless communications service providers"
},
{
"operation": "Now on",
"time": "230.746",
"value": "List of United States wireless communications service providers"
},
{
"operation": "Highlighting",
"time": "283.492",
"value": "List of United States wireless communications service providers_sentence_15"
},
{
"operation": "Highlighting",
"time": "324.923",
"value": "List of United States wireless communications service providers_sentence_16"
},
{
"operation": "finish",
"time": "348.859",
"value": "finish"
}
]
}
|
exfever_4104
|
The Man in the High Castle is an alternative historical novel in which the Axis Powers win in World War II, and the story describes the political intrigues of the historical nation-state and great power that existed from the Meiji Restoration until the enactment of the post-World War II 1957 constitution and the subsequent formation of modern Japan.
|
{
"evidence": [
"The Man in the High Castle",
"Empire of Japan"
],
"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 Man in the High Castle is an alternative historical novel in which the Axis Powers win in World War II, and the story describes the political intrigues of the historical nation-state and great power that existed from the Meiji Restoration until the enactment of the post-World War II 1957 constitution and the subsequent formation of modern Japan.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Man in the High Castle",
"Empire of Japan"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5535,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7788
|
David Angell died heading home from their vacation on Cape Cod aboard a domestic passenger flight that was hijacked by five al-Qaeda terrorists on September 11, 2001.
|
{
"evidence": [
"David Angell",
"American Airlines Flight 11"
],
"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:\nDavid Angell died heading home from their vacation on Cape Cod aboard a domestic passenger flight that was hijacked by five al-Qaeda terrorists on September 11, 2001.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"David Angell",
"American Airlines Flight 11"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5536,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6704
|
In J. K. Rowling's Harry Potter series, Hogwarts School of Witchcraft and Wizardry is a fictional British institution where children live within premises while receiving formal education.
|
{
"evidence": [
"Hogwarts",
"Boarding school"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn J. K. Rowling's Harry Potter series, Hogwarts School of Witchcraft and Wizardry is a fictional British institution where children live within premises while receiving formal education.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hogwarts",
"Boarding school"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5537,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2302
|
Where the Wild Things Are is a 2009 fantasy adventure drama film written by Jonze and a Canadian writer, editor, and publisher who wrote the best-selling memoir A Heartbreaking Work of Staggering Genius.
|
{
"evidence": [
"Where the Wild Things Are (film)",
"Dave Eggers"
],
"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:\nWhere the Wild Things Are is a 2009 fantasy adventure drama film written by Jonze and a Canadian writer, editor, and publisher who wrote the best-selling memoir A Heartbreaking Work of Staggering Genius.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Where the Wild Things Are (film)",
"Dave Eggers"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5538,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_364
|
Maid in Manhattan is a 2002 American romantic comedy-drama film based on a story by an American filmmaker who directed a 1983 American road comedy film.
|
{
"evidence": [
"Maid in Manhattan",
"John Hughes (filmmaker)",
"National Lampoon's Vacation"
],
"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:\nMaid in Manhattan is a 2002 American romantic comedy-drama film based on a story by an American filmmaker who directed a 1983 American road comedy film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Maid in Manhattan",
"John Hughes (filmmaker)",
"National Lampoon's Vacation"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5539,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
36675
|
After Aerosmith abandoned the song, Jacks and his wife recorded "Seasons in the Sun" in 1973 which he released on his own label, Goldfish Records.
|
{
"evidence": [
"Seasons in the Sun_sentence_12",
"Seasons in the Sun_sentence_14"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAfter Aerosmith abandoned the song, Jacks and his wife recorded \"Seasons in the Sun\" in 1973 which he released on his own label, Goldfish Records.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Seasons in the Sun_sentence_12",
"Seasons in the Sun_sentence_14"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5540,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.391",
"value": "?search="
},
{
"operation": "search",
"time": "26.608",
"value": "Seasons in the Sun"
},
{
"operation": "Now on",
"time": "29.717",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "33.043",
"value": "Seasons in the Sun"
},
{
"operation": "Now on",
"time": "33.83",
"value": "Seasons in the Sun"
},
{
"operation": "Highlighting",
"time": "58.553",
"value": "Seasons in the Sun_sentence_2"
},
{
"operation": "Page search",
"time": "65.161",
"value": "aero"
},
{
"operation": "Page search",
"time": "65.312",
"value": "aero"
},
{
"operation": "Page search",
"time": "65.447",
"value": "aero"
},
{
"operation": "Page search",
"time": "81.474",
"value": "wife"
},
{
"operation": "Page search",
"time": "81.64",
"value": "wife"
},
{
"operation": "Page search",
"time": "81.832",
"value": "wife"
},
{
"operation": "Page search",
"time": "81.993",
"value": "wife"
},
{
"operation": "Highlighting",
"time": "111.109",
"value": "Seasons in the Sun_sentence_12"
},
{
"operation": "Highlighting deleted",
"time": "195.98",
"value": "Seasons in the Sun_sentence_2"
},
{
"operation": "Page search",
"time": "215.736",
"value": "label"
},
{
"operation": "Page search",
"time": "215.943",
"value": "label"
},
{
"operation": "Page search",
"time": "216.159",
"value": "label"
},
{
"operation": "Highlighting",
"time": "225.213",
"value": "Seasons in the Sun_sentence_14"
},
{
"operation": "Highlighting",
"time": "251.106",
"value": "Seasons in the Sun_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "260.689",
"value": "Seasons in the Sun_sentence_1"
},
{
"operation": "Page search",
"time": "349.907",
"value": "aerosmith"
},
{
"operation": "Page search",
"time": "350.132",
"value": "aerosmith"
},
{
"operation": "Page search",
"time": "350.329",
"value": "aerosmith"
},
{
"operation": "hyperlink",
"time": "357.404",
"value": "Jacques Brel"
},
{
"operation": "Now on",
"time": "358.881",
"value": "Jacques Brel"
},
{
"operation": "Page search",
"time": "365.767",
"value": "aerosmith"
},
{
"operation": "Page search",
"time": "365.963",
"value": "aerosmith"
},
{
"operation": "Page search",
"time": "366.17",
"value": "aerosmith"
},
{
"operation": "Page search",
"time": "366.378",
"value": "aerosmith"
},
{
"operation": "Page search",
"time": "366.53",
"value": "aerosmith"
},
{
"operation": "back-button-clicked",
"time": "406.894",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "407.422",
"value": "Seasons in the Sun"
},
{
"operation": "finish",
"time": "418.864",
"value": "finish"
}
]
}
|
exfever_2249
|
Glenn Frey was an English musician who sang lead vocals on hits such as the second single from their album One of These Nights, reaching No. 2 on the Billboard Hot 100 chart and No. 8 on the Billboard Country chart.
|
{
"evidence": [
"Glenn Frey",
"Lyin' Eyes"
],
"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:\nGlenn Frey was an English musician who sang lead vocals on hits such as the second single from their album One of These Nights, reaching No. 2 on the Billboard Hot 100 chart and No. 8 on the Billboard Country chart.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Glenn Frey",
"Lyin' Eyes"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5541,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
54078
|
Albert Lorey Groll was elected into the National Academy of Design in 1910 and kept a studio in New York.
|
{
"evidence": [
"Albert Lorey Groll_sentence_2",
"Albert Lorey Groll_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:\nAlbert Lorey Groll was elected into the National Academy of Design in 1910 and kept a studio in New York.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Albert Lorey Groll_sentence_2",
"Albert Lorey Groll_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5542,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.715",
"value": "?search="
},
{
"operation": "search",
"time": "9.98",
"value": "Albert Lorey Groll"
},
{
"operation": "Now on",
"time": "10.964",
"value": "Albert Lorey Groll"
},
{
"operation": "Highlighting",
"time": "14.848",
"value": "Albert Lorey Groll_sentence_2"
},
{
"operation": "Highlighting",
"time": "19.519",
"value": "Albert Lorey Groll_sentence_3"
},
{
"operation": "finish",
"time": "29.592",
"value": "finish"
}
]
}
|
10347
|
Tasmania has a climate similar to places like the United Kingdom, New Zealand, and the Pacific Northwest region of the United States, and Canada, it was heavily influenced by the Roaring Forties with strong westerly winds in the southern hemisphere, which cool the rising air mass and lower air temperatures in this region.
|
{
"evidence": [
"Climate of Tasmania_sentence_8"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nTasmania has a climate similar to places like the United Kingdom, New Zealand, and the Pacific Northwest region of the United States, and Canada, it was heavily influenced by the Roaring Forties with strong westerly winds in the southern hemisphere, which cool the rising air mass and lower air temperatures in this region.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Climate of Tasmania_sentence_8"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5543,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.546",
"value": "?search="
},
{
"operation": "search",
"time": "36.059",
"value": "Contains...climate in tasmania"
},
{
"operation": "Now on",
"time": "37.156",
"value": "undefined"
},
{
"operation": "search",
"time": "59.301",
"value": "tasmania"
},
{
"operation": "Now on",
"time": "59.902",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "65.576",
"value": "Premier of Tasmania"
},
{
"operation": "Now on",
"time": "66.603",
"value": "Premier of Tasmania"
},
{
"operation": "back-button-clicked",
"time": "78.958",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "79.579",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "85.605",
"value": "Local government areas of Tasmania"
},
{
"operation": "Now on",
"time": "86.514",
"value": "Local government areas of Tasmania"
},
{
"operation": "hyperlink",
"time": "107.215",
"value": "Government of Tasmania"
},
{
"operation": "Now on",
"time": "107.913",
"value": "Tasmanian Government"
},
{
"operation": "search",
"time": "119.955",
"value": "tasmanian climate"
},
{
"operation": "Now on",
"time": "120.801",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "133.967",
"value": "Climate Deloraine, Tasmania_section_5"
},
{
"operation": "Now on",
"time": "134.481",
"value": "Deloraine, Tasmania#Climate Deloraine.2C Tasmania section 5"
},
{
"operation": "search",
"time": "179.576",
"value": "tasmania climate"
},
{
"operation": "Now on",
"time": "180.094",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "183.95",
"value": "Climate of Tasmania"
},
{
"operation": "Now on",
"time": "184.48",
"value": "Climate of Tasmania"
},
{
"operation": "Highlighting",
"time": "195.004",
"value": "Climate of Tasmania_sentence_8"
},
{
"operation": "Page search",
"time": "213.977",
"value": "roaring"
},
{
"operation": "Page search",
"time": "280.657",
"value": "westerly"
},
{
"operation": "hyperlink",
"time": "311.403",
"value": "8 See also Climate of Tasmania_section_13"
},
{
"operation": "finish",
"time": "349.585",
"value": "finish"
}
]
}
|
89703
|
The 1950 Duquesne Dukes football team played for Duquesne University under head coach Doc Skender, after which the team was disbanded until 1969.
|
{
"evidence": [
"1950 Duquesne Dukes football team_sentence_0",
"1950 Duquesne Dukes football team_sentence_1",
"1950 Duquesne Dukes football team_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:\nThe 1950 Duquesne Dukes football team played for Duquesne University under head coach Doc Skender, after which the team was disbanded until 1969.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"1950 Duquesne Dukes football team_sentence_0",
"1950 Duquesne Dukes football team_sentence_1",
"1950 Duquesne Dukes football team_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5544,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.989",
"value": "?search="
},
{
"operation": "search",
"time": "14.102",
"value": "1950 Duquesne Dukes football team"
},
{
"operation": "Now on",
"time": "14.973",
"value": "1950 Duquesne Dukes football team"
},
{
"operation": "Highlighting",
"time": "17.642",
"value": "1950 Duquesne Dukes football team_sentence_0"
},
{
"operation": "Highlighting",
"time": "24.671",
"value": "1950 Duquesne Dukes football team_sentence_1"
},
{
"operation": "Highlighting",
"time": "29.302",
"value": "1950 Duquesne Dukes football team_sentence_5"
},
{
"operation": "finish",
"time": "35.296",
"value": "finish"
}
]
}
|
exfever_6434
|
Boromir is a fictional character appears in the second volume of J. R. R. Tolkien's high fantasy novel The Lord of the Rings.
|
{
"evidence": [
"Boromir",
"The Two Towers"
],
"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:\nBoromir is a fictional character appears in the second volume of J. R. R. Tolkien's high fantasy novel The Lord of the Rings.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Boromir",
"The Two Towers"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5545,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
37068
|
Oswald Houston and his family moved to Tennessee in 1833, then he and his wife raised their family, then he became the first treasurer of the city of Atlanta (who was responsible for liquidity risk management, cash management, issuing and repaying debt, and interest rate risk, and oversight of pension investment management) after its inception in 1847 to the mid-1850s.
|
{
"evidence": [
"Oswald Houston_sentence_6",
"Oswald Houston_sentence_0",
"Oswald Houston_sentence_8",
"Oswald Houston_sentence_11"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nOswald Houston and his family moved to Tennessee in 1833, then he and his wife raised their family, then he became the first treasurer of the city of Atlanta (who was responsible for liquidity risk management, cash management, issuing and repaying debt, and interest rate risk, and oversight of pension investment management) after its inception in 1847 to the mid-1850s.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Oswald Houston_sentence_6",
"Oswald Houston_sentence_0",
"Oswald Houston_sentence_8",
"Oswald Houston_sentence_11"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5546,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.599",
"value": "?search="
},
{
"operation": "search",
"time": "16.735",
"value": "Oswald Houston"
},
{
"operation": "Now on",
"time": "17.323",
"value": "Oswald Houston"
},
{
"operation": "Highlighting",
"time": "30.838",
"value": "Oswald Houston_sentence_6"
},
{
"operation": "Highlighting",
"time": "41.554",
"value": "Oswald Houston_sentence_0"
},
{
"operation": "Highlighting",
"time": "52.246",
"value": "Oswald Houston_sentence_8"
},
{
"operation": "Highlighting",
"time": "77.518",
"value": "Oswald Houston_sentence_11"
},
{
"operation": "search",
"time": "155.997",
"value": "Treasurer City of Atlanta"
},
{
"operation": "Now on",
"time": "157.562",
"value": "undefined"
},
{
"operation": "search",
"time": "171.418",
"value": "Atlanta Treasurer"
},
{
"operation": "Now on",
"time": "171.918",
"value": "undefined"
},
{
"operation": "finish",
"time": "210.873",
"value": "finish"
}
]
}
|
exfever_8159
|
The Himalayas or Himalayas are a mountain range in Asia, separating the plains of the physiographical region in Northern Asia from the Tibetan Plateau.
|
{
"evidence": [
"Himalayas",
"Indian subcontinent"
],
"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 Himalayas or Himalayas are a mountain range in Asia, separating the plains of the physiographical region in Northern Asia from the Tibetan Plateau.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Himalayas",
"Indian subcontinent"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5547,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
40176
|
In the girls badminton at the 2018 Summer Youth Olympics, Wang Zhiyi of China was ranked higher than Wei Yaxin, also of China.
|
{
"evidence": [
"Wang Zhiyi_sentence_1",
"Wang Zhiyi_sentence_6",
"Wei Yaxin_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn the girls badminton at the 2018 Summer Youth Olympics, Wang Zhiyi of China was ranked higher than Wei Yaxin, also of China.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Wang Zhiyi_sentence_1",
"Wang Zhiyi_sentence_6",
"Wei Yaxin_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5548,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.003",
"value": "?search="
},
{
"operation": "search",
"time": "15.961",
"value": "2018 Summer Youth Olympics"
},
{
"operation": "Now on",
"time": "17.151",
"value": "2018 Summer Youth Olympics"
},
{
"operation": "search",
"time": "32.638",
"value": "Contains...girls badminton at the 2018 summer olymics"
},
{
"operation": "Now on",
"time": "33.556",
"value": "undefined"
},
{
"operation": "search",
"time": "65.431",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles"
},
{
"operation": "Now on",
"time": "66.81",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles"
},
{
"operation": "Highlighting",
"time": "107.81",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_0_0"
},
{
"operation": "Highlighting",
"time": "107.811",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_title"
},
{
"operation": "Highlighting",
"time": "107.812",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_1"
},
{
"operation": "Highlighting",
"time": "107.813",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_2"
},
{
"operation": "Highlighting",
"time": "107.813",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_10"
},
{
"operation": "Highlighting",
"time": "109.994",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_cell_8_2_0"
},
{
"operation": "Highlighting",
"time": "110",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_0_0"
},
{
"operation": "Highlighting",
"time": "111.858",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_0_1"
},
{
"operation": "Highlighting",
"time": "113.631",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_0"
},
{
"operation": "Highlighting",
"time": "114.231",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_cell_8_2_1"
},
{
"operation": "Highlighting",
"time": "114.236",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_0"
},
{
"operation": "Highlighting",
"time": "114.24",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_0_1"
},
{
"operation": "Highlighting",
"time": "114.818",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_1"
},
{
"operation": "Highlighting",
"time": "115.329",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_cell_8_2_2"
},
{
"operation": "Highlighting",
"time": "115.337",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_1"
},
{
"operation": "Highlighting",
"time": "115.866",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_2"
},
{
"operation": "Highlighting",
"time": "116.488",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_cell_8_2_3"
},
{
"operation": "Highlighting",
"time": "116.497",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_2"
},
{
"operation": "Highlighting deleted",
"time": "126.059",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_cell_8_2_1"
},
{
"operation": "Highlighting deleted",
"time": "126.061",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_0"
},
{
"operation": "Highlighting deleted",
"time": "126.673",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_cell_8_2_2"
},
{
"operation": "Highlighting deleted",
"time": "126.68",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_1"
},
{
"operation": "Highlighting deleted",
"time": "127.141",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_cell_8_2_3"
},
{
"operation": "Highlighting deleted",
"time": "127.146",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_2"
},
{
"operation": "Highlighting deleted",
"time": "127.147",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_0_1"
},
{
"operation": "Highlighting deleted",
"time": "128.167",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_cell_8_2_0"
},
{
"operation": "Highlighting deleted",
"time": "128.169",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_title"
},
{
"operation": "Highlighting deleted",
"time": "128.172",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_1"
},
{
"operation": "Highlighting deleted",
"time": "128.174",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_2"
},
{
"operation": "Highlighting deleted",
"time": "128.175",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_10"
},
{
"operation": "Highlighting deleted",
"time": "128.177",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_0_0"
},
{
"operation": "Highlighting",
"time": "170.762",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_0_0"
},
{
"operation": "Highlighting",
"time": "170.764",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_title"
},
{
"operation": "Highlighting",
"time": "170.765",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_1"
},
{
"operation": "Highlighting",
"time": "170.765",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_2"
},
{
"operation": "Highlighting",
"time": "170.766",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_10"
},
{
"operation": "Highlighting",
"time": "172.061",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_cell_8_2_0"
},
{
"operation": "Highlighting",
"time": "172.067",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_0_0"
},
{
"operation": "Highlighting",
"time": "173.297",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_0_1"
},
{
"operation": "Highlighting",
"time": "174.038",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_0"
},
{
"operation": "Highlighting deleted",
"time": "175.137",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_0"
},
{
"operation": "Highlighting deleted",
"time": "175.713",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_0_1"
},
{
"operation": "Highlighting",
"time": "176.322",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_0_1"
},
{
"operation": "Highlighting",
"time": "176.809",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_0"
},
{
"operation": "Highlighting",
"time": "177.285",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_cell_8_2_1"
},
{
"operation": "Highlighting",
"time": "177.292",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_0"
},
{
"operation": "Highlighting",
"time": "177.295",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_0_1"
},
{
"operation": "Highlighting",
"time": "177.764",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_1"
},
{
"operation": "Highlighting",
"time": "178.22",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_cell_8_2_2"
},
{
"operation": "Highlighting",
"time": "178.23",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_1"
},
{
"operation": "Highlighting",
"time": "178.765",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_2"
},
{
"operation": "Highlighting",
"time": "179.254",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_cell_8_3_3"
},
{
"operation": "Highlighting",
"time": "179.263",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_1_2"
},
{
"operation": "hyperlink",
"time": "180.167",
"value": "CHN"
},
{
"operation": "Now on",
"time": "181.719",
"value": "China at the 2018 Summer Youth Olympics"
},
{
"operation": "back-button-clicked",
"time": "182.064",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "182.196",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles"
},
{
"operation": "hyperlink",
"time": "183.595",
"value": "Wang Zhiyi"
},
{
"operation": "Now on",
"time": "184.154",
"value": "Wang Zhiyi"
},
{
"operation": "Highlighting",
"time": "202.935",
"value": "Wang Zhiyi_sentence_1"
},
{
"operation": "Highlighting",
"time": "206.617",
"value": "Wang Zhiyi_sentence_6"
},
{
"operation": "Highlighting",
"time": "227.347",
"value": "Wang Zhiyi_sentence_1"
},
{
"operation": "Highlighting",
"time": "228.815",
"value": "Wang Zhiyi_sentence_6"
},
{
"operation": "search",
"time": "237.634",
"value": "Wei Yaxin"
},
{
"operation": "Now on",
"time": "238.359",
"value": "Wei Yaxin"
},
{
"operation": "Highlighting",
"time": "250.283",
"value": "Wei Yaxin_sentence_1"
},
{
"operation": "back-button-clicked",
"time": "277.807",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "277.881",
"value": "Wang Zhiyi"
},
{
"operation": "back-button-clicked",
"time": "279.601",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "279.606",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "279.705",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles"
},
{
"operation": "Highlighting",
"time": "282.558",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_0_0"
},
{
"operation": "Highlighting",
"time": "282.56",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_title"
},
{
"operation": "Highlighting",
"time": "282.561",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_1"
},
{
"operation": "Highlighting",
"time": "282.561",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_2"
},
{
"operation": "Highlighting",
"time": "282.562",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_10"
},
{
"operation": "Highlighting deleted",
"time": "283.352",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_0_0"
},
{
"operation": "Highlighting deleted",
"time": "283.352",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_title"
},
{
"operation": "Highlighting deleted",
"time": "283.354",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_1"
},
{
"operation": "Highlighting deleted",
"time": "283.355",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_2"
},
{
"operation": "Highlighting deleted",
"time": "283.356",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_10"
},
{
"operation": "Highlighting",
"time": "284.266",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_0_0"
},
{
"operation": "Highlighting",
"time": "284.268",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_title"
},
{
"operation": "Highlighting",
"time": "284.268",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_1"
},
{
"operation": "Highlighting",
"time": "284.268",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_2"
},
{
"operation": "Highlighting",
"time": "284.269",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_section_10"
},
{
"operation": "Highlighting",
"time": "284.832",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_cell_8_2_0"
},
{
"operation": "Highlighting",
"time": "284.841",
"value": "Badminton at the 2018 Summer Youth Olympics – Girls' singles_header_cell_8_0_0"
},
{
"operation": "finish",
"time": "380.592",
"value": "finish"
}
]
}
|
exfever_4368
|
Maya Khabira Rudolph is an actress who played supporting roles in a 2006 American musical comedy film featuring an American actor who received an Academy Award and three Tony Awards.
|
{
"evidence": [
"Maya Rudolph",
"A Prairie Home Companion (film)",
"Kevin Kline"
],
"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:\nMaya Khabira Rudolph is an actress who played supporting roles in a 2006 American musical comedy film featuring an American actor who received an Academy Award and three Tony Awards.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Maya Rudolph",
"A Prairie Home Companion (film)",
"Kevin Kline"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5549,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
86500
|
Wuhu ranked first in terms of GDP, with the Primary Industry being its biggest contributor, among the 16 prefectures in Anhui Province.
|
{
"evidence": [
"Anhui_sentence_69",
"Anhui_sentence_90",
"Anhui_sentence_91"
],
"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:\nWuhu ranked first in terms of GDP, with the Primary Industry being its biggest contributor, among the 16 prefectures in Anhui Province.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Anhui_sentence_69",
"Anhui_sentence_90",
"Anhui_sentence_91"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5550,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.735",
"value": "?search="
},
{
"operation": "search",
"time": "17.697",
"value": "Anhui Province"
},
{
"operation": "Now on",
"time": "19.673",
"value": "Anhui"
},
{
"operation": "hyperlink",
"time": "42.961",
"value": "6.1.3 Wuhu Economic and Technological Development Zone Anhui_section_10"
},
{
"operation": "search",
"time": "118.887",
"value": "Wuhu"
},
{
"operation": "Now on",
"time": "119.818",
"value": "WUHU"
},
{
"operation": "hyperlink",
"time": "123.867",
"value": "Wuhu"
},
{
"operation": "Now on",
"time": "124.246",
"value": "Wuhu (disambiguation)"
},
{
"operation": "Highlighting",
"time": "129.421",
"value": "Wuhu (disambiguation)_item_0_0"
},
{
"operation": "Highlighting",
"time": "129.424",
"value": "Wuhu (disambiguation)_title"
},
{
"operation": "Highlighting deleted",
"time": "139.961",
"value": "Wuhu (disambiguation)_item_0_0"
},
{
"operation": "Highlighting deleted",
"time": "139.962",
"value": "Wuhu (disambiguation)_title"
},
{
"operation": "search",
"time": "149.483",
"value": "Contains...Wuhu ranked first in terms of GDP"
},
{
"operation": "Now on",
"time": "151.383",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "163.015",
"value": "Wuhu Economic and Technological Development Zone Anhui_section_10"
},
{
"operation": "Now on",
"time": "164.698",
"value": "Anhui#Wuhu Economic and Technological Development Zone Anhui section 10"
},
{
"operation": "hyperlink",
"time": "176.866",
"value": "6.1.3 Wuhu Economic and Technological Development Zone Anhui_section_10"
},
{
"operation": "search",
"time": "259.786",
"value": "Contains...he Primary Industry being its biggest contributor"
},
{
"operation": "Now on",
"time": "262.417",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "273.842",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "274.02",
"value": "Anhui#Wuhu Economic and Technological Development Zone Anhui section 10"
},
{
"operation": "hyperlink",
"time": "303.63",
"value": "6.1 Major economic and technological development zones Anhui_section_7"
},
{
"operation": "Highlighting",
"time": "351.038",
"value": "Anhui_sentence_69"
},
{
"operation": "Highlighting",
"time": "443.641",
"value": "Anhui_sentence_90"
},
{
"operation": "Highlighting",
"time": "454.04",
"value": "Anhui_sentence_91"
},
{
"operation": "finish",
"time": "490.939",
"value": "finish"
}
]
}
|
exfever_4560
|
Pablo F. Fenjves is a Los Angeles-based screenwriter and a writer who writes literary or journalistic works, speeches, or other texts officially credited to another person as the author.
|
{
"evidence": [
"Pablo Fenjves",
"Ghostwriter"
],
"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:\nPablo F. Fenjves is a Los Angeles-based screenwriter and a writer who writes literary or journalistic works, speeches, or other texts officially credited to another person as the author.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Pablo Fenjves",
"Ghostwriter"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5551,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
5440
|
The Hong Kong government has several policies for Dai Pai Dongs which centers on the tightening of licenses transfer and and the improvement of of quality of Dai Pai Dongs with the Food and Environmental Hygiene Department providing funds to stalls located in Central.
|
{
"evidence": [
"Dai pai dong_sentence_24",
"Dai pai dong_sentence_6",
"Dai pai dong_sentence_9"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Hong Kong government has several policies for Dai Pai Dongs which centers on the tightening of licenses transfer and and the improvement of of quality of Dai Pai Dongs with the Food and Environmental Hygiene Department providing funds to stalls located in Central.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Dai pai dong_sentence_24",
"Dai pai dong_sentence_6",
"Dai pai dong_sentence_9"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5552,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.584",
"value": "undefined"
},
{
"operation": "Now on",
"time": "16.959",
"value": "undefined"
},
{
"operation": "search",
"time": "118.821",
"value": "Dai pai dong"
},
{
"operation": "Now on",
"time": "119.814",
"value": "Dai pai dong"
},
{
"operation": "Highlighting",
"time": "212.115",
"value": "Dai pai dong_sentence_24"
},
{
"operation": "Highlighting",
"time": "234.546",
"value": "Dai pai dong_sentence_6"
},
{
"operation": "Highlighting",
"time": "242.152",
"value": "Dai pai dong_sentence_9"
},
{
"operation": "hyperlink",
"time": "249.138",
"value": "Food and Environmental Hygiene Department"
},
{
"operation": "Now on",
"time": "249.885",
"value": "Food and Environmental Hygiene Department"
},
{
"operation": "back-button-clicked",
"time": "260.49",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "260.559",
"value": "Dai pai dong"
},
{
"operation": "finish",
"time": "376.242",
"value": "finish"
}
]
}
|
exfever_7233
|
Murder on the Orient Express is a 2017 mystery film directed by a British actor and filmmaker.
|
{
"evidence": [
"Murder on the Orient Express (2017 film)",
"Kenneth Branagh"
],
"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:\nMurder on the Orient Express is a 2017 mystery film directed by a British actor and filmmaker.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Murder on the Orient Express (2017 film)",
"Kenneth Branagh"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5553,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
43281
|
In 2011, Evans signed with the Cincinnati Bengals after going undrafted in the NFL draft; but in November 2011, Evans was suspended for four games.
|
{
"evidence": [
"DeQuin Evans_sentence_13",
"DeQuin Evans_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:\nIn 2011, Evans signed with the Cincinnati Bengals after going undrafted in the NFL draft; but in November 2011, Evans was suspended for four games.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"DeQuin Evans_sentence_13",
"DeQuin Evans_sentence_14"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5554,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.194",
"value": "?search="
},
{
"operation": "search",
"time": "23.2",
"value": "Evans signed with the Cincinnati Bengals"
},
{
"operation": "Now on",
"time": "25.451",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "30.264",
"value": "DeQuin Evans"
},
{
"operation": "Now on",
"time": "30.848",
"value": "DeQuin Evans"
},
{
"operation": "Highlighting",
"time": "36.377",
"value": "DeQuin Evans_sentence_13"
},
{
"operation": "Highlighting",
"time": "47.19",
"value": "DeQuin Evans_sentence_14"
},
{
"operation": "finish",
"time": "65.307",
"value": "finish"
}
]
}
|
27365
|
Renault Scénic was mechanically identical to the Mégane hatchback and marketed as a multi purpose vehicle by Renault, the first to be lablled as such in Europe.
|
{
"evidence": [
"Renault Scénic_sentence_6",
"Renault Scénic_sentence_8",
"Renault Scénic_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:\nRenault Scénic was mechanically identical to the Mégane hatchback and marketed as a multi purpose vehicle by Renault, the first to be lablled as such in Europe.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Renault Scénic_sentence_6",
"Renault Scénic_sentence_8",
"Renault Scénic_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5555,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.638",
"value": "?search="
},
{
"operation": "search",
"time": "5.828",
"value": "Renault Scénic"
},
{
"operation": "Now on",
"time": "6.605",
"value": "Renault Scénic"
},
{
"operation": "Highlighting",
"time": "21.078",
"value": "Renault Scénic_sentence_6"
},
{
"operation": "Highlighting",
"time": "24.649",
"value": "Renault Scénic_sentence_8"
},
{
"operation": "hyperlink",
"time": "48.253",
"value": "French"
},
{
"operation": "Now on",
"time": "52.922",
"value": "Renault Scénic"
},
{
"operation": "Highlighting",
"time": "53.651",
"value": "Renault Scénic_sentence_0"
},
{
"operation": "finish",
"time": "66.624",
"value": "finish"
}
]
}
|
exfever_6082
|
Wayne Rooney is a former football player who has the most goals scored for the?national football team that has represented Japan.
|
{
"evidence": [
"Wayne Rooney",
"England national football team"
],
"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:\nWayne Rooney is a former football player who has the most goals scored for the?national football team that has represented Japan.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Wayne Rooney",
"England national football team"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5556,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8095
|
Saving Private Ryan is a 1998 American epic war film written by Robert Rodat that is set during Operation Overlord, the Allied operation that launched the successful invasion of German-occupied Western Europe during World War I.
|
{
"evidence": [
"Saving Private Ryan",
"Invasion of Normandy"
],
"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:\nSaving Private Ryan is a 1998 American epic war film written by Robert Rodat that is set during Operation Overlord, the Allied operation that launched the successful invasion of German-occupied Western Europe during World War I.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Saving Private Ryan",
"Invasion of Normandy"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5557,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1071
|
The Incredibles is a 2004 computer-animated film that focuses on the actions of superheroes.
|
{
"evidence": [
"The Incredibles",
"Superhero film"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Incredibles is a 2004 computer-animated film that focuses on the actions of superheroes.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Incredibles",
"Superhero film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5558,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
51764
|
Jorunn Ringstad, born on September 3, 1943, was a Norwegian politician for the Centre Party (neither an agrarian nor centrist political party), and was elected to the Norwegian Parliament from Sogn og Fjordane in 1993.
|
{
"evidence": [
"Jorunn Ringstad_sentence_0",
"Centre Party (Norway)_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:\nJorunn Ringstad, born on September 3, 1943, was a Norwegian politician for the Centre Party (neither an agrarian nor centrist political party), and was elected to the Norwegian Parliament from Sogn og Fjordane in 1993.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Jorunn Ringstad_sentence_0",
"Centre Party (Norway)_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5559,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.473",
"value": "?search="
},
{
"operation": "search",
"time": "9.847",
"value": "Jorunn Ringstad"
},
{
"operation": "Now on",
"time": "11.235",
"value": "Jorunn Ringstad"
},
{
"operation": "Highlighting",
"time": "16.58",
"value": "Jorunn Ringstad_sentence_0"
},
{
"operation": "hyperlink",
"time": "20.717",
"value": "Centre Party"
},
{
"operation": "Now on",
"time": "23.274",
"value": "Centre Party (Norway)"
},
{
"operation": "Highlighting",
"time": "25.729",
"value": "Centre Party (Norway)_sentence_0"
},
{
"operation": "finish",
"time": "37.342",
"value": "finish"
}
]
}
|
92356
|
Richard I of England, son of King Henry II of England and brother-in-law of William II of Sicily through the marriage of the later with his sister Joan of England, would prove to be just as adventurous as his kinsman.
|
{
"evidence": [
"Richard I of England_sentence_1",
"Richard I of England_sentence_3",
"Joan of England, Queen of Sicily_sentence_3"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRichard I of England, son of King Henry II of England and brother-in-law of William II of Sicily through the marriage of the later with his sister Joan of England, would prove to be just as adventurous as his kinsman.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Richard I of England_sentence_1",
"Richard I of England_sentence_3",
"Joan of England, Queen of Sicily_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5560,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.693",
"value": "undefined"
},
{
"operation": "search",
"time": "6.955",
"value": "Richard I of England"
},
{
"operation": "Now on",
"time": "9.099",
"value": "Richard I of England"
},
{
"operation": "Highlighting",
"time": "11.444",
"value": "Richard I of England_sentence_1"
},
{
"operation": "Highlighting",
"time": "23.204",
"value": "Richard I of England_sentence_3"
},
{
"operation": "Page search",
"time": "54.061",
"value": "William II"
},
{
"operation": "Page search",
"time": "109.306",
"value": "brother-in-law"
},
{
"operation": "Page search",
"time": "129.41",
"value": "joan"
},
{
"operation": "Page search",
"time": "130.068",
"value": "joan"
},
{
"operation": "hyperlink",
"time": "141.776",
"value": "Queen Joan of Sicily"
},
{
"operation": "Now on",
"time": "142.526",
"value": "Joan of England, Queen of Sicily"
},
{
"operation": "Highlighting",
"time": "150.562",
"value": "Joan of England, Queen of Sicily_sentence_3"
},
{
"operation": "back-button-clicked",
"time": "160.723",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "160.854",
"value": "Richard I of England"
},
{
"operation": "Page search",
"time": "180.127",
"value": "adventu"
},
{
"operation": "Page search",
"time": "180.769",
"value": "adventu"
},
{
"operation": "hyperlink",
"time": "209.459",
"value": "Queen Joan of Sicily"
},
{
"operation": "Now on",
"time": "209.764",
"value": "Joan of England, Queen of Sicily"
},
{
"operation": "back-button-clicked",
"time": "346.001",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "346.132",
"value": "Richard I of England"
},
{
"operation": "Page search",
"time": "354.456",
"value": "kinsman"
},
{
"operation": "Page search",
"time": "372.601",
"value": "kinsman"
},
{
"operation": "Page search",
"time": "401.7",
"value": "adventurous"
},
{
"operation": "finish",
"time": "584.79",
"value": "finish"
}
]
}
|
74000
|
The only team to receive tobacco sponsorship for the 2007 season, Ferrari had Philip Morris International as its principal sponsor.
|
{
"evidence": [
"Philip Morris International_sentence_88"
],
"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 only team to receive tobacco sponsorship for the 2007 season, Ferrari had Philip Morris International as its principal sponsor.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Philip Morris International_sentence_88"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5561,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.76",
"value": "?search="
},
{
"operation": "search",
"time": "15.163",
"value": "Ferrari Phillip Morris"
},
{
"operation": "Now on",
"time": "16.842",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "21.98",
"value": "Philip Morris International"
},
{
"operation": "Now on",
"time": "23.322",
"value": "Philip Morris International"
},
{
"operation": "Highlighting",
"time": "30.676",
"value": "Philip Morris International_sentence_88"
},
{
"operation": "search",
"time": "57.98",
"value": "2007 Ferrari"
},
{
"operation": "Now on",
"time": "59.318",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "88.75",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "88.893",
"value": "Philip Morris International"
},
{
"operation": "search",
"time": "107.064",
"value": "Phillip Morris tabacco sponsership"
},
{
"operation": "Now on",
"time": "108.145",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "116.583",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "116.791",
"value": "Philip Morris International"
},
{
"operation": "search",
"time": "169.614",
"value": "Ferrari Formula One"
},
{
"operation": "Now on",
"time": "171.641",
"value": "undefined"
},
{
"operation": "search",
"time": "188.121",
"value": "2007 Ferrari season"
},
{
"operation": "Now on",
"time": "189.886",
"value": "undefined"
},
{
"operation": "search",
"time": "208.674",
"value": "Ferrari Formula One"
},
{
"operation": "Now on",
"time": "209.58",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "212.894",
"value": "Ferrari"
},
{
"operation": "Now on",
"time": "214.474",
"value": "Ferrari"
},
{
"operation": "hyperlink",
"time": "247.994",
"value": "Formula One"
},
{
"operation": "Now on",
"time": "256.59",
"value": "Formula One"
},
{
"operation": "finish",
"time": "282.24",
"value": "finish"
}
]
}
|
exfever_4058
|
The comedy series Community (2009–2015) won a Primetime Emmy Award.
|
{
"evidence": [
"Alison Brie",
"Mad Men"
],
"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 comedy series Community (2009–2015) won a Primetime Emmy Award.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Alison Brie",
"Mad Men"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5562,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
1965
|
Kinase insert domain receptor, a VEGF receptor encoded by the human gene KDR, has been shown to interact with SHC2.
|
{
"evidence": [
"Kinase insert domain receptor_sentence_0",
"Kinase insert domain receptor_sentence_1",
"Kinase insert domain receptor_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:\nKinase insert domain receptor, a VEGF receptor encoded by the human gene KDR, has been shown to interact with SHC2.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Kinase insert domain receptor_sentence_0",
"Kinase insert domain receptor_sentence_1",
"Kinase insert domain receptor_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5563,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "22.821",
"value": "Kinase insert domain receptor"
},
{
"operation": "Highlighting",
"time": "30.463",
"value": "Kinase insert domain receptor_sentence_0"
},
{
"operation": "Highlighting",
"time": "39.316",
"value": "Kinase insert domain receptor_sentence_1"
},
{
"operation": "Highlighting",
"time": "44.633",
"value": "Kinase insert domain receptor_sentence_5"
},
{
"operation": "finish",
"time": "61.529",
"value": "finish"
}
]
}
|
exfever_4752
|
Naturi Cora Maria Naughton is an American author best known as one-third of an American girl group that enjoyed a number of modest hits during the early 2000s.
|
{
"evidence": [
"Naturi Naughton",
"3LW"
],
"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:\nNaturi Cora Maria Naughton is an American author best known as one-third of an American girl group that enjoyed a number of modest hits during the early 2000s.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Naturi Naughton",
"3LW"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5564,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
79379
|
Though he did not take official credit, Thomas Muntzer is thought to have been partially responsible for creating the first document regarding civil liberties after the Roman Empire, potentially due to his 1523 opposition to Martin Luther.
|
{
"evidence": [
"Thomas Müntzer_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThough he did not take official credit, Thomas Muntzer is thought to have been partially responsible for creating the first document regarding civil liberties after the Roman Empire, potentially due to his 1523 opposition to Martin Luther.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Thomas Müntzer_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5565,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.746",
"value": "?search="
},
{
"operation": "search",
"time": "19.556",
"value": "Thomas Müntzer"
},
{
"operation": "Now on",
"time": "20.765",
"value": "Thomas Müntzer"
},
{
"operation": "Page search",
"time": "74.614",
"value": "official credit"
},
{
"operation": "Page search",
"time": "136.095",
"value": "1523"
},
{
"operation": "Page search",
"time": "188.51",
"value": "civil liberties"
},
{
"operation": "Highlighting",
"time": "236.817",
"value": "Thomas Müntzer_sentence_1"
},
{
"operation": "Page search",
"time": "253.11",
"value": "credit"
},
{
"operation": "search",
"time": "308.684",
"value": "Contains...Thomas Muntzer is thought to have been partially responsible for creating the first document regarding civil liberties"
},
{
"operation": "Now on",
"time": "310.645",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "322.987",
"value": "Collision with the Roman Empire History of Germany_section_4"
},
{
"operation": "Now on",
"time": "324.399",
"value": "History of Germany#Collision with the Roman Empire History of Germany section 4"
},
{
"operation": "Page search",
"time": "343.146",
"value": "Thomas"
},
{
"operation": "Page search",
"time": "370.514",
"value": "civil liberties"
},
{
"operation": "finish",
"time": "404.955",
"value": "finish"
}
]
}
|
64723
|
After graduating from Princeton University in 1977, Henry Farber joined the Massachusetts Institute of Technology (a school in Cambridge) as an assistant professor of economics; then in 1991, he left MIT in order to return to Princeton University as a professor of economics, before being endowed with the Hughes-Rogers professorship in 1995.
|
{
"evidence": [
"Henry Farber_sentence_7",
"Massachusetts Institute of Technology_sentence_2",
"Henry Farber_sentence_8"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAfter graduating from Princeton University in 1977, Henry Farber joined the Massachusetts Institute of Technology (a school in Cambridge) as an assistant professor of economics; then in 1991, he left MIT in order to return to Princeton University as a professor of economics, before being endowed with the Hughes-Rogers professorship in 1995.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Henry Farber_sentence_7",
"Massachusetts Institute of Technology_sentence_2",
"Henry Farber_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5566,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.689",
"value": "?search="
},
{
"operation": "search",
"time": "68.492",
"value": "Henry Farber"
},
{
"operation": "Now on",
"time": "69.394",
"value": "Henry Farber"
},
{
"operation": "Highlighting",
"time": "291.041",
"value": "Henry Farber_sentence_4"
},
{
"operation": "Highlighting",
"time": "291.041",
"value": "Henry Farber_section_0"
},
{
"operation": "Highlighting",
"time": "291.819",
"value": "Henry Farber_sentence_5"
},
{
"operation": "Highlighting",
"time": "298.51",
"value": "Henry Farber_sentence_2"
},
{
"operation": "Highlighting",
"time": "300.038",
"value": "Henry Farber_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "303.826",
"value": "Henry Farber_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "304.363",
"value": "Henry Farber_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "304.763",
"value": "Henry Farber_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "305.164",
"value": "Henry Farber_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "305.166",
"value": "Henry Farber_section_0"
},
{
"operation": "Highlighting",
"time": "306.368",
"value": "Henry Farber_sentence_7"
},
{
"operation": "hyperlink",
"time": "313.092",
"value": "Massachusetts Institute of Technology"
},
{
"operation": "Now on",
"time": "315.57",
"value": "Massachusetts Institute of Technology"
},
{
"operation": "Highlighting",
"time": "317.609",
"value": "Massachusetts Institute of Technology_sentence_2"
},
{
"operation": "back-button-clicked",
"time": "319.283",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "319.315",
"value": "Henry Farber"
},
{
"operation": "Highlighting",
"time": "324.075",
"value": "Henry Farber_sentence_8"
},
{
"operation": "finish",
"time": "340.254",
"value": "finish"
}
]
}
|
65623
|
Mode Media was founded by nine people in 2003 and was originally a newspaper that offered fashion and beauty content and blog content.
|
{
"evidence": [
"Mode Media_sentence_6"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMode Media was founded by nine people in 2003 and was originally a newspaper that offered fashion and beauty content and blog content.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Mode Media_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5567,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.124",
"value": "?search="
},
{
"operation": "search",
"time": "8.266",
"value": "Mode Media"
},
{
"operation": "Now on",
"time": "9.154",
"value": "Mode Media"
},
{
"operation": "Highlighting",
"time": "11.559",
"value": "Mode Media_cell_0_2_1"
},
{
"operation": "Highlighting",
"time": "11.565",
"value": "Mode Media_header_cell_0_2_0"
},
{
"operation": "Highlighting",
"time": "13.5",
"value": "Mode Media_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "20.424",
"value": "Mode Media_cell_0_3_1"
},
{
"operation": "Highlighting",
"time": "20.43",
"value": "Mode Media_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "57.06",
"value": "Mode Media_sentence_5"
},
{
"operation": "Highlighting",
"time": "58.062",
"value": "Mode Media_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "58.831",
"value": "Mode Media_sentence_5"
},
{
"operation": "Page search",
"time": "68.586",
"value": "newspaper"
},
{
"operation": "Highlighting deleted",
"time": "72.853",
"value": "Mode Media_cell_0_3_1"
},
{
"operation": "Highlighting deleted",
"time": "72.855",
"value": "Mode Media_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "73.308",
"value": "Mode Media_cell_0_3_1"
},
{
"operation": "Highlighting",
"time": "73.311",
"value": "Mode Media_header_cell_0_3_0"
},
{
"operation": "Highlighting deleted",
"time": "74.146",
"value": "Mode Media_cell_0_2_1"
},
{
"operation": "Highlighting deleted",
"time": "74.149",
"value": "Mode Media_header_cell_0_2_0"
},
{
"operation": "Highlighting deleted",
"time": "74.52",
"value": "Mode Media_cell_0_3_1"
},
{
"operation": "Highlighting deleted",
"time": "74.523",
"value": "Mode Media_header_cell_0_3_0"
},
{
"operation": "finish",
"time": "82.978",
"value": "finish"
}
]
}
|
65922
|
John Edward Costigan's oil and watercolor paintings featuring his wife and child made him a member of the honorary association of American artists, the National Academy of Design.
|
{
"evidence": [
"John Edward Costigan_sentence_4",
"John Edward Costigan_sentence_6",
"John Edward Costigan_sentence_7",
"National Academy of Design_sentence_0",
"National Academy of Design_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJohn Edward Costigan's oil and watercolor paintings featuring his wife and child made him a member of the honorary association of American artists, the National Academy of Design.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"John Edward Costigan_sentence_4",
"John Edward Costigan_sentence_6",
"John Edward Costigan_sentence_7",
"National Academy of Design_sentence_0",
"National Academy of Design_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5568,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.806",
"value": "?search="
},
{
"operation": "search",
"time": "16.316",
"value": "John Edward Costigan"
},
{
"operation": "Now on",
"time": "39.43",
"value": "?search="
},
{
"operation": "search",
"time": "49.429",
"value": "John Edward Costigan"
},
{
"operation": "Now on",
"time": "49.949",
"value": "John Edward Costigan"
},
{
"operation": "Highlighting",
"time": "62.236",
"value": "John Edward Costigan_sentence_4"
},
{
"operation": "Highlighting",
"time": "67.739",
"value": "John Edward Costigan_sentence_6"
},
{
"operation": "Highlighting",
"time": "70.386",
"value": "John Edward Costigan_sentence_7"
},
{
"operation": "hyperlink",
"time": "112.649",
"value": "National Academy of Design"
},
{
"operation": "Now on",
"time": "113.944",
"value": "National Academy of Design"
},
{
"operation": "Highlighting",
"time": "115.697",
"value": "National Academy of Design_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "117.258",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "117.302",
"value": "John Edward Costigan"
},
{
"operation": "hyperlink",
"time": "125.001",
"value": "National Academy of Design"
},
{
"operation": "Now on",
"time": "125.283",
"value": "National Academy of Design"
},
{
"operation": "Highlighting",
"time": "133.065",
"value": "National Academy of Design_sentence_1"
},
{
"operation": "finish",
"time": "138.928",
"value": "finish"
}
]
}
|
exfever_7884
|
Johnny Mnemonic is a 1995 cyberpunk film directed by an American filmmaker and a person engaged in an activity related to creating art.
|
{
"evidence": [
"Johnny Mnemonic (film)",
"Robert Longo",
"Artist"
],
"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:\nJohnny Mnemonic is a 1995 cyberpunk film directed by an American filmmaker and a person engaged in an activity related to creating art.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Johnny Mnemonic (film)",
"Robert Longo",
"Artist"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5569,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_234
|
Until Dawn is an interactive drama horror video game developed by a British video game developer based in Guildford, Surrey, that is best known for developing horror games such as Until Dawn for the PlayStation 4.
|
{
"evidence": [
"Until Dawn",
"Supermassive Games"
],
"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:\nUntil Dawn is an interactive drama horror video game developed by a British video game developer based in Guildford, Surrey, that is best known for developing horror games such as Until Dawn for the PlayStation 4.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Until Dawn",
"Supermassive Games"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5570,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
4203
|
Best known for her paintings of horses and cavalry, Magda died in1929 at the age of 71.
|
{
"evidence": [
"Francesca Stuart Sindici_sentence_5",
"Francesca Stuart Sindici_sentence_4"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBest known for her paintings of horses and cavalry, Magda died in1929 at the age of 71.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Francesca Stuart Sindici_sentence_5",
"Francesca Stuart Sindici_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5571,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "0.115",
"value": "magda"
},
{
"operation": "search",
"time": "0.165",
"value": "magda henderson"
},
{
"operation": "search",
"time": "0.216",
"value": "maria magda"
},
{
"operation": "search",
"time": "0.225",
"value": "Magda died in 1929"
},
{
"operation": "search",
"time": "0.264",
"value": "Francesca Stuart Sindici"
},
{
"operation": "Highlighting",
"time": "0.274",
"value": "Francesca Stuart Sindici_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "0.281",
"value": "Francesca Stuart Sindici_sentence_0"
},
{
"operation": "Highlighting",
"time": "0.281",
"value": "Francesca Stuart Sindici_sentence_5"
},
{
"operation": "Highlighting",
"time": "0.282",
"value": "Francesca Stuart Sindici_sentence_4"
},
{
"operation": "finish",
"time": "0.294",
"value": "finish"
},
{
"operation": "finish",
"time": "1618806411.303",
"value": "finish"
}
]
}
|
46402
|
The Machine Question's first chapter, titled Moral Agency, breaks down the history of said agency based on what is included and excluded in various parts of history; it also discusses the morality of humans toward objects and the theory of the philosophy of technology, a theory that was first used in the late 19th century by German-born philosopher and geographer Ernst Kapp.
|
{
"evidence": [
"The Machine Question_sentence_3",
"The Machine Question_sentence_4",
"Philosophy of technology_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 Machine Question's first chapter, titled Moral Agency, breaks down the history of said agency based on what is included and excluded in various parts of history; it also discusses the morality of humans toward objects and the theory of the philosophy of technology, a theory that was first used in the late 19th century by German-born philosopher and geographer Ernst Kapp.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"The Machine Question_sentence_3",
"The Machine Question_sentence_4",
"Philosophy of technology_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5572,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.621",
"value": "?search="
},
{
"operation": "search",
"time": "21.159",
"value": "Machine Question's first chapter"
},
{
"operation": "Now on",
"time": "23.183",
"value": "undefined"
},
{
"operation": "search",
"time": "38.032",
"value": "The Machine Question"
},
{
"operation": "Now on",
"time": "39.005",
"value": "The Machine Question"
},
{
"operation": "Page search",
"time": "51.525",
"value": "Moral Agency"
},
{
"operation": "Highlighting",
"time": "59.421",
"value": "The Machine Question_sentence_3"
},
{
"operation": "Highlighting",
"time": "75.168",
"value": "The Machine Question_sentence_4"
},
{
"operation": "hyperlink",
"time": "97.79",
"value": "philosophy of technology"
},
{
"operation": "Now on",
"time": "98.453",
"value": "Philosophy of technology"
},
{
"operation": "Highlighting",
"time": "104.548",
"value": "Philosophy of technology_sentence_2"
},
{
"operation": "finish",
"time": "167.848",
"value": "finish"
}
]
}
|
57855
|
William Raggio served as Senate Minority Floor Leader, Majority Floor Leader and Chairman of the Interim Senate Finance Committee.
|
{
"evidence": [
"William Raggio_sentence_16",
"William Raggio_sentence_17",
"William Raggio_sentence_18"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWilliam Raggio served as Senate Minority Floor Leader, Majority Floor Leader and Chairman of the Interim Senate Finance Committee.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"William Raggio_sentence_16",
"William Raggio_sentence_17",
"William Raggio_sentence_18"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5573,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.54",
"value": "?search="
},
{
"operation": "search",
"time": "7.985",
"value": "William Raggio"
},
{
"operation": "Now on",
"time": "8.951",
"value": "William Raggio"
},
{
"operation": "Highlighting",
"time": "73.658",
"value": "William Raggio_sentence_16"
},
{
"operation": "Highlighting",
"time": "75.345",
"value": "William Raggio_sentence_17"
},
{
"operation": "Highlighting",
"time": "77.179",
"value": "William Raggio_sentence_18"
},
{
"operation": "finish",
"time": "84.14",
"value": "finish"
}
]
}
|
41419
|
2007 Baku terrorist plot prompted the closure of the American and British embassies, while some institutions, such as Norwegian Statoil and American McDermott oil companies, limited their activity; then on October 29, 2007, Azerbaijani law enforcement agencies reported that they had detained a group of Salafis armed with grenade launchers.
|
{
"evidence": [
"2007 Baku terrorist plot_sentence_1",
"2007 Baku terrorist plot_sentence_2",
"2007 Baku terrorist plot_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:\n2007 Baku terrorist plot prompted the closure of the American and British embassies, while some institutions, such as Norwegian Statoil and American McDermott oil companies, limited their activity; then on October 29, 2007, Azerbaijani law enforcement agencies reported that they had detained a group of Salafis armed with grenade launchers.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"2007 Baku terrorist plot_sentence_1",
"2007 Baku terrorist plot_sentence_2",
"2007 Baku terrorist plot_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5574,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.23",
"value": "?search="
},
{
"operation": "search",
"time": "27.396",
"value": "2007 Baku terrorist plot prompted the closure of the American and British embassies"
},
{
"operation": "Now on",
"time": "29.922",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "32.626",
"value": "2007 Baku terrorist plot"
},
{
"operation": "Now on",
"time": "33.353",
"value": "2007 Baku terrorist plot"
},
{
"operation": "Highlighting",
"time": "42.439",
"value": "2007 Baku terrorist plot_sentence_1"
},
{
"operation": "Highlighting",
"time": "49.111",
"value": "2007 Baku terrorist plot_sentence_2"
},
{
"operation": "Highlighting",
"time": "58.393",
"value": "2007 Baku terrorist plot_sentence_3"
},
{
"operation": "finish",
"time": "101.003",
"value": "finish"
}
]
}
|
40219
|
Agnes Jebet Tirop, a Kenyan long-distance runner mainly competing in 5000-meter races and cross country running, won the Kenyan cross country junior title at the 2014 Kenyan Cross Country Championships after being second place during the 2013 African Cross Country Championships.
|
{
"evidence": [
"Agnes Jebet Tirop_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:\nAgnes Jebet Tirop, a Kenyan long-distance runner mainly competing in 5000-meter races and cross country running, won the Kenyan cross country junior title at the 2014 Kenyan Cross Country Championships after being second place during the 2013 African Cross Country Championships.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Agnes Jebet Tirop_sentence_9"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5575,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0",
"value": "?search="
},
{
"operation": "search",
"time": "0.023",
"value": "Agnes Jebet Tirop"
},
{
"operation": "Now on",
"time": "0.024",
"value": "Agnes Jebet Tirop"
},
{
"operation": "Highlighting",
"time": "0.028",
"value": "Agnes Jebet Tirop_sentence_0"
},
{
"operation": "Page search",
"time": "0.051",
"value": "2014"
},
{
"operation": "Highlighting",
"time": "0.086",
"value": "Agnes Jebet Tirop_sentence_12"
},
{
"operation": "Page search",
"time": "0.102",
"value": "2013"
},
{
"operation": "Highlighting",
"time": "0.111",
"value": "Agnes Jebet Tirop_sentence_9"
},
{
"operation": "Highlighting deleted",
"time": "0.14",
"value": "Agnes Jebet Tirop_sentence_9"
},
{
"operation": "Page search",
"time": "0.187",
"value": "2013"
},
{
"operation": "Highlighting",
"time": "0.202",
"value": "Agnes Jebet Tirop_sentence_9"
},
{
"operation": "Highlighting deleted",
"time": "0.211",
"value": "Agnes Jebet Tirop_sentence_12"
},
{
"operation": "Highlighting",
"time": "0.214",
"value": "Agnes Jebet Tirop_item_0_0"
},
{
"operation": "Highlighting",
"time": "0.214",
"value": "Agnes Jebet Tirop_section_1"
},
{
"operation": "Highlighting deleted",
"time": "0.215",
"value": "Agnes Jebet Tirop_item_0_0"
},
{
"operation": "Highlighting deleted",
"time": "0.215",
"value": "Agnes Jebet Tirop_section_1"
},
{
"operation": "Highlighting deleted",
"time": "0.219",
"value": "Agnes Jebet Tirop_sentence_0"
},
{
"operation": "finish",
"time": "0.26",
"value": "finish"
},
{
"operation": "Now on",
"time": "1620680513.5",
"value": "Agnes Jebet Tirop"
},
{
"operation": "finish",
"time": "1620680545.9",
"value": "finish"
}
]
}
|
exfever_1728
|
Ellen Kathleen Pompeo is a Singaporean actress and producer who was born in a city in Middlesex County of most populous state in the New England region of the United States.
|
{
"evidence": [
"Ellen Pompeo",
"Everett, Massachusetts",
"Massachusetts"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nEllen Kathleen Pompeo is a Singaporean actress and producer who was born in a city in Middlesex County of most populous state in the New England region of the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ellen Pompeo",
"Everett, Massachusetts",
"Massachusetts"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5576,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
54290
|
Oenomaus, a native neither of Gadara nor of Petra, is listed in the Chronicle of Jerome as flourishing in the 224th Olympiad.
|
{
"evidence": [
"Oenomaus of Gadara_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:\nOenomaus, a native neither of Gadara nor of Petra, is listed in the Chronicle of Jerome as flourishing in the 224th Olympiad.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Oenomaus of Gadara_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5577,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.691",
"value": "?search="
},
{
"operation": "search",
"time": "16.86",
"value": "Oenomaus"
},
{
"operation": "Now on",
"time": "18.769",
"value": "Oenomaus"
},
{
"operation": "search",
"time": "76.471",
"value": "Oenomaus of Gadara"
},
{
"operation": "Now on",
"time": "77.672",
"value": "Oenomaus of Gadara"
},
{
"operation": "Highlighting",
"time": "94.464",
"value": "Oenomaus of Gadara_sentence_2"
},
{
"operation": "finish",
"time": "108.884",
"value": "finish"
}
]
}
|
exfever_7026
|
Robert Francis Kennedy was an American lawyer and politician who was born into a wealthy political family in a town in Norfolk County, Massachusetts, in the United Kingdom.
|
{
"evidence": [
"Robert F. Kennedy",
"Brookline, Massachusetts"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRobert Francis Kennedy was an American lawyer and politician who was born into a wealthy political family in a town in Norfolk County, Massachusetts, in the United Kingdom.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Robert F. Kennedy",
"Brookline, Massachusetts"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5578,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_257
|
Pearl Jam is an American rock band formed in Seattle had a member who later joined in 1998 and used to be the lead guitarist for Seattle-based rock band Soundgarden.
|
{
"evidence": [
"Pearl Jam",
"Matt Cameron"
],
"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:\nPearl Jam is an American rock band formed in Seattle had a member who later joined in 1998 and used to be the lead guitarist for Seattle-based rock band Soundgarden.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Pearl Jam",
"Matt Cameron"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5579,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
77959
|
Władysław Oporowski(1395-11 March 1453) was a Polish medieval political and religious leader who served as Deputy Chancellor of Poland, Bishop of Kujawy and archbishop of Gniezno and primate of Poland from 1449 to 1453.
|
{
"evidence": [
"Władysław Oporowski_sentence_0",
"Władysław Oporowski_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:\nWładysław Oporowski(1395-11 March 1453) was a Polish medieval political and religious leader who served as Deputy Chancellor of Poland, Bishop of Kujawy and archbishop of Gniezno and primate of Poland from 1449 to 1453.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Władysław Oporowski_sentence_0",
"Władysław Oporowski_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5580,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.768",
"value": "?search="
},
{
"operation": "search",
"time": "12.981",
"value": "Władysław Oporowski"
},
{
"operation": "Now on",
"time": "13.639",
"value": "Władysław Oporowski"
},
{
"operation": "Highlighting",
"time": "15.94",
"value": "Władysław Oporowski_sentence_0"
},
{
"operation": "Highlighting",
"time": "18.196",
"value": "Władysław Oporowski_sentence_1"
},
{
"operation": "finish",
"time": "44.923",
"value": "finish"
}
]
}
|
exfever_500
|
The Big Chill is an American fantasy-drama film starring an American-Canadian actress and writer who was nominated for the Academy Award for Best Supporting Actress.
|
{
"evidence": [
"The Big Chill (film)",
"Meg Tilly"
],
"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 Big Chill is an American fantasy-drama film starring an American-Canadian actress and writer who was nominated for the Academy Award for Best Supporting Actress.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Big Chill (film)",
"Meg Tilly"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5581,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4403
|
Silicon is a chemical element with the symbol Al and atomic number 13 and has a great affinity toward the chemical element with the symbol O and atomic number 8.
|
{
"evidence": [
"Aluminium",
"Oxygen"
],
"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:\nSilicon is a chemical element with the symbol Al and atomic number 13 and has a great affinity toward the chemical element with the symbol O and atomic number 8.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Aluminium",
"Oxygen"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5582,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
34899
|
Animals Australia conducted an investigation of the treatment of exported Australian cattle in Indonesia worth $328 million, animals had their throats cut while fully conscious and sparked mass protests across Australia.
|
{
"evidence": [
"Australian live export industry_sentence_11",
"Australian live export industry_sentence_12",
"Australian live export industry_sentence_13"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAnimals Australia conducted an investigation of the treatment of exported Australian cattle in Indonesia worth $328 million, animals had their throats cut while fully conscious and sparked mass protests across Australia.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Australian live export industry_sentence_11",
"Australian live export industry_sentence_12",
"Australian live export industry_sentence_13"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5583,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.493",
"value": "?search="
},
{
"operation": "search",
"time": "32.86",
"value": "Animals Australia"
},
{
"operation": "Now on",
"time": "34.709",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "37.087",
"value": "Animals Australia"
},
{
"operation": "Now on",
"time": "38.107",
"value": "Animals Australia"
},
{
"operation": "Page search",
"time": "54.516",
"value": "protest"
},
{
"operation": "Page search",
"time": "54.774",
"value": "protest"
},
{
"operation": "Page search",
"time": "54.995",
"value": "protest"
},
{
"operation": "Page search",
"time": "55.174",
"value": "protest"
},
{
"operation": "search",
"time": "74.227",
"value": "Animals Australia conducted an investigation of the treatment of exported Australian cattle"
},
{
"operation": "Now on",
"time": "77.115",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "84.224",
"value": "Australian live export industry"
},
{
"operation": "Now on",
"time": "85.159",
"value": "Australian live export industry"
},
{
"operation": "Page search",
"time": "88.458",
"value": "Animals Australia conducted an investigation of the treatment of exported Australian cattle"
},
{
"operation": "Page search",
"time": "88.679",
"value": "Animals Australia conducted an investigation of the treatment of exported Australian cattle"
},
{
"operation": "Page search",
"time": "88.833",
"value": "Animals Australia conducted an investigation of the treatment of exported Australian cattle"
},
{
"operation": "Page search",
"time": "89.008",
"value": "Animals Australia conducted an investigation of the treatment of exported Australian cattle"
},
{
"operation": "Page search",
"time": "92.577",
"value": "exported Australian cattle"
},
{
"operation": "Page search",
"time": "92.922",
"value": "exported Australian cattle"
},
{
"operation": "Page search",
"time": "93.176",
"value": "exported Australian cattle"
},
{
"operation": "Page search",
"time": "93.272",
"value": "exported Australian cattle"
},
{
"operation": "Highlighting",
"time": "99.283",
"value": "Australian live export industry_sentence_11"
},
{
"operation": "Highlighting",
"time": "121.935",
"value": "Australian live export industry_sentence_12"
},
{
"operation": "Highlighting",
"time": "130.248",
"value": "Australian live export industry_sentence_13"
},
{
"operation": "Page search",
"time": "144.166",
"value": "328"
},
{
"operation": "Page search",
"time": "144.386",
"value": "328"
},
{
"operation": "Page search",
"time": "144.593",
"value": "328"
},
{
"operation": "Highlighting deleted",
"time": "156.957",
"value": "Australian live export industry_sentence_11"
},
{
"operation": "Highlighting",
"time": "157.771",
"value": "Australian live export industry_sentence_11"
},
{
"operation": "Highlighting deleted",
"time": "159.806",
"value": "Australian live export industry_sentence_13"
},
{
"operation": "Highlighting deleted",
"time": "160.198",
"value": "Australian live export industry_sentence_12"
},
{
"operation": "Highlighting deleted",
"time": "160.556",
"value": "Australian live export industry_sentence_11"
},
{
"operation": "Highlighting",
"time": "161.274",
"value": "Australian live export industry_sentence_11"
},
{
"operation": "Highlighting",
"time": "161.816",
"value": "Australian live export industry_sentence_12"
},
{
"operation": "Highlighting",
"time": "162.353",
"value": "Australian live export industry_sentence_13"
},
{
"operation": "finish",
"time": "220.722",
"value": "finish"
}
]
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.