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 |
|---|---|---|---|---|---|---|---|---|
32515
|
Ram Rai Pur is situated on the right bank of the Sutlej also known as Satadree which is one of the oldest villages in Punjab and consists of 4 large Sikh family clans.
|
{
"evidence": [
"Ram Rai Pur_sentence_2",
"Sutlej_sentence_2",
"Ram Rai Pur_sentence_3",
"Ram Rai Pur_sentence_4"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nRam Rai Pur is situated on the right bank of the Sutlej also known as Satadree which is one of the oldest villages in Punjab and consists of 4 large Sikh family clans.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Ram Rai Pur_sentence_2",
"Sutlej_sentence_2",
"Ram Rai Pur_sentence_3",
"Ram Rai Pur_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7984,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.991",
"value": "?search="
},
{
"operation": "search",
"time": "6.525",
"value": "Ram Rai Pur"
},
{
"operation": "Now on",
"time": "7.062",
"value": "Ram Rai Pur"
},
{
"operation": "Page search",
"time": "16.076",
"value": "Sutlej"
},
{
"operation": "Highlighting",
"time": "17.703",
"value": "Ram Rai Pur_sentence_2"
},
{
"operation": "hyperlink",
"time": "23.796",
"value": "Sutlej"
},
{
"operation": "Now on",
"time": "24.232",
"value": "Sutlej"
},
{
"operation": "Page search",
"time": "28.787",
"value": "Satadree"
},
{
"operation": "Highlighting",
"time": "30.411",
"value": "Sutlej_sentence_2"
},
{
"operation": "back-button-clicked",
"time": "31.642",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "31.708",
"value": "Ram Rai Pur"
},
{
"operation": "Highlighting",
"time": "44.816",
"value": "Ram Rai Pur_sentence_3"
},
{
"operation": "Highlighting",
"time": "55.655",
"value": "Ram Rai Pur_sentence_4"
},
{
"operation": "finish",
"time": "90.165",
"value": "finish"
}
]
}
|
78274
|
CrossLink International was founded in November 1996, it was inspired when a Baptist Sunday school class visited a hospital in Monaco which was established in 1147.
|
{
"evidence": [
"CrossLink International_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nCrossLink International was founded in November 1996, it was inspired when a Baptist Sunday school class visited a hospital in Monaco which was established in 1147.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"CrossLink International_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7985,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.828",
"value": "?search="
},
{
"operation": "search",
"time": "7.996",
"value": "CrossLink International"
},
{
"operation": "Now on",
"time": "8.817",
"value": "CrossLink International"
},
{
"operation": "Highlighting",
"time": "11.453",
"value": "CrossLink International_sentence_4"
},
{
"operation": "Highlighting",
"time": "18.313",
"value": "CrossLink International_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "38.148",
"value": "CrossLink International_sentence_4"
},
{
"operation": "finish",
"time": "42.602",
"value": "finish"
}
]
}
|
56088
|
In the 2020 Belgian Cup Final, the defending champions, Mechelen (2019) lost their title to the Antwerp football team.
|
{
"evidence": [
"2020 Belgian Cup Final_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nIn the 2020 Belgian Cup Final, the defending champions, Mechelen (2019) lost their title to the Antwerp football team.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"2020 Belgian Cup Final_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7986,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.12",
"value": "undefined"
},
{
"operation": "search",
"time": "27.599",
"value": "2020 Belgian Cup Final"
},
{
"operation": "Now on",
"time": "28.856",
"value": "2020 Belgian Cup Final"
},
{
"operation": "Highlighting",
"time": "38.604",
"value": "2020 Belgian Cup Final_sentence_5"
},
{
"operation": "finish",
"time": "57.925",
"value": "finish"
}
]
}
|
exfever_4529
|
Friends with Benefits was the highest-grossing film in 2011.
|
{
"evidence": [
"Friends with Benefits (film)",
"Justin Timberlake"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nFriends with Benefits was the highest-grossing film in 2011.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Friends with Benefits (film)",
"Justin Timberlake"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7987,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
44071
|
The courtside reporter in the 2015β16 Fox Sports North was neither Marney Gellner nor Anthony LaPanta; while the studio host was Tom Hanneman.
|
{
"evidence": [
"Tom Hanneman_sentence_5"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nThe courtside reporter in the 2015β16 Fox Sports North was neither Marney Gellner nor Anthony LaPanta; while the studio host was Tom Hanneman.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Tom Hanneman_sentence_5"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7988,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.676",
"value": "?search="
},
{
"operation": "search",
"time": "35.482",
"value": "2015β16 Fox Sports North"
},
{
"operation": "Now on",
"time": "38.639",
"value": "undefined"
},
{
"operation": "search",
"time": "55.65",
"value": "courtside reporter in the 2015β16 Fox Sports North"
},
{
"operation": "Now on",
"time": "57.546",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "71.676",
"value": "Sports commentator"
},
{
"operation": "Now on",
"time": "72.767",
"value": "Sports commentator"
},
{
"operation": "Page search",
"time": "79.091",
"value": "fox"
},
{
"operation": "search",
"time": "123.538",
"value": "Marney Gellner"
},
{
"operation": "Now on",
"time": "124.97",
"value": "undefined"
},
{
"operation": "search",
"time": "171.075",
"value": "\"Marney Gellner\" \"fox\" \"Anthony LaPanta\""
},
{
"operation": "Now on",
"time": "171.953",
"value": "undefined"
},
{
"operation": "search",
"time": "186.423",
"value": "Anthony LaPanta"
},
{
"operation": "Now on",
"time": "187.861",
"value": "undefined"
},
{
"operation": "search",
"time": "208.154",
"value": "\"Anthony LaPanta\" \"fox\""
},
{
"operation": "Now on",
"time": "208.829",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "211.758",
"value": "List of Minnesota Wild broadcasters"
},
{
"operation": "Now on",
"time": "212.79",
"value": "List of Minnesota Wild broadcasters"
},
{
"operation": "Page search",
"time": "223.034",
"value": "LaPanta"
},
{
"operation": "search",
"time": "267.261",
"value": "Tom Hanneman"
},
{
"operation": "Now on",
"time": "268.567",
"value": "Tom Hanneman"
},
{
"operation": "Page search",
"time": "283.668",
"value": "fox"
},
{
"operation": "search",
"time": "339.944",
"value": "2015β16 Fox Sports North"
},
{
"operation": "Now on",
"time": "341.521",
"value": "undefined"
},
{
"operation": "search",
"time": "387.982",
"value": "\"2015β16 Fox Sports North\" \"gellner\""
},
{
"operation": "Now on",
"time": "388.675",
"value": "undefined"
},
{
"operation": "search",
"time": "411.931",
"value": "Fox Sports North 2015-16"
},
{
"operation": "Now on",
"time": "413.292",
"value": "undefined"
},
{
"operation": "search",
"time": "450.198",
"value": "Marney Gellner"
},
{
"operation": "Now on",
"time": "450.904",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "454.896",
"value": "Minnesota Lynx"
},
{
"operation": "Now on",
"time": "457.86",
"value": "Minnesota Lynx"
},
{
"operation": "Page search",
"time": "461.805",
"value": "Marney Gellner"
},
{
"operation": "back-button-clicked",
"time": "501.161",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "502.21",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "524.601",
"value": "Tom Hanneman"
},
{
"operation": "Now on",
"time": "524.967",
"value": "Tom Hanneman"
},
{
"operation": "Page search",
"time": "530.127",
"value": "Marney Gellner"
},
{
"operation": "Highlighting",
"time": "535.382",
"value": "Tom Hanneman_sentence_5"
},
{
"operation": "Page search",
"time": "602.713",
"value": "lapanta"
},
{
"operation": "search",
"time": "647.878",
"value": "\"Anthony LaPanta\" \"Hanneman\""
},
{
"operation": "Now on",
"time": "649.238",
"value": "undefined"
},
{
"operation": "search",
"time": "672.274",
"value": "\"Anthony LaPanta\" \"fox\""
},
{
"operation": "Now on",
"time": "672.953",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "675.997",
"value": "List of Minnesota Wild broadcasters"
},
{
"operation": "Now on",
"time": "676.671",
"value": "List of Minnesota Wild broadcasters"
},
{
"operation": "Page search",
"time": "683.337",
"value": "lapanta"
},
{
"operation": "finish",
"time": "709.952",
"value": "finish"
}
]
}
|
42761
|
Trumaine Sykes is a professional American Football defensive end who signed as an undrafted free agent meaning a player is eligible to sign with other clubs or franchises and is not under contract to any specific team.
|
{
"evidence": [
"Trumaine Sykes_sentence_0",
"Trumaine Sykes_sentence_1",
"Free agent_sentence_1"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nTrumaine Sykes is a professional American Football defensive end who signed as an undrafted free agent meaning a player is eligible to sign with other clubs or franchises and is not under contract to any specific team.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Trumaine Sykes_sentence_0",
"Trumaine Sykes_sentence_1",
"Free agent_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7989,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.74",
"value": "?search="
},
{
"operation": "search",
"time": "27.863",
"value": "Trumaine Sykes"
},
{
"operation": "Now on",
"time": "28.668",
"value": "Trumaine Sykes"
},
{
"operation": "Highlighting",
"time": "33.905",
"value": "Trumaine Sykes_sentence_0"
},
{
"operation": "Highlighting",
"time": "46.032",
"value": "Trumaine Sykes_sentence_1"
},
{
"operation": "hyperlink",
"time": "49.866",
"value": "undrafted free agent"
},
{
"operation": "hyperlink",
"time": "53.633",
"value": "undrafted free agent"
},
{
"operation": "Now on",
"time": "57.896",
"value": "Free agent"
},
{
"operation": "Highlighting",
"time": "63.236",
"value": "Free agent_sentence_1"
},
{
"operation": "finish",
"time": "99.777",
"value": "finish"
}
]
}
|
exfever_4847
|
Bonnie Hunt is an American actress known for her voice work in the 2006 American computer-animated action-adventure buddy cop comedy film produced by Walt Disney Animation Studios.
|
{
"evidence": [
"Bonnie Hunt",
"Zootopia"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nBonnie Hunt is an American actress known for her voice work in the 2006 American computer-animated action-adventure buddy cop comedy film produced by Walt Disney Animation Studios.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Bonnie Hunt",
"Zootopia"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7990,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8566
|
Azerbaijan is a part of the South Caucasus region and is bounded to the east by the world's largest inland body of water.
|
{
"evidence": [
"Azerbaijan",
"Caspian Sea"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nAzerbaijan is a part of the South Caucasus region and is bounded to the east by the world's largest inland body of water.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Azerbaijan",
"Caspian Sea"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7991,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8786
|
Jared Leto is an American actor who then began focusing increasingly on music, returning to acting with a 2002 American thriller film directed by David Fincher.
|
{
"evidence": [
"Jared Leto",
"Panic Room"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nJared Leto is an American actor who then began focusing increasingly on music, returning to acting with a 2002 American thriller film directed by David Fincher.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jared Leto",
"Panic Room"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7992,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
32484
|
The Royal Marines is part of Her Majesty's Naval Service and has 99 Vikings which an amphibious armoured fighting vehicle .
|
{
"evidence": [
"BvS 10_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nThe Royal Marines is part of Her Majesty's Naval Service and has 99 Vikings which an amphibious armoured fighting vehicle .\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"BvS 10_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7993,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.016",
"value": "?search="
},
{
"operation": "search",
"time": "22.868",
"value": "list of royal marine equipment"
},
{
"operation": "Now on",
"time": "25.157",
"value": "undefined"
},
{
"operation": "search",
"time": "33.784",
"value": "Royal Marines"
},
{
"operation": "Now on",
"time": "34.572",
"value": "Royal Marines"
},
{
"operation": "Highlighting",
"time": "55.01",
"value": "Royal Marines_sentence_4"
},
{
"operation": "Highlighting",
"time": "55.012",
"value": "Royal Marines_title"
},
{
"operation": "hyperlink",
"time": "55.777",
"value": "Royal Navy"
},
{
"operation": "Now on",
"time": "57.159",
"value": "Royal Navy"
},
{
"operation": "back-button-clicked",
"time": "61.728",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "61.924",
"value": "Royal Marines"
},
{
"operation": "hyperlink",
"time": "72.935",
"value": "10 See also Royal Marines_section_25"
},
{
"operation": "hyperlink",
"time": "75.899",
"value": "List of active Royal Marines military watercraft"
},
{
"operation": "Now on",
"time": "76.733",
"value": "List of active Royal Marines military watercraft"
},
{
"operation": "search",
"time": "141.176",
"value": "Viking armoured vehicle"
},
{
"operation": "Now on",
"time": "141.962",
"value": "BvS 10"
},
{
"operation": "Highlighting",
"time": "179.099",
"value": "BvS 10_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "179.11",
"value": "BvS 10_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "179.111",
"value": "BvS 10_header_cell_0_0_0"
},
{
"operation": "hyperlink",
"time": "196.214",
"value": "1.1 Royal Marines BvS 10_section_1"
},
{
"operation": "Highlighting",
"time": "205.633",
"value": "BvS 10_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "211.468",
"value": "BvS 10_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "211.479",
"value": "BvS 10_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "211.481",
"value": "BvS 10_header_cell_0_0_0"
},
{
"operation": "finish",
"time": "274.716",
"value": "finish"
}
]
}
|
72853
|
According to a census in 2011, there were a total of 124 households in Sennavaram.
|
{
"evidence": [
"Sennavaram_sentence_4"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nAccording to a census in 2011, there were a total of 124 households in Sennavaram.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Sennavaram_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7994,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.479",
"value": "?search="
},
{
"operation": "search",
"time": "10.161",
"value": "Sennavaram"
},
{
"operation": "Now on",
"time": "11.25",
"value": "Sennavaram"
},
{
"operation": "Highlighting",
"time": "16.725",
"value": "Sennavaram_table_caption_1"
},
{
"operation": "Highlighting deleted",
"time": "19.635",
"value": "Sennavaram_table_caption_1"
},
{
"operation": "Highlighting",
"time": "20.634",
"value": "Sennavaram_sentence_4"
},
{
"operation": "finish",
"time": "34.092",
"value": "finish"
}
]
}
|
42948
|
Mandatory signs are subset of the regulatory sign group and must have at least 90 centimetres across rural roads, they are used to set the obligations of all traffic that uses a specific area of road and most road signs are circular.
|
{
"evidence": [
"Mandatory sign_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nMandatory signs are subset of the regulatory sign group and must have at least 90 centimetres across rural roads, they are used to set the obligations of all traffic that uses a specific area of road and most road signs are circular.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Mandatory sign_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7995,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "3.16",
"value": "undefined"
},
{
"operation": "search",
"time": "18.228",
"value": "Mandatory signs are subset of the regulatory sign group"
},
{
"operation": "Now on",
"time": "20.855",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "24.622",
"value": "Mandatory sign"
},
{
"operation": "Now on",
"time": "25.828",
"value": "Mandatory sign"
},
{
"operation": "Page search",
"time": "29.527",
"value": "90"
},
{
"operation": "Highlighting",
"time": "40.466",
"value": "Mandatory sign_sentence_5"
},
{
"operation": "finish",
"time": "62.715",
"value": "finish"
}
]
}
|
exfever_4595
|
42 is a 2013 American biographical sports film about an American professional baseball player who became the first African American to play in Major League Baseball in the modern era.
|
{
"evidence": [
"42 (film)",
"Jackie Robinson"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\n42 is a 2013 American biographical sports film about an American professional baseball player who became the first African American to play in Major League Baseball in the modern era.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"42 (film)",
"Jackie Robinson"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7996,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
76987
|
In 1911, Alfred Laubmann joined the Zoological Museum at Munich to work on the skins collected by Gottfried Merzbacher, and also worked under Carl Hellmayr (an Austrian ornithologist).
|
{
"evidence": [
"Alfred Laubmann_sentence_7",
"Alfred Laubmann_sentence_8",
"Carl Eduard Hellmayr_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nIn 1911, Alfred Laubmann joined the Zoological Museum at Munich to work on the skins collected by Gottfried Merzbacher, and also worked under Carl Hellmayr (an Austrian ornithologist).\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Alfred Laubmann_sentence_7",
"Alfred Laubmann_sentence_8",
"Carl Eduard Hellmayr_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7997,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.002",
"value": "?search="
},
{
"operation": "search",
"time": "17.457",
"value": "Alfred Laubmann"
},
{
"operation": "Now on",
"time": "17.984",
"value": "Alfred Laubmann"
},
{
"operation": "Highlighting",
"time": "34.457",
"value": "Alfred Laubmann_sentence_7"
},
{
"operation": "Highlighting",
"time": "45.96",
"value": "Alfred Laubmann_sentence_8"
},
{
"operation": "hyperlink",
"time": "53.808",
"value": "Carl Hellmayr"
},
{
"operation": "Now on",
"time": "54.206",
"value": "Carl Eduard Hellmayr"
},
{
"operation": "Highlighting",
"time": "57.897",
"value": "Carl Eduard Hellmayr_sentence_0"
},
{
"operation": "finish",
"time": "154.01",
"value": "finish"
}
]
}
|
exfever_1314
|
Mauritius is an island nation in the Atlantic Ocean about 2,000 kilometres off the southeast coast of the African continent, east of the world's second-largest island country, after Indonesia.
|
{
"evidence": [
"Mauritius",
"Madagascar"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nMauritius is an island nation in the Atlantic Ocean about 2,000 kilometres off the southeast coast of the African continent, east of the world's second-largest island country, after Indonesia.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mauritius",
"Madagascar"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7998,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2493
|
Naga Chaitanya is an Indian actor who made his acting debut with a 2009 Indian Telugu-language action film directed by Mani Ratnam.
|
{
"evidence": [
"Naga Chaitanya",
"Josh (2009 Telugu film)"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nNaga Chaitanya is an Indian actor who made his acting debut with a 2009 Indian Telugu-language action film directed by Mani Ratnam.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Naga Chaitanya",
"Josh (2009 Telugu film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7999,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
55394
|
A pure land is the celestial realm of a buddha, and in the Lotus Sutra Buddha's close followers would also have their own pure lands.
|
{
"evidence": [
"Pure land_sentence_2",
"Pure land_sentence_8"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nA pure land is the celestial realm of a buddha, and in the Lotus Sutra Buddha's close followers would also have their own pure lands.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Pure land_sentence_2",
"Pure land_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8000,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.324",
"value": "?search="
},
{
"operation": "search",
"time": "17.706",
"value": "celestial realm of a buddha"
},
{
"operation": "Now on",
"time": "19.616",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "39.565",
"value": "Field of Merit Pure land_section_6"
},
{
"operation": "Now on",
"time": "40.252",
"value": "Pure land#Field of Merit Pure land section 6"
},
{
"operation": "Highlighting",
"time": "65.184",
"value": "Pure land_sentence_2"
},
{
"operation": "Highlighting",
"time": "82.695",
"value": "Pure land_sentence_8"
},
{
"operation": "finish",
"time": "93.565",
"value": "finish"
}
]
}
|
exfever_5389
|
Tale of Tales is a 2015 fantasy horror film starring an American actor, comedian, screenwriter, musician, and producer.
|
{
"evidence": [
"Tale of Tales (2015 film)",
"John C. Reilly"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nTale of Tales is a 2015 fantasy horror film starring an American actor, comedian, screenwriter, musician, and producer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tale of Tales (2015 film)",
"John C. Reilly"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8001,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
38673
|
LΓΈkken IF was originally two clubs known as IF Thor LΓΈkken AIL; then IF Thor changed federation to be part of Red Sport (founded by the Communist Party of Norway in 1931), Workers' Confederation of Sports, to create a new team called SK Fjell.
|
{
"evidence": [
"Kampforbundet for RΓΈd Sportsenhet_sentence_2",
"LΓΈkken IF_sentence_4",
"LΓΈkken IF_sentence_5",
"LΓΈkken IF_sentence_6",
"LΓΈkken IF_sentence_7"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nLΓΈkken IF was originally two clubs known as IF Thor LΓΈkken AIL; then IF Thor changed federation to be part of Red Sport (founded by the Communist Party of Norway in 1931), Workers' Confederation of Sports, to create a new team called SK Fjell.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Kampforbundet for RΓΈd Sportsenhet_sentence_2",
"LΓΈkken IF_sentence_4",
"LΓΈkken IF_sentence_5",
"LΓΈkken IF_sentence_6",
"LΓΈkken IF_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8002,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.471",
"value": "undefined"
},
{
"operation": "search",
"time": "47.569",
"value": "LΓΈkken IF"
},
{
"operation": "Now on",
"time": "48.292",
"value": "LΓΈkken IF"
},
{
"operation": "Page search",
"time": "52.878",
"value": "two"
},
{
"operation": "Highlighting",
"time": "55.657",
"value": "LΓΈkken IF_sentence_4"
},
{
"operation": "Highlighting",
"time": "66.753",
"value": "LΓΈkken IF_sentence_5"
},
{
"operation": "Highlighting",
"time": "87.149",
"value": "LΓΈkken IF_sentence_6"
},
{
"operation": "hyperlink",
"time": "91.698",
"value": "Red Sport"
},
{
"operation": "Now on",
"time": "92.275",
"value": "Kampforbundet for RΓΈd Sportsenhet"
},
{
"operation": "Highlighting",
"time": "94.32",
"value": "Kampforbundet for RΓΈd Sportsenhet_sentence_2"
},
{
"operation": "back-button-clicked",
"time": "95.923",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "96.011",
"value": "LΓΈkken IF"
},
{
"operation": "Highlighting",
"time": "103.534",
"value": "LΓΈkken IF_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "235.979",
"value": "LΓΈkken IF_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "236.829",
"value": "LΓΈkken IF_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "238.396",
"value": "LΓΈkken IF_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "244.19",
"value": "LΓΈkken IF_sentence_7"
},
{
"operation": "Highlighting",
"time": "246.626",
"value": "LΓΈkken IF_sentence_7"
},
{
"operation": "search",
"time": "252.708",
"value": "SK Fjell"
},
{
"operation": "Now on",
"time": "253.249",
"value": "LΓΈkken IF"
},
{
"operation": "Highlighting deleted",
"time": "276.575",
"value": "LΓΈkken IF_sentence_7"
},
{
"operation": "Highlighting",
"time": "280.63",
"value": "LΓΈkken IF_sentence_4"
},
{
"operation": "Highlighting",
"time": "281.384",
"value": "LΓΈkken IF_sentence_5"
},
{
"operation": "Highlighting",
"time": "282.072",
"value": "LΓΈkken IF_sentence_6"
},
{
"operation": "Highlighting",
"time": "283.066",
"value": "LΓΈkken IF_sentence_7"
},
{
"operation": "finish",
"time": "350.011",
"value": "finish"
}
]
}
|
exfever_1707
|
Pacific Blue is an American crime drama series that ran for five seasons on an American basic cable channel owned by the NBCUniversal Television.
|
{
"evidence": [
"Pacific Blue (TV series)",
"USA Network"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nPacific Blue is an American crime drama series that ran for five seasons on an American basic cable channel owned by the NBCUniversal Television.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Pacific Blue (TV series)",
"USA Network"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8003,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
77466
|
The Russian Fedora 11's final release was on June 9, 2009, and the Russian Fedora Remix 12's final release was on November 17, 2009.
|
{
"evidence": [
"Russian Fedora Remix_sentence_19",
"Russian Fedora Remix_sentence_23"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nThe Russian Fedora 11's final release was on June 9, 2009, and the Russian Fedora Remix 12's final release was on November 17, 2009.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Russian Fedora Remix_sentence_19",
"Russian Fedora Remix_sentence_23"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8004,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.226",
"value": "?search="
},
{
"operation": "search",
"time": "20.659",
"value": "Russian Fedora"
},
{
"operation": "Now on",
"time": "21.985",
"value": "Russian Fedora"
},
{
"operation": "hyperlink",
"time": "34.697",
"value": "2.5 Russian Fedora Remix 11 Russian Fedora Remix_section_6"
},
{
"operation": "Highlighting",
"time": "40.577",
"value": "Russian Fedora Remix_sentence_19"
},
{
"operation": "Highlighting",
"time": "43.676",
"value": "Russian Fedora Remix_sentence_23"
},
{
"operation": "finish",
"time": "55.81",
"value": "finish"
}
]
}
|
exfever_3891
|
Brian Douglas Wilson is an American songwriter who is often called a genius for his novel approaches to a genre with wide appeal that is typically distributed to large audiences through the music industry that originated in its modern form during the mid-1960s in the United States and the United Kingdom.
|
{
"evidence": [
"Brian Wilson",
"Pop music",
"Popular music"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nBrian Douglas Wilson is an American songwriter who is often called a genius for his novel approaches to a genre with wide appeal that is typically distributed to large audiences through the music industry that originated in its modern form during the mid-1960s in the United States and the United Kingdom.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Brian Wilson",
"Pop music",
"Popular music"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8005,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
39692
|
Pavle MladenoviΔ was a Serbian Chetnik Commander which was a revolutionary organization with the aim of liberation of Old Serbia from the Ottoman Empire.
|
{
"evidence": [
"Pavle MladenovicΜ_sentence_0",
"Serbian Chetnik Organization_sentence_1"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nPavle MladenoviΔ was a Serbian Chetnik Commander which was a revolutionary organization with the aim of liberation of Old Serbia from the Ottoman Empire.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Pavle MladenovicΜ_sentence_0",
"Serbian Chetnik Organization_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8006,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.883",
"value": "?search="
},
{
"operation": "search",
"time": "4.883",
"value": "Pavle MladenoviΔ"
},
{
"operation": "Now on",
"time": "5.598",
"value": "Pavle MladenovicΜ"
},
{
"operation": "Highlighting",
"time": "98.562",
"value": "Pavle MladenovicΜ_sentence_0"
},
{
"operation": "hyperlink",
"time": "115.917",
"value": "Serbian Chetnik commander"
},
{
"operation": "Now on",
"time": "117.2",
"value": "Serbian Chetnik Organization"
},
{
"operation": "Highlighting",
"time": "118.789",
"value": "Serbian Chetnik Organization_sentence_1"
},
{
"operation": "finish",
"time": "126.492",
"value": "finish"
}
]
}
|
67105
|
Orville Wright was appointed to NACA's board on January 29, 1920.
|
{
"evidence": [
"National Advisory Committee for Aeronautics_sentence_23"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nOrville Wright was appointed to NACA's board on January 29, 1920.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"National Advisory Committee for Aeronautics_sentence_23"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8007,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "4.42",
"value": "undefined"
},
{
"operation": "search",
"time": "7.15",
"value": "Orville Wright"
},
{
"operation": "Now on",
"time": "27.367",
"value": "undefined"
},
{
"operation": "search",
"time": "30.562",
"value": "Orville Wright"
},
{
"operation": "Now on",
"time": "32.177",
"value": "Wright brothers"
},
{
"operation": "search",
"time": "115.904",
"value": "\"NACA's board\""
},
{
"operation": "Now on",
"time": "117.976",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "120.589",
"value": "National Advisory Committee for Aeronautics"
},
{
"operation": "Now on",
"time": "121.093",
"value": "National Advisory Committee for Aeronautics"
},
{
"operation": "Page search",
"time": "128.857",
"value": "NACA's board"
},
{
"operation": "Highlighting",
"time": "132.114",
"value": "National Advisory Committee for Aeronautics_sentence_23"
},
{
"operation": "finish",
"time": "158.636",
"value": "finish"
}
]
}
|
exfever_3753
|
Mars is the next planet to cradle life after the third planet from the Sun.
|
{
"evidence": [
"South America",
"Southern Hemisphere",
"Earth"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nMars is the next planet to cradle life after the third planet from the Sun.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"South America",
"Southern Hemisphere",
"Earth"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8008,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
27859
|
Lodewijk van Bylandt(1718 β Hoeven, 28 December 1793) was a Dutch lieutenant-admiral who first served as a gunner then a captain of West Stellingwerff in 1756 and acting-commander of a Dutch naval expedition against Morocco.
|
{
"evidence": [
"Lodewijk van Bylandt_sentence_0",
"Lodewijk van Bylandt_sentence_10",
"Lodewijk van Bylandt_sentence_11"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nLodewijk van Bylandt(1718 β Hoeven, 28 December 1793) was a Dutch lieutenant-admiral who first served as a gunner then a captain of West Stellingwerff in 1756 and acting-commander of a Dutch naval expedition against Morocco.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Lodewijk van Bylandt_sentence_0",
"Lodewijk van Bylandt_sentence_10",
"Lodewijk van Bylandt_sentence_11"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8009,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.762",
"value": "undefined"
},
{
"operation": "search",
"time": "11.541",
"value": "Lodewijk van Bylandt(1718 β Hoeven, 28 December 1793)"
},
{
"operation": "Now on",
"time": "13.643",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "16.706",
"value": "Lodewijk van Bylandt"
},
{
"operation": "Now on",
"time": "17.281",
"value": "Lodewijk van Bylandt"
},
{
"operation": "Highlighting",
"time": "20.134",
"value": "Lodewijk van Bylandt_sentence_0"
},
{
"operation": "Page search",
"time": "33.076",
"value": "gunner"
},
{
"operation": "Page search",
"time": "58.95",
"value": "west"
},
{
"operation": "Highlighting",
"time": "67.767",
"value": "Lodewijk van Bylandt_sentence_10"
},
{
"operation": "Page search",
"time": "153.48",
"value": "morocco"
},
{
"operation": "Highlighting",
"time": "160.829",
"value": "Lodewijk van Bylandt_sentence_11"
},
{
"operation": "finish",
"time": "179.526",
"value": "finish"
}
]
}
|
81686
|
Heroes of Jin Yong first published in 1996, the player takes on the role of present-day protagonist and to return to modern society, he must find all the fourteen novels by Jin Yong whose family name is Cha (Zha)
|
{
"evidence": [
"Heroes of Jin Yong_sentence_0",
"Heroes of Jin Yong_sentence_1",
"Heroes of Jin Yong_sentence_2",
"Jin Yong_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nHeroes of Jin Yong first published in 1996, the player takes on the role of present-day protagonist and to return to modern society, he must find all the fourteen novels by Jin Yong whose family name is Cha (Zha)\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Heroes of Jin Yong_sentence_0",
"Heroes of Jin Yong_sentence_1",
"Heroes of Jin Yong_sentence_2",
"Jin Yong_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8010,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.241",
"value": "?search="
},
{
"operation": "search",
"time": "13.93",
"value": "Heroes of Jin Yong"
},
{
"operation": "Now on",
"time": "14.584",
"value": "Heroes of Jin Yong"
},
{
"operation": "Highlighting",
"time": "24.26",
"value": "Heroes of Jin Yong_sentence_0"
},
{
"operation": "Highlighting",
"time": "31.659",
"value": "Heroes of Jin Yong_sentence_1"
},
{
"operation": "Highlighting",
"time": "44.21",
"value": "Heroes of Jin Yong_sentence_2"
},
{
"operation": "search",
"time": "93.345",
"value": "Jin Yong"
},
{
"operation": "Now on",
"time": "94.434",
"value": "Jin Yong"
},
{
"operation": "Highlighting",
"time": "104.664",
"value": "Jin Yong_sentence_0"
},
{
"operation": "finish",
"time": "114.631",
"value": "finish"
}
]
}
|
exfever_759
|
Baylor University is one of the first educational institutions west of the second-largest drainage system on the North American continent, second only to a drainage system located in a large body of saltwater in northeastern Canada.
|
{
"evidence": [
"Baylor University",
"Mississippi River",
"Hudson Bay"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nBaylor University is one of the first educational institutions west of the second-largest drainage system on the North American continent, second only to a drainage system located in a large body of saltwater in northeastern Canada.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Baylor University",
"Mississippi River",
"Hudson Bay"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8011,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
4585
|
McCornack flew from California to Wisconsin in 1979 before being invited to fly an aircraft to the Atlantic coast on ethanol.
|
{
"evidence": [
"Pterodactyl Ascender_sentence_19",
"Pterodactyl Ascender_sentence_20"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nMcCornack flew from California to Wisconsin in 1979 before being invited to fly an aircraft to the Atlantic coast on ethanol.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Pterodactyl Ascender_sentence_19",
"Pterodactyl Ascender_sentence_20"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8012,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "17.537",
"value": "McCormack"
},
{
"operation": "search",
"time": "34",
"value": "pilot mccormack"
},
{
"operation": "search",
"time": "58.774",
"value": "McCornack"
},
{
"operation": "hyperlink",
"time": "71.44",
"value": "McCornack"
},
{
"operation": "back-button-clicked",
"time": "78.145",
"value": "back-button-clicked"
},
{
"operation": "Highlighting",
"time": "87.879",
"value": "Walter McCornack_sentence_0"
},
{
"operation": "Highlighting",
"time": "87.882",
"value": "Walter McCornack_title"
},
{
"operation": "Highlighting",
"time": "98.487",
"value": "Walter McCornack_sentence_9"
},
{
"operation": "Highlighting",
"time": "98.494",
"value": "Walter McCornack_section_0"
},
{
"operation": "search",
"time": "109.543",
"value": "pilot mccornack"
},
{
"operation": "hyperlink",
"time": "126.408",
"value": "Pterodactyl Ascender"
},
{
"operation": "Highlighting",
"time": "136.696",
"value": "Pterodactyl Ascender_header_cell_0_4_0"
},
{
"operation": "Highlighting",
"time": "137.746",
"value": "Pterodactyl Ascender_cell_0_4_1"
},
{
"operation": "Highlighting",
"time": "137.805",
"value": "Pterodactyl Ascender_header_cell_0_4_0"
},
{
"operation": "Highlighting",
"time": "137.821",
"value": "Pterodactyl Ascender_header_cell_0_0_0"
},
{
"operation": "Page search",
"time": "156.578",
"value": "california"
},
{
"operation": "Highlighting",
"time": "171.435",
"value": "Pterodactyl Ascender_sentence_19"
},
{
"operation": "Highlighting",
"time": "186.257",
"value": "Pterodactyl Ascender_sentence_20"
},
{
"operation": "Highlighting deleted",
"time": "205.803",
"value": "Pterodactyl Ascender_cell_0_4_1"
},
{
"operation": "Highlighting deleted",
"time": "205.854",
"value": "Pterodactyl Ascender_header_cell_0_4_0"
},
{
"operation": "Highlighting deleted",
"time": "205.875",
"value": "Pterodactyl Ascender_header_cell_0_0_0"
},
{
"operation": "finish",
"time": "242.112",
"value": "finish"
}
]
}
|
exfever_8592
|
Paul Bettany was an award-winning English actor for portraying Vision in the Marvel Cinematic Universe films Iron Man.
|
{
"evidence": [
"Paul Bettany",
"Avengers: Infinity War"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nPaul Bettany was an award-winning English actor for portraying Vision in the Marvel Cinematic Universe films Iron Man.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Paul Bettany",
"Avengers: Infinity War"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8013,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
83709
|
Despite cold War tensions Mikhail Chumakov, a Soviet microbiologist and virologist, traveled to the U.S. to study the Salk polio vaccine.
|
{
"evidence": [
"Mikhail Chumakov_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nDespite cold War tensions Mikhail Chumakov, a Soviet microbiologist and virologist, traveled to the U.S. to study the Salk polio vaccine.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Mikhail Chumakov_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8014,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.594",
"value": "undefined"
},
{
"operation": "search",
"time": "5.606",
"value": "Mikhail Chumakov,"
},
{
"operation": "Now on",
"time": "6.338",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "14.324",
"value": "Mikhail Chumakov"
},
{
"operation": "Now on",
"time": "14.847",
"value": "Mikhail Chumakov"
},
{
"operation": "Highlighting",
"time": "22.773",
"value": "Mikhail Chumakov_sentence_0"
},
{
"operation": "Page search",
"time": "31.741",
"value": "united states"
},
{
"operation": "Page search",
"time": "56.787",
"value": "travel"
},
{
"operation": "Page search",
"time": "62.235",
"value": "salk"
},
{
"operation": "Page search",
"time": "82.46",
"value": "polio"
},
{
"operation": "hyperlink",
"time": "132.534",
"value": "Polio Vaccine"
},
{
"operation": "Now on",
"time": "133.947",
"value": "Polio vaccine"
},
{
"operation": "Page search",
"time": "137.868",
"value": "Chumakov"
},
{
"operation": "search",
"time": "157.554",
"value": "Salk polio vaccine"
},
{
"operation": "Now on",
"time": "158.726",
"value": "Polio vaccine"
},
{
"operation": "Page search",
"time": "164.141",
"value": "Salk"
},
{
"operation": "hyperlink",
"time": "175.137",
"value": "4 History Polio vaccine_section_10"
},
{
"operation": "Page search",
"time": "217.225",
"value": "Chumakov"
},
{
"operation": "finish",
"time": "248.821",
"value": "finish"
}
]
}
|
34663
|
In January 2016, there are 264,980,280 grains of rice donated in Freerice (discontinued by British academic and Japanologist John Breen), which doubled the next month at 525,172,070.
|
{
"evidence": [
"Freerice_sentence_6"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nIn January 2016, there are 264,980,280 grains of rice donated in Freerice (discontinued by British academic and Japanologist John Breen), which doubled the next month at 525,172,070.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Freerice_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8015,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.134",
"value": "?search="
},
{
"operation": "search",
"time": "751.396",
"value": "rice in Freerice"
},
{
"operation": "Now on",
"time": "752.405",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "754.745",
"value": "Freerice"
},
{
"operation": "Now on",
"time": "755.278",
"value": "Freerice"
},
{
"operation": "Page search",
"time": "761.096",
"value": "rice"
},
{
"operation": "page-search-reset",
"time": "767.457",
"value": "page-search-reset"
},
{
"operation": "Highlighting",
"time": "798.17",
"value": "Freerice_cell_9_2_0"
},
{
"operation": "Highlighting",
"time": "798.172",
"value": "Freerice_section_7"
},
{
"operation": "Highlighting",
"time": "798.173",
"value": "Freerice_header_cell_9_1_0"
},
{
"operation": "Highlighting",
"time": "798.173",
"value": "Freerice_header_cell_9_0_0"
},
{
"operation": "Highlighting",
"time": "800.681",
"value": "Freerice_cell_9_2_1"
},
{
"operation": "Highlighting",
"time": "800.684",
"value": "Freerice_header_cell_9_1_1"
},
{
"operation": "Page search",
"time": "885.279",
"value": "John Breen"
},
{
"operation": "Page search",
"time": "928.118",
"value": "discontinued"
},
{
"operation": "search",
"time": "953.364",
"value": "John Breen"
},
{
"operation": "Now on",
"time": "954.06",
"value": "John Breen"
},
{
"operation": "hyperlink",
"time": "970.847",
"value": "John Breen (scholar)"
},
{
"operation": "Now on",
"time": "971.473",
"value": "John Breen (scholar)"
},
{
"operation": "Highlighting",
"time": "1035.9",
"value": "John Breen (scholar)_sentence_0"
},
{
"operation": "Highlighting",
"time": "1035.9",
"value": "John Breen (scholar)_title"
},
{
"operation": "Highlighting",
"time": "1036.308",
"value": "John Breen (scholar)_sentence_1"
},
{
"operation": "Highlighting",
"time": "1036.845",
"value": "John Breen (scholar)_sentence_2"
},
{
"operation": "Highlighting",
"time": "1037.384",
"value": "John Breen (scholar)_sentence_3"
},
{
"operation": "Highlighting",
"time": "1037.385",
"value": "John Breen (scholar)_section_0"
},
{
"operation": "Highlighting",
"time": "1037.84",
"value": "John Breen (scholar)_sentence_4"
},
{
"operation": "Highlighting",
"time": "1041.694",
"value": "John Breen (scholar)_sentence_5"
},
{
"operation": "Highlighting",
"time": "1041.695",
"value": "John Breen (scholar)_section_1"
},
{
"operation": "Highlighting",
"time": "1042.495",
"value": "John Breen (scholar)_sentence_6"
},
{
"operation": "Highlighting",
"time": "1043.483",
"value": "John Breen (scholar)_sentence_7"
},
{
"operation": "back-button-clicked",
"time": "1044.809",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "1044.852",
"value": "John Breen"
},
{
"operation": "back-button-clicked",
"time": "1046.245",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "1046.249",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "1046.392",
"value": "Freerice"
},
{
"operation": "Highlighting",
"time": "1075.357",
"value": "Freerice_sentence_6"
},
{
"operation": "finish",
"time": "1151.42",
"value": "finish"
}
]
}
|
exfever_1295
|
War of the Worlds is a 2005 American science fiction action film directed by an American film director, producer, and screenwriter, who began his career in the period in American film history from the mid-1960s to the early 1980s when a new generation of young filmmakers rose to prominence.
|
{
"evidence": [
"War of the Worlds (2005 film)",
"Steven Spielberg",
"New Hollywood"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nWar of the Worlds is a 2005 American science fiction action film directed by an American film director, producer, and screenwriter, who began his career in the period in American film history from the mid-1960s to the early 1980s when a new generation of young filmmakers rose to prominence.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"War of the Worlds (2005 film)",
"Steven Spielberg",
"New Hollywood"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8016,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
39275
|
Superoxide dismutase mimetics convert the superoxide anion, which is a reactive oxygen species, into hydrogen peroxide while being a prime interest in therapeutic treatment of oxidative stress.
|
{
"evidence": [
"Superoxide dismutase mimetics_sentence_1",
"Superoxide dismutase mimetics_sentence_3"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nSuperoxide dismutase mimetics convert the superoxide anion, which is a reactive oxygen species, into hydrogen peroxide while being a prime interest in therapeutic treatment of oxidative stress.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Superoxide dismutase mimetics_sentence_1",
"Superoxide dismutase mimetics_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8017,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.699",
"value": "?search="
},
{
"operation": "search",
"time": "21.018",
"value": "Superoxide dismutase"
},
{
"operation": "Now on",
"time": "22.134",
"value": "Superoxide dismutase"
},
{
"operation": "search",
"time": "26.121",
"value": "Superoxide dismutase mimetics"
},
{
"operation": "Now on",
"time": "26.872",
"value": "Superoxide dismutase mimetics"
},
{
"operation": "Highlighting",
"time": "41.104",
"value": "Superoxide dismutase mimetics_sentence_1"
},
{
"operation": "Highlighting",
"time": "77.541",
"value": "Superoxide dismutase mimetics_sentence_3"
},
{
"operation": "finish",
"time": "97.62",
"value": "finish"
}
]
}
|
exfever_2167
|
"Bohemian Rhapsody" is a song by the British rock band that consists of several sections including heavy rock, a loosely defined subgenre of rock music typified by the heavy use of aggressive vocals.
|
{
"evidence": [
"Bohemian Rhapsody",
"Hard rock"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\n\"Bohemian Rhapsody\" is a song by the British rock band that consists of several sections including heavy rock, a loosely defined subgenre of rock music typified by the heavy use of aggressive vocals.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Bohemian Rhapsody",
"Hard rock"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8018,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
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>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β 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": 8019,
"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"
}
]
}
|
exfever_1768
|
Usher Raymond IV was born in the tenth most populous city in the U.S. and the third-largest in Texas after Houston and San Antonio.
|
{
"evidence": [
"Usher (musician)",
"Dallas"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nUsher Raymond IV was born in the tenth most populous city in the U.S. and the third-largest in Texas after Houston and San Antonio.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Usher (musician)",
"Dallas"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8020,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8267
|
Ellen Burstyn earned a nomination for Best Actress at an Academy Award for her appearance in the 1973 American supernatural horror film also starring a Swedish-French actor who appeared in more than 150 films and several television series in many languages.
|
{
"evidence": [
"Ellen Burstyn",
"The Exorcist (film)",
"Max von Sydow"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nEllen Burstyn earned a nomination for Best Actress at an Academy Award for her appearance in the 1973 American supernatural horror film also starring a Swedish-French actor who appeared in more than 150 films and several television series in many languages.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ellen Burstyn",
"The Exorcist (film)",
"Max von Sydow"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8021,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
40280
|
Wenfeng District is located in China which is the world's most populous country.
|
{
"evidence": [
"Wenfeng District_sentence_0",
"China_sentence_3"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nWenfeng District is located in China which is the world's most populous country.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Wenfeng District_sentence_0",
"China_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8022,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.907",
"value": "?search="
},
{
"operation": "search",
"time": "9.561",
"value": "Wenfeng District"
},
{
"operation": "Now on",
"time": "10.384",
"value": "Wenfeng District"
},
{
"operation": "Highlighting",
"time": "22.977",
"value": "Wenfeng District_sentence_0"
},
{
"operation": "hyperlink",
"time": "25.215",
"value": "People's Republic of China"
},
{
"operation": "Now on",
"time": "26.654",
"value": "China"
},
{
"operation": "Highlighting",
"time": "29.502",
"value": "China_sentence_3"
},
{
"operation": "finish",
"time": "38.501",
"value": "finish"
}
]
}
|
exfever_268
|
Buddhism described evil as a supernatural force.
|
{
"evidence": [
"Evil",
"Natural evil"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nBuddhism described evil as a supernatural force.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Evil",
"Natural evil"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8023,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
88280
|
AeronΓ‘utica, which was an airline established in 2001 that provided consumer services for airlines however it has been banned from operating due to safety concerns.
|
{
"evidence": [
"AeronaΜutica (Angola)_sentence_1",
"AeronaΜutica (Angola)_sentence_2"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nAeronΓ‘utica, which was an airline established in 2001 that provided consumer services for airlines however it has been banned from operating due to safety concerns.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"AeronaΜutica (Angola)_sentence_1",
"AeronaΜutica (Angola)_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8024,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.918",
"value": "?search="
},
{
"operation": "search",
"time": "7.981",
"value": "AeronΓ‘utica"
},
{
"operation": "Now on",
"time": "8.823",
"value": "Aeronautica"
},
{
"operation": "search",
"time": "588.508",
"value": "ATA Airlines (Iran)"
},
{
"operation": "Now on",
"time": "589.479",
"value": "ATA Airlines (Iran)"
},
{
"operation": "search",
"time": "673.249",
"value": "AeronΓ‘utica"
},
{
"operation": "Now on",
"time": "674.203",
"value": "Aeronautica"
},
{
"operation": "hyperlink",
"time": "676.478",
"value": "AeronΓ‘utica (Angola)"
},
{
"operation": "Now on",
"time": "676.841",
"value": "AeronaΜutica (Angola)"
},
{
"operation": "Highlighting",
"time": "684.364",
"value": "AeronaΜutica (Angola)_sentence_1"
},
{
"operation": "Highlighting",
"time": "696.103",
"value": "AeronaΜutica (Angola)_sentence_2"
},
{
"operation": "Highlighting",
"time": "703.994",
"value": "AeronaΜutica (Angola)_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "708.945",
"value": "AeronaΜutica (Angola)_sentence_3"
},
{
"operation": "finish",
"time": "738.627",
"value": "finish"
}
]
}
|
62570
|
Victor Mansaray has never played soccer professionally.
|
{
"evidence": [
"Victor Mansaray_sentence_0"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nVictor Mansaray has never played soccer professionally.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Victor Mansaray_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8025,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.772",
"value": "?search="
},
{
"operation": "search",
"time": "25.876",
"value": "Victor Mansaray"
},
{
"operation": "Now on",
"time": "27.103",
"value": "Victor Mansaray"
},
{
"operation": "Highlighting",
"time": "30.041",
"value": "Victor Mansaray_sentence_0"
},
{
"operation": "finish",
"time": "46.318",
"value": "finish"
}
]
}
|
exfever_3963
|
Logan is the 7th film directed by James Mangold.
|
{
"evidence": [
"Logan (film)",
"Old Man Logan",
"Ongoing series"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nLogan is the 7th film directed by James Mangold.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Logan (film)",
"Old Man Logan",
"Ongoing series"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8026,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7112
|
Mischa Barton is an English model who starred in a 2015 film by the same team that wrote the first film adaptation of another of Green's novels that was adapted into a film in 2014.
|
{
"evidence": [
"Cara Delevingne",
"Paper Towns (film)",
"The Fault in Our Stars (film)"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nMischa Barton is an English model who starred in a 2015 film by the same team that wrote the first film adaptation of another of Green's novels that was adapted into a film in 2014.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Cara Delevingne",
"Paper Towns (film)",
"The Fault in Our Stars (film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8027,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3595
|
Sean Michael Spicer served as the White House Communications Director under the 2017 head of state and government of the United States of America.
|
{
"evidence": [
"Sean Spicer",
"President of the United States"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nSean Michael Spicer served as the White House Communications Director under the 2017 head of state and government of the United States of America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sean Spicer",
"President of the United States"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8028,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1657
|
Rachel Karen Green was the most famous character in the American sitcom Friends.
|
{
"evidence": [
"Rachel Green",
"Jennifer Aniston"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nRachel Karen Green was the most famous character in the American sitcom Friends.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Rachel Green",
"Jennifer Aniston"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8029,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1376
|
The Man in the Iron Mask is a 1998 Canadian drama film that starred Jeremy Irons as a fictional character in the novel The Three Musketeers in 1844.
|
{
"evidence": [
"The Man in the Iron Mask (1998 film)",
"Aramis"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nThe Man in the Iron Mask is a 1998 Canadian drama film that starred Jeremy Irons as a fictional character in the novel The Three Musketeers in 1844.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Man in the Iron Mask (1998 film)",
"Aramis"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8030,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
17106
|
Pratt was born in Houston, taught music and drama at Virginia and directed the Virginia Glee Club which was founded in 1871.
|
{
"evidence": [
"Harry Rogers Pratt_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nPratt was born in Houston, taught music and drama at Virginia and directed the Virginia Glee Club which was founded in 1871.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Harry Rogers Pratt_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8031,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.127",
"value": "?search="
},
{
"operation": "search",
"time": "92.02",
"value": "Harry Rogers Pratt"
},
{
"operation": "Now on",
"time": "92.922",
"value": "Harry Rogers Pratt"
},
{
"operation": "Highlighting",
"time": "95.846",
"value": "Harry Rogers Pratt_sentence_0"
},
{
"operation": "Highlighting",
"time": "98.865",
"value": "Harry Rogers Pratt_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "100.62",
"value": "Harry Rogers Pratt_sentence_0"
},
{
"operation": "finish",
"time": "115.349",
"value": "finish"
}
]
}
|
20979
|
The opera "Maria" by Polish composer Roman Statkowski (a Harvard University graduate) was performed in various Polish cities between 1919 and 1965.
|
{
"evidence": [
"Roman Statkowski_sentence_2",
"Maria (opera)_sentence_2",
"Maria (opera)_sentence_0"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nThe opera \"Maria\" by Polish composer Roman Statkowski (a Harvard University graduate) was performed in various Polish cities between 1919 and 1965.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Roman Statkowski_sentence_2",
"Maria (opera)_sentence_2",
"Maria (opera)_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8032,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.119",
"value": "?search="
},
{
"operation": "search",
"time": "107.027",
"value": "opera maria"
},
{
"operation": "Now on",
"time": "108.141",
"value": "undefined"
},
{
"operation": "search",
"time": "125.296",
"value": "Roman Statkowski"
},
{
"operation": "Now on",
"time": "125.852",
"value": "Roman Statkowski"
},
{
"operation": "Highlighting",
"time": "185.724",
"value": "Roman Statkowski_sentence_2"
},
{
"operation": "hyperlink",
"time": "196.93",
"value": "Maria"
},
{
"operation": "Now on",
"time": "197.66",
"value": "Maria (opera)"
},
{
"operation": "Highlighting",
"time": "259.7",
"value": "Maria (opera)_sentence_2"
},
{
"operation": "Highlighting",
"time": "263.901",
"value": "Maria (opera)_sentence_0"
},
{
"operation": "finish",
"time": "284.649",
"value": "finish"
}
]
}
|
exfever_6981
|
Cory Allan Michael Monteith (; May 11, 1982 β July 13, 2013) was a Canadian actor and musician best known for his role as a fictional character from the HBO musical comedy-drama series Glee.
|
{
"evidence": [
"Cory Monteith",
"Finn Hudson"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nCory Allan Michael Monteith (; May 11, 1982 β July 13, 2013) was a Canadian actor and musician best known for his role as a fictional character from the HBO musical comedy-drama series Glee.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Cory Monteith",
"Finn Hudson"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8033,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7420
|
Little Big Shots is a variety television series created and produced by an American comedian, television host, actress, writer, and producer.
|
{
"evidence": [
"Little Big Shots",
"Ellen DeGeneres"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nLittle Big Shots is a variety television series created and produced by an American comedian, television host, actress, writer, and producer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Little Big Shots",
"Ellen DeGeneres"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8034,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
78169
|
In 1628, Willem van Haecht destroyed the art collection owned by Cornelis van der Geest, aΒ spiceΒ merchant from Antwerp, who used his wealth to support the Antwerp artists and to grow his art collection.
|
{
"evidence": [
"Cornelis van der Geest_sentence_0",
"Willem van Haecht_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nIn 1628, Willem van Haecht destroyed the art collection owned by Cornelis van der Geest, aΒ spiceΒ merchant from Antwerp, who used his wealth to support the Antwerp artists and to grow his art collection.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Cornelis van der Geest_sentence_0",
"Willem van Haecht_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8035,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.89",
"value": "?search="
},
{
"operation": "search",
"time": "170.044",
"value": "Willem van Haecht"
},
{
"operation": "Now on",
"time": "170.925",
"value": "Willem van Haecht"
},
{
"operation": "search",
"time": "214.857",
"value": "Cornelis van der Geest"
},
{
"operation": "Now on",
"time": "215.885",
"value": "Cornelis van der Geest"
},
{
"operation": "Highlighting",
"time": "257.196",
"value": "Cornelis van der Geest_sentence_0"
},
{
"operation": "hyperlink",
"time": "293.083",
"value": "Willem van Haecht"
},
{
"operation": "Now on",
"time": "293.988",
"value": "Willem van Haecht"
},
{
"operation": "search",
"time": "328.652",
"value": "Contains...1628, Willem van Haecht destroyed the art collection owned"
},
{
"operation": "Now on",
"time": "330.929",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "335.495",
"value": "1621 willem van haecht destroyer the art collection owned"
},
{
"operation": "Now on",
"time": "336.094",
"value": "undefined"
},
{
"operation": "search",
"time": "355.316",
"value": "Willem van Haecht"
},
{
"operation": "Now on",
"time": "356.012",
"value": "Willem van Haecht"
},
{
"operation": "Highlighting",
"time": "360.032",
"value": "Willem van Haecht_sentence_1"
},
{
"operation": "Highlighting",
"time": "360.034",
"value": "Willem van Haecht_section_0"
},
{
"operation": "Highlighting",
"time": "363.503",
"value": "Willem van Haecht_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "364.263",
"value": "Willem van Haecht_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "364.264",
"value": "Willem van Haecht_section_0"
},
{
"operation": "finish",
"time": "389.661",
"value": "finish"
}
]
}
|
84809
|
European Universities Taekwondo Championships are coordinated by European University Sports Association along with the 18 other sports on the program of the European universities championships which are university sports competitions and in 2019 new sports were added that include kickboxing, orienteering, beach handball, and water polo.
|
{
"evidence": [
"European Universities Taekwondo Championships_sentence_1",
"European Universities Championships_sentence_0",
"European Universities Championships_sentence_3"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nEuropean Universities Taekwondo Championships are coordinated by European University Sports Association along with the 18 other sports on the program of the European universities championships which are university sports competitions and in 2019 new sports were added that include kickboxing, orienteering, beach handball, and water polo.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"European Universities Taekwondo Championships_sentence_1",
"European Universities Championships_sentence_0",
"European Universities Championships_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8036,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.893",
"value": "?search="
},
{
"operation": "search",
"time": "14.119",
"value": "European Universities Taekwondo Championships"
},
{
"operation": "Now on",
"time": "14.874",
"value": "European Universities Taekwondo Championships"
},
{
"operation": "Highlighting",
"time": "19.924",
"value": "European Universities Taekwondo Championships_sentence_1"
},
{
"operation": "hyperlink",
"time": "39.19",
"value": "European universities championships"
},
{
"operation": "Now on",
"time": "40.245",
"value": "European Universities Championships"
},
{
"operation": "Highlighting",
"time": "41.383",
"value": "European Universities Championships_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "43.948",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "44.031",
"value": "European Universities Taekwondo Championships"
},
{
"operation": "hyperlink",
"time": "49.108",
"value": "European universities championships"
},
{
"operation": "Now on",
"time": "49.641",
"value": "European universities championships"
},
{
"operation": "Highlighting",
"time": "52.246",
"value": "European Universities Championships_sentence_3"
},
{
"operation": "finish",
"time": "61.285",
"value": "finish"
}
]
}
|
54543
|
Ronald ("the Beast") Ashton (born on May 8, 1954) is a former WHA left winger for the Winnipeg Jets, and is best known neither as a fourth line player nor as an enforcer.
|
{
"evidence": [
"Ron Ashton_sentence_4"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nRonald (\"the Beast\") Ashton (born on May 8, 1954) is a former WHA left winger for the Winnipeg Jets, and is best known neither as a fourth line player nor as an enforcer.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ron Ashton_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8037,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.785",
"value": "?search="
},
{
"operation": "search",
"time": "14.893",
"value": "ronald ashton"
},
{
"operation": "Now on",
"time": "15.822",
"value": "Ron Ashton"
},
{
"operation": "Highlighting",
"time": "20.852",
"value": "Ron Ashton_sentence_2"
},
{
"operation": "Highlighting",
"time": "28.916",
"value": "Ron Ashton_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "31.388",
"value": "Ron Ashton_sentence_2"
},
{
"operation": "finish",
"time": "41.963",
"value": "finish"
}
]
}
|
40419
|
Harold Jarman father was permanently blinded, he attended Penpole School and he also showed an aptitude for both cricket and football
|
{
"evidence": [
"Harold Jarman_sentence_7",
"Harold Jarman_sentence_8",
"Harold Jarman_sentence_10"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nHarold Jarman father was permanently blinded, he attended Penpole School and he also showed an aptitude for both cricket and football\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Harold Jarman_sentence_7",
"Harold Jarman_sentence_8",
"Harold Jarman_sentence_10"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8038,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.084",
"value": "?search="
},
{
"operation": "search",
"time": "8.662",
"value": "Harold Jarman"
},
{
"operation": "Now on",
"time": "9.778",
"value": "Harold Jarman"
},
{
"operation": "Highlighting",
"time": "40.229",
"value": "Harold Jarman_sentence_7"
},
{
"operation": "Highlighting",
"time": "58.989",
"value": "Harold Jarman_sentence_8"
},
{
"operation": "Highlighting",
"time": "63.413",
"value": "Harold Jarman_sentence_10"
},
{
"operation": "finish",
"time": "76.428",
"value": "finish"
}
]
}
|
exfever_3869
|
William Jennings Bryan was a well-known American orator and politician that opposed Darwinism on religious and humanitarian grounds.
|
{
"evidence": [
"William Jennings Bryan",
"Nebraska"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nWilliam Jennings Bryan was a well-known American orator and politician that opposed Darwinism on religious and humanitarian grounds.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"William Jennings Bryan",
"Nebraska"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8039,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
58033
|
When cover models have a solo cover appearance, the unveiling often causes a media event thay happens to be largely reported among the press.
|
{
"evidence": [
"List of Sports Illustrated Swimsuit Issue cover models_sentence_7",
"List of Sports Illustrated Swimsuit Issue cover models_sentence_8"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nWhen cover models have a solo cover appearance, the unveiling often causes a media event thay happens to be largely reported among the press.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"List of Sports Illustrated Swimsuit Issue cover models_sentence_7",
"List of Sports Illustrated Swimsuit Issue cover models_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8040,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.413",
"value": "?search="
},
{
"operation": "search",
"time": "74.686",
"value": "cover model"
},
{
"operation": "Now on",
"time": "76.091",
"value": "Cover model"
},
{
"operation": "search",
"time": "122.629",
"value": "List of Sports Illustrated Swimsuit Issue cover models"
},
{
"operation": "Now on",
"time": "126.772",
"value": "List of Sports Illustrated Swimsuit Issue cover models"
},
{
"operation": "Highlighting",
"time": "133.142",
"value": "List of Sports Illustrated Swimsuit Issue cover models_sentence_7"
},
{
"operation": "hyperlink",
"time": "134.374",
"value": "media event"
},
{
"operation": "Now on",
"time": "136.789",
"value": "List of Sports Illustrated Swimsuit Issue cover models"
},
{
"operation": "Highlighting",
"time": "138.265",
"value": "List of Sports Illustrated Swimsuit Issue cover models_sentence_8"
},
{
"operation": "finish",
"time": "144.353",
"value": "finish"
}
]
}
|
exfever_7699
|
Joe Hart is a goalkeeper for a football organization that had sixteen clubs since its creation on August 2, 2013 and concluded on May 11, 2014.
|
{
"evidence": [
"Joe Hart",
"Scottish Premiership",
"2013β14 Scottish Premiership"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nJoe Hart is a goalkeeper for a football organization that had sixteen clubs since its creation on August 2, 2013 and concluded on May 11, 2014.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Joe Hart",
"Scottish Premiership",
"2013β14 Scottish Premiership"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8041,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8284
|
Estelle Getty is an English actress and comedian best known for her portrayal in The Golden Girls as a character who also appeared in episodes of Blossom and Nurses.
|
{
"evidence": [
"Estelle Getty",
"Sophia Petrillo"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nEstelle Getty is an English actress and comedian best known for her portrayal in The Golden Girls as a character who also appeared in episodes of Blossom and Nurses.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Estelle Getty",
"Sophia Petrillo"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8042,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1532
|
The Bostonians is a 1984 romantic drama period film directed by an American film director who also directed the 1985 British romance film written by Ruth Prawer Jhabvala.
|
{
"evidence": [
"The Bostonians (film)",
"James Ivory",
"A Room with a View (1985 film)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nThe Bostonians is a 1984 romantic drama period film directed by an American film director who also directed the 1985 British romance film written by Ruth Prawer Jhabvala.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Bostonians (film)",
"James Ivory",
"A Room with a View (1985 film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8043,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
68883
|
McLane Stadium features a large high-definition LED video board with an area of 5,029 square feet, ranking it the largest college football video board.
|
{
"evidence": [
"McLane Stadium_sentence_15"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nMcLane Stadium features a large high-definition LED video board with an area of 5,029 square feet, ranking it the largest college football video board.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"McLane Stadium_sentence_15"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8044,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.125",
"value": "?search="
},
{
"operation": "search",
"time": "9.524",
"value": "McLane Stadium"
},
{
"operation": "Now on",
"time": "101.391",
"value": "?search="
},
{
"operation": "search",
"time": "104.943",
"value": "McLane Stadium"
},
{
"operation": "Now on",
"time": "106.2",
"value": "McLane Stadium"
},
{
"operation": "hyperlink",
"time": "115.875",
"value": "3 Video board McLane Stadium_section_2"
},
{
"operation": "Highlighting",
"time": "120.107",
"value": "McLane Stadium_sentence_14"
},
{
"operation": "Highlighting",
"time": "124.837",
"value": "McLane Stadium_sentence_15"
},
{
"operation": "Highlighting deleted",
"time": "135.239",
"value": "McLane Stadium_sentence_14"
},
{
"operation": "finish",
"time": "152.029",
"value": "finish"
}
]
}
|
25337
|
Massachusetts Route 68 heads into Hubbardston before bending into the city of Gardner.
|
{
"evidence": [
"Massachusetts Route 68_sentence_6"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nMassachusetts Route 68 heads into Hubbardston before bending into the city of Gardner.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Massachusetts Route 68_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8045,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.694",
"value": "?search="
},
{
"operation": "search",
"time": "29.478",
"value": "Route 68 Massachusetts"
},
{
"operation": "Now on",
"time": "31.476",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "35.72",
"value": "Massachusetts Route 68"
},
{
"operation": "Now on",
"time": "36.277",
"value": "Massachusetts Route 68"
},
{
"operation": "Highlighting",
"time": "54.684",
"value": "Massachusetts Route 68_sentence_6"
},
{
"operation": "finish",
"time": "66.518",
"value": "finish"
}
]
}
|
exfever_8224
|
Back to the Future Part II is a 1989 American science fiction film starring an American actress best known for her role in the 2010 American 3D horror comedy film that serves as a loose remake of the Adventure horror film Piranha.
|
{
"evidence": [
"Back to the Future Part II",
"Elisabeth Shue",
"Piranha 3D"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nBack to the Future Part II is a 1989 American science fiction film starring an American actress best known for her role in the 2010 American 3D horror comedy film that serves as a loose remake of the Adventure horror film Piranha.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Back to the Future Part II",
"Elisabeth Shue",
"Piranha 3D"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8046,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8794
|
Borrelia afzelii was first named after the Swedish dermatologist, Arvid Afzelius.
|
{
"evidence": [
"Borrelia afzelii",
"Vertebrate"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nBorrelia afzelii was first named after the Swedish dermatologist, Arvid Afzelius.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Borrelia afzelii",
"Vertebrate"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8047,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
92507
|
Keno is a turnover tax that is apart of the major forms of gambling taxation in Australia.
|
{
"evidence": [
"Keno_sentence_1",
"Lottery_sentence_14",
"Lottery_sentence_80",
"Lottery_sentence_81",
"Lottery_sentence_82",
"Lottery_sentence_83",
"Lottery_sentence_84"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nKeno is a turnover tax that is apart of the major forms of gambling taxation in Australia.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Keno_sentence_1",
"Lottery_sentence_14",
"Lottery_sentence_80",
"Lottery_sentence_81",
"Lottery_sentence_82",
"Lottery_sentence_83",
"Lottery_sentence_84"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8048,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.618",
"value": "?search="
},
{
"operation": "search",
"time": "10.627",
"value": "Keno"
},
{
"operation": "Now on",
"time": "11.693",
"value": "Keno"
},
{
"operation": "Highlighting",
"time": "18.214",
"value": "Keno_sentence_1"
},
{
"operation": "search",
"time": "33.169",
"value": "Contains...gambling taxation in Australia"
},
{
"operation": "Now on",
"time": "35.127",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "42.256",
"value": "Lottery"
},
{
"operation": "Now on",
"time": "43.555",
"value": "Lottery"
},
{
"operation": "hyperlink",
"time": "56.081",
"value": "5.1 Australia Lottery_section_10"
},
{
"operation": "hyperlink",
"time": "67.359",
"value": "Keno"
},
{
"operation": "Now on",
"time": "68.16",
"value": "Keno"
},
{
"operation": "back-button-clicked",
"time": "80.448",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "81.02",
"value": "Lottery#Australia Lottery section 10"
},
{
"operation": "Highlighting",
"time": "119.502",
"value": "Lottery_sentence_14"
},
{
"operation": "hyperlink",
"time": "127.063",
"value": "5.1 Australia Lottery_section_10"
},
{
"operation": "Highlighting",
"time": "128.733",
"value": "Lottery_sentence_80"
},
{
"operation": "Highlighting",
"time": "129.229",
"value": "Lottery_sentence_81"
},
{
"operation": "Highlighting",
"time": "130.467",
"value": "Lottery_sentence_82"
},
{
"operation": "Highlighting",
"time": "130.977",
"value": "Lottery_sentence_83"
},
{
"operation": "Highlighting",
"time": "131.552",
"value": "Lottery_sentence_84"
},
{
"operation": "finish",
"time": "138.822",
"value": "finish"
}
]
}
|
73605
|
Mundia Sikatana was appointed foreign minister on 9 August 2006 under President Levy Mwanawasa but got fired on the 22nd day of the same month due to his allegedly declining health.
|
{
"evidence": [
"Mundia Sikatana_sentence_4"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nMundia Sikatana was appointed foreign minister on 9 August 2006 under President Levy Mwanawasa but got fired on the 22nd day of the same month due to his allegedly declining health.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Mundia Sikatana_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8049,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "15.674",
"value": "?search="
},
{
"operation": "search",
"time": "142.089",
"value": "Mundia Sikatana"
},
{
"operation": "Now on",
"time": "142.988",
"value": "Mundia Sikatana"
},
{
"operation": "Highlighting",
"time": "157.446",
"value": "Mundia Sikatana_sentence_3"
},
{
"operation": "Highlighting",
"time": "180.742",
"value": "Mundia Sikatana_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "234.325",
"value": "Mundia Sikatana_sentence_3"
},
{
"operation": "finish",
"time": "246.16",
"value": "finish"
}
]
}
|
exfever_2169
|
Life After Death is the second and final studio album by an American rapper and songwriter who was rooted in the New York rap scene and gangsta rap traditions.
|
{
"evidence": [
"Life After Death",
"The Notorious B.I.G."
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nLife After Death is the second and final studio album by an American rapper and songwriter who was rooted in the New York rap scene and gangsta rap traditions.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Life After Death",
"The Notorious B.I.G."
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8050,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
95061
|
In the 2000-2001 Polish Volleyball League, Czestochowa achieved third title of Polish Champion, Stolarka Wolomin was promoted to the league, and three players played in CEV contests.
|
{
"evidence": [
"2000β01 Polish Volleyball League_sentence_3"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nIn the 2000-2001 Polish Volleyball League, Czestochowa achieved third title of Polish Champion, Stolarka Wolomin was promoted to the league, and three players played in CEV contests.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"2000β01 Polish Volleyball League_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8051,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.007",
"value": "?search="
},
{
"operation": "search",
"time": "143.644",
"value": "2000-2001 Polish Volleyball League"
},
{
"operation": "Now on",
"time": "146.382",
"value": "undefined"
},
{
"operation": "search",
"time": "157.048",
"value": "CzΔstochowa"
},
{
"operation": "Now on",
"time": "161.19",
"value": "Częstochowa"
},
{
"operation": "search",
"time": "187.024",
"value": "2000β01 Polish Volleyball League"
},
{
"operation": "Now on",
"time": "187.951",
"value": "2000β01 Polish Volleyball League"
},
{
"operation": "Highlighting",
"time": "218.945",
"value": "2000β01 Polish Volleyball League_sentence_3"
},
{
"operation": "finish",
"time": "266.733",
"value": "finish"
}
]
}
|
exfever_923
|
On Stranger Tides is a 1979 historical fantasy novel that features an English sea captain known as the captain of the vessel that rescued marooned Alexander Selkirk, whose plight is generally believed to have inspired a novel first published on 25 April 1719.
|
{
"evidence": [
"On Stranger Tides",
"Woodes Rogers",
"Robinson Crusoe"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nOn Stranger Tides is a 1979 historical fantasy novel that features an English sea captain known as the captain of the vessel that rescued marooned Alexander Selkirk, whose plight is generally believed to have inspired a novel first published on 25 April 1719.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"On Stranger Tides",
"Woodes Rogers",
"Robinson Crusoe"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8052,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6537
|
DJ Khaled is an American DJ and record producer who sang the third single from the 2011 album Victory.
|
{
"evidence": [
"DJ Khaled",
"All I Do Is Win"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nDJ Khaled is an American DJ and record producer who sang the third single from the 2011 album Victory.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"DJ Khaled",
"All I Do Is Win"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8053,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4151
|
The Little Prince is a 2015 animated fantasy adventure drama film directed by Mark Osborne and based on the 1943 The Little Prince by French aristocrat, writer, and military aviator Antoine de Saint-ExupΓ©ry.
|
{
"evidence": [
"The Little Prince (2015 film)",
"The Little Prince"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nThe Little Prince is a 2015 animated fantasy adventure drama film directed by Mark Osborne and based on the 1943 The Little Prince by French aristocrat, writer, and military aviator Antoine de Saint-ExupΓ©ry.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Little Prince (2015 film)",
"The Little Prince"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8054,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_180
|
The Big Trail was the third early widescreen film.
|
{
"evidence": [
"The Big Trail",
"Raoul Walsh"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nThe Big Trail was the third early widescreen film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Big Trail",
"Raoul Walsh"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8055,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2951
|
Eric Clapton was the leader of the supergroup Blind Faith.
|
{
"evidence": [
"Blind Faith",
"Steve Winwood"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nEric Clapton was the leader of the supergroup Blind Faith.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Blind Faith",
"Steve Winwood"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8056,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
68825
|
The The 1921 Michigan Agricultural Aggies football team represented Michigan Agriculture College in their first year under head coach Albert Barron.
|
{
"evidence": [
"1921 Michigan Agricultural Aggies football team_sentence_0",
"1921 Michigan Agricultural Aggies football team_sentence_1"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nThe The 1921 Michigan Agricultural Aggies football team represented Michigan Agriculture College in their first year under head coach Albert Barron.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"1921 Michigan Agricultural Aggies football team_sentence_0",
"1921 Michigan Agricultural Aggies football team_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8057,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.638",
"value": "?search="
},
{
"operation": "search",
"time": "10.913",
"value": "1921 Michigan Agricultural Aggies football team"
},
{
"operation": "Now on",
"time": "15.461",
"value": "1921 Michigan Agricultural Aggies football team"
},
{
"operation": "Highlighting",
"time": "17.991",
"value": "1921 Michigan Agricultural Aggies football team_sentence_0"
},
{
"operation": "Highlighting",
"time": "25.977",
"value": "1921 Michigan Agricultural Aggies football team_sentence_1"
},
{
"operation": "finish",
"time": "46.509",
"value": "finish"
}
]
}
|
exfever_3802
|
Derrick Rose is an American professional basketball player who became the youngest player in 2011 to win the annual National Basketball Association award given since the 1955β56 season.
|
{
"evidence": [
"Derrick Rose",
"NBA Most Valuable Player Award"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nDerrick Rose is an American professional basketball player who became the youngest player in 2011 to win the annual National Basketball Association award given since the 1955β56 season.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Derrick Rose",
"NBA Most Valuable Player Award"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8058,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_142
|
"See You Again" by Miley Cyrus was produced by Tim James and a Canadian songwriter, record producer, music producer, and businesswoman.
|
{
"evidence": [
"See You Again (Miley Cyrus song)",
"Antonina Armato"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\n\"See You Again\" by Miley Cyrus was produced by Tim James and a Canadian songwriter, record producer, music producer, and businesswoman.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"See You Again (Miley Cyrus song)",
"Antonina Armato"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8059,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
13033
|
Pinguicula filifolia has leaves that are approximately 4β6 mm long and 1-1.5 mm in width with flat or sub-erect leaves while seedlings that arise from resettes that are 8-10mm in diameter where a single rosette typically has 4-6 blades sprouting from it.
|
{
"evidence": [
"Pinguicula filifolia_sentence_15",
"Pinguicula filifolia_sentence_16",
"Pinguicula filifolia_sentence_17"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nPinguicula filifolia has leaves that are approximately 4β6 mm long and 1-1.5 mm in width with flat or sub-erect leaves while seedlings that arise from resettes that are 8-10mm in diameter where a single rosette typically has 4-6 blades sprouting from it.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Pinguicula filifolia_sentence_15",
"Pinguicula filifolia_sentence_16",
"Pinguicula filifolia_sentence_17"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8060,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.873",
"value": "?search="
},
{
"operation": "search",
"time": "11.72",
"value": "Pinguicula filifolia"
},
{
"operation": "Now on",
"time": "12.575",
"value": "Pinguicula filifolia"
},
{
"operation": "Highlighting",
"time": "31.724",
"value": "Pinguicula filifolia_sentence_15"
},
{
"operation": "Highlighting",
"time": "57.276",
"value": "Pinguicula filifolia_sentence_16"
},
{
"operation": "Highlighting",
"time": "62.403",
"value": "Pinguicula filifolia_sentence_17"
},
{
"operation": "finish",
"time": "90.149",
"value": "finish"
}
]
}
|
81338
|
TypeCon is a typography-focused convention run by the Society of Typographic Experts focused on typography, type design, and calligraphy.
|
{
"evidence": [
"TypeCon_sentence_0",
"TypeCon_sentence_1"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nTypeCon is a typography-focused convention run by the Society of Typographic Experts focused on typography, type design, and calligraphy.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"TypeCon_sentence_0",
"TypeCon_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8061,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.348",
"value": "?search="
},
{
"operation": "search",
"time": "11.862",
"value": "TypeCon"
},
{
"operation": "Now on",
"time": "12.681",
"value": "TypeCon"
},
{
"operation": "Highlighting",
"time": "22.491",
"value": "TypeCon_sentence_0"
},
{
"operation": "Highlighting",
"time": "30.867",
"value": "TypeCon_sentence_1"
},
{
"operation": "finish",
"time": "44.502",
"value": "finish"
}
]
}
|
88454
|
The gene CDX2 regulates several essential processes in the development and function of the lower gastrointestinal tract, as well as in the vertebrate embryonic development (that starts with the fertilization of an egg cell by a sperm cell) where it becomes active in endodermal cells that are posterior to the developing stomach.
|
{
"evidence": [
"CDX2_sentence_4",
"CDX2_sentence_5",
"Embryonic development_sentence_4"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nThe gene CDX2 regulates several essential processes in the development and function of the lower gastrointestinal tract, as well as in the vertebrate embryonic development (that starts with the fertilization of an egg cell by a sperm cell) where it becomes active in endodermal cells that are posterior to the developing stomach.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"CDX2_sentence_4",
"CDX2_sentence_5",
"Embryonic development_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8062,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.694",
"value": "?search="
},
{
"operation": "search",
"time": "10.719",
"value": "CDX2"
},
{
"operation": "Now on",
"time": "11.983",
"value": "CDX2"
},
{
"operation": "Page search",
"time": "76.299",
"value": "regul"
},
{
"operation": "Highlighting",
"time": "78.88",
"value": "CDX2_sentence_4"
},
{
"operation": "Highlighting",
"time": "86.084",
"value": "CDX2_sentence_5"
},
{
"operation": "hyperlink",
"time": "296.112",
"value": "embryonic development"
},
{
"operation": "Now on",
"time": "297.375",
"value": "Embryonic development"
},
{
"operation": "Highlighting",
"time": "299.653",
"value": "Embryonic development_sentence_4"
},
{
"operation": "back-button-clicked",
"time": "301.421",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "301.543",
"value": "CDX2"
},
{
"operation": "finish",
"time": "311.462",
"value": "finish"
}
]
}
|
48755
|
John Lindsay set the world record for the Men's 200m TW3 event at the ninth Summer Paralympics event in 1992.
|
{
"evidence": [
"John Lindsay (Paralympian)_sentence_0",
"John Lindsay (Paralympian)_sentence_4",
"John Lindsay (Paralympian)_sentence_2",
"1992 Summer Paralympics_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nJohn Lindsay set the world record for the Men's 200m TW3 event at the ninth Summer Paralympics event in 1992.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"John Lindsay (Paralympian)_sentence_0",
"John Lindsay (Paralympian)_sentence_4",
"John Lindsay (Paralympian)_sentence_2",
"1992 Summer Paralympics_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8063,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.272",
"value": "?search="
},
{
"operation": "search",
"time": "8.586",
"value": "John Lindsay"
},
{
"operation": "Now on",
"time": "9.606",
"value": "John Lindsay"
},
{
"operation": "hyperlink",
"time": "13.565",
"value": "John Lindsay (disambiguation)"
},
{
"operation": "Now on",
"time": "14.186",
"value": "John Lindsay (disambiguation)"
},
{
"operation": "hyperlink",
"time": "19.006",
"value": "3 Sports John Lindsay (disambiguation)_section_2"
},
{
"operation": "hyperlink",
"time": "26.537",
"value": "John Lindsay (Paralympian)"
},
{
"operation": "Now on",
"time": "27.458",
"value": "John Lindsay (Paralympian)"
},
{
"operation": "Highlighting",
"time": "32.22",
"value": "John Lindsay (Paralympian)_sentence_0"
},
{
"operation": "Highlighting",
"time": "67.67",
"value": "John Lindsay (Paralympian)_sentence_4"
},
{
"operation": "Highlighting",
"time": "99.82",
"value": "John Lindsay (Paralympian)_sentence_2"
},
{
"operation": "hyperlink",
"time": "111.866",
"value": "1992 Barcelona Games"
},
{
"operation": "Now on",
"time": "112.67",
"value": "1992 Summer Paralympics"
},
{
"operation": "Highlighting",
"time": "115.12",
"value": "1992 Summer Paralympics_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "140.111",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "140.459",
"value": "John Lindsay (Paralympian)"
},
{
"operation": "finish",
"time": "169.858",
"value": "finish"
}
]
}
|
exfever_895
|
Alfred Hitchcock was an English filmmaker who produced an American television series that spans different genres and presents a different story and a different set of characters in each episode, season, segment, or short.
|
{
"evidence": [
"Alfred Hitchcock",
"Alfred Hitchcock Presents",
"Anthology series"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nAlfred Hitchcock was an English filmmaker who produced an American television series that spans different genres and presents a different story and a different set of characters in each episode, season, segment, or short.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Alfred Hitchcock",
"Alfred Hitchcock Presents",
"Anthology series"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8064,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7477
|
Ghostbusters is the tenth movie directed by Paul Feig.
|
{
"evidence": [
"Ghostbusters (2016 film)",
"Comedy film",
"Theatre"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nGhostbusters is the tenth movie directed by Paul Feig.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ghostbusters (2016 film)",
"Comedy film",
"Theatre"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8065,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
61944
|
Flappers and Mods have has a strong influence on youth culture.
|
{
"evidence": [
"Youth culture_sentence_27",
"Youth culture_sentence_28",
"Youth culture_sentence_35"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nFlappers and Mods have has a strong influence on youth culture.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Youth culture_sentence_27",
"Youth culture_sentence_28",
"Youth culture_sentence_35"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8066,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.217",
"value": "?search="
},
{
"operation": "search",
"time": "10.729",
"value": "Flappers"
},
{
"operation": "Now on",
"time": "13.324",
"value": "Flapper"
},
{
"operation": "search",
"time": "30.148",
"value": "Mods"
},
{
"operation": "Now on",
"time": "32.001",
"value": "Mod"
},
{
"operation": "search",
"time": "52.152",
"value": "flappers and mods"
},
{
"operation": "Now on",
"time": "53.663",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "57.049",
"value": "Youth culture"
},
{
"operation": "Now on",
"time": "58.299",
"value": "Youth culture"
},
{
"operation": "Page search",
"time": "61.396",
"value": "flapper"
},
{
"operation": "Highlighting",
"time": "68.648",
"value": "Youth culture_sentence_27"
},
{
"operation": "Highlighting",
"time": "71.927",
"value": "Youth culture_sentence_28"
},
{
"operation": "Highlighting",
"time": "80.636",
"value": "Youth culture_sentence_35"
},
{
"operation": "finish",
"time": "105.696",
"value": "finish"
}
]
}
|
45315
|
The 33rd Network Warfare Squadron was conferred the Distinguished Unit Citation, now Presidential Unit Citation, in 1944.
|
{
"evidence": [
"33rd Network Warfare Squadron_sentence_19",
"Presidential Unit Citation (United States)_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nThe 33rd Network Warfare Squadron was conferred the Distinguished Unit Citation, now Presidential Unit Citation, in 1944.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"33rd Network Warfare Squadron_sentence_19",
"Presidential Unit Citation (United States)_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8067,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.797",
"value": "?search="
},
{
"operation": "search",
"time": "10.81",
"value": "33rd Network Warfare Squadron"
},
{
"operation": "Now on",
"time": "11.756",
"value": "33rd Network Warfare Squadron"
},
{
"operation": "Page search",
"time": "18.666",
"value": "distinguished"
},
{
"operation": "Highlighting",
"time": "36.134",
"value": "33rd Network Warfare Squadron_sentence_19"
},
{
"operation": "hyperlink",
"time": "39.96",
"value": "Distinguished Unit Citation"
},
{
"operation": "Now on",
"time": "43.921",
"value": "Presidential Unit Citation (United States)"
},
{
"operation": "Highlighting",
"time": "45.512",
"value": "Presidential Unit Citation (United States)_sentence_0"
},
{
"operation": "Page search",
"time": "146.027",
"value": "33rd Network Warfare Squadron"
},
{
"operation": "back-button-clicked",
"time": "191.897",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "192.06",
"value": "33rd Network Warfare Squadron"
},
{
"operation": "hyperlink",
"time": "203.64",
"value": "Distinguished Unit Citation"
},
{
"operation": "Now on",
"time": "204.943",
"value": "Presidential Unit Citation (United States)"
},
{
"operation": "Page search",
"time": "207.992",
"value": "1944"
},
{
"operation": "finish",
"time": "287.332",
"value": "finish"
}
]
}
|
exfever_7148
|
Ryder is an American transportation and logistics company that has its headquarters in a coastal metropolis located in Miami-Dade County in northeastern Florida.
|
{
"evidence": [
"Ryder",
"Miami"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nRyder is an American transportation and logistics company that has its headquarters in a coastal metropolis located in Miami-Dade County in northeastern Florida.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ryder",
"Miami"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8068,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8120
|
Michael Edward Love co-founded an American rock band that formed in Hawthorne, California, in 1961.
|
{
"evidence": [
"Mike Love",
"The Beach Boys"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nMichael Edward Love co-founded an American rock band that formed in Hawthorne, California, in 1961.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mike Love",
"The Beach Boys"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8069,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
58706
|
Bhillama II, a ruler of the Seuna (Yadava) dynasty of Deccan region in India, played a key role in Chalukya-Paramara war which resulted to the fall of a Paramara king and was rewarded the present-day Ahmednagar area as his hereditary fief.
|
{
"evidence": [
"Bhillama II_sentence_0",
"Bhillama II_sentence_7",
"Bhillama II_sentence_9"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nBhillama II, a ruler of the Seuna (Yadava) dynasty of Deccan region in India, played a key role in Chalukya-Paramara war which resulted to the fall of a Paramara king and was rewarded the present-day Ahmednagar area as his hereditary fief.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Bhillama II_sentence_0",
"Bhillama II_sentence_7",
"Bhillama II_sentence_9"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8070,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.845",
"value": "?search="
},
{
"operation": "search",
"time": "12.063",
"value": "Bhillama II"
},
{
"operation": "Now on",
"time": "12.739",
"value": "Bhillama II"
},
{
"operation": "Highlighting",
"time": "24.04",
"value": "Bhillama II_sentence_0"
},
{
"operation": "Highlighting",
"time": "29.84",
"value": "Bhillama II_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "34.586",
"value": "Bhillama II_sentence_1"
},
{
"operation": "Highlighting",
"time": "58.999",
"value": "Bhillama II_sentence_7"
},
{
"operation": "Highlighting",
"time": "74.737",
"value": "Bhillama II_sentence_9"
},
{
"operation": "finish",
"time": "84.361",
"value": "finish"
}
]
}
|
11822
|
Social inequality happens when resources in a society are unevenly distributed, but this rarely includes the quantitative variables of income and wealth.
|
{
"evidence": [
"Social inequality_sentence_128"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nSocial inequality happens when resources in a society are unevenly distributed, but this rarely includes the quantitative variables of income and wealth.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Social inequality_sentence_128"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8071,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.807",
"value": "?search="
},
{
"operation": "search",
"time": "17.001",
"value": "Contains...Social inequality happens when resources in a society are unevenly distributed"
},
{
"operation": "Now on",
"time": "18.444",
"value": "undefined"
},
{
"operation": "search",
"time": "48.412",
"value": "Contains...quantitative variables of income and wealth"
},
{
"operation": "Now on",
"time": "50.221",
"value": "undefined"
},
{
"operation": "search",
"time": "71.65",
"value": "Social inequality"
},
{
"operation": "Now on",
"time": "72.887",
"value": "Social inequality"
},
{
"operation": "Highlighting",
"time": "83.785",
"value": "Social inequality_sentence_2"
},
{
"operation": "Page search",
"time": "105.945",
"value": "quantitative"
},
{
"operation": "Page search",
"time": "110.965",
"value": "quantitative"
},
{
"operation": "Highlighting",
"time": "161.529",
"value": "Social inequality_sentence_128"
},
{
"operation": "Highlighting",
"time": "199.178",
"value": "Social inequality_sentence_2"
},
{
"operation": "finish",
"time": "232.675",
"value": "finish"
}
]
}
|
exfever_5833
|
School is a 2015 South Korean television series written as εΈζ ‘ in the Korean name for a traditional writing system that consists of logograms developed for the writing of Chinese that has been adapted to write other East Asian languages.
|
{
"evidence": [
"School 2013",
"Hanja",
"Chinese characters"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nSchool is a 2015 South Korean television series written as εΈζ ‘ in the Korean name for a traditional writing system that consists of logograms developed for the writing of Chinese that has been adapted to write other East Asian languages.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"School 2013",
"Hanja",
"Chinese characters"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8072,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
64731
|
Bruce Petway traveled with the 1913 Chicago American Giants when they toured the United States and he remained with the team until he moved to the Detroit Stars in 1919.
|
{
"evidence": [
"Bruce Petway_sentence_7",
"Bruce Petway_sentence_8"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nBruce Petway traveled with the 1913 Chicago American Giants when they toured the United States and he remained with the team until he moved to the Detroit Stars in 1919.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bruce Petway_sentence_7",
"Bruce Petway_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8073,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.063",
"value": "?search="
},
{
"operation": "search",
"time": "157.795",
"value": "Bruce Petway"
},
{
"operation": "Now on",
"time": "158.924",
"value": "Bruce Petway"
},
{
"operation": "Highlighting",
"time": "168.588",
"value": "Bruce Petway_sentence_7"
},
{
"operation": "Highlighting",
"time": "171.319",
"value": "Bruce Petway_sentence_8"
},
{
"operation": "finish",
"time": "231.668",
"value": "finish"
}
]
}
|
exfever_7381
|
Dracula is a novel by an Irish author.
|
{
"evidence": [
"Dracula",
"Bram Stoker"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nDracula is a novel by an Irish author.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dracula",
"Bram Stoker"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8074,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2644
|
Edward J. Noble is the founder of The American Broadcasting Company.
|
{
"evidence": [
"American Broadcasting Company",
"Walt Disney Studios (Burbank)"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nEdward J. Noble is the founder of The American Broadcasting Company.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"American Broadcasting Company",
"Walt Disney Studios (Burbank)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8075,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1518
|
Don MacLaughlin likes to paint images that reveal emotion and expression.
|
{
"evidence": [
"As the World Turns",
"Irna Phillips"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nDon MacLaughlin likes to paint images that reveal emotion and expression.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"As the World Turns",
"Irna Phillips"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8076,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
11301
|
The Felice Brothers, originally hail from Palenville, New York, lived in a little apartment in Brooklyn and would play in the subway stations at 42nd Street and Union Square and in Greenwich Village (a neighborhood on the east side of Manhattan in New York City).
|
{
"evidence": [
"Greenwich Village_sentence_3"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nThe Felice Brothers, originally hail from Palenville, New York, lived in a little apartment in Brooklyn and would play in the subway stations at 42nd Street and Union Square and in Greenwich Village (a neighborhood on the east side of Manhattan in New York City).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Greenwich Village_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8077,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.824",
"value": "?search="
},
{
"operation": "search",
"time": "10.753",
"value": "Felice brothers"
},
{
"operation": "Now on",
"time": "13.732",
"value": "The Felice Brothers"
},
{
"operation": "Highlighting",
"time": "24.428",
"value": "The Felice Brothers_sentence_3"
},
{
"operation": "Highlighting",
"time": "24.429",
"value": "The Felice Brothers_title"
},
{
"operation": "Highlighting",
"time": "24.43",
"value": "The Felice Brothers_section_0"
},
{
"operation": "Highlighting",
"time": "28.646",
"value": "The Felice Brothers_sentence_2"
},
{
"operation": "hyperlink",
"time": "38.693",
"value": "Greenwich Village"
},
{
"operation": "Now on",
"time": "40.823",
"value": "Greenwich Village"
},
{
"operation": "Highlighting",
"time": "45.607",
"value": "Greenwich Village_sentence_3"
},
{
"operation": "back-button-clicked",
"time": "48.686",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "49.611",
"value": "The Felice Brothers"
},
{
"operation": "Highlighting deleted",
"time": "50.757",
"value": "The Felice Brothers_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "51.368",
"value": "The Felice Brothers_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "51.37",
"value": "The Felice Brothers_title"
},
{
"operation": "Highlighting deleted",
"time": "51.372",
"value": "The Felice Brothers_section_0"
},
{
"operation": "finish",
"time": "71.249",
"value": "finish"
}
]
}
|
87172
|
Baijiu has been in production since dynastic times and is bottled for domestic consumption and for shipping internationally.
|
{
"evidence": [
"Baijiu_sentence_10"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nBaijiu has been in production since dynastic times and is bottled for domestic consumption and for shipping internationally.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Baijiu_sentence_10"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8078,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.255",
"value": "?search="
},
{
"operation": "search",
"time": "9.706",
"value": "Baijiu"
},
{
"operation": "Now on",
"time": "11.42",
"value": "Baijiu"
},
{
"operation": "Page search",
"time": "24.48",
"value": "production"
},
{
"operation": "Page search",
"time": "35.869",
"value": "dynastic"
},
{
"operation": "Highlighting",
"time": "72.148",
"value": "Baijiu_sentence_10"
},
{
"operation": "Highlighting deleted",
"time": "90.057",
"value": "Baijiu_sentence_10"
},
{
"operation": "Highlighting",
"time": "98.001",
"value": "Baijiu_item_0_1"
},
{
"operation": "hyperlink",
"time": "118.443",
"value": "3.1 Global baijiu market Baijiu_section_6"
},
{
"operation": "Page search",
"time": "146.198",
"value": "domestic"
},
{
"operation": "Page search",
"time": "163.999",
"value": "bottled"
},
{
"operation": "Page search",
"time": "190.336",
"value": "shipping"
},
{
"operation": "Page search",
"time": "214.202",
"value": "international"
},
{
"operation": "Page search",
"time": "265.208",
"value": "production"
},
{
"operation": "Highlighting deleted",
"time": "313.424",
"value": "Baijiu_item_0_1"
},
{
"operation": "Highlighting",
"time": "321.643",
"value": "Baijiu_sentence_10"
},
{
"operation": "finish",
"time": "330.081",
"value": "finish"
}
]
}
|
exfever_7932
|
Sir Ridley Scott was an English film director of the 2000 epic historical drama film written by David Franzoni.
|
{
"evidence": [
"Ridley Scott",
"Gladiator (2000 film)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nSir Ridley Scott was an English film director of the 2000 epic historical drama film written by David Franzoni.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ridley Scott",
"Gladiator (2000 film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8079,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5399
|
Brittany Anne Murphy-Monjack made her stage debut in a Broadway production of a play by a Canadian playwright Arthur Miller.
|
{
"evidence": [
"Brittany Murphy",
"A View from the Bridge"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nBrittany Anne Murphy-Monjack made her stage debut in a Broadway production of a play by a Canadian playwright Arthur Miller.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Brittany Murphy",
"A View from the Bridge"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 8080,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5486
|
Super Bowl XLVI was the first American football game played at Lucas Oil Stadium.
|
{
"evidence": [
"Super Bowl XLVI",
"American Football Conference"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nSuper Bowl XLVI was the first American football game played at Lucas Oil Stadium.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Super Bowl XLVI",
"American Football Conference"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 8081,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
68396
|
s a multi-purpose stadium in Seoul, South Korea and it is the main stadium built for the 1988 Summer Olympics and the 10th Asian Games
|
{
"evidence": [
"Seoul Olympic Stadium_sentence_0",
"Seoul Olympic Stadium_sentence_1"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\ns a multi-purpose stadium in Seoul, South Korea and it is the main stadium built for the 1988 Summer Olympics and the 10th Asian Games\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Seoul Olympic Stadium_sentence_0",
"Seoul Olympic Stadium_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8082,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.296",
"value": "?search="
},
{
"operation": "search",
"time": "139.058",
"value": "multi-purpose stadium in Seoul, South Korea and it is the main stadium built for the 1988 Summer Olympics and the 10th Asian Games"
},
{
"operation": "Now on",
"time": "141.433",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "143.106",
"value": "Seoul Olympic Stadium"
},
{
"operation": "Now on",
"time": "143.76",
"value": "Seoul Olympic Stadium"
},
{
"operation": "Highlighting",
"time": "150.567",
"value": "Seoul Olympic Stadium_sentence_0"
},
{
"operation": "Highlighting",
"time": "151.496",
"value": "Seoul Olympic Stadium_sentence_1"
},
{
"operation": "finish",
"time": "184.237",
"value": "finish"
}
]
}
|
exfever_5979
|
Ryan O'Neal landed a role on the ABC nighttime soap opera Peyton Place as a fictional character in the 1956 Grace Metalious novel.
|
{
"evidence": [
"Ryan O'Neal",
"Rodney Harrington"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>β¦</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>β¦</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, donβt repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\nβ¦\n</information>\n- You will be given claim related information in the format above.\n\n<think>β¦</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]β¦\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan β (search β information β think) repeat until conclusion β answer\n\nVerify the claim:\nRyan O'Neal landed a role on the ABC nighttime soap opera Peyton Place as a fictional character in the 1956 Grace Metalious novel.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ryan O'Neal",
"Rodney Harrington"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 8083,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.