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 |
|---|---|---|---|---|---|---|---|---|
85880
|
Palazzo Gherardi is a 15th-century building in the Santa croce quarter of Florence, Tuscany,with an austere design and built with a series of square stone windows with grills and ten large, arched windows on the first and second floor.
|
{
"evidence": [
"Palazzo Gherardi_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:\nPalazzo Gherardi is a 15th-century building in the Santa croce quarter of Florence, Tuscany,with an austere design and built with a series of square stone windows with grills and ten large, arched windows on the first and second floor.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Palazzo Gherardi_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2784,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.998",
"value": "?search="
},
{
"operation": "search",
"time": "17.098",
"value": "Palazzo Gherardi"
},
{
"operation": "Now on",
"time": "19.773",
"value": "Palazzo Gherardi"
},
{
"operation": "Highlighting",
"time": "22.171",
"value": "Palazzo Gherardi_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "52.615",
"value": "Palazzo Gherardi_sentence_0"
},
{
"operation": "search",
"time": "62.105",
"value": "Palazzo Gherardi"
},
{
"operation": "Now on",
"time": "62.486",
"value": "Palazzo Gherardi"
},
{
"operation": "hyperlink",
"time": "69.495",
"value": "Italy"
},
{
"operation": "Now on",
"time": "72.386",
"value": "Palazzo Gherardi"
},
{
"operation": "Highlighting",
"time": "73.695",
"value": "Palazzo Gherardi_sentence_0"
},
{
"operation": "hyperlink",
"time": "75.413",
"value": "Florence"
},
{
"operation": "Now on",
"time": "76.837",
"value": "Florence"
},
{
"operation": "finish",
"time": "86.401",
"value": "finish"
}
]
}
|
23780
|
Chersotis fimbriola is a butterfly from the Noctuidae family, order Lepidoptera of the kingdom Animalia.
|
{
"evidence": [
"Chersotis fimbriola_sentence_0",
"Moth_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:\nChersotis fimbriola is a butterfly from the Noctuidae family, order Lepidoptera of the kingdom Animalia.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Chersotis fimbriola_sentence_0",
"Moth_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2785,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.543",
"value": "?search="
},
{
"operation": "search",
"time": "6.738",
"value": "Chersotis fimbriola"
},
{
"operation": "Now on",
"time": "7.396",
"value": "Chersotis fimbriola"
},
{
"operation": "Highlighting",
"time": "14.661",
"value": "Chersotis fimbriola_sentence_0"
},
{
"operation": "hyperlink",
"time": "17.329",
"value": "moth"
},
{
"operation": "Now on",
"time": "17.86",
"value": "Moth"
},
{
"operation": "Highlighting",
"time": "33.068",
"value": "Moth_sentence_1"
},
{
"operation": "finish",
"time": "44.061",
"value": "finish"
}
]
}
|
36300
|
In race two of nine of the 1962 World Championship of Drivers, the Monaco Grand Prix, Bruce McLaren qualified at number one with BRM while Jim Clark qualified at number two with Ferrari.
|
{
"evidence": [
"Bruce McLaren_sentence_22"
],
"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 race two of nine of the 1962 World Championship of Drivers, the Monaco Grand Prix, Bruce McLaren qualified at number one with BRM while Jim Clark qualified at number two with Ferrari.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Bruce McLaren_sentence_22"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2786,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.762",
"value": "?search="
},
{
"operation": "search",
"time": "27.016",
"value": "1962 World Championship of Drivers"
},
{
"operation": "Now on",
"time": "28.935",
"value": "1962 Formula One season"
},
{
"operation": "Highlighting",
"time": "49.249",
"value": "1962 Formula One season_sentence_1"
},
{
"operation": "Highlighting",
"time": "49.253",
"value": "1962 Formula One season_title"
},
{
"operation": "Highlighting",
"time": "103.648",
"value": "1962 Formula One season_cell_0_2_6"
},
{
"operation": "Highlighting",
"time": "103.653",
"value": "1962 Formula One season_section_1"
},
{
"operation": "Highlighting",
"time": "103.667",
"value": "1962 Formula One season_header_cell_0_2_0"
},
{
"operation": "Highlighting",
"time": "103.668",
"value": "1962 Formula One season_header_cell_0_0_6"
},
{
"operation": "search",
"time": "278.705",
"value": "Bruce McLaren"
},
{
"operation": "Now on",
"time": "280.399",
"value": "Bruce McLaren"
},
{
"operation": "Highlighting",
"time": "339.438",
"value": "Bruce McLaren_sentence_22"
},
{
"operation": "search",
"time": "375.933",
"value": "Jim Clark"
},
{
"operation": "Now on",
"time": "377.817",
"value": "Jim Clark"
},
{
"operation": "finish",
"time": "566.391",
"value": "finish"
}
]
}
|
exfever_3685
|
"Two Hearts" is a song first recorded by electronic duo Kish Mauve, which was later re-recorded by Kylie Minogue, and was released on November 9, 2007, by a German–British record label, founded in Germany in 1896.
|
{
"evidence": [
"2 Hearts (Kylie Minogue song)",
"Parlophone"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n\"Two Hearts\" is a song first recorded by electronic duo Kish Mauve, which was later re-recorded by Kylie Minogue, and was released on November 9, 2007, by a German–British record label, founded in Germany in 1896.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"2 Hearts (Kylie Minogue song)",
"Parlophone"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2787,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7545
|
Sharon Marie Tate Polanski was an American actress who appeared in the 1966 British horror film that revolves around the solution of a problem or a crime.
|
{
"evidence": [
"Sharon Tate",
"Eye of the Devil",
"Mystery 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:\nSharon Marie Tate Polanski was an American actress who appeared in the 1966 British horror film that revolves around the solution of a problem or a crime.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sharon Tate",
"Eye of the Devil",
"Mystery film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2788,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3835
|
Jared Padalecki is an African actor who is best known for playing the role of a fictional character and one of the protagonists in the American drama television series Supernatural along with his older brother, Dean.
|
{
"evidence": [
"Jared Padalecki",
"Sam Winchester"
],
"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:\nJared Padalecki is an African actor who is best known for playing the role of a fictional character and one of the protagonists in the American drama television series Supernatural along with his older brother, Dean.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jared Padalecki",
"Sam Winchester"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2789,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1877
|
Paul Newman is the first American actor who had numerous awards.
|
{
"evidence": [
"Paul Newman",
"Primetime Emmy Award",
"Daytime Emmy Award"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPaul Newman is the first American actor who had numerous awards.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Paul Newman",
"Primetime Emmy Award",
"Daytime Emmy Award"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2790,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1685
|
Free Guy is an American comedy film directed by Leonard Nimoy starring an American actor who played the lead character Sam Malone on the NBC sitcom Cheers.
|
{
"evidence": [
"Three Men and a Baby",
"Ted Danson"
],
"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:\nFree Guy is an American comedy film directed by Leonard Nimoy starring an American actor who played the lead character Sam Malone on the NBC sitcom Cheers.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Three Men and a Baby",
"Ted Danson"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2791,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7281
|
Ontario is home to the capital city of Canada.
|
{
"evidence": [
"Ontario",
"Ottawa"
],
"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:\nOntario is home to the capital city of Canada.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ontario",
"Ottawa"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2792,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3766
|
50 First Dates is a 2003 American romantic comedy film and was shot on location in the third-largest of the Hawaiian Islands and is within the capital and largest city of the U.S. state of Hawaii.
|
{
"evidence": [
"50 First Dates",
"Oahu",
"Honolulu"
],
"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:\n50 First Dates is a 2003 American romantic comedy film and was shot on location in the third-largest of the Hawaiian Islands and is within the capital and largest city of the U.S. state of Hawaii.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"50 First Dates",
"Oahu",
"Honolulu"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2793,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
7012
|
Tung Sing was promoted to second division in the 2014-15 Hong Kong football season and Lung Moon was eliminated from the league.
|
{
"evidence": [
"Tung Sing FC_sentence_22",
"Tung Sing FC_sentence_24"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nTung Sing was promoted to second division in the 2014-15 Hong Kong football season and Lung Moon was eliminated from the league.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Tung Sing FC_sentence_22",
"Tung Sing FC_sentence_24"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2794,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.726",
"value": "?search="
},
{
"operation": "search",
"time": "35.761",
"value": "Tung Sing FC"
},
{
"operation": "Now on",
"time": "36.571",
"value": "Tung Sing FC"
},
{
"operation": "Highlighting",
"time": "49.1",
"value": "Tung Sing FC_sentence_22"
},
{
"operation": "Highlighting",
"time": "57.931",
"value": "Tung Sing FC_sentence_24"
},
{
"operation": "Page search",
"time": "85.523",
"value": "Lung Moon"
},
{
"operation": "search",
"time": "96.405",
"value": "Contains...Lung Moon"
},
{
"operation": "Now on",
"time": "97.691",
"value": "undefined"
},
{
"operation": "finish",
"time": "109.834",
"value": "finish"
}
]
}
|
36443
|
James Gallander (born February 9, 1970) was part of some films includes Due South (1997), Reluctant Angel (1999), Mutant X (2002) and others.
|
{
"evidence": [
"James Gallanders_sentence_9"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJames Gallander (born February 9, 1970) was part of some films includes Due South (1997), Reluctant Angel (1999), Mutant X (2002) and others.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"James Gallanders_sentence_9"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2795,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.007",
"value": "?search="
},
{
"operation": "search",
"time": "72.253",
"value": "James Gallander"
},
{
"operation": "Now on",
"time": "72.788",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "105.206",
"value": "James Gallanders"
},
{
"operation": "Now on",
"time": "105.937",
"value": "James Gallanders"
},
{
"operation": "Highlighting",
"time": "107.766",
"value": "James Gallanders_sentence_0"
},
{
"operation": "Page search",
"time": "136.433",
"value": "due south"
},
{
"operation": "Highlighting",
"time": "143.725",
"value": "James Gallanders_sentence_9"
},
{
"operation": "Highlighting deleted",
"time": "169.036",
"value": "James Gallanders_sentence_0"
},
{
"operation": "finish",
"time": "182.087",
"value": "finish"
}
]
}
|
exfever_7560
|
Sylvester Stallone is an Indian actor, screenwriter, director, and producer who uses the he 72 stone steps leading up to the entrance of the Philadelphia Museum of Art, in Philadelphia, Pennsylvania, from the film Rocky.
|
{
"evidence": [
"Sylvester Stallone",
"Rocky Steps"
],
"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:\nSylvester Stallone is an Indian actor, screenwriter, director, and producer who uses the he 72 stone steps leading up to the entrance of the Philadelphia Museum of Art, in Philadelphia, Pennsylvania, from the film Rocky.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sylvester Stallone",
"Rocky Steps"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2796,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
83827
|
Carodnia, an extinct genus of South American ungulate, is the largest mammal known from the Eocene of South America which would have been a tapir-sized animal that bore strong resemblances to dinoceratans.
|
{
"evidence": [
"Carodnia_sentence_0",
"Carodnia_sentence_2",
"Carodnia_sentence_5",
"Carodnia_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:\nCarodnia, an extinct genus of South American ungulate, is the largest mammal known from the Eocene of South America which would have been a tapir-sized animal that bore strong resemblances to dinoceratans.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Carodnia_sentence_0",
"Carodnia_sentence_2",
"Carodnia_sentence_5",
"Carodnia_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2797,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.249",
"value": "?search="
},
{
"operation": "search",
"time": "94.483",
"value": "Carodnia"
},
{
"operation": "Now on",
"time": "95.411",
"value": "Carodnia"
},
{
"operation": "Highlighting",
"time": "99.755",
"value": "Carodnia_sentence_0"
},
{
"operation": "Highlighting",
"time": "108.043",
"value": "Carodnia_sentence_2"
},
{
"operation": "Highlighting",
"time": "120.616",
"value": "Carodnia_sentence_5"
},
{
"operation": "Highlighting",
"time": "121.361",
"value": "Carodnia_sentence_4"
},
{
"operation": "finish",
"time": "127.943",
"value": "finish"
}
]
}
|
11158
|
Christianity, in its early development, accommodated many traditional practices which at different times, accepted omens and practiced forms of ritual sacrifice; and concepts of "Divine Grace" or "Blessing", as they are described by believers, closely resemble what is referred to as "luck" (the phenomenon and belief that defines the experience of notably positive, negative, or improbable events) by others.
|
{
"evidence": [
"Luck_sentence_2",
"Divine grace_sentence_18"
],
"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:\nChristianity, in its early development, accommodated many traditional practices which at different times, accepted omens and practiced forms of ritual sacrifice; and concepts of \"Divine Grace\" or \"Blessing\", as they are described by believers, closely resemble what is referred to as \"luck\" (the phenomenon and belief that defines the experience of notably positive, negative, or improbable events) by others.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Luck_sentence_2",
"Divine grace_sentence_18"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2798,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.063",
"value": "undefined"
},
{
"operation": "search",
"time": "112.979",
"value": "Contains...\"Divine Grace\""
},
{
"operation": "Now on",
"time": "115.285",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "128.873",
"value": "Merit (Christianity)"
},
{
"operation": "Now on",
"time": "130.556",
"value": "Merit (Christianity)"
},
{
"operation": "search",
"time": "259.772",
"value": "Luck"
},
{
"operation": "Now on",
"time": "261.839",
"value": "Luck"
},
{
"operation": "Highlighting",
"time": "268.02",
"value": "Luck_sentence_2"
},
{
"operation": "hyperlink",
"time": "293.979",
"value": "4 In religion and mythology Luck_section_12"
},
{
"operation": "search",
"time": "340.808",
"value": "sacrifice"
},
{
"operation": "Now on",
"time": "343.369",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "361.885",
"value": "Salvation in Christianity"
},
{
"operation": "Now on",
"time": "363.091",
"value": "Salvation in Christianity"
},
{
"operation": "back-button-clicked",
"time": "415.849",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "416.72",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "422.749",
"value": "Isaac"
},
{
"operation": "Now on",
"time": "423.702",
"value": "Isaac"
},
{
"operation": "hyperlink",
"time": "457.557",
"value": "mercy of God"
},
{
"operation": "Now on",
"time": "458.667",
"value": "Atonement in Judaism"
},
{
"operation": "Highlighting",
"time": "466.215",
"value": "Atonement in Judaism_sentence_3"
},
{
"operation": "Highlighting",
"time": "466.216",
"value": "Atonement in Judaism_title"
},
{
"operation": "back-button-clicked",
"time": "488.751",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "488.835",
"value": "Isaac"
},
{
"operation": "hyperlink",
"time": "514.414",
"value": "Jewish"
},
{
"operation": "Now on",
"time": "516.64",
"value": "Jewish"
},
{
"operation": "hyperlink",
"time": "540.285",
"value": "4.1 Babylon and Rome Jews_section_4"
},
{
"operation": "search",
"time": "581.155",
"value": "divine grace"
},
{
"operation": "Now on",
"time": "582.29",
"value": "Divine grace"
},
{
"operation": "hyperlink",
"time": "585.813",
"value": "2 Christianity Divine grace_section_1"
},
{
"operation": "Highlighting",
"time": "602.352",
"value": "Divine grace_sentence_18"
},
{
"operation": "hyperlink",
"time": "636.716",
"value": "Grace in Christianity"
},
{
"operation": "Now on",
"time": "637.782",
"value": "Grace in Christianity"
},
{
"operation": "back-button-clicked",
"time": "641.596",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "641.692",
"value": "Divine grace#Christianity Divine grace section 1"
},
{
"operation": "search",
"time": "651.304",
"value": "blessing"
},
{
"operation": "Now on",
"time": "652.948",
"value": "Blessing"
},
{
"operation": "search",
"time": "705.955",
"value": "Sacrifice in Christianity"
},
{
"operation": "Now on",
"time": "707.823",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "726.229",
"value": "Early Christianity Christianity_section_24"
},
{
"operation": "Now on",
"time": "727.663",
"value": "Christianity#Early Christianity Christianity section 24"
},
{
"operation": "hyperlink",
"time": "769.16",
"value": "Proto-orthodox Christianity"
},
{
"operation": "Now on",
"time": "770.315",
"value": "Proto-orthodox Christianity"
},
{
"operation": "back-button-clicked",
"time": "790.816",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "791.104",
"value": "Christianity#Early Christianity Christianity section 24"
},
{
"operation": "hyperlink",
"time": "840.408",
"value": "Massacre of Verden"
},
{
"operation": "Now on",
"time": "841.932",
"value": "Massacre of Verden"
},
{
"operation": "back-button-clicked",
"time": "875.523",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "875.711",
"value": "Christianity#Early Christianity Christianity section 24"
},
{
"operation": "finish",
"time": "908.319",
"value": "finish"
}
]
}
|
exfever_5330
|
Gone Girl is a 2014 Canadian psychological thriller film directed by an American film director whose films are mostly psychological thrillers and biographical dramas.
|
{
"evidence": [
"Gone Girl (film)",
"David Fincher"
],
"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:\nGone Girl is a 2014 Canadian psychological thriller film directed by an American film director whose films are mostly psychological thrillers and biographical dramas.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Gone Girl (film)",
"David Fincher"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2799,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
47057
|
Jacques Rogge was the chairman of the International Olympic Committee from 1973 to 2002 and also served as president of the Court of Arbitration for Sport, the highest Olympic tribunal.
|
{
"evidence": [
"President of the International Olympic Committee_sentence_79",
"Jacques Rogge_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:\nJacques Rogge was the chairman of the International Olympic Committee from 1973 to 2002 and also served as president of the Court of Arbitration for Sport, the highest Olympic tribunal.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"President of the International Olympic Committee_sentence_79",
"Jacques Rogge_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2800,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.745",
"value": "?search="
},
{
"operation": "search",
"time": "8.546",
"value": "Jacques Rogge"
},
{
"operation": "Now on",
"time": "9.422",
"value": "Jacques Rogge"
},
{
"operation": "Page search",
"time": "93.206",
"value": "chairman"
},
{
"operation": "search",
"time": "127.92",
"value": "chairman of the international olympic committee"
},
{
"operation": "Now on",
"time": "130.387",
"value": "undefined"
},
{
"operation": "search",
"time": "162.838",
"value": "chairman of the international olympic committee jacques rogge"
},
{
"operation": "Now on",
"time": "164.125",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "176.889",
"value": "chairman of the international olympic committee jacques rouge"
},
{
"operation": "Now on",
"time": "178.019",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "182.751",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "183.47",
"value": "undefined"
},
{
"operation": "search",
"time": "189.116",
"value": "jacques roggee"
},
{
"operation": "Now on",
"time": "190.01",
"value": "undefined"
},
{
"operation": "search",
"time": "217.516",
"value": "Jacques Rogge"
},
{
"operation": "Now on",
"time": "218.442",
"value": "Jacques Rogge"
},
{
"operation": "Page search",
"time": "288.795",
"value": "1973"
},
{
"operation": "search",
"time": "328.381",
"value": "list of chairmen of the olympic committee"
},
{
"operation": "Now on",
"time": "330.086",
"value": "undefined"
},
{
"operation": "search",
"time": "355.904",
"value": "International Olympic Committee Presidents"
},
{
"operation": "Now on",
"time": "356.851",
"value": "President of the International Olympic Committee"
},
{
"operation": "Page search",
"time": "366.604",
"value": "jacques rogge"
},
{
"operation": "Highlighting",
"time": "388.093",
"value": "President of the International Olympic Committee_sentence_79"
},
{
"operation": "hyperlink",
"time": "404.845",
"value": "Jacques Rogge"
},
{
"operation": "Now on",
"time": "405.277",
"value": "Jacques Rogge"
},
{
"operation": "Highlighting",
"time": "414.252",
"value": "Jacques Rogge_sentence_0"
},
{
"operation": "finish",
"time": "427.194",
"value": "finish"
}
]
}
|
exfever_1204
|
Glee season 6 was a high-rating Fox musical comedy-drama series.
|
{
"evidence": [
"Glee (season 6)",
"Sue Sylvester"
],
"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:\nGlee season 6 was a high-rating Fox musical comedy-drama series.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Glee (season 6)",
"Sue Sylvester"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2801,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5177
|
Chris Evans is an American actor who appeared in an American teen parody film directed by Joel Gallen and features multiple principal actors and performers who are typically assigned roughly equal amounts of screen time.
|
{
"evidence": [
"Chris Evans (actor)",
"Not Another Teen Movie",
"Ensemble cast"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nChris Evans is an American actor who appeared in an American teen parody film directed by Joel Gallen and features multiple principal actors and performers who are typically assigned roughly equal amounts of screen time.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Chris Evans (actor)",
"Not Another Teen Movie",
"Ensemble cast"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2802,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
17744
|
In the 18th century, Bernardo Bellotto arrived in Kraków from Italy in April 1863. where he later became an independent hero in both Italy and Poland.
|
{
"evidence": [
"Garibaldi Legion (Poland)_sentence_6",
"18th century_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:\nIn the 18th century, Bernardo Bellotto arrived in Kraków from Italy in April 1863. where he later became an independent hero in both Italy and Poland.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Garibaldi Legion (Poland)_sentence_6",
"18th century_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2803,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.158",
"value": "?search="
},
{
"operation": "search",
"time": "10.584",
"value": "Bernardo Bellotto"
},
{
"operation": "Now on",
"time": "12.003",
"value": "Bernardo Bellotto"
},
{
"operation": "Page search",
"time": "34.997",
"value": "arrived"
},
{
"operation": "Page search",
"time": "47.059",
"value": "krakow"
},
{
"operation": "Page search",
"time": "56.455",
"value": "1863"
},
{
"operation": "Page search",
"time": "71.677",
"value": "independent hero"
},
{
"operation": "search",
"time": "88.301",
"value": "\"arrived in Kraków from Italy in April 1863\""
},
{
"operation": "Now on",
"time": "89.114",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "94.845",
"value": "Garibaldi Legion (Poland)"
},
{
"operation": "Now on",
"time": "95.891",
"value": "Garibaldi Legion (Poland)"
},
{
"operation": "Highlighting",
"time": "104.102",
"value": "Garibaldi Legion (Poland)_sentence_6"
},
{
"operation": "search",
"time": "133.554",
"value": "18th century"
},
{
"operation": "Now on",
"time": "137.475",
"value": "18th century"
},
{
"operation": "Highlighting",
"time": "139.449",
"value": "18th century_sentence_1"
},
{
"operation": "back-button-clicked",
"time": "142.805",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "142.925",
"value": "Garibaldi Legion (Poland)"
},
{
"operation": "search",
"time": "151.71",
"value": "\"he later became an independent hero in both Italy and Poland\""
},
{
"operation": "Now on",
"time": "152.855",
"value": "undefined"
},
{
"operation": "finish",
"time": "170.179",
"value": "finish"
}
]
}
|
exfever_3360
|
Con Man is an American comedy web series that is co-produced by a Canadian born producer and science fiction author who developed a school presentation program in which he discusses The Softwire books, astronomy, and other science fiction and a systematic enterprise that builds and organizes knowledge in the form of testable explanations.
|
{
"evidence": [
"Con Man (web series)",
"PJ Haarsma",
"Science"
],
"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:\nCon Man is an American comedy web series that is co-produced by a Canadian born producer and science fiction author who developed a school presentation program in which he discusses The Softwire books, astronomy, and other science fiction and a systematic enterprise that builds and organizes knowledge in the form of testable explanations.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Con Man (web series)",
"PJ Haarsma",
"Science"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2804,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
87924
|
Wreaths were given out during the 2004 Summer Olympics in Athens in honor of the ancient Olympic Games
|
{
"evidence": [
"Olive wreath_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:\nWreaths were given out during the 2004 Summer Olympics in Athens in honor of the ancient Olympic Games\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Olive wreath_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2805,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.874",
"value": "?search="
},
{
"operation": "search",
"time": "11.907",
"value": "2004 Summer Olympics"
},
{
"operation": "Now on",
"time": "12.949",
"value": "2004 Summer Olympics"
},
{
"operation": "Page search",
"time": "27.113",
"value": "wreath"
},
{
"operation": "Page search",
"time": "46.544",
"value": "wreaths"
},
{
"operation": "Page search",
"time": "60.552",
"value": "ancient Olympic Games"
},
{
"operation": "search",
"time": "169.97",
"value": "olive wreath"
},
{
"operation": "Now on",
"time": "170.867",
"value": "Olive wreath"
},
{
"operation": "Highlighting",
"time": "173.336",
"value": "Olive wreath_sentence_8"
},
{
"operation": "finish",
"time": "198.839",
"value": "finish"
}
]
}
|
exfever_2965
|
Margot Frank was the youngest daughter of Otto Frank and Edith Frank, who died in?a prisoner of war camp.
|
{
"evidence": [
"Margot Frank",
"Bergen-Belsen concentration camp"
],
"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:\nMargot Frank was the youngest daughter of Otto Frank and Edith Frank, who died in?a prisoner of war camp.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Margot Frank",
"Bergen-Belsen concentration camp"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2806,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
33162
|
Single albums includes the "The Lost Get Found" (2009) by Britt Harvey, "Walk on the Water" (2009), "Hanging On" (2010) and "Headphones" (2011).
|
{
"evidence": [
"The Lost Get Found_sentence_0",
"Walk on the Water (song)_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:\nSingle albums includes the \"The Lost Get Found\" (2009) by Britt Harvey, \"Walk on the Water\" (2009), \"Hanging On\" (2010) and \"Headphones\" (2011).\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"The Lost Get Found_sentence_0",
"Walk on the Water (song)_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2807,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.982",
"value": "?search="
},
{
"operation": "search",
"time": "19.083",
"value": "The Lost Get Found"
},
{
"operation": "Now on",
"time": "19.954",
"value": "The Lost Get Found"
},
{
"operation": "Highlighting",
"time": "23.607",
"value": "The Lost Get Found_sentence_0"
},
{
"operation": "search",
"time": "44.278",
"value": "walk on the water"
},
{
"operation": "Now on",
"time": "45.944",
"value": "Walk On the Water"
},
{
"operation": "search",
"time": "56.504",
"value": "Walk on the Water (album)"
},
{
"operation": "Now on",
"time": "57.415",
"value": "Walk on the Water (album)"
},
{
"operation": "search",
"time": "68.951",
"value": "Walk on the Water (song)"
},
{
"operation": "Now on",
"time": "70.315",
"value": "Walk on the Water (song)"
},
{
"operation": "Highlighting",
"time": "77.828",
"value": "Walk on the Water (song)_sentence_1"
},
{
"operation": "search",
"time": "96.747",
"value": "Hanging On"
},
{
"operation": "Now on",
"time": "97.611",
"value": "Hanging On"
},
{
"operation": "search",
"time": "110.448",
"value": "Headphones (song)"
},
{
"operation": "Now on",
"time": "111.427",
"value": "Headphones (song)"
},
{
"operation": "search",
"time": "125.175",
"value": "Headphones (Britt Nicole song)"
},
{
"operation": "Now on",
"time": "126.318",
"value": "Britt Nicole"
},
{
"operation": "hyperlink",
"time": "139.088",
"value": "4 Discography Britt Nicole_section_5"
},
{
"operation": "hyperlink",
"time": "140.986",
"value": "Britt Nicole discography"
},
{
"operation": "Now on",
"time": "141.708",
"value": "Britt Nicole discography"
},
{
"operation": "finish",
"time": "200.971",
"value": "finish"
}
]
}
|
exfever_2614
|
Ryan Gosling is a well-known actor worldwide.
|
{
"evidence": [
"Ryan Gosling",
"Are You Afraid of the Dark?"
],
"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:\nRyan Gosling is a well-known actor worldwide.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ryan Gosling",
"Are You Afraid of the Dark?"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2808,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5414
|
The Menace II Society is a 1993 American film directed by the Hughes Brothers, whose genre originated in the United States and features aspects of urban African American culture.
|
{
"evidence": [
"Menace II Society",
"Hood 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 Menace II Society is a 1993 American film directed by the Hughes Brothers, whose genre originated in the United States and features aspects of urban African American culture.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Menace II Society",
"Hood film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2809,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
7521
|
Thomas Hardcastle Sykes, married to Mary Platt, daughter of John Platt, an English manufacturer of textile machinery and Liberal politician, was a captain in the 20th Cheshire Royal Volunteers.
|
{
"evidence": [
"Thomas Hardcastle Sykes_sentence_9",
"John Platt (MP)_sentence_1",
"Thomas Hardcastle Sykes_sentence_6"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThomas Hardcastle Sykes, married to Mary Platt, daughter of John Platt, an English manufacturer of textile machinery and Liberal politician, was a captain in the 20th Cheshire Royal Volunteers.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Thomas Hardcastle Sykes_sentence_9",
"John Platt (MP)_sentence_1",
"Thomas Hardcastle Sykes_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2810,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.791",
"value": "?search="
},
{
"operation": "search",
"time": "9.647",
"value": "Thomas Hardcastle Sykes"
},
{
"operation": "Now on",
"time": "10.911",
"value": "Thomas Hardcastle Sykes"
},
{
"operation": "Highlighting",
"time": "19.986",
"value": "Thomas Hardcastle Sykes_sentence_9"
},
{
"operation": "hyperlink",
"time": "27.886",
"value": "John Platt"
},
{
"operation": "Now on",
"time": "28.722",
"value": "John Platt (MP)"
},
{
"operation": "Highlighting",
"time": "32.127",
"value": "John Platt (MP)_sentence_1"
},
{
"operation": "back-button-clicked",
"time": "33.689",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "33.79",
"value": "Thomas Hardcastle Sykes"
},
{
"operation": "Highlighting",
"time": "37.919",
"value": "Thomas Hardcastle Sykes_sentence_6"
},
{
"operation": "finish",
"time": "44.745",
"value": "finish"
}
]
}
|
17350
|
Lessard completed his junior year of high school, passed the GED examinations in the summer of 1991, and was accepted into the Jazz Studies program at Virginia Commonwealth University.
|
{
"evidence": [
"Stefan Lessard_sentence_10",
"Stefan Lessard_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:\nLessard completed his junior year of high school, passed the GED examinations in the summer of 1991, and was accepted into the Jazz Studies program at Virginia Commonwealth University.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Stefan Lessard_sentence_10",
"Stefan Lessard_sentence_9"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2811,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.095",
"value": "?search="
},
{
"operation": "search",
"time": "26.515",
"value": "Lessard"
},
{
"operation": "Now on",
"time": "27.787",
"value": "Lessard"
},
{
"operation": "hyperlink",
"time": "42.603",
"value": "Stefan Lessard"
},
{
"operation": "Now on",
"time": "43.338",
"value": "Stefan Lessard"
},
{
"operation": "Page search",
"time": "56.52",
"value": "GED"
},
{
"operation": "Highlighting",
"time": "60.041",
"value": "Stefan Lessard_sentence_10"
},
{
"operation": "Highlighting",
"time": "63.84",
"value": "Stefan Lessard_sentence_9"
},
{
"operation": "finish",
"time": "87.901",
"value": "finish"
}
]
}
|
33557
|
Maria Evelia Marmolejo is a performance artist who often used self-harm as a means of self-expression to shock her audience into paying attention to the violence happening around them.
|
{
"evidence": [
"María Evelia Marmolejo_sentence_0",
"María Evelia Marmolejo_sentence_29",
"María Evelia Marmolejo_sentence_28"
],
"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:\nMaria Evelia Marmolejo is a performance artist who often used self-harm as a means of self-expression to shock her audience into paying attention to the violence happening around them.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"María Evelia Marmolejo_sentence_0",
"María Evelia Marmolejo_sentence_29",
"María Evelia Marmolejo_sentence_28"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2812,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.061",
"value": "?search="
},
{
"operation": "search",
"time": "7.571",
"value": "Maria Evelia Marmolejo"
},
{
"operation": "Now on",
"time": "9.372",
"value": "María Evelia Marmolejo"
},
{
"operation": "Highlighting",
"time": "12.077",
"value": "María Evelia Marmolejo_sentence_0"
},
{
"operation": "Page search",
"time": "37.403",
"value": "expression"
},
{
"operation": "Page search",
"time": "52.099",
"value": "self"
},
{
"operation": "Highlighting",
"time": "71.189",
"value": "María Evelia Marmolejo_sentence_29"
},
{
"operation": "Page search",
"time": "102.517",
"value": "around them"
},
{
"operation": "Highlighting",
"time": "109.131",
"value": "María Evelia Marmolejo_sentence_28"
},
{
"operation": "finish",
"time": "117.531",
"value": "finish"
}
]
}
|
exfever_3905
|
The border between Canada and the United States also has one land dispute over Machias Seal Island.
|
{
"evidence": [
"Electric chair",
"Buffalo, New York",
"Canada–United States border"
],
"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 border between Canada and the United States also has one land dispute over Machias Seal Island.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Electric chair",
"Buffalo, New York",
"Canada–United States border"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2813,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
64331
|
There is no record of tropical cyclones in the Arabian Peninsula.
|
{
"evidence": [
"North Indian Ocean tropical cyclone_sentence_15"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThere is no record of tropical cyclones in the Arabian Peninsula.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"North Indian Ocean tropical cyclone_sentence_15"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2814,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.564",
"value": "?search="
},
{
"operation": "search",
"time": "78.814",
"value": "Tropical cyclones in the Arabian Sea"
},
{
"operation": "Now on",
"time": "80.294",
"value": "North Indian Ocean tropical cyclone"
},
{
"operation": "Page search",
"time": "87.034",
"value": "penins"
},
{
"operation": "Highlighting",
"time": "98.966",
"value": "North Indian Ocean tropical cyclone_sentence_15"
},
{
"operation": "finish",
"time": "111.432",
"value": "finish"
}
]
}
|
exfever_7885
|
Central America is bordered by the largest division of the body of salt water that covers approximately 64.8% of the surface of Earth.
|
{
"evidence": [
"Central America",
"Pacific Ocean",
"World Ocean"
],
"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:\nCentral America is bordered by the largest division of the body of salt water that covers approximately 64.8% of the surface of Earth.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Central America",
"Pacific Ocean",
"World Ocean"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2815,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
32799
|
Jan Kratochvíl, born on February 10, 1959, studied at Charles University in Prague (the oldest and largest university in the Czech Republic), and became a professor there.
|
{
"evidence": [
"Jan Kratochvíl_sentence_0",
"Jan Kratochvíl_sentence_2",
"Charles University_sentence_1",
"Jan Kratochvíl_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:\nJan Kratochvíl, born on February 10, 1959, studied at Charles University in Prague (the oldest and largest university in the Czech Republic), and became a professor there.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Jan Kratochvíl_sentence_0",
"Jan Kratochvíl_sentence_2",
"Charles University_sentence_1",
"Jan Kratochvíl_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2816,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.554",
"value": "?search="
},
{
"operation": "search",
"time": "9.324",
"value": "Jan Kratochvíl"
},
{
"operation": "Now on",
"time": "10.393",
"value": "Jan Kratochvíl"
},
{
"operation": "Highlighting",
"time": "15.169",
"value": "Jan Kratochvíl_sentence_0"
},
{
"operation": "Highlighting",
"time": "22.488",
"value": "Jan Kratochvíl_sentence_2"
},
{
"operation": "hyperlink",
"time": "26.368",
"value": "Charles University in Prague"
},
{
"operation": "Now on",
"time": "27.174",
"value": "Charles University"
},
{
"operation": "Highlighting",
"time": "31.276",
"value": "Charles University_sentence_1"
},
{
"operation": "back-button-clicked",
"time": "32.753",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "32.818",
"value": "Jan Kratochvíl"
},
{
"operation": "Highlighting",
"time": "40.203",
"value": "Jan Kratochvíl_sentence_3"
},
{
"operation": "finish",
"time": "68.359",
"value": "finish"
}
]
}
|
81938
|
Bill Paxon (born 1954) is a politician from New York.
|
{
"evidence": [
"Bill Paxon_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:\nBill Paxon (born 1954) is a politician from New York.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bill Paxon_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2817,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.143",
"value": "?search="
},
{
"operation": "search",
"time": "6.726",
"value": "Bill Paxon"
},
{
"operation": "Now on",
"time": "7.432",
"value": "Bill Paxon"
},
{
"operation": "Highlighting",
"time": "22.1",
"value": "Bill Paxon_sentence_0"
},
{
"operation": "hyperlink",
"time": "125.822",
"value": "lobbyist"
},
{
"operation": "Now on",
"time": "126.381",
"value": "Lobbying"
},
{
"operation": "back-button-clicked",
"time": "166.997",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "167.07",
"value": "Bill Paxon"
},
{
"operation": "finish",
"time": "291.754",
"value": "finish"
}
]
}
|
35415
|
Al Sabith was offered to do the role of Lord Ayyappan in Uppum Mulakum serial telecasting on Flowers TV, and after the success of Uppum Mulakum, he got numerous offers from the film industry
|
{
"evidence": [
"Al Sabith_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:\nAl Sabith was offered to do the role of Lord Ayyappan in Uppum Mulakum serial telecasting on Flowers TV, and after the success of Uppum Mulakum, he got numerous offers from the film industry\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Al Sabith_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2818,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.537",
"value": "?search="
},
{
"operation": "search",
"time": "15.36",
"value": "Al Sabith"
},
{
"operation": "Now on",
"time": "17.415",
"value": "Al Sabith"
},
{
"operation": "Highlighting",
"time": "24.284",
"value": "Al Sabith_sentence_1"
},
{
"operation": "Highlighting",
"time": "46.706",
"value": "Al Sabith_sentence_2"
},
{
"operation": "hyperlink",
"time": "56.817",
"value": "Lord Ayyappan"
},
{
"operation": "Now on",
"time": "58.216",
"value": "Ayyappan"
},
{
"operation": "back-button-clicked",
"time": "59.22",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "59.419",
"value": "Al Sabith"
},
{
"operation": "Highlighting",
"time": "62.089",
"value": "Al Sabith_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "65.191",
"value": "Al Sabith_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "70.69",
"value": "Al Sabith_sentence_1"
},
{
"operation": "finish",
"time": "85.9",
"value": "finish"
}
]
}
|
exfever_3041
|
Rhona Mitra is a Canadian actress, model, and singer who is known for her role as Detective Kit McGraw in the third season of an American medical drama television series created by Ryan Murphy.
|
{
"evidence": [
"Rhona Mitra",
"Nip/Tuck"
],
"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:\nRhona Mitra is a Canadian actress, model, and singer who is known for her role as Detective Kit McGraw in the third season of an American medical drama television series created by Ryan Murphy.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Rhona Mitra",
"Nip/Tuck"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2819,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_678
|
The Strain is an Afghan horror drama television series whose pilot episode, "Night Zero," premiered at the ATX Television Festival in the capital city, as well as the seat and largest city of Travis County.
|
{
"evidence": [
"The Strain (TV series)",
"Austin, Texas"
],
"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 Strain is an Afghan horror drama television series whose pilot episode, \"Night Zero,\" premiered at the ATX Television Festival in the capital city, as well as the seat and largest city of Travis County.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Strain (TV series)",
"Austin, Texas"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2820,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
49074
|
Chow Chun Fai was born in Hong Kong in 1980 and graduated from the New Asia College of the Chinese University of Hong Kong (CUHK), where he earned a BA and an MFA from the Department of Fine Arts.
|
{
"evidence": [
"Chow Chun Fai_sentence_4",
"Chow Chun Fai_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:\nChow Chun Fai was born in Hong Kong in 1980 and graduated from the New Asia College of the Chinese University of Hong Kong (CUHK), where he earned a BA and an MFA from the Department of Fine Arts.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Chow Chun Fai_sentence_4",
"Chow Chun Fai_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2821,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.751",
"value": "?search="
},
{
"operation": "search",
"time": "20.867",
"value": "Chow Chun Fai"
},
{
"operation": "Now on",
"time": "21.668",
"value": "Chow Chun Fai"
},
{
"operation": "Highlighting",
"time": "32.88",
"value": "Chow Chun Fai_sentence_0"
},
{
"operation": "Page search",
"time": "42.808",
"value": "CUHK"
},
{
"operation": "Highlighting",
"time": "47.381",
"value": "Chow Chun Fai_sentence_4"
},
{
"operation": "Highlighting",
"time": "54.095",
"value": "Chow Chun Fai_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "56.185",
"value": "Chow Chun Fai_sentence_0"
},
{
"operation": "finish",
"time": "76.984",
"value": "finish"
}
]
}
|
exfever_390
|
Ginger Baker gained early fame alongside a bassist who?gained popularity as the co-lead vocalist of the British rock band Cream.?
|
{
"evidence": [
"Ginger Baker",
"Jack Bruce"
],
"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:\nGinger Baker gained early fame alongside a bassist who?gained popularity as the co-lead vocalist of the British rock band Cream.?\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ginger Baker",
"Jack Bruce"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2822,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8899
|
Morgan Fairchild also appeared in a video game as herself.
|
{
"evidence": [
"Morgan Fairchild",
"Jenna Wade"
],
"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:\nMorgan Fairchild also appeared in a video game as herself.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Morgan Fairchild",
"Jenna Wade"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2823,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1247
|
Tanya Roberts was a Canadian actress, producer, and model best known for playing Stacey Sutton in the 1985 spy film and the fourteenth in the James Bond series produced by Eon Productions.
|
{
"evidence": [
"Tanya Roberts",
"A View to a Kill"
],
"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:\nTanya Roberts was a Canadian actress, producer, and model best known for playing Stacey Sutton in the 1985 spy film and the fourteenth in the James Bond series produced by Eon Productions.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tanya Roberts",
"A View to a Kill"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2824,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
8978
|
Greg Stafford is an American game designer, applying design and aesthetics to create a game, born 1948 and died in 2018.
|
{
"evidence": [
"Greg Stafford_sentence_1",
"Game design_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:\nGreg Stafford is an American game designer, applying design and aesthetics to create a game, born 1948 and died in 2018.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Greg Stafford_sentence_1",
"Game design_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2825,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.6",
"value": "?search="
},
{
"operation": "search",
"time": "99.363",
"value": "Greg Stafford"
},
{
"operation": "Now on",
"time": "100.344",
"value": "Greg Stafford"
},
{
"operation": "Highlighting",
"time": "107.617",
"value": "Greg Stafford_sentence_1"
},
{
"operation": "Page search",
"time": "123.961",
"value": "design"
},
{
"operation": "Page search",
"time": "134.868",
"value": "aesthetics"
},
{
"operation": "hyperlink",
"time": "150.252",
"value": "shamanism"
},
{
"operation": "Now on",
"time": "151.103",
"value": "Shamanism"
},
{
"operation": "back-button-clicked",
"time": "163.811",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "163.926",
"value": "Greg Stafford"
},
{
"operation": "hyperlink",
"time": "169.998",
"value": "Game designer"
},
{
"operation": "Now on",
"time": "171.121",
"value": "Game designer"
},
{
"operation": "Highlighting",
"time": "173.871",
"value": "Game design_sentence_1"
},
{
"operation": "back-button-clicked",
"time": "180.674",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "180.789",
"value": "Greg Stafford"
},
{
"operation": "finish",
"time": "188.487",
"value": "finish"
}
]
}
|
45008
|
Barbara Stauffacher Solomon was born in 1928 as a third-generation San Franciscan and studied at San Francisco Art Institute, a private college of contemporary art in San Francisco, California.
|
{
"evidence": [
"Barbara Stauffacher Solomon_sentence_2",
"Barbara Stauffacher Solomon_sentence_3",
"San Francisco Art Institute_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:\nBarbara Stauffacher Solomon was born in 1928 as a third-generation San Franciscan and studied at San Francisco Art Institute, a private college of contemporary art in San Francisco, California.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Barbara Stauffacher Solomon_sentence_2",
"Barbara Stauffacher Solomon_sentence_3",
"San Francisco Art Institute_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2826,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.077",
"value": "?search="
},
{
"operation": "search",
"time": "16.705",
"value": "Barbara Stauffacher Solomon"
},
{
"operation": "Now on",
"time": "17.393",
"value": "Barbara Stauffacher Solomon"
},
{
"operation": "Highlighting",
"time": "21.179",
"value": "Barbara Stauffacher Solomon_sentence_0"
},
{
"operation": "Highlighting",
"time": "31.982",
"value": "Barbara Stauffacher Solomon_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "34.547",
"value": "Barbara Stauffacher Solomon_sentence_0"
},
{
"operation": "Highlighting",
"time": "42.756",
"value": "Barbara Stauffacher Solomon_sentence_3"
},
{
"operation": "hyperlink",
"time": "47.711",
"value": "San Francisco Art Institute"
},
{
"operation": "Now on",
"time": "48.565",
"value": "San Francisco Art Institute"
},
{
"operation": "Highlighting",
"time": "53.083",
"value": "San Francisco Art Institute_sentence_1"
},
{
"operation": "finish",
"time": "62.383",
"value": "finish"
}
]
}
|
78195
|
Annie Dillard not only received bachelor's and master's degrees at Hollin's College but also met her husband, creative writing professor Richard Henry Wilde Dillard, who she claimed taught her everything she knew about writing.
|
{
"evidence": [
"Annie Dillard_sentence_18",
"Annie Dillard_sentence_24",
"Annie Dillard_sentence_92"
],
"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:\nAnnie Dillard not only received bachelor's and master's degrees at Hollin's College but also met her husband, creative writing professor Richard Henry Wilde Dillard, who she claimed taught her everything she knew about writing.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Annie Dillard_sentence_18",
"Annie Dillard_sentence_24",
"Annie Dillard_sentence_92"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2827,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.823",
"value": "?search="
},
{
"operation": "search",
"time": "19.172",
"value": "Annie Dillard"
},
{
"operation": "Now on",
"time": "20.025",
"value": "Annie Dillard"
},
{
"operation": "Highlighting",
"time": "69.742",
"value": "Annie Dillard_sentence_18"
},
{
"operation": "Highlighting",
"time": "75.104",
"value": "Annie Dillard_sentence_24"
},
{
"operation": "Page search",
"time": "171.942",
"value": "everything she knew"
},
{
"operation": "Highlighting",
"time": "209.446",
"value": "Annie Dillard_sentence_92"
},
{
"operation": "Page search",
"time": "230.185",
"value": "everything"
},
{
"operation": "Highlighting",
"time": "295.055",
"value": "Annie Dillard_sentence_21"
},
{
"operation": "Highlighting deleted",
"time": "305.168",
"value": "Annie Dillard_sentence_21"
},
{
"operation": "Highlighting",
"time": "341.695",
"value": "Annie Dillard_sentence_22"
},
{
"operation": "Highlighting deleted",
"time": "364.274",
"value": "Annie Dillard_sentence_22"
},
{
"operation": "finish",
"time": "385.614",
"value": "finish"
}
]
}
|
28631
|
In the 2016 season of the Minnesota Vikings, Michael Griffin was signed in a one year.three million dollar contract, previously playing on the Detroit Lions.
|
{
"evidence": [
"Michael Griffin (American football)_sentence_70"
],
"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 2016 season of the Minnesota Vikings, Michael Griffin was signed in a one year.three million dollar contract, previously playing on the Detroit Lions.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Michael Griffin (American football)_sentence_70"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2828,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.408",
"value": "?search="
},
{
"operation": "search",
"time": "210.353",
"value": "2016 Minnesota Vikings season"
},
{
"operation": "Now on",
"time": "211.557",
"value": "2016 Minnesota Vikings season"
},
{
"operation": "Page search",
"time": "215.647",
"value": "griffin"
},
{
"operation": "hyperlink",
"time": "231.663",
"value": "Michael Griffin"
},
{
"operation": "Now on",
"time": "232.616",
"value": "Michael Griffin (American football)"
},
{
"operation": "Page search",
"time": "238.942",
"value": "contract"
},
{
"operation": "Highlighting",
"time": "257.775",
"value": "Michael Griffin (American football)_sentence_70"
},
{
"operation": "Page search",
"time": "289.281",
"value": "1.3"
},
{
"operation": "Page search",
"time": "308.534",
"value": "lions"
},
{
"operation": "Now on",
"time": "328.749",
"value": "Michael Griffin (American football)"
},
{
"operation": "Now on",
"time": "379.825",
"value": "Michael Griffin (American football)"
},
{
"operation": "finish",
"time": "387.975",
"value": "finish"
}
]
}
|
88337
|
Terri Priest was surrounded with multiple controversies about corruption and infidelity but was a supporter of the Worcester Art Museum (houses over 38,000 works of art dating from antiquity to the present day).
|
{
"evidence": [
"Terri Priest_sentence_10"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nTerri Priest was surrounded with multiple controversies about corruption and infidelity but was a supporter of the Worcester Art Museum (houses over 38,000 works of art dating from antiquity to the present day).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Terri Priest_sentence_10"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2829,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.85",
"value": "?search="
},
{
"operation": "search",
"time": "10.683",
"value": "Terri Priest"
},
{
"operation": "Now on",
"time": "11.299",
"value": "Terri Priest"
},
{
"operation": "Page search",
"time": "68.392",
"value": "corruption"
},
{
"operation": "Page search",
"time": "111.352",
"value": "Worcester Art Museum"
},
{
"operation": "Highlighting",
"time": "120.183",
"value": "Terri Priest_sentence_10"
},
{
"operation": "finish",
"time": "184.464",
"value": "finish"
}
]
}
|
14984
|
Horowpothana Divisional Secretariat is on the the administrative sub-unites in the districts of Sri Lanka that were originially based on the feudal countries, the korales and ratas.
|
{
"evidence": [
"Horowpothana Divisional Secretariat_sentence_0",
"Divisional Secretariats of Sri Lanka_sentence_0",
"Divisional Secretariats of Sri Lanka_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:\nHorowpothana Divisional Secretariat is on the the administrative sub-unites in the districts of Sri Lanka that were originially based on the feudal countries, the korales and ratas.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Horowpothana Divisional Secretariat_sentence_0",
"Divisional Secretariats of Sri Lanka_sentence_0",
"Divisional Secretariats of Sri Lanka_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2830,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.844",
"value": "?search="
},
{
"operation": "search",
"time": "14.606",
"value": "Horowpothana Divisional Secretariat"
},
{
"operation": "Now on",
"time": "15.492",
"value": "Horowpothana Divisional Secretariat"
},
{
"operation": "hyperlink",
"time": "55.866",
"value": "Divisional Secretariat"
},
{
"operation": "Now on",
"time": "56.411",
"value": "Divisional Secretariats of Sri Lanka"
},
{
"operation": "Highlighting",
"time": "64.87",
"value": "Divisional Secretariats of Sri Lanka_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "73.06",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "73.133",
"value": "Horowpothana Divisional Secretariat"
},
{
"operation": "Highlighting",
"time": "75.579",
"value": "Horowpothana Divisional Secretariat_sentence_0"
},
{
"operation": "hyperlink",
"time": "90.856",
"value": "Divisional Secretariat"
},
{
"operation": "Now on",
"time": "91.285",
"value": "Divisional Secretariats of Sri Lanka"
},
{
"operation": "hyperlink",
"time": "93.173",
"value": "counties"
},
{
"operation": "Now on",
"time": "93.946",
"value": "County"
},
{
"operation": "back-button-clicked",
"time": "96.338",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "96.44",
"value": "Divisional Secretariats of Sri Lanka"
},
{
"operation": "Highlighting",
"time": "99.463",
"value": "Divisional Secretariats of Sri Lanka_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "103.046",
"value": "Divisional Secretariats of Sri Lanka_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "104.414",
"value": "Divisional Secretariats of Sri Lanka_sentence_0"
},
{
"operation": "Highlighting",
"time": "106.017",
"value": "Divisional Secretariats of Sri Lanka_sentence_0"
},
{
"operation": "Highlighting",
"time": "107.066",
"value": "Divisional Secretariats of Sri Lanka_sentence_1"
},
{
"operation": "finish",
"time": "139.533",
"value": "finish"
}
]
}
|
exfever_4369
|
Limbu is language spoken by the Limbu people of Nepal and Northeastern India, a family of more than 500 languages in which other languages are spoken in the mountain range in Asia, separating the plains of the Indian subcontinent from the Tibetan Plateau.
|
{
"evidence": [
"Limbu language",
"Sino-Tibetan languages",
"Himalayas"
],
"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:\nLimbu is language spoken by the Limbu people of Nepal and Northeastern India, a family of more than 500 languages in which other languages are spoken in the mountain range in Asia, separating the plains of the Indian subcontinent from the Tibetan Plateau.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Limbu language",
"Sino-Tibetan languages",
"Himalayas"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2831,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_603
|
Jonah Hill is a British comedian known for his comedic roles in films including a 2007 American coming-of-age teen buddy comedy film directed by Greg Mottola.
|
{
"evidence": [
"Jonah Hill",
"Superbad (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:\nJonah Hill is a British comedian known for his comedic roles in films including a 2007 American coming-of-age teen buddy comedy film directed by Greg Mottola.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jonah Hill",
"Superbad (film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2832,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
67202
|
Singer, songwriter and rapper Coco Jones released four singles in 2009 that made it to the top ten charts.
|
{
"evidence": [
"Coco Jones discography_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:\nSinger, songwriter and rapper Coco Jones released four singles in 2009 that made it to the top ten charts.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Coco Jones discography_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2833,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.134",
"value": "?search="
},
{
"operation": "search",
"time": "13.793",
"value": "Coco Jones"
},
{
"operation": "Now on",
"time": "16.598",
"value": "Coco Jones"
},
{
"operation": "hyperlink",
"time": "157.75",
"value": "Coco Jones discography"
},
{
"operation": "Now on",
"time": "158.386",
"value": "Coco Jones discography"
},
{
"operation": "Highlighting",
"time": "259.277",
"value": "Coco Jones discography_sentence_0"
},
{
"operation": "finish",
"time": "262.356",
"value": "finish"
}
]
}
|
90143
|
According to the 2006 census, there were seven hundred twenty-eight families in Barzul, the capital of Zarrin Dasht in the Iranian province of Hamadan.
|
{
"evidence": [
"Barzul_sentence_1",
"Barzul_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:\nAccording to the 2006 census, there were seven hundred twenty-eight families in Barzul, the capital of Zarrin Dasht in the Iranian province of Hamadan.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Barzul_sentence_1",
"Barzul_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2834,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.177",
"value": "?search="
},
{
"operation": "search",
"time": "11.145",
"value": "Barzul"
},
{
"operation": "Now on",
"time": "12.241",
"value": "Barzul"
},
{
"operation": "Highlighting",
"time": "15.339",
"value": "Barzul_sentence_1"
},
{
"operation": "Highlighting",
"time": "27.15",
"value": "Barzul_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "46.431",
"value": "Barzul_sentence_0"
},
{
"operation": "Highlighting",
"time": "50.013",
"value": "Barzul_sentence_0"
},
{
"operation": "finish",
"time": "72.248",
"value": "finish"
}
]
}
|
exfever_2398
|
King Kong is a 2005 adventure monster film directed by Peter Jackson that features a large-scale, sweeping scope, and spectacle.
|
{
"evidence": [
"King Kong (2005 film)",
"Epic 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:\nKing Kong is a 2005 adventure monster film directed by Peter Jackson that features a large-scale, sweeping scope, and spectacle.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"King Kong (2005 film)",
"Epic film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2835,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
45769
|
The Bristol Development Corporation (established in 1989) built 650 housing units and created 4,825 new jobs during its lifetime.
|
{
"evidence": [
"Bristol Development Corporation_sentence_3"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Bristol Development Corporation (established in 1989) built 650 housing units and created 4,825 new jobs during its lifetime.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bristol Development Corporation_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2836,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.9",
"value": "?search="
},
{
"operation": "search",
"time": "12.525",
"value": "Bristol Development Corporation"
},
{
"operation": "Now on",
"time": "13.846",
"value": "Bristol Development Corporation"
},
{
"operation": "Highlighting",
"time": "17.295",
"value": "Bristol Development Corporation_sentence_0"
},
{
"operation": "Highlighting",
"time": "34.863",
"value": "Bristol Development Corporation_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "37.14",
"value": "Bristol Development Corporation_sentence_0"
},
{
"operation": "finish",
"time": "65.061",
"value": "finish"
}
]
}
|
exfever_3592
|
Superman is Christopher Reeve's first film.
|
{
"evidence": [
"Superman (1978 film)",
"Superhero film",
"Origin story"
],
"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:\nSuperman is Christopher Reeve's first film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Superman (1978 film)",
"Superhero film",
"Origin story"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2837,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
22614
|
The 2016 Pasig local elections were held on May 9, 2016 and the voters elected the elective local posts in the city.
|
{
"evidence": [
"2016 Pasig local elections_sentence_0",
"2016 Pasig local elections_sentence_1"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe 2016 Pasig local elections were held on May 9, 2016 and the voters elected the elective local posts in the city.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"2016 Pasig local elections_sentence_0",
"2016 Pasig local elections_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2838,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.127",
"value": "?search="
},
{
"operation": "search",
"time": "17.173",
"value": "2016 Pasig local elections"
},
{
"operation": "Now on",
"time": "18.056",
"value": "2016 Pasig local elections"
},
{
"operation": "Highlighting",
"time": "26.702",
"value": "2016 Pasig local elections_sentence_0"
},
{
"operation": "Highlighting",
"time": "35.533",
"value": "2016 Pasig local elections_sentence_1"
},
{
"operation": "finish",
"time": "55.022",
"value": "finish"
}
]
}
|
77821
|
When Rappers Attack was the first studio album released by Tame One (which contained a hit single with the same name) in 2003 which made it to number 74 on the "100 Best Indie Hip-Hop Records of All Time" list.
|
{
"evidence": [
"When Rappers Attack_sentence_0",
"When Rappers Attack_sentence_1",
"When Rappers Attack_sentence_4"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWhen Rappers Attack was the first studio album released by Tame One (which contained a hit single with the same name) in 2003 which made it to number 74 on the \"100 Best Indie Hip-Hop Records of All Time\" list.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"When Rappers Attack_sentence_0",
"When Rappers Attack_sentence_1",
"When Rappers Attack_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2839,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.093",
"value": "?search="
},
{
"operation": "search",
"time": "14.154",
"value": "When Rappers Attack"
},
{
"operation": "Now on",
"time": "15.059",
"value": "When Rappers Attack"
},
{
"operation": "Highlighting",
"time": "21.922",
"value": "When Rappers Attack_sentence_0"
},
{
"operation": "Highlighting",
"time": "43.818",
"value": "When Rappers Attack_sentence_1"
},
{
"operation": "Highlighting",
"time": "49.258",
"value": "When Rappers Attack_sentence_4"
},
{
"operation": "finish",
"time": "54.206",
"value": "finish"
}
]
}
|
exfever_8776
|
Tim Rice is best known for his collaborations with an Italian composer and impresario of musical theatre
|
{
"evidence": [
"Tim Rice",
"Andrew Lloyd Webber"
],
"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:\nTim Rice is best known for his collaborations with an Italian composer and impresario of musical theatre\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tim Rice",
"Andrew Lloyd Webber"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2840,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
7287
|
A Tonic for the Troops is the second album by Irish rock band The Boomtown Rats, a band that originally formed in Dublin in 1975.
|
{
"evidence": [
"A Tonic for the Troops_sentence_0",
"The Boomtown Rats_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:\nA Tonic for the Troops is the second album by Irish rock band The Boomtown Rats, a band that originally formed in Dublin in 1975.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"A Tonic for the Troops_sentence_0",
"The Boomtown Rats_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2841,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.415",
"value": "?search="
},
{
"operation": "search",
"time": "15.884",
"value": "Tonic for the Troops"
},
{
"operation": "Now on",
"time": "16.736",
"value": "A Tonic for the Troops"
},
{
"operation": "Highlighting",
"time": "27.283",
"value": "A Tonic for the Troops_sentence_0"
},
{
"operation": "hyperlink",
"time": "28.259",
"value": "The Boomtown Rats"
},
{
"operation": "Now on",
"time": "28.84",
"value": "The Boomtown Rats"
},
{
"operation": "Highlighting",
"time": "32.939",
"value": "The Boomtown Rats_sentence_0"
},
{
"operation": "finish",
"time": "49.482",
"value": "finish"
}
]
}
|
exfever_2893
|
The Punisher is the second character created by writer Gerry Conway.
|
{
"evidence": [
"Punisher",
"Antihero",
"Courage"
],
"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 Punisher is the second character created by writer Gerry Conway.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Punisher",
"Antihero",
"Courage"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2842,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7871
|
Gordon James Ramsay is a British chef who became one of the best-known and most influential chefs in a sovereign country in north-western Europe that have been separated by the sea of people who spoke a Goidelic language of the Insular Celtic branch of the Celtic language family, which is a part of the Indo-European language family.
|
{
"evidence": [
"Gordon Ramsay",
"United Kingdom",
"Irish language"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nGordon James Ramsay is a British chef who became one of the best-known and most influential chefs in a sovereign country in north-western Europe that have been separated by the sea of people who spoke a Goidelic language of the Insular Celtic branch of the Celtic language family, which is a part of the Indo-European language family.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Gordon Ramsay",
"United Kingdom",
"Irish language"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2843,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
30836
|
Gongzhuling South railway station is a railway station of the Harbin–Dalian section of the Beijing–Harbin High-Speed Railway.
|
{
"evidence": [
"Gongzhuling South railway station_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:\nGongzhuling South railway station is a railway station of the Harbin–Dalian section of the Beijing–Harbin High-Speed Railway.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Gongzhuling South railway station_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2844,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.982",
"value": "?search="
},
{
"operation": "search",
"time": "8.693",
"value": "Gongzhuling South railway station"
},
{
"operation": "Now on",
"time": "9.352",
"value": "Gongzhuling South railway station"
},
{
"operation": "Highlighting",
"time": "13.492",
"value": "Gongzhuling South railway station_sentence_0"
},
{
"operation": "finish",
"time": "21.258",
"value": "finish"
}
]
}
|
exfever_1831
|
Sherlock is a British television series based on a British writer and physician story, Holmes detective.
|
{
"evidence": [
"Sherlock (TV series)",
"Arthur Conan Doyle"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSherlock is a British television series based on a British writer and physician story, Holmes detective.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sherlock (TV series)",
"Arthur Conan Doyle"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2845,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3489
|
War of the Worlds is a 2005 Canadian science fiction action film produced by an American independent film production company founded by Tom Cruise and Paula Wagner.
|
{
"evidence": [
"War of the Worlds (2005 film)",
"Cruise/Wagner Productions"
],
"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:\nWar of the Worlds is a 2005 Canadian science fiction action film produced by an American independent film production company founded by Tom Cruise and Paula Wagner.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"War of the Worlds (2005 film)",
"Cruise/Wagner Productions"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2846,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3855
|
Super Street Fighter II Turbo became Japan's highest-grossing arcade game in 1994.
|
{
"evidence": [
"Super Street Fighter II Turbo",
"Fighting game"
],
"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:\nSuper Street Fighter II Turbo became Japan's highest-grossing arcade game in 1994.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Super Street Fighter II Turbo",
"Fighting game"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2847,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
30422
|
Fort Totten station is part of the 20 million passenger stop Washington Metro and was part of a 36 month trial, being made permanent in 2008.
|
{
"evidence": [
"Washington Metro_sentence_13",
"Fort Totten station_sentence_14"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nFort Totten station is part of the 20 million passenger stop Washington Metro and was part of a 36 month trial, being made permanent in 2008.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Washington Metro_sentence_13",
"Fort Totten station_sentence_14"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2848,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.85",
"value": "?search="
},
{
"operation": "search",
"time": "16.079",
"value": "Fort Totten station"
},
{
"operation": "Now on",
"time": "17.231",
"value": "Fort Totten station"
},
{
"operation": "Highlighting",
"time": "81.482",
"value": "Fort Totten station_sentence_0"
},
{
"operation": "hyperlink",
"time": "82.635",
"value": "Washington Metro"
},
{
"operation": "Now on",
"time": "84.938",
"value": "Washington Metro"
},
{
"operation": "hyperlink",
"time": "99.726",
"value": "3.4 Passenger information systems Washington Metro_section_14"
},
{
"operation": "Page search",
"time": "121.865",
"value": "20 million"
},
{
"operation": "back-button-clicked",
"time": "179.844",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "180.12",
"value": "Fort Totten station"
},
{
"operation": "Page search",
"time": "185.588",
"value": "20 million"
},
{
"operation": "hyperlink",
"time": "201.001",
"value": "Washington Metro"
},
{
"operation": "Now on",
"time": "202.538",
"value": "Washington Metro"
},
{
"operation": "Highlighting",
"time": "206.538",
"value": "Washington Metro_sentence_13"
},
{
"operation": "Highlighting deleted",
"time": "208.976",
"value": "Washington Metro_sentence_13"
},
{
"operation": "Highlighting",
"time": "213.174",
"value": "Washington Metro_sentence_13"
},
{
"operation": "back-button-clicked",
"time": "214.081",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "214.331",
"value": "Fort Totten station"
},
{
"operation": "Page search",
"time": "242.59",
"value": "trial"
},
{
"operation": "Highlighting",
"time": "247.628",
"value": "Fort Totten station_sentence_14"
},
{
"operation": "Highlighting deleted",
"time": "256.519",
"value": "Fort Totten station_sentence_0"
},
{
"operation": "finish",
"time": "258.883",
"value": "finish"
}
]
}
|
83918
|
Launch Party's opening scene showed Michael Scott at Dunder Mifflin preparing a party for the launch of their new website.
|
{
"evidence": [
"Launch Party_sentence_7"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLaunch Party's opening scene showed Michael Scott at Dunder Mifflin preparing a party for the launch of their new website.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Launch Party_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2849,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.194",
"value": "?search="
},
{
"operation": "search",
"time": "16.008",
"value": "Launch Party"
},
{
"operation": "Now on",
"time": "16.841",
"value": "Launch Party"
},
{
"operation": "Highlighting",
"time": "45.163",
"value": "Launch Party_sentence_7"
},
{
"operation": "finish",
"time": "98.478",
"value": "finish"
}
]
}
|
exfever_9
|
Elementary is a procedural drama television series that presented a contemporary update of the character Sherlock Holmes, created by a Canadian writer and physician.
|
{
"evidence": [
"Elementary (TV series)",
"Arthur Conan Doyle"
],
"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:\nElementary is a procedural drama television series that presented a contemporary update of the character Sherlock Holmes, created by a Canadian writer and physician.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Elementary (TV series)",
"Arthur Conan Doyle"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2850,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
81035
|
During his early years, John Hambrock, born and raised in Fort Wayne, Indiana and creator of the comic strip The Brilliant Mind of Edison Lee, liked marine biology and did not think about being a cartoonist as a career.
|
{
"evidence": [
"John Hambrock_sentence_2",
"John Hambrock_sentence_0",
"John Hambrock_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:\nDuring his early years, John Hambrock, born and raised in Fort Wayne, Indiana and creator of the comic strip The Brilliant Mind of Edison Lee, liked marine biology and did not think about being a cartoonist as a career.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"John Hambrock_sentence_2",
"John Hambrock_sentence_0",
"John Hambrock_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2851,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.771",
"value": "?search="
},
{
"operation": "search",
"time": "171.379",
"value": "John Hambrock"
},
{
"operation": "Now on",
"time": "172.125",
"value": "John Hambrock"
},
{
"operation": "Highlighting",
"time": "277.708",
"value": "John Hambrock_sentence_2"
},
{
"operation": "Highlighting",
"time": "301.527",
"value": "John Hambrock_sentence_0"
},
{
"operation": "Highlighting",
"time": "307.694",
"value": "John Hambrock_sentence_4"
},
{
"operation": "Highlighting",
"time": "308.804",
"value": "John Hambrock_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "329.84",
"value": "John Hambrock_sentence_5"
},
{
"operation": "hyperlink",
"time": "337.546",
"value": "The Brilliant Mind of Edison Lee"
},
{
"operation": "Now on",
"time": "338.385",
"value": "The Brilliant Mind of Edison Lee"
},
{
"operation": "finish",
"time": "372.281",
"value": "finish"
}
]
}
|
51966
|
Marie Dubas (3 September 1894 – 21 February 1972) was a French music-hall singer, and one of her songs, "Mon légionnaire", was written by Raymond Asso (2 June 1901 – 24 October 1968), who was a French lyricist, and Marie recorded it in 1936, and then in 1939 she toured the United States.
|
{
"evidence": [
"Marie Dubas_sentence_0",
"Marie Dubas_sentence_4",
"Marie Dubas_sentence_5",
"Raymond Asso_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:\nMarie Dubas (3 September 1894 – 21 February 1972) was a French music-hall singer, and one of her songs, \"Mon légionnaire\", was written by Raymond Asso (2 June 1901 – 24 October 1968), who was a French lyricist, and Marie recorded it in 1936, and then in 1939 she toured the United States.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Marie Dubas_sentence_0",
"Marie Dubas_sentence_4",
"Marie Dubas_sentence_5",
"Raymond Asso_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2852,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.655",
"value": "?search="
},
{
"operation": "search",
"time": "11.686",
"value": "Marie Dubas"
},
{
"operation": "Now on",
"time": "12.378",
"value": "Marie Dubas"
},
{
"operation": "Highlighting",
"time": "19.829",
"value": "Marie Dubas_sentence_0"
},
{
"operation": "Page search",
"time": "30.205",
"value": "mon"
},
{
"operation": "Highlighting",
"time": "35.389",
"value": "Marie Dubas_sentence_4"
},
{
"operation": "Highlighting",
"time": "44.324",
"value": "Marie Dubas_sentence_5"
},
{
"operation": "hyperlink",
"time": "45.085",
"value": "Raymond Asso"
},
{
"operation": "Now on",
"time": "45.73",
"value": "Raymond Asso"
},
{
"operation": "Highlighting",
"time": "55.82",
"value": "Raymond Asso_sentence_0"
},
{
"operation": "finish",
"time": "62.284",
"value": "finish"
}
]
}
|
exfever_3188
|
Steffi Graf is a former German player in a racket sport, a modern game that originated in a country that is part of the United States.
|
{
"evidence": [
"Steffi Graf",
"Tennis",
"England"
],
"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:\nSteffi Graf is a former German player in a racket sport, a modern game that originated in a country that is part of the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Steffi Graf",
"Tennis",
"England"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2853,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2025
|
Patrick Dempsey is one of the most popular American race car drivers.
|
{
"evidence": [
"Patrick Dempsey",
"Loverboy (1989 film)"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPatrick Dempsey is one of the most popular American race car drivers.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Patrick Dempsey",
"Loverboy (1989 film)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2854,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_684
|
Southport is the eleventh most populated settlement in one of nine official regions of England whose administrative counties include a historic county.
|
{
"evidence": [
"Southport",
"North West England",
"Cheshire"
],
"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:\nSouthport is the eleventh most populated settlement in one of nine official regions of England whose administrative counties include a historic county.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Southport",
"North West England",
"Cheshire"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2855,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
46127
|
Australia's truple world champion Jack Brabham was in the first position at the Classification, finishing 80 laps with a time of 1:49:34.6
|
{
"evidence": [
"Jack Brabham_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:\nAustralia's truple world champion Jack Brabham was in the first position at the Classification, finishing 80 laps with a time of 1:49:34.6\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Jack Brabham_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2856,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.588",
"value": "?search="
},
{
"operation": "search",
"time": "53.659",
"value": "Jack Brabham"
},
{
"operation": "Now on",
"time": "55.529",
"value": "Jack Brabham"
},
{
"operation": "Page search",
"time": "75.527",
"value": "triple world champion"
},
{
"operation": "hyperlink",
"time": "160.856",
"value": "1960"
},
{
"operation": "Now on",
"time": "162.272",
"value": "1960 Formula One season"
},
{
"operation": "back-button-clicked",
"time": "188.023",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "188.284",
"value": "Jack Brabham"
},
{
"operation": "Highlighting",
"time": "207.774",
"value": "Jack Brabham_header_cell_0_7_0"
},
{
"operation": "Highlighting",
"time": "208.517",
"value": "Jack Brabham_cell_0_7_1"
},
{
"operation": "Highlighting",
"time": "208.529",
"value": "Jack Brabham_header_cell_0_7_0"
},
{
"operation": "Highlighting",
"time": "208.53",
"value": "Jack Brabham_header_cell_0_2_0"
},
{
"operation": "Page search",
"time": "225.025",
"value": "Classification"
},
{
"operation": "Highlighting",
"time": "250.253",
"value": "Jack Brabham_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "253.909",
"value": "Jack Brabham_cell_0_7_1"
},
{
"operation": "Highlighting deleted",
"time": "253.912",
"value": "Jack Brabham_header_cell_0_7_0"
},
{
"operation": "Highlighting deleted",
"time": "253.913",
"value": "Jack Brabham_header_cell_0_2_0"
},
{
"operation": "search",
"time": "274.116",
"value": "1"
},
{
"operation": "Now on",
"time": "276.257",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "305.899",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "306.161",
"value": "Jack Brabham"
},
{
"operation": "finish",
"time": "325.978",
"value": "finish"
}
]
}
|
exfever_7058
|
Rock music is a broad genre of popular music that originated as "rock and roll" in Europe, which is a style that drew directly from the genre of popular music that originated in African-American communities in the 1940s.
|
{
"evidence": [
"Rock music",
"Rhythm and blues"
],
"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:\nRock music is a broad genre of popular music that originated as \"rock and roll\" in Europe, which is a style that drew directly from the genre of popular music that originated in African-American communities in the 1940s.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Rock music",
"Rhythm and blues"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2857,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
48501
|
Christian Christophersen Sehested became a nobleman Hofjunker in 1677 and was Foreign Minister and later replace Jens Juel
|
{
"evidence": [
"Christian Christophersen Sehested_sentence_8"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nChristian Christophersen Sehested became a nobleman Hofjunker in 1677 and was Foreign Minister and later replace Jens Juel\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Christian Christophersen Sehested_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2858,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.277",
"value": "?search="
},
{
"operation": "search",
"time": "9.689",
"value": "Christian Christophersen Sehested"
},
{
"operation": "Now on",
"time": "10.704",
"value": "Christian Christophersen Sehested"
},
{
"operation": "Page search",
"time": "27.724",
"value": "nobleman"
},
{
"operation": "Highlighting",
"time": "29.687",
"value": "Christian Christophersen Sehested_sentence_8"
},
{
"operation": "finish",
"time": "41.165",
"value": "finish"
}
]
}
|
exfever_4699
|
Ed Wood is a 1994 film directed and produced by an American film director, producer, screenwriter, animator, and artist.
|
{
"evidence": [
"Ed Wood (film)",
"Tim Burton"
],
"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:\nEd Wood is a 1994 film directed and produced by an American film director, producer, screenwriter, animator, and artist.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ed Wood (film)",
"Tim Burton"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2859,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
23349
|
In the Ramsar Convention, an international treaty or formal and binding written agreement, the members of the Standing Committee, an intersessional executive body, are elected in the meetings of the Conference of the Contracting Parties to serve for three years.
|
{
"evidence": [
"Ramsar Convention_sentence_0",
"Treaty_sentence_3",
"Ramsar Convention_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:\nIn the Ramsar Convention, an international treaty or formal and binding written agreement, the members of the Standing Committee, an intersessional executive body, are elected in the meetings of the Conference of the Contracting Parties to serve for three years.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Ramsar Convention_sentence_0",
"Treaty_sentence_3",
"Ramsar Convention_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2860,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.37",
"value": "?search="
},
{
"operation": "search",
"time": "15.53",
"value": "Ramsar Convention"
},
{
"operation": "Now on",
"time": "16.65",
"value": "Ramsar Convention"
},
{
"operation": "Page search",
"time": "96.31",
"value": "binding"
},
{
"operation": "hyperlink",
"time": "112.654",
"value": "treaty"
},
{
"operation": "Now on",
"time": "113.595",
"value": "Treaty"
},
{
"operation": "back-button-clicked",
"time": "138.09",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "138.664",
"value": "Ramsar Convention"
},
{
"operation": "Highlighting",
"time": "184.542",
"value": "Ramsar Convention_sentence_0"
},
{
"operation": "hyperlink",
"time": "196.657",
"value": "treaty"
},
{
"operation": "Now on",
"time": "197.712",
"value": "Treaty"
},
{
"operation": "Highlighting",
"time": "202.728",
"value": "Treaty_sentence_3"
},
{
"operation": "Page search",
"time": "236.015",
"value": "standing"
},
{
"operation": "Page search",
"time": "279.356",
"value": "standing"
},
{
"operation": "Page search",
"time": "281.235",
"value": "standing"
},
{
"operation": "Page search",
"time": "281.522",
"value": "standing"
},
{
"operation": "Page search",
"time": "321.017",
"value": "committee"
},
{
"operation": "Highlighting",
"time": "336.395",
"value": "Treaty_item_1_12"
},
{
"operation": "Highlighting",
"time": "336.405",
"value": "Treaty_section_26"
},
{
"operation": "Highlighting deleted",
"time": "339.1",
"value": "Treaty_item_1_12"
},
{
"operation": "Highlighting deleted",
"time": "339.109",
"value": "Treaty_section_26"
},
{
"operation": "Page search",
"time": "361.793",
"value": "serve"
},
{
"operation": "Page search",
"time": "412.565",
"value": "Conference of the Contracting Parties"
},
{
"operation": "search",
"time": "436.56",
"value": "Conference of the Contracting Parties"
},
{
"operation": "Now on",
"time": "439.4",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "450.43",
"value": "Ramsar Convention"
},
{
"operation": "Now on",
"time": "451.483",
"value": "Ramsar Convention"
},
{
"operation": "Page search",
"time": "457.331",
"value": "Conference of the Contracting Parties"
},
{
"operation": "Highlighting",
"time": "472.826",
"value": "Ramsar Convention_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "557.577",
"value": "Ramsar Convention_sentence_3"
},
{
"operation": "Highlighting",
"time": "558.798",
"value": "Ramsar Convention_sentence_0"
},
{
"operation": "hyperlink",
"time": "559.92",
"value": "treaty"
},
{
"operation": "Now on",
"time": "561.025",
"value": "Treaty"
},
{
"operation": "Highlighting deleted",
"time": "562.924",
"value": "Treaty_sentence_3"
},
{
"operation": "Highlighting",
"time": "564.064",
"value": "Treaty_sentence_3"
},
{
"operation": "back-button-clicked",
"time": "566.318",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "566.86",
"value": "Ramsar Convention"
},
{
"operation": "Highlighting",
"time": "570.632",
"value": "Ramsar Convention_sentence_3"
},
{
"operation": "finish",
"time": "580.48",
"value": "finish"
}
]
}
|
69290
|
In 1911, Columbia Records recorded a duet of Aloha ʻOe by Nani Alapai and Henry N. Clark, and a 1913 score of the song is in the Levy Sheet Music Collection.
|
{
"evidence": [
"Aloha ʻOe_sentence_13",
"Aloha ʻOe_sentence_12"
],
"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 1911, Columbia Records recorded a duet of Aloha ʻOe by Nani Alapai and Henry N. Clark, and a 1913 score of the song is in the Levy Sheet Music Collection.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Aloha ʻOe_sentence_13",
"Aloha ʻOe_sentence_12"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2861,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.21",
"value": "?search="
},
{
"operation": "search",
"time": "16.166",
"value": "In 1911, Columbia Records recorded a duet of Aloha ʻOe by Nani Alapai and Henry N. Clark, and a 1913 score of the song is in the Levy Sheet Music Collection."
},
{
"operation": "Now on",
"time": "17.598",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "27.099",
"value": "Aloha ʻOe"
},
{
"operation": "Now on",
"time": "27.675",
"value": "Aloha ʻOe"
},
{
"operation": "Page search",
"time": "43.302",
"value": "1913 score"
},
{
"operation": "Highlighting",
"time": "47.9",
"value": "Aloha ʻOe_sentence_13"
},
{
"operation": "Highlighting",
"time": "71.396",
"value": "Aloha ʻOe_sentence_12"
},
{
"operation": "finish",
"time": "78.288",
"value": "finish"
}
]
}
|
27735
|
Leonardus Achates worked as a compositor but was never established to have been a printer.
|
{
"evidence": [
"Leonardus Achates_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:\nLeonardus Achates worked as a compositor but was never established to have been a printer.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Leonardus Achates_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2862,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.303",
"value": "?search="
},
{
"operation": "search",
"time": "5.886",
"value": "Leonardus Achates"
},
{
"operation": "Now on",
"time": "7.003",
"value": "Leonardus Achates"
},
{
"operation": "Highlighting",
"time": "10.256",
"value": "Leonardus Achates_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "22.678",
"value": "Leonardus Achates_sentence_0"
},
{
"operation": "Highlighting",
"time": "24.36",
"value": "Leonardus Achates_sentence_1"
},
{
"operation": "finish",
"time": "40.789",
"value": "finish"
}
]
}
|
68241
|
Humphry Marshall was a Quaker (a historically Christian denomination known formally as the Religious Society of Friends or Friends Church) and was cousins with John and William Bartram.
|
{
"evidence": [
"Humphry Marshall_sentence_4",
"Humphry Marshall_sentence_3",
"Quakers_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:\nHumphry Marshall was a Quaker (a historically Christian denomination known formally as the Religious Society of Friends or Friends Church) and was cousins with John and William Bartram.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Humphry Marshall_sentence_4",
"Humphry Marshall_sentence_3",
"Quakers_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2863,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.08",
"value": "?search="
},
{
"operation": "search",
"time": "10.724",
"value": "Humphry Marshall"
},
{
"operation": "Now on",
"time": "11.31",
"value": "Humphry Marshall"
},
{
"operation": "Highlighting",
"time": "18.996",
"value": "Humphry Marshall_sentence_4"
},
{
"operation": "Highlighting",
"time": "32.686",
"value": "Humphry Marshall_sentence_3"
},
{
"operation": "hyperlink",
"time": "38.405",
"value": "Quaker"
},
{
"operation": "Now on",
"time": "40.387",
"value": "Quakers"
},
{
"operation": "Highlighting",
"time": "47.932",
"value": "Quakers_sentence_4"
},
{
"operation": "finish",
"time": "70.282",
"value": "finish"
}
]
}
|
23132
|
Bach used the BACH motif in Canonic Variations on "Vom Himmel hoch da komm' ich her" and in The Art of Fugue (an incomplete musical work of unspecified instrumentation).
|
{
"evidence": [
"BACH motif_sentence_11",
"BACH motif_sentence_10",
"The Art of Fugue_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:\nBach used the BACH motif in Canonic Variations on \"Vom Himmel hoch da komm' ich her\" and in The Art of Fugue (an incomplete musical work of unspecified instrumentation).\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"BACH motif_sentence_11",
"BACH motif_sentence_10",
"The Art of Fugue_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2864,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.606",
"value": "?search="
},
{
"operation": "search",
"time": "19.659",
"value": "Bach"
},
{
"operation": "Now on",
"time": "21.07",
"value": "Johann Sebastian Bach"
},
{
"operation": "Page search",
"time": "28.156",
"value": "motif"
},
{
"operation": "search",
"time": "78.151",
"value": "Vom Himmel hoch da komm ich her"
},
{
"operation": "Now on",
"time": "78.992",
"value": "Vom Himmel hoch da komm ich her"
},
{
"operation": "Page search",
"time": "105.383",
"value": "motif"
},
{
"operation": "Page search",
"time": "129.939",
"value": "bach motif"
},
{
"operation": "search",
"time": "150.774",
"value": "Bach-Werke-Verzeichnis"
},
{
"operation": "Now on",
"time": "153.031",
"value": "Bach-Werke-Verzeichnis"
},
{
"operation": "Page search",
"time": "174.861",
"value": "canonic"
},
{
"operation": "search",
"time": "206.283",
"value": "BACH motif"
},
{
"operation": "Now on",
"time": "207.291",
"value": "BACH motif"
},
{
"operation": "Page search",
"time": "213.317",
"value": "vom himmel"
},
{
"operation": "Highlighting",
"time": "216.844",
"value": "BACH motif_sentence_11"
},
{
"operation": "Page search",
"time": "226.146",
"value": "art of"
},
{
"operation": "Highlighting",
"time": "233.755",
"value": "BACH motif_sentence_10"
},
{
"operation": "hyperlink",
"time": "235.013",
"value": "The Art of Fugue"
},
{
"operation": "Now on",
"time": "235.522",
"value": "The Art of Fugue"
},
{
"operation": "Highlighting",
"time": "241.211",
"value": "The Art of Fugue_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "287.936",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "287.991",
"value": "BACH motif"
},
{
"operation": "finish",
"time": "310.45",
"value": "finish"
}
]
}
|
exfever_6098
|
The Association of Tennis Professionals is also headquartered in St. Petersburg, Florida.
|
{
"evidence": [
"John McEnroe",
"Grand Slam (tennis)",
"Women's Tennis Association"
],
"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 Association of Tennis Professionals is also headquartered in St. Petersburg, Florida.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"John McEnroe",
"Grand Slam (tennis)",
"Women's Tennis Association"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2865,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
20454
|
Vogt joined the British Pharmacological Society and began work with Sir Henry Dale and Wilhelm Feldberg in 1936 at the National Institute for Medical Research, London.
|
{
"evidence": [
"British Journal of Pharmacology_sentence_0",
"Henry Hallett Dale_sentence_0",
"Henry Hallett Dale_sentence_9",
"Wilhelm Feldberg_sentence_0",
"Wilhelm Feldberg_sentence_5",
"Wilhelm Feldberg_sentence_13",
"Wilhelm Feldberg_sentence_16"
],
"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:\nVogt joined the British Pharmacological Society and began work with Sir Henry Dale and Wilhelm Feldberg in 1936 at the National Institute for Medical Research, London.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"British Journal of Pharmacology_sentence_0",
"Henry Hallett Dale_sentence_0",
"Henry Hallett Dale_sentence_9",
"Wilhelm Feldberg_sentence_0",
"Wilhelm Feldberg_sentence_5",
"Wilhelm Feldberg_sentence_13",
"Wilhelm Feldberg_sentence_16"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2866,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.813",
"value": "?search="
},
{
"operation": "search",
"time": "44.126",
"value": "Vogt"
},
{
"operation": "Now on",
"time": "45.292",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "49.803",
"value": "A. E. van Vogt"
},
{
"operation": "Now on",
"time": "51.178",
"value": "A. E. van Vogt"
},
{
"operation": "back-button-clicked",
"time": "58.596",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "59.424",
"value": "undefined"
},
{
"operation": "search",
"time": "85.897",
"value": "National Institute for Medical Research, London."
},
{
"operation": "Now on",
"time": "88.1",
"value": "undefined"
},
{
"operation": "search",
"time": "105.808",
"value": "National Institute for Medical Research"
},
{
"operation": "Now on",
"time": "106.984",
"value": "National Institute for Medical Research"
},
{
"operation": "Page search",
"time": "136.899",
"value": "Vogt"
},
{
"operation": "Page search",
"time": "163.184",
"value": "Sir Henry Dale and Wilhelm Feldberg i"
},
{
"operation": "search",
"time": "192.973",
"value": "British Pharmacological Society"
},
{
"operation": "Now on",
"time": "194.713",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "213.052",
"value": "Annette Dolphin"
},
{
"operation": "Now on",
"time": "214.208",
"value": "Annette Dolphin"
},
{
"operation": "search",
"time": "260.192",
"value": "Vogt Prize |"
},
{
"operation": "Now on",
"time": "261.744",
"value": "undefined"
},
{
"operation": "search",
"time": "276.059",
"value": "Vogt Prize"
},
{
"operation": "Now on",
"time": "276.943",
"value": "undefined"
},
{
"operation": "search",
"time": "305.463",
"value": "British Pharmacological Society"
},
{
"operation": "Now on",
"time": "306.166",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "308.701",
"value": "British Journal of Pharmacology"
},
{
"operation": "Now on",
"time": "309.503",
"value": "British Journal of Pharmacology"
},
{
"operation": "Highlighting",
"time": "311.179",
"value": "British Journal of Pharmacology_sentence_0"
},
{
"operation": "Page search",
"time": "341.543",
"value": "Vogt"
},
{
"operation": "search",
"time": "373.656",
"value": "vogt prize"
},
{
"operation": "Now on",
"time": "374.464",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "385.863",
"value": "next 20"
},
{
"operation": "Now on",
"time": "386.992",
"value": "undefined"
},
{
"operation": "search",
"time": "405.927",
"value": "Sir Henry Dale"
},
{
"operation": "Now on",
"time": "407.219",
"value": "Henry Hallett Dale"
},
{
"operation": "Highlighting",
"time": "422.703",
"value": "Henry Hallett Dale_sentence_0"
},
{
"operation": "Page search",
"time": "434.273",
"value": "vogt"
},
{
"operation": "Page search",
"time": "447.983",
"value": "vogt"
},
{
"operation": "Page search",
"time": "453.403",
"value": "Wilhelm Feldberg"
},
{
"operation": "Highlighting",
"time": "461.673",
"value": "Henry Hallett Dale_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "472.024",
"value": "Henry Hallett Dale_header_cell_0_0_0"
},
{
"operation": "Page search",
"time": "484.111",
"value": "e National Institute for Medical Research"
},
{
"operation": "Highlighting",
"time": "491.164",
"value": "Henry Hallett Dale_sentence_9"
},
{
"operation": "Page search",
"time": "514.436",
"value": "Wilhelm Feldberg"
},
{
"operation": "search",
"time": "519.744",
"value": "Wilhelm Feldberg"
},
{
"operation": "Now on",
"time": "521.333",
"value": "Wilhelm Feldberg"
},
{
"operation": "Highlighting",
"time": "523.875",
"value": "Wilhelm Feldberg_sentence_0"
},
{
"operation": "search",
"time": "541.025",
"value": "National Institute for Medical Research"
},
{
"operation": "Now on",
"time": "541.622",
"value": "National Institute for Medical Research"
},
{
"operation": "back-button-clicked",
"time": "545.436",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "545.504",
"value": "Wilhelm Feldberg"
},
{
"operation": "Page search",
"time": "548.742",
"value": "National Institute for Medical Research"
},
{
"operation": "Highlighting",
"time": "550.623",
"value": "Wilhelm Feldberg_sentence_5"
},
{
"operation": "Highlighting",
"time": "557.228",
"value": "Wilhelm Feldberg_sentence_13"
},
{
"operation": "Highlighting",
"time": "564.572",
"value": "Wilhelm Feldberg_sentence_16"
},
{
"operation": "Page search",
"time": "590.422",
"value": "Nvogt"
},
{
"operation": "Page search",
"time": "594.721",
"value": "vogt"
},
{
"operation": "search",
"time": "609.5",
"value": "British Pharmacological Society prizes"
},
{
"operation": "Now on",
"time": "610.962",
"value": "undefined"
},
{
"operation": "finish",
"time": "654.197",
"value": "finish"
}
]
}
|
49923
|
The majority of secondary school students in Ukraine have always spoken Russian.
|
{
"evidence": [
"Russian language in Ukraine_sentence_55",
"Russian language in Ukraine_sentence_62",
"Russian language in Ukraine_sentence_149"
],
"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 majority of secondary school students in Ukraine have always spoken Russian.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Russian language in Ukraine_sentence_55",
"Russian language in Ukraine_sentence_62",
"Russian language in Ukraine_sentence_149"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2867,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "4.382",
"value": "undefined"
},
{
"operation": "search",
"time": "14.191",
"value": "Contains...students in Ukraine have always spoken Russian"
},
{
"operation": "Now on",
"time": "18.095",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "46.532",
"value": "Multilingualism"
},
{
"operation": "Now on",
"time": "48.12",
"value": "Multilingualism"
},
{
"operation": "Page search",
"time": "61.883",
"value": "Russian"
},
{
"operation": "Highlighting",
"time": "71.9",
"value": "Multilingualism_sentence_31"
},
{
"operation": "Highlighting",
"time": "71.903",
"value": "Multilingualism_title"
},
{
"operation": "Highlighting",
"time": "71.904",
"value": "Multilingualism_section_0"
},
{
"operation": "Highlighting deleted",
"time": "73.457",
"value": "Multilingualism_sentence_31"
},
{
"operation": "Highlighting deleted",
"time": "73.46",
"value": "Multilingualism_title"
},
{
"operation": "Highlighting deleted",
"time": "73.463",
"value": "Multilingualism_section_0"
},
{
"operation": "Highlighting",
"time": "74.937",
"value": "Multilingualism_sentence_28"
},
{
"operation": "Highlighting",
"time": "74.939",
"value": "Multilingualism_title"
},
{
"operation": "Highlighting",
"time": "74.94",
"value": "Multilingualism_section_0"
},
{
"operation": "Highlighting deleted",
"time": "78.506",
"value": "Multilingualism_sentence_28"
},
{
"operation": "Highlighting deleted",
"time": "78.51",
"value": "Multilingualism_title"
},
{
"operation": "Highlighting deleted",
"time": "78.512",
"value": "Multilingualism_section_0"
},
{
"operation": "Highlighting",
"time": "123.691",
"value": "Multilingualism_sentence_209"
},
{
"operation": "Highlighting",
"time": "123.692",
"value": "Multilingualism_title"
},
{
"operation": "Highlighting",
"time": "123.693",
"value": "Multilingualism_section_21"
},
{
"operation": "Highlighting deleted",
"time": "124.963",
"value": "Multilingualism_sentence_209"
},
{
"operation": "Highlighting deleted",
"time": "124.965",
"value": "Multilingualism_title"
},
{
"operation": "Highlighting deleted",
"time": "124.969",
"value": "Multilingualism_section_21"
},
{
"operation": "Highlighting",
"time": "126.858",
"value": "Multilingualism_sentence_210"
},
{
"operation": "Highlighting",
"time": "126.86",
"value": "Multilingualism_title"
},
{
"operation": "Highlighting",
"time": "126.861",
"value": "Multilingualism_section_21"
},
{
"operation": "Highlighting deleted",
"time": "131.051",
"value": "Multilingualism_sentence_210"
},
{
"operation": "Highlighting deleted",
"time": "131.054",
"value": "Multilingualism_title"
},
{
"operation": "Highlighting deleted",
"time": "131.056",
"value": "Multilingualism_section_21"
},
{
"operation": "Highlighting",
"time": "142.515",
"value": "Multilingualism_sentence_219"
},
{
"operation": "Highlighting",
"time": "142.517",
"value": "Multilingualism_title"
},
{
"operation": "Highlighting",
"time": "142.518",
"value": "Multilingualism_section_22"
},
{
"operation": "Highlighting deleted",
"time": "146.099",
"value": "Multilingualism_sentence_219"
},
{
"operation": "Highlighting deleted",
"time": "146.102",
"value": "Multilingualism_title"
},
{
"operation": "Highlighting deleted",
"time": "146.105",
"value": "Multilingualism_section_22"
},
{
"operation": "hyperlink",
"time": "183.504",
"value": "Cultural diversity"
},
{
"operation": "Now on",
"time": "184.295",
"value": "Cultural diversity"
},
{
"operation": "Page search",
"time": "190.244",
"value": "Russian"
},
{
"operation": "search",
"time": "221.729",
"value": "List of multilingual countries and regions"
},
{
"operation": "Now on",
"time": "223.726",
"value": "List of multilingual countries and regions"
},
{
"operation": "Page search",
"time": "242.496",
"value": "students"
},
{
"operation": "Page search",
"time": "274.765",
"value": "Russian"
},
{
"operation": "search",
"time": "318.998",
"value": "Language in Ukraine"
},
{
"operation": "Now on",
"time": "323.518",
"value": "Language in Ukraine"
},
{
"operation": "Page search",
"time": "332.088",
"value": "students"
},
{
"operation": "hyperlink",
"time": "346.103",
"value": "Russian language in Ukraine"
},
{
"operation": "Now on",
"time": "347.617",
"value": "Russian language in Ukraine"
},
{
"operation": "Page search",
"time": "353.435",
"value": "students"
},
{
"operation": "Highlighting",
"time": "384.846",
"value": "Russian language in Ukraine_sentence_149"
},
{
"operation": "Page search",
"time": "414.549",
"value": "majority"
},
{
"operation": "hyperlink",
"time": "419.27",
"value": "2 Current usage statistics Russian language in Ukraine_section_5"
},
{
"operation": "Highlighting",
"time": "436.142",
"value": "Russian language in Ukraine_sentence_55"
},
{
"operation": "Highlighting",
"time": "449.758",
"value": "Russian language in Ukraine_sentence_62"
},
{
"operation": "hyperlink",
"time": "508.736",
"value": "Higher education"
},
{
"operation": "Now on",
"time": "510.019",
"value": "Higher education in Ukraine"
},
{
"operation": "Page search",
"time": "516.43",
"value": "majority"
},
{
"operation": "Page search",
"time": "546.347",
"value": "Russian"
},
{
"operation": "search",
"time": "598.95",
"value": "Contains...Multilingual Education in Ukraine"
},
{
"operation": "Now on",
"time": "613.152",
"value": "undefined"
},
{
"operation": "search",
"time": "635.721",
"value": "Russian language in Ukraine"
},
{
"operation": "Now on",
"time": "637.037",
"value": "Russian language in Ukraine"
},
{
"operation": "Page search",
"time": "647.049",
"value": "students"
},
{
"operation": "Highlighting deleted",
"time": "657.863",
"value": "Russian language in Ukraine_sentence_149"
},
{
"operation": "Highlighting",
"time": "659.117",
"value": "Russian language in Ukraine_sentence_149"
},
{
"operation": "finish",
"time": "673.069",
"value": "finish"
}
]
}
|
exfever_4495
|
Pixar is a British computer animation company with funding from the American entrepreneur and co-founder of Apple.
|
{
"evidence": [
"Pixar",
"Steve Jobs"
],
"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:\nPixar is a British computer animation company with funding from the American entrepreneur and co-founder of Apple.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Pixar",
"Steve Jobs"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2868,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5574
|
Harold Ramis was an American actor, comedian, director, and writer who directed a 1998 American mafia comedy film, for which he co-wrote the screenplay with playwright Kenneth Lonergan and Peter Tolan.
|
{
"evidence": [
"Harold Ramis",
"Analyze This"
],
"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:\nHarold Ramis was an American actor, comedian, director, and writer who directed a 1998 American mafia comedy film, for which he co-wrote the screenplay with playwright Kenneth Lonergan and Peter Tolan.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Harold Ramis",
"Analyze This"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2869,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
12152
|
O Deza is a comarca in the northwest corner of Galician which covers an area of 1,024.7 sq.km and composed of 6 municipalities.
|
{
"evidence": [
"O Deza_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:\nO Deza is a comarca in the northwest corner of Galician which covers an area of 1,024.7 sq.km and composed of 6 municipalities.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"O Deza_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2870,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.602",
"value": "undefined"
},
{
"operation": "Now on",
"time": "10.543",
"value": "undefined"
},
{
"operation": "search",
"time": "24.899",
"value": "O Deza"
},
{
"operation": "Now on",
"time": "25.687",
"value": "O Deza"
},
{
"operation": "Highlighting",
"time": "63.398",
"value": "O Deza_sentence_0"
},
{
"operation": "finish",
"time": "87.323",
"value": "finish"
}
]
}
|
exfever_1982
|
Guinevere betrayed the king and was sentenced to die.
|
{
"evidence": [
"Guinevere",
"Great Britain",
"Island"
],
"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:\nGuinevere betrayed the king and was sentenced to die.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Guinevere",
"Great Britain",
"Island"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2871,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5138
|
Rashida Jones appeared as Karen Filippelli in an Australian mockumentary sitcom television series that aired on an English-language commercial broadcast television and radio network.
|
{
"evidence": [
"Rashida Jones",
"The Office (U.S. TV series)",
"NBC"
],
"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:\nRashida Jones appeared as Karen Filippelli in an Australian mockumentary sitcom television series that aired on an English-language commercial broadcast television and radio network.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Rashida Jones",
"The Office (U.S. TV series)",
"NBC"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2872,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
26052
|
Zhakyp Kozhamberdi started playing on June 14, 2013 for Taraz and appeared in all of the games for that season.
|
{
"evidence": [
"Zhakyp Kozhamberdi_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:\nZhakyp Kozhamberdi started playing on June 14, 2013 for Taraz and appeared in all of the games for that season.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Zhakyp Kozhamberdi_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2873,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.625",
"value": "?search="
},
{
"operation": "search",
"time": "12.303",
"value": "Zhakyp Kozhamberdi"
},
{
"operation": "Now on",
"time": "13.395",
"value": "Zhakyp Kozhamberdi"
},
{
"operation": "Highlighting",
"time": "34.742",
"value": "Zhakyp Kozhamberdi_sentence_1"
},
{
"operation": "Page search",
"time": "60.016",
"value": "all"
},
{
"operation": "Page search",
"time": "60.194",
"value": "all"
},
{
"operation": "Page search",
"time": "60.437",
"value": "all"
},
{
"operation": "Page search",
"time": "60.652",
"value": "all"
},
{
"operation": "finish",
"time": "221.999",
"value": "finish"
}
]
}
|
53553
|
Nate Prosser was born on May 7, 1986, and plays hockey for the Philadelphia Flyers and the Lehigh Valley Phantoms.
|
{
"evidence": [
"Nate Prosser_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:\nNate Prosser was born on May 7, 1986, and plays hockey for the Philadelphia Flyers and the Lehigh Valley Phantoms.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Nate Prosser_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2874,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.234",
"value": "?search="
},
{
"operation": "search",
"time": "39.083",
"value": "Nate Prosser"
},
{
"operation": "Now on",
"time": "40.676",
"value": "Nate Prosser"
},
{
"operation": "Highlighting",
"time": "48.526",
"value": "Nate Prosser_sentence_0"
},
{
"operation": "finish",
"time": "85.263",
"value": "finish"
}
]
}
|
exfever_7214
|
Teen Wolf is an Australian supernatural teen drama television series based on the 1985 film of the same name, a coming-of-age romantic fantasy comedy film directed by Rod Daniel.
|
{
"evidence": [
"Teen Wolf (2011 TV series)",
"Teen Wolf"
],
"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:\nTeen Wolf is an Australian supernatural teen drama television series based on the 1985 film of the same name, a coming-of-age romantic fantasy comedy film directed by Rod Daniel.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Teen Wolf (2011 TV series)",
"Teen Wolf"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2875,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
38889
|
During the 1953 Indianapolis championship, Alberto Ascari had more points than Luigi Villoresi and Art Cross.
|
{
"evidence": [
"1953 Formula One season_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:\nDuring the 1953 Indianapolis championship, Alberto Ascari had more points than Luigi Villoresi and Art Cross.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"1953 Formula One season_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2876,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.634",
"value": "?search="
},
{
"operation": "search",
"time": "22.88",
"value": "Contains...1953 Indianapolis championship"
},
{
"operation": "Now on",
"time": "24.074",
"value": "undefined"
},
{
"operation": "search",
"time": "48.361",
"value": "Alberto Ascari"
},
{
"operation": "Now on",
"time": "49.705",
"value": "Alberto Ascari"
},
{
"operation": "Page search",
"time": "65.395",
"value": "1953"
},
{
"operation": "Page search",
"time": "118.375",
"value": "alberto"
},
{
"operation": "Page search",
"time": "126.697",
"value": "luigi"
},
{
"operation": "Page search",
"time": "193.485",
"value": "indianapolis"
},
{
"operation": "Page search",
"time": "249.999",
"value": "cross"
},
{
"operation": "Page search",
"time": "277.855",
"value": "1953 500"
},
{
"operation": "Page search",
"time": "312.332",
"value": "Indianapolis"
},
{
"operation": "hyperlink",
"time": "315.59",
"value": "Indianapolis 500"
},
{
"operation": "Now on",
"time": "316.5",
"value": "Indianapolis 500"
},
{
"operation": "Page search",
"time": "324.157",
"value": "1953"
},
{
"operation": "hyperlink",
"time": "332.697",
"value": "Indianapolis Motor Speedway Radio Network"
},
{
"operation": "Now on",
"time": "333.522",
"value": "Indianapolis Motor Speedway Radio Network"
},
{
"operation": "back-button-clicked",
"time": "337.647",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "337.807",
"value": "Indianapolis 500"
},
{
"operation": "Page search",
"time": "343.403",
"value": "1953"
},
{
"operation": "search",
"time": "483.786",
"value": "Alberto Ascari"
},
{
"operation": "Now on",
"time": "484.634",
"value": "Alberto Ascari"
},
{
"operation": "hyperlink",
"time": "489.661",
"value": "1953"
},
{
"operation": "Now on",
"time": "490.75",
"value": "1953 Formula One season"
},
{
"operation": "Page search",
"time": "496.114",
"value": "lugi"
},
{
"operation": "Page search",
"time": "499.208",
"value": "luigi"
},
{
"operation": "Page search",
"time": "547.815",
"value": "ascari"
},
{
"operation": "Highlighting",
"time": "556.645",
"value": "1953 Formula One season_sentence_3"
},
{
"operation": "finish",
"time": "564.854",
"value": "finish"
}
]
}
|
exfever_4650
|
William McKinley was the 25th president of the United States from 1897 until his death in 1901, marking him the third president to be murdered, following the 16th president of the United States in 1865.
|
{
"evidence": [
"William McKinley",
"Assassination of William McKinley",
"Abraham Lincoln"
],
"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:\nWilliam McKinley was the 25th president of the United States from 1897 until his death in 1901, marking him the third president to be murdered, following the 16th president of the United States in 1865.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"William McKinley",
"Assassination of William McKinley",
"Abraham Lincoln"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2877,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
72720
|
Dai pi dong is a kind of traditional food stall in Hong Kong, a metropolitan area and special administrative region of the People's Republic of China.
|
{
"evidence": [
"Hong Kong_sentence_2",
"Rejuvenation of dai pai dong_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:\nDai pi dong is a kind of traditional food stall in Hong Kong, a metropolitan area and special administrative region of the People's Republic of China.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Hong Kong_sentence_2",
"Rejuvenation of dai pai dong_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 2878,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.5",
"value": "undefined"
},
{
"operation": "search",
"time": "21.053",
"value": "Dai pi"
},
{
"operation": "Now on",
"time": "22.457",
"value": "undefined"
},
{
"operation": "search",
"time": "33.141",
"value": "Dai pi dong"
},
{
"operation": "Now on",
"time": "34.634",
"value": "undefined"
},
{
"operation": "search",
"time": "54.372",
"value": "Dai pi dong food stall"
},
{
"operation": "Now on",
"time": "55.145",
"value": "undefined"
},
{
"operation": "search",
"time": "74.147",
"value": "traditional food stall in Hong Kong"
},
{
"operation": "Now on",
"time": "75.838",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "88.539",
"value": "Hong Kong street food"
},
{
"operation": "Now on",
"time": "89.374",
"value": "Hong Kong street food"
},
{
"operation": "Page search",
"time": "95.995",
"value": "Dai pi dong"
},
{
"operation": "hyperlink",
"time": "117.69",
"value": "Mobile stalls in Hong Kong"
},
{
"operation": "Now on",
"time": "118.093",
"value": "Mobile stalls in Hong Kong"
},
{
"operation": "Page search",
"time": "129.257",
"value": "dai pi dong"
},
{
"operation": "hyperlink",
"time": "143.616",
"value": "Hong Kong"
},
{
"operation": "Now on",
"time": "148.219",
"value": "Hong Kong"
},
{
"operation": "Highlighting",
"time": "159.009",
"value": "Hong Kong_sentence_2"
},
{
"operation": "search",
"time": "187.863",
"value": "Dai pi dong"
},
{
"operation": "Now on",
"time": "188.53",
"value": "undefined"
},
{
"operation": "search",
"time": "214.134",
"value": "\"dai pi dong\""
},
{
"operation": "Now on",
"time": "214.372",
"value": "undefined"
},
{
"operation": "search",
"time": "228.637",
"value": "daipi dong"
},
{
"operation": "Now on",
"time": "229.442",
"value": "undefined"
},
{
"operation": "search",
"time": "240.997",
"value": "food stall in hong kong"
},
{
"operation": "Now on",
"time": "241.798",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "246.957",
"value": "Rejuvenation of dai pai dong"
},
{
"operation": "Now on",
"time": "247.47",
"value": "Rejuvenation of dai pai dong"
},
{
"operation": "Highlighting",
"time": "252.653",
"value": "Rejuvenation of dai pai dong_sentence_0"
},
{
"operation": "finish",
"time": "261.676",
"value": "finish"
}
]
}
|
60438
|
Dutch leaders Louis and Henry of Nassau and their older brother William the Silent, Prince of Orange all survived in the Battle of Mookerheyde after winning convincingly against the Spanish forces.
|
{
"evidence": [
"Battle of Mookerheyde_sentence_0",
"Battle of Mookerheyde_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:\nDutch leaders Louis and Henry of Nassau and their older brother William the Silent, Prince of Orange all survived in the Battle of Mookerheyde after winning convincingly against the Spanish forces.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Battle of Mookerheyde_sentence_0",
"Battle of Mookerheyde_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2879,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.402",
"value": "?search="
},
{
"operation": "search",
"time": "27.82",
"value": "Battle of Mookerheyde"
},
{
"operation": "Now on",
"time": "28.609",
"value": "Battle of Mookerheyde"
},
{
"operation": "Highlighting",
"time": "38.76",
"value": "Battle of Mookerheyde_sentence_0"
},
{
"operation": "Highlighting",
"time": "56.978",
"value": "Battle of Mookerheyde_sentence_1"
},
{
"operation": "finish",
"time": "91.115",
"value": "finish"
}
]
}
|
exfever_7851
|
Denial is a 2016 biographical film directed by Mick Jackson that dramatizes a case in English law against American historian Deborah Lipstadt and her publisher Penguin Books, filed in the High Court of Justice by the British author David Irving in 2000.
|
{
"evidence": [
"Denial (2016 film)",
"Irving v Penguin Books Ltd"
],
"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:\nDenial is a 2016 biographical film directed by Mick Jackson that dramatizes a case in English law against American historian Deborah Lipstadt and her publisher Penguin Books, filed in the High Court of Justice by the British author David Irving in 2000.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Denial (2016 film)",
"Irving v Penguin Books Ltd"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2880,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5321
|
The New York Giants were the first team to have 29 Hall of Fame players.
|
{
"evidence": [
"New York Giants",
"1925 NFL season"
],
"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 New York Giants were the first team to have 29 Hall of Fame players.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"New York Giants",
"1925 NFL season"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 2881,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
40402
|
After being named Secretary of the Republican Party of Texas (which was founded in 1867) in 2010, Tom Mechler resigned from the party.
|
{
"evidence": [
"Tom Mechler_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:\nAfter being named Secretary of the Republican Party of Texas (which was founded in 1867) in 2010, Tom Mechler resigned from the party.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Tom Mechler_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2882,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.718",
"value": "?search="
},
{
"operation": "search",
"time": "18.058",
"value": "Tom Mechler"
},
{
"operation": "Now on",
"time": "18.755",
"value": "Tom Mechler"
},
{
"operation": "Page search",
"time": "54.815",
"value": "secretary"
},
{
"operation": "search",
"time": "117.552",
"value": "Contains...Secretary of the Republican Party of Texas"
},
{
"operation": "Now on",
"time": "120.448",
"value": "undefined"
},
{
"operation": "search",
"time": "150.675",
"value": "Republican Party of Texas"
},
{
"operation": "Now on",
"time": "151.491",
"value": "Republican Party of Texas"
},
{
"operation": "Page search",
"time": "168.341",
"value": "secretary"
},
{
"operation": "back-button-clicked",
"time": "189.908",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "190.726",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "192.948",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "192.953",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "193.054",
"value": "Tom Mechler"
},
{
"operation": "Highlighting",
"time": "234.806",
"value": "Tom Mechler_sentence_19"
},
{
"operation": "Highlighting deleted",
"time": "239.213",
"value": "Tom Mechler_sentence_19"
},
{
"operation": "Highlighting",
"time": "244.195",
"value": "Tom Mechler_sentence_19"
},
{
"operation": "Highlighting",
"time": "271.498",
"value": "Tom Mechler_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "275.373",
"value": "Tom Mechler_sentence_19"
},
{
"operation": "finish",
"time": "296.003",
"value": "finish"
}
]
}
|
92642
|
Sphedanolestes cingulatus are black and red but has weak femurs.
|
{
"evidence": [
"Sphedanolestes cingulatus_sentence_8"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSphedanolestes cingulatus are black and red but has weak femurs.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sphedanolestes cingulatus_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 2883,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.137",
"value": "?search="
},
{
"operation": "search",
"time": "9.956",
"value": "Sphedanolestes cingulatus"
},
{
"operation": "Now on",
"time": "10.794",
"value": "Sphedanolestes cingulatus"
},
{
"operation": "Highlighting",
"time": "27.006",
"value": "Sphedanolestes cingulatus_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "31.693",
"value": "Sphedanolestes cingulatus_sentence_7"
},
{
"operation": "Highlighting",
"time": "32.255",
"value": "Sphedanolestes cingulatus_sentence_8"
},
{
"operation": "finish",
"time": "38.81",
"value": "finish"
}
]
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.