id
stringlengths 1
12
| question
stringlengths 22
566
| golden_answers
dict | data_source
stringclasses 2
values | prompt
listlengths 1
1
| ability
stringclasses 6
values | reward_model
dict | extra_info
dict | metadata
dict |
|---|---|---|---|---|---|---|---|---|
exfever_3752
|
Tropic Thunder is a 2008 action-comedy film directed by an American actor, comedian, producer, director, and screenwriter with a 2013 American adventure comedy-drama film credited to him.
|
{
"evidence": [
"Tropic Thunder",
"Ben Stiller",
"The Secret Life of Walter Mitty (2013 film)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nTropic Thunder is a 2008 action-comedy film directed by an American actor, comedian, producer, director, and screenwriter with a 2013 American adventure comedy-drama film credited to him.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tropic Thunder",
"Ben Stiller",
"The Secret Life of Walter Mitty (2013 film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 388,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
15451
|
Moon Ballad(foaled 4 March 1999) is a chestnut Thoroughbred racehorse sired Singspiel, Solow who is owned by Godolphin Racing and trained by Saeed bin Suroor.
|
{
"evidence": [
"Moon Ballad_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMoon Ballad(foaled 4 March 1999) is a chestnut Thoroughbred racehorse sired Singspiel, Solow who is owned by Godolphin Racing and trained by Saeed bin Suroor.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Moon Ballad_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 389,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.957",
"value": "?search="
},
{
"operation": "search",
"time": "41.913",
"value": "moon ballad"
},
{
"operation": "Now on",
"time": "42.86",
"value": "Moon Ballad"
},
{
"operation": "Highlighting",
"time": "134.71",
"value": "Moon Ballad_sentence_0"
},
{
"operation": "finish",
"time": "157.813",
"value": "finish"
}
]
}
|
75237
|
An average San Francisco Giants player has a Run/s Batted In or RBI record of more than 1,000 from 1926-1947.
|
{
"evidence": [
"San Francisco Giants_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAn average San Francisco Giants player has a Run/s Batted In or RBI record of more than 1,000 from 1926-1947.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"San Francisco Giants_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 390,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.136",
"value": "?search="
},
{
"operation": "Now on",
"time": "15.189",
"value": "?search="
},
{
"operation": "search",
"time": "33.352",
"value": "Contains...average San Francisco Giants player"
},
{
"operation": "Now on",
"time": "35.711",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "38.427",
"value": "average san francisco giants players"
},
{
"operation": "Now on",
"time": "39.451",
"value": "undefined"
},
{
"operation": "search",
"time": "75.38",
"value": "San Francisoc Giants rbi average"
},
{
"operation": "Now on",
"time": "76.523",
"value": "undefined"
},
{
"operation": "search",
"time": "91.432",
"value": "San Francisco giants"
},
{
"operation": "Now on",
"time": "97.266",
"value": "San Francisco Giants"
},
{
"operation": "Highlighting",
"time": "106.931",
"value": "San Francisco Giants_sentence_0"
},
{
"operation": "finish",
"time": "154.591",
"value": "finish"
}
]
}
|
93885
|
In February 2007 Domenici indicated his intention to run for re-election, changed his mind in October and in 2008 Pearce and Lujan won their respective nomination contests.
|
{
"evidence": [
"2008 United States Senate election in New Mexico_sentence_10"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn February 2007 Domenici indicated his intention to run for re-election, changed his mind in October and in 2008 Pearce and Lujan won their respective nomination contests.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"2008 United States Senate election in New Mexico_sentence_10"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 391,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.047",
"value": "?search="
},
{
"operation": "search",
"time": "14.715",
"value": "In February 2007 Domenici indicated"
},
{
"operation": "Now on",
"time": "16.87",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "23.232",
"value": "Background 2008 United States Senate election in New Mexico_section_0"
},
{
"operation": "Now on",
"time": "24.584",
"value": "2008 United States Senate election in New Mexico#Background 2008 United States Senate election in New Mexico section 0"
},
{
"operation": "Highlighting",
"time": "26.358",
"value": "2008 United States Senate election in New Mexico_sentence_7"
},
{
"operation": "Highlighting",
"time": "28.76",
"value": "2008 United States Senate election in New Mexico_sentence_8"
},
{
"operation": "Highlighting",
"time": "38.913",
"value": "2008 United States Senate election in New Mexico_sentence_10"
},
{
"operation": "Highlighting deleted",
"time": "39.6",
"value": "2008 United States Senate election in New Mexico_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "40.012",
"value": "2008 United States Senate election in New Mexico_sentence_7"
},
{
"operation": "finish",
"time": "47.679",
"value": "finish"
}
]
}
|
37669
|
"David and Goliath," which aired on BBC One, a British free-to-air television network on February 13th, of 2014 had 4.72 million viewers in the UK.
|
{
"evidence": [
"BBC One_sentence_3",
"David and Goliath (1960 film)_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n\"David and Goliath,\" which aired on BBC One, a British free-to-air television network on February 13th, of 2014 had 4.72 million viewers in the UK.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"BBC One_sentence_3",
"David and Goliath (1960 film)_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 392,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.642",
"value": "?search="
},
{
"operation": "search",
"time": "19.237",
"value": "David and Goliath"
},
{
"operation": "Now on",
"time": "21.27",
"value": "Goliath"
},
{
"operation": "search",
"time": "45.934",
"value": "David and Goliath (1960 film)"
},
{
"operation": "Now on",
"time": "46.53",
"value": "David and Goliath (1960 film)"
},
{
"operation": "search",
"time": "60.997",
"value": "David and Goliath 2014"
},
{
"operation": "Now on",
"time": "62.698",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "65.865",
"value": "Davids and Goliath (2014 film)"
},
{
"operation": "Now on",
"time": "66.306",
"value": "Human Harvest (film)"
},
{
"operation": "hyperlink",
"time": "151.5",
"value": "Davids and Goliath (2014 film)"
},
{
"operation": "Now on",
"time": "152.048",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "153.55",
"value": "Human Harvest (film)"
},
{
"operation": "Now on",
"time": "153.851",
"value": "Human Harvest (film)"
},
{
"operation": "search",
"time": "162.71",
"value": "David and Goliath BBC"
},
{
"operation": "Now on",
"time": "164.004",
"value": "undefined"
},
{
"operation": "search",
"time": "185.684",
"value": "David and Goliath (disambiguation)"
},
{
"operation": "Now on",
"time": "186.688",
"value": "David and Goliath (disambiguation)"
},
{
"operation": "Highlighting",
"time": "200.41",
"value": "David and Goliath (disambiguation)_item_0_1"
},
{
"operation": "Highlighting",
"time": "200.412",
"value": "David and Goliath (disambiguation)_title"
},
{
"operation": "Highlighting deleted",
"time": "201.94",
"value": "David and Goliath (disambiguation)_item_0_1"
},
{
"operation": "Highlighting deleted",
"time": "201.942",
"value": "David and Goliath (disambiguation)_title"
},
{
"operation": "hyperlink",
"time": "216.931",
"value": "David and Goliath (1960 film)"
},
{
"operation": "Now on",
"time": "217.311",
"value": "David and Goliath (1960 film)"
},
{
"operation": "search",
"time": "251.081",
"value": "Contains...David and Goliath BBC one"
},
{
"operation": "Now on",
"time": "251.788",
"value": "undefined"
},
{
"operation": "search",
"time": "273.812",
"value": "BBC One"
},
{
"operation": "Now on",
"time": "274.92",
"value": "BBC One"
},
{
"operation": "hyperlink",
"time": "279.825",
"value": "BBC Radio 1"
},
{
"operation": "Now on",
"time": "280.595",
"value": "BBC Radio 1"
},
{
"operation": "back-button-clicked",
"time": "298.125",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "298.243",
"value": "BBC One"
},
{
"operation": "Highlighting",
"time": "308.221",
"value": "BBC One_sentence_3"
},
{
"operation": "search",
"time": "430.896",
"value": "David and Goliath (Artemisia Gentileschi)"
},
{
"operation": "Now on",
"time": "431.555",
"value": "David and Goliath (Artemisia Gentileschi)"
},
{
"operation": "search",
"time": "442.741",
"value": "Contains...David and Goliath"
},
{
"operation": "Now on",
"time": "443.557",
"value": "undefined"
},
{
"operation": "search",
"time": "462.443",
"value": "David and Goliath"
},
{
"operation": "Now on",
"time": "462.906",
"value": "David and Goliath"
},
{
"operation": "search",
"time": "470.904",
"value": "David and Goliath (1960 film)"
},
{
"operation": "Now on",
"time": "471.668",
"value": "David and Goliath (1960 film)"
},
{
"operation": "Highlighting",
"time": "476.857",
"value": "David and Goliath (1960 film)_sentence_0"
},
{
"operation": "finish",
"time": "518.295",
"value": "finish"
}
]
}
|
52530
|
Arsim Gashi was born on 4 December 1983, and his height is 6 ft. 1 in.
|
{
"evidence": [
"Arsim Gashi_sentence_0"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nArsim Gashi was born on 4 December 1983, and his height is 6 ft. 1 in.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Arsim Gashi_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 393,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.11",
"value": "?search="
},
{
"operation": "search",
"time": "10.042",
"value": "Arsim Gashi"
},
{
"operation": "Now on",
"time": "10.834",
"value": "Arsim Gashi"
},
{
"operation": "Highlighting",
"time": "20.247",
"value": "Arsim Gashi_sentence_0"
},
{
"operation": "finish",
"time": "30.484",
"value": "finish"
}
]
}
|
exfever_6116
|
Colm Feore received an honorary Doctor of Humanities degree from the University of America.
|
{
"evidence": [
"Colm Feore",
"Stratford Festival"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nColm Feore received an honorary Doctor of Humanities degree from the University of America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Colm Feore",
"Stratford Festival"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 394,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_7185
|
In Secret is an American erotic thriller romance film starring an English actor who is best known for playing Harry Potter in the film adaptations of the Harry Potter fantasy novels by J. K. Rowling.
|
{
"evidence": [
"In Secret",
"Tom Felton"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn Secret is an American erotic thriller romance film starring an English actor who is best known for playing Harry Potter in the film adaptations of the Harry Potter fantasy novels by J. K. Rowling.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"In Secret",
"Tom Felton"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 395,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
70934
|
Murray Canal is a canal in the municipalities of Quinte West and Brighton, Canada which is a country in the northern part of North America.
|
{
"evidence": [
"Murray Canal_sentence_1",
"Canada_sentence_2"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMurray Canal is a canal in the municipalities of Quinte West and Brighton, Canada which is a country in the northern part of North America.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Murray Canal_sentence_1",
"Canada_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 396,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.937",
"value": "?search="
},
{
"operation": "search",
"time": "13.47",
"value": "Murray Canal"
},
{
"operation": "Now on",
"time": "15.142",
"value": "Murray Canal"
},
{
"operation": "Highlighting",
"time": "19.762",
"value": "Murray Canal_sentence_1"
},
{
"operation": "hyperlink",
"time": "31.312",
"value": "Canada"
},
{
"operation": "Now on",
"time": "42.049",
"value": "Canada"
},
{
"operation": "hyperlink",
"time": "42.335",
"value": "Canada (disambiguation)"
},
{
"operation": "Now on",
"time": "42.986",
"value": "Canada (disambiguation)"
},
{
"operation": "back-button-clicked",
"time": "59.366",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "59.536",
"value": "Canada"
},
{
"operation": "Highlighting",
"time": "60.786",
"value": "Canada_sentence_2"
},
{
"operation": "finish",
"time": "66.307",
"value": "finish"
}
]
}
|
exfever_2254
|
Malaysia is the third federal constitutional monarchy in Southeast Asia.
|
{
"evidence": [
"Malaysia",
"Kuala Lumpur"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMalaysia is the third federal constitutional monarchy in Southeast Asia.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Malaysia",
"Kuala Lumpur"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 397,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
52369
|
Tuan Ibrahim, an alim (Islamic religious scholar), contested and lost the Pahang State Legislative Assembly seat of Jengka in the 1999 election.
|
{
"evidence": [
"Tuan Ibrahim_sentence_12"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nTuan Ibrahim, an alim (Islamic religious scholar), contested and lost the Pahang State Legislative Assembly seat of Jengka in the 1999 election.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Tuan Ibrahim_sentence_12"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 398,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.806",
"value": "?search="
},
{
"operation": "search",
"time": "21.434",
"value": "Tuan Ibrahim"
},
{
"operation": "Now on",
"time": "22.638",
"value": "Tuan Ibrahim"
},
{
"operation": "Page search",
"time": "38.251",
"value": "alim"
},
{
"operation": "Highlighting",
"time": "46.421",
"value": "Tuan Ibrahim_sentence_11"
},
{
"operation": "Highlighting",
"time": "56.688",
"value": "Tuan Ibrahim_sentence_12"
},
{
"operation": "Highlighting deleted",
"time": "60.381",
"value": "Tuan Ibrahim_sentence_11"
},
{
"operation": "finish",
"time": "239.619",
"value": "finish"
}
]
}
|
69638
|
Mark Whalberg won three professional boxing matches, two of which via TKO, in 1993.
|
{
"evidence": [
"Mark Wahlberg_sentence_64",
"Mark Wahlberg_sentence_4"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMark Whalberg won three professional boxing matches, two of which via TKO, in 1993.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Mark Wahlberg_sentence_64",
"Mark Wahlberg_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 399,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.05",
"value": "?search="
},
{
"operation": "search",
"time": "12.197",
"value": "Mark Whalberg"
},
{
"operation": "Now on",
"time": "14.819",
"value": "Mark Whalberg"
},
{
"operation": "Highlighting",
"time": "50.71",
"value": "Mark Wahlberg_sentence_64"
},
{
"operation": "Highlighting",
"time": "80.419",
"value": "Mark Wahlberg_sentence_4"
},
{
"operation": "finish",
"time": "92.138",
"value": "finish"
}
]
}
|
2420
|
In 1996, Edmond Plawczyk set the third fastest world snowboard speed performance at 168 km/hr (40 km/hr faster than a running cheetah, the fastest land animal), then in 1999, he completed the mountain Surf Challenge.
|
{
"evidence": [
"Edmond Plawczyk_sentence_8",
"Edmond Plawczyk_sentence_7",
"Cheetah_sentence_3"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn 1996, Edmond Plawczyk set the third fastest world snowboard speed performance at 168 km/hr (40 km/hr faster than a running cheetah, the fastest land animal), then in 1999, he completed the mountain Surf Challenge.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Edmond Plawczyk_sentence_8",
"Edmond Plawczyk_sentence_7",
"Cheetah_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 400,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "12.572",
"value": "Edmond Plawczyk"
},
{
"operation": "Highlighting",
"time": "25.135",
"value": "Edmond Plawczyk_sentence_8"
},
{
"operation": "Highlighting",
"time": "39.895",
"value": "Edmond Plawczyk_sentence_7"
},
{
"operation": "search",
"time": "47.646",
"value": "Cheetah"
},
{
"operation": "Highlighting",
"time": "54.519",
"value": "Cheetah_sentence_3"
},
{
"operation": "finish",
"time": "114.895",
"value": "finish"
}
]
}
|
13622
|
David Bar Katz is a Primetime Emmy Award for Outstanding Variety, Music or Comedy Special nominated British screenwriter, playwright, author, director and philanthropist who is a member of Israel Horovitz's New York Playwright's Lab and New York City's LAByrinth Theater Company.
|
{
"evidence": [
"David Bar Katz_sentence_0"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDavid Bar Katz is a Primetime Emmy Award for Outstanding Variety, Music or Comedy Special nominated British screenwriter, playwright, author, director and philanthropist who is a member of Israel Horovitz's New York Playwright's Lab and New York City's LAByrinth Theater Company.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"David Bar Katz_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 401,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.044",
"value": "?search="
},
{
"operation": "search",
"time": "13.329",
"value": "David Bar Katz"
},
{
"operation": "Now on",
"time": "14.231",
"value": "David Bar Katz"
},
{
"operation": "Highlighting",
"time": "81.385",
"value": "David Bar Katz_sentence_0"
},
{
"operation": "finish",
"time": "90.023",
"value": "finish"
}
]
}
|
10103
|
Canadian professional wrestler Bobby Roode debuted in 1998 for WWE but was in his first video game TNA Impact ten years later and appeared in 6 other video games since then.
|
{
"evidence": [
"Bobby Roode_sentence_8"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nCanadian professional wrestler Bobby Roode debuted in 1998 for WWE but was in his first video game TNA Impact ten years later and appeared in 6 other video games since then.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bobby Roode_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 402,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.526",
"value": "?search="
},
{
"operation": "search",
"time": "5.592",
"value": "Bobby Roode"
},
{
"operation": "Now on",
"time": "7.086",
"value": "Bobby Roode"
},
{
"operation": "Page search",
"time": "9.585",
"value": "debute"
},
{
"operation": "Highlighting",
"time": "13.305",
"value": "Bobby Roode_sentence_8"
},
{
"operation": "finish",
"time": "64.818",
"value": "finish"
}
]
}
|
exfever_1867
|
Star Wars directed in 1977 by George Lucas is a TV show that has a style of filmmaking with large-scale, sweeping scope and spectacle.
|
{
"evidence": [
"Star Wars (film)",
"Epic film"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nStar Wars directed in 1977 by George Lucas is a TV show that has a style of filmmaking with large-scale, sweeping scope and spectacle.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Star Wars (film)",
"Epic film"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 403,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_1088
|
Gemini is an Indian Tamil-language crime action film that features Vikram and an Indian actress who acted in Tamil, Hindi, Malayalam, Telugu, and Kannada language films.
|
{
"evidence": [
"Gemini (2002 Tamil film)",
"Kiran Rathod"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nGemini is an Indian Tamil-language crime action film that features Vikram and an Indian actress who acted in Tamil, Hindi, Malayalam, Telugu, and Kannada language films.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Gemini (2002 Tamil film)",
"Kiran Rathod"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 404,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
82055
|
Old Carbrook State School served as a location for German church services and nursery for German children.
|
{
"evidence": [
"Old Carbrook State School_sentence_27"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nOld Carbrook State School served as a location for German church services and nursery for German children.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Old Carbrook State School_sentence_27"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 405,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.567",
"value": "?search="
},
{
"operation": "search",
"time": "23.724",
"value": "Old Carbrook State School"
},
{
"operation": "Now on",
"time": "26.77",
"value": "Old Carbrook State School"
},
{
"operation": "Highlighting",
"time": "131.532",
"value": "Old Carbrook State School_sentence_27"
},
{
"operation": "Highlighting deleted",
"time": "159.258",
"value": "Old Carbrook State School_sentence_27"
},
{
"operation": "Highlighting",
"time": "208.157",
"value": "Old Carbrook State School_sentence_27"
},
{
"operation": "Highlighting",
"time": "262.093",
"value": "Old Carbrook State School_sentence_65"
},
{
"operation": "Highlighting deleted",
"time": "267.396",
"value": "Old Carbrook State School_sentence_65"
},
{
"operation": "finish",
"time": "331.276",
"value": "finish"
}
]
}
|
exfever_4869
|
James Blake was the first English musician to release a solo album in America.
|
{
"evidence": [
"James Blake (musician)",
"CMYK (EP)",
"United States"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJames Blake was the first English musician to release a solo album in America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"James Blake (musician)",
"CMYK (EP)",
"United States"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 406,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
48038
|
Faugli Mawa is a village in Dhar Kalan, a village and administrative headquarter of Dhar Kalan tehsil in Pathankot district of Punjab State, India, has 65 children under the age of 6 years which is 15.12% of total population of the village, while most of the people are from Schedule Caste which constitutes 33.49% of total population in the village, but does not have any Schedule Tribe so far.
|
{
"evidence": [
"Faugli Mawa_sentence_0",
"Dhar Kalan_sentence_0",
"Faugli Mawa_sentence_5",
"Faugli Mawa_sentence_6",
"Faugli Mawa_sentence_7"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nFaugli Mawa is a village in Dhar Kalan, a village and administrative headquarter of Dhar Kalan tehsil in Pathankot district of Punjab State, India, has 65 children under the age of 6 years which is 15.12% of total population of the village, while most of the people are from Schedule Caste which constitutes 33.49% of total population in the village, but does not have any Schedule Tribe so far.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Faugli Mawa_sentence_0",
"Dhar Kalan_sentence_0",
"Faugli Mawa_sentence_5",
"Faugli Mawa_sentence_6",
"Faugli Mawa_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 407,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.137",
"value": "?search="
},
{
"operation": "search",
"time": "15.382",
"value": "Faugli Mawa"
},
{
"operation": "Now on",
"time": "16.056",
"value": "Faugli Mawa"
},
{
"operation": "Highlighting",
"time": "54.37",
"value": "Faugli Mawa_sentence_0"
},
{
"operation": "hyperlink",
"time": "61.643",
"value": "Dhar Kalan"
},
{
"operation": "Now on",
"time": "62.163",
"value": "Dhar Kalan"
},
{
"operation": "Highlighting",
"time": "63.819",
"value": "Dhar Kalan_sentence_0"
},
{
"operation": "search",
"time": "75.453",
"value": "Faugli Mawa"
},
{
"operation": "Now on",
"time": "76.405",
"value": "Faugli Mawa"
},
{
"operation": "Highlighting",
"time": "83.74",
"value": "Faugli Mawa_sentence_5"
},
{
"operation": "Highlighting",
"time": "93.227",
"value": "Faugli Mawa_sentence_6"
},
{
"operation": "Highlighting",
"time": "96.531",
"value": "Faugli Mawa_sentence_7"
},
{
"operation": "finish",
"time": "113.09",
"value": "finish"
}
]
}
|
78399
|
The Oshawa Generals and the Trail Smoke Eaters competed in the 1944 Memorial Cup, with Oshawa winning all four games of the series.
|
{
"evidence": [
"1944 Memorial Cup_sentence_3",
"1944 Memorial Cup_sentence_4"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Oshawa Generals and the Trail Smoke Eaters competed in the 1944 Memorial Cup, with Oshawa winning all four games of the series.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"1944 Memorial Cup_sentence_3",
"1944 Memorial Cup_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 408,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.091",
"value": "List of cities and towns in Saint Kitts and Nevis"
},
{
"operation": "search",
"time": "56.534",
"value": "Contains...Oshawa Generals"
},
{
"operation": "Now on",
"time": "58.495",
"value": "undefined"
},
{
"operation": "search",
"time": "69.631",
"value": "1944 Memorial Cup"
},
{
"operation": "Now on",
"time": "70.813",
"value": "1944 Memorial Cup"
},
{
"operation": "Highlighting",
"time": "91.694",
"value": "1944 Memorial Cup_sentence_3"
},
{
"operation": "Highlighting",
"time": "104.299",
"value": "1944 Memorial Cup_sentence_4"
},
{
"operation": "finish",
"time": "332.691",
"value": "finish"
}
]
}
|
66263
|
Sheree Gray is an American soccer defender, she played for Sky Blue FC.
|
{
"evidence": [
"Sheree Gray_sentence_0",
"Sheree Gray_sentence_1"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSheree Gray is an American soccer defender, she played for Sky Blue FC.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sheree Gray_sentence_0",
"Sheree Gray_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 409,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.121",
"value": "?search="
},
{
"operation": "search",
"time": "14.946",
"value": "Sheree Gray"
},
{
"operation": "Now on",
"time": "20.054",
"value": "Sheree Gray"
},
{
"operation": "Highlighting",
"time": "26.236",
"value": "Sheree Gray_sentence_0"
},
{
"operation": "Highlighting",
"time": "36.972",
"value": "Sheree Gray_sentence_1"
},
{
"operation": "finish",
"time": "94.508",
"value": "finish"
}
]
}
|
13495
|
Model bacteria have been used in the study of SUPT5H function and male and female animals underwent a standardized phenotypic screen to determine the effects of deletion.
|
{
"evidence": [
"SUPT5H_sentence_2",
"Model organism_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nModel bacteria have been used in the study of SUPT5H function and male and female animals underwent a standardized phenotypic screen to determine the effects of deletion.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"SUPT5H_sentence_2",
"Model organism_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 410,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.535",
"value": "?search="
},
{
"operation": "search",
"time": "20.19",
"value": "SUPT5H"
},
{
"operation": "Now on",
"time": "20.872",
"value": "SUPT5H"
},
{
"operation": "Highlighting",
"time": "31.528",
"value": "SUPT5H_sentence_2"
},
{
"operation": "Highlighting",
"time": "37.54",
"value": "SUPT5H_sentence_4"
},
{
"operation": "hyperlink",
"time": "51.533",
"value": "Model organisms"
},
{
"operation": "Now on",
"time": "52.291",
"value": "Model organism"
},
{
"operation": "Highlighting",
"time": "69.714",
"value": "Model organism_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "77.513",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "77.611",
"value": "SUPT5H"
},
{
"operation": "Highlighting deleted",
"time": "80.661",
"value": "SUPT5H_sentence_4"
},
{
"operation": "finish",
"time": "92.758",
"value": "finish"
}
]
}
|
26885
|
The comprehension approach refers to a method of learning a new language through the process of understanding the meaning of words and expressions in the language, and it usually involves a silent period when the learner tries to assimilate the various meanings of the words.
|
{
"evidence": [
"Comprehension approach_sentence_7",
"Comprehension approach_sentence_10"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe comprehension approach refers to a method of learning a new language through the process of understanding the meaning of words and expressions in the language, and it usually involves a silent period when the learner tries to assimilate the various meanings of the words.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Comprehension approach_sentence_7",
"Comprehension approach_sentence_10"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 411,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.928",
"value": "?search="
},
{
"operation": "search",
"time": "11.095",
"value": "comprehension approach"
},
{
"operation": "Now on",
"time": "12.64",
"value": "Comprehension approach"
},
{
"operation": "Highlighting",
"time": "74.597",
"value": "Comprehension approach_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "89.781",
"value": "Comprehension approach_sentence_0"
},
{
"operation": "Highlighting",
"time": "97.655",
"value": "Comprehension approach_sentence_7"
},
{
"operation": "Highlighting",
"time": "126.041",
"value": "Comprehension approach_sentence_10"
},
{
"operation": "finish",
"time": "857.052",
"value": "finish"
}
]
}
|
21448
|
The Missyaf Fault, which is located north of Yammouneh Fault, is the location where the 115 Antioch earthquake, which occurred on December 115 and killed more than 250,000 people, was identified and is most likely responsible for the 1170 Syria Earthquake.
|
{
"evidence": [
"Dead Sea Transform_sentence_57",
"115 Antioch earthquake_sentence_0",
"115 Antioch earthquake_sentence_9",
"115 Antioch earthquake_sentence_21"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Missyaf Fault, which is located north of Yammouneh Fault, is the location where the 115 Antioch earthquake, which occurred on December 115 and killed more than 250,000 people, was identified and is most likely responsible for the 1170 Syria Earthquake.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Dead Sea Transform_sentence_57",
"115 Antioch earthquake_sentence_0",
"115 Antioch earthquake_sentence_9",
"115 Antioch earthquake_sentence_21"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 412,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.659",
"value": "?search="
},
{
"operation": "search",
"time": "15.531",
"value": "The Missyaf Fault"
},
{
"operation": "Now on",
"time": "16.072",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "34.011",
"value": "Missyaf fault Dead Sea Transform_section_15"
},
{
"operation": "Now on",
"time": "34.737",
"value": "Dead Sea Transform#Missyaf fault Dead Sea Transform section 15"
},
{
"operation": "Highlighting",
"time": "69.211",
"value": "Dead Sea Transform_sentence_57"
},
{
"operation": "search",
"time": "124.467",
"value": "115 Antioch earthquake"
},
{
"operation": "Now on",
"time": "125.462",
"value": "115 Antioch earthquake"
},
{
"operation": "Highlighting",
"time": "135.539",
"value": "115 Antioch earthquake_sentence_0"
},
{
"operation": "Highlighting",
"time": "157.938",
"value": "115 Antioch earthquake_sentence_9"
},
{
"operation": "Highlighting",
"time": "185.186",
"value": "115 Antioch earthquake_sentence_21"
},
{
"operation": "finish",
"time": "195.57",
"value": "finish"
}
]
}
|
12393
|
In La frousse aux trousses, Spirou and Fantasio never tried to travel to Touboutt-Chan but the two journalists have trouble finding the funds until Doctor Placebo funded the trip.
|
{
"evidence": [
"La frousse aux trousses_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn La frousse aux trousses, Spirou and Fantasio never tried to travel to Touboutt-Chan but the two journalists have trouble finding the funds until Doctor Placebo funded the trip.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"La frousse aux trousses_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 413,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.731",
"value": "?search="
},
{
"operation": "search",
"time": "10.544",
"value": "La frousse aux trousses"
},
{
"operation": "Now on",
"time": "11.812",
"value": "La frousse aux trousses"
},
{
"operation": "Highlighting",
"time": "15.625",
"value": "La frousse aux trousses_sentence_2"
},
{
"operation": "finish",
"time": "55.959",
"value": "finish"
}
]
}
|
59206
|
Jade Record explains that after death the dead forgets their memories on Earth through the "Mirror of Reflection" and then they will be delivered to the Earth again to have the same lives.
|
{
"evidence": [
"Jade Record_sentence_4"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJade Record explains that after death the dead forgets their memories on Earth through the \"Mirror of Reflection\" and then they will be delivered to the Earth again to have the same lives.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Jade Record_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 414,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.302",
"value": "?search="
},
{
"operation": "search",
"time": "21.174",
"value": "Jade Record"
},
{
"operation": "Now on",
"time": "21.745",
"value": "Jade Record"
},
{
"operation": "Page search",
"time": "31.463",
"value": "earth"
},
{
"operation": "Page search",
"time": "31.885",
"value": "earth"
},
{
"operation": "Page search",
"time": "42.246",
"value": "memories"
},
{
"operation": "Page search",
"time": "43.585",
"value": "memorie"
},
{
"operation": "Page search",
"time": "45.409",
"value": "memory"
},
{
"operation": "Page search",
"time": "47.913",
"value": "memories"
},
{
"operation": "Page search",
"time": "62.631",
"value": "Mirror of Reflection"
},
{
"operation": "Highlighting",
"time": "66.63",
"value": "Jade Record_sentence_4"
},
{
"operation": "finish",
"time": "85.954",
"value": "finish"
}
]
}
|
45313
|
All of Bni Bouayach's inhabitants are of Berber or Amazigh origin.
|
{
"evidence": [
"Bni Bouayach_sentence_2"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAll of Bni Bouayach's inhabitants are of Berber or Amazigh origin.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bni Bouayach_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 415,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.654",
"value": "?search="
},
{
"operation": "search",
"time": "17.152",
"value": "Bni Bouayach"
},
{
"operation": "Now on",
"time": "18.276",
"value": "Bni Bouayach"
},
{
"operation": "Highlighting",
"time": "28.092",
"value": "Bni Bouayach_sentence_2"
},
{
"operation": "Page search",
"time": "58.04",
"value": "Berber"
},
{
"operation": "finish",
"time": "134.897",
"value": "finish"
}
]
}
|
exfever_4959
|
Michael J. Fox is a Canadian-American retired actor who rose to prominence as Alex P. Keaton in the sitcom Family Ties, which was produced by an American English-language commercial broadcast television and radio network headquartered in the Rockefeller Center skyscraper in Midtown Brooklyn.
|
{
"evidence": [
"Michael J. Fox",
"NBC",
"30 Rockefeller Plaza"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMichael J. Fox is a Canadian-American retired actor who rose to prominence as Alex P. Keaton in the sitcom Family Ties, which was produced by an American English-language commercial broadcast television and radio network headquartered in the Rockefeller Center skyscraper in Midtown Brooklyn.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Michael J. Fox",
"NBC",
"30 Rockefeller Plaza"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 416,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_428
|
Jerry Lewis was the second national chairman of the Muscular Dystrophy Association.
|
{
"evidence": [
"Jerry Lewis",
"Dean Martin"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJerry Lewis was the second national chairman of the Muscular Dystrophy Association.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jerry Lewis",
"Dean Martin"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 417,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_5928
|
Quartet is a 2012 British comedy-drama film based on the play Quartet by a South African-born British author, playwright, and screenwriter, best known for his plays for the British stage as well as the screenplays for The Dresser.
|
{
"evidence": [
"Quartet (2012 film)",
"Ronald Harwood"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nQuartet is a 2012 British comedy-drama film based on the play Quartet by a South African-born British author, playwright, and screenwriter, best known for his plays for the British stage as well as the screenplays for The Dresser.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Quartet (2012 film)",
"Ronald Harwood"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 418,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
51746
|
Karyn Kupcinet played the role of Amy in the 1963 series Going My Way, an American comedy-drama series starring dancer and actor Gene Kelly.
|
{
"evidence": [
"Going My Way (TV series)_sentence_1",
"Karyn Kupcinet_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nKaryn Kupcinet played the role of Amy in the 1963 series Going My Way, an American comedy-drama series starring dancer and actor Gene Kelly.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Going My Way (TV series)_sentence_1",
"Karyn Kupcinet_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 419,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.855",
"value": "?search="
},
{
"operation": "search",
"time": "11.649",
"value": "Karyn Kupcinet"
},
{
"operation": "Now on",
"time": "12.938",
"value": "Karyn Kupcinet"
},
{
"operation": "hyperlink",
"time": "65.427",
"value": "Going My Way"
},
{
"operation": "Now on",
"time": "66.417",
"value": "Going My Way (TV series)"
},
{
"operation": "hyperlink",
"time": "107.271",
"value": "Going My Way"
},
{
"operation": "Now on",
"time": "108.283",
"value": "Going My Way"
},
{
"operation": "hyperlink",
"time": "112.408",
"value": "Going My Way (TV series)"
},
{
"operation": "Now on",
"time": "112.794",
"value": "Going My Way (TV series)"
},
{
"operation": "back-button-clicked",
"time": "114.212",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "114.293",
"value": "Going My Way"
},
{
"operation": "back-button-clicked",
"time": "114.882",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "114.9",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "114.978",
"value": "Going My Way (TV series)"
},
{
"operation": "back-button-clicked",
"time": "115.892",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "115.901",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "115.967",
"value": "Karyn Kupcinet"
},
{
"operation": "Highlighting",
"time": "152.797",
"value": "Karyn Kupcinet_sentence_15"
},
{
"operation": "Highlighting",
"time": "152.801",
"value": "Karyn Kupcinet_section_1"
},
{
"operation": "hyperlink",
"time": "185.107",
"value": "Going My Way"
},
{
"operation": "Now on",
"time": "185.5",
"value": "Going My Way (TV series)"
},
{
"operation": "Highlighting",
"time": "190.056",
"value": "Going My Way (TV series)_sentence_1"
},
{
"operation": "search",
"time": "240.232",
"value": "Going My Way"
},
{
"operation": "Now on",
"time": "240.817",
"value": "Going My Way"
},
{
"operation": "back-button-clicked",
"time": "261.118",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "261.218",
"value": "Going My Way (TV series)"
},
{
"operation": "back-button-clicked",
"time": "270.353",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "270.375",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "270.458",
"value": "Karyn Kupcinet"
},
{
"operation": "Highlighting",
"time": "316.888",
"value": "Karyn Kupcinet_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "326.319",
"value": "Karyn Kupcinet_sentence_15"
},
{
"operation": "Highlighting deleted",
"time": "326.324",
"value": "Karyn Kupcinet_section_1"
},
{
"operation": "finish",
"time": "332.039",
"value": "finish"
}
]
}
|
exfever_5605
|
The Silence of the Lambs is a 1991 American psychological horror film directed by an American filmmaker.
|
{
"evidence": [
"The Silence of the Lambs (film)",
"Jonathan Demme"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Silence of the Lambs is a 1991 American psychological horror film directed by an American filmmaker.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Silence of the Lambs (film)",
"Jonathan Demme"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 420,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
47676
|
In the ITTF World Tour, Doha, Qatar had Rd of 35 in Men's singles (2015) while Spała, Poland (a village in the administrative district of Gmina Inowłódz) was the winner on the same year.
|
{
"evidence": [
"ITTF World Tour_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn the ITTF World Tour, Doha, Qatar had Rd of 35 in Men's singles (2015) while Spała, Poland (a village in the administrative district of Gmina Inowłódz) was the winner on the same year.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"ITTF World Tour_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 421,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.694",
"value": "?search="
},
{
"operation": "search",
"time": "11.237",
"value": "ITTF World Tour"
},
{
"operation": "Now on",
"time": "13.323",
"value": "ITTF World Tour"
},
{
"operation": "hyperlink",
"time": "106.784",
"value": "Qatar"
},
{
"operation": "Now on",
"time": "109.637",
"value": "Qatar"
},
{
"operation": "back-button-clicked",
"time": "110.577",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "110.814",
"value": "ITTF World Tour"
},
{
"operation": "hyperlink",
"time": "126.524",
"value": "ITTF World Tour Grand Finals"
},
{
"operation": "Now on",
"time": "127.509",
"value": "ITTF World Tour Grand Finals"
},
{
"operation": "hyperlink",
"time": "165.016",
"value": "ITTF World Tour"
},
{
"operation": "Now on",
"time": "165.518",
"value": "ITTF World Tour"
},
{
"operation": "back-button-clicked",
"time": "174.237",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "174.384",
"value": "ITTF World Tour Grand Finals"
},
{
"operation": "hyperlink",
"time": "201.144",
"value": "List of ITTF World Tour Grand Finals medalists"
},
{
"operation": "Now on",
"time": "203.233",
"value": "List of ITTF World Tour Grand Finals medalists"
},
{
"operation": "Highlighting",
"time": "229.42",
"value": "List of ITTF World Tour Grand Finals medalists_item_0_0"
},
{
"operation": "Highlighting",
"time": "229.422",
"value": "List of ITTF World Tour Grand Finals medalists_title"
},
{
"operation": "Highlighting",
"time": "229.423",
"value": "List of ITTF World Tour Grand Finals medalists_section_6"
},
{
"operation": "Highlighting deleted",
"time": "230.815",
"value": "List of ITTF World Tour Grand Finals medalists_item_0_0"
},
{
"operation": "Highlighting deleted",
"time": "230.818",
"value": "List of ITTF World Tour Grand Finals medalists_title"
},
{
"operation": "Highlighting deleted",
"time": "230.82",
"value": "List of ITTF World Tour Grand Finals medalists_section_6"
},
{
"operation": "search",
"time": "274.176",
"value": "Spała, Poland"
},
{
"operation": "Now on",
"time": "275.175",
"value": "Spała, Poland"
},
{
"operation": "search",
"time": "297.529",
"value": "ITTF World Tour"
},
{
"operation": "Now on",
"time": "298.639",
"value": "ITTF World Tour"
},
{
"operation": "search",
"time": "305.981",
"value": "ITTF World Tour German Open"
},
{
"operation": "Now on",
"time": "308.061",
"value": "German Open (table tennis)"
},
{
"operation": "back-button-clicked",
"time": "321.643",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "321.862",
"value": "ITTF World Tour"
},
{
"operation": "Highlighting",
"time": "324.587",
"value": "ITTF World Tour_sentence_0"
},
{
"operation": "hyperlink",
"time": "346.482",
"value": "2016"
},
{
"operation": "Now on",
"time": "347.985",
"value": "2016 ITTF World Tour"
},
{
"operation": "back-button-clicked",
"time": "374.745",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "374.976",
"value": "ITTF World Tour"
},
{
"operation": "finish",
"time": "385.478",
"value": "finish"
}
]
}
|
60932
|
Savage, an English minister and dissenting tutor, first thought of medicine, and spent a year or two with his Uncle Toulmin, an apothecary, in Old Gravel Lane, Wapping; then through the influence of Isaac Watts, he entered the Fund Academy.
|
{
"evidence": [
"Samuel Morton Savage_sentence_0",
"Samuel Morton Savage_sentence_4",
"Samuel Morton Savage_sentence_5"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSavage, an English minister and dissenting tutor, first thought of medicine, and spent a year or two with his Uncle Toulmin, an apothecary, in Old Gravel Lane, Wapping; then through the influence of Isaac Watts, he entered the Fund Academy.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Samuel Morton Savage_sentence_0",
"Samuel Morton Savage_sentence_4",
"Samuel Morton Savage_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 422,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.961",
"value": "?search="
},
{
"operation": "search",
"time": "11.257",
"value": "Savage, an English minister and dissenting tutor"
},
{
"operation": "Now on",
"time": "13.014",
"value": "undefined"
},
{
"operation": "search",
"time": "96.456",
"value": "Uncle Toulmin"
},
{
"operation": "Now on",
"time": "97.918",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "102.207",
"value": "Samuel Morton Savage"
},
{
"operation": "Now on",
"time": "102.655",
"value": "Samuel Morton Savage"
},
{
"operation": "Highlighting",
"time": "232.694",
"value": "Samuel Morton Savage_sentence_0"
},
{
"operation": "Highlighting",
"time": "237.415",
"value": "Samuel Morton Savage_sentence_4"
},
{
"operation": "Highlighting",
"time": "244.201",
"value": "Samuel Morton Savage_sentence_5"
},
{
"operation": "finish",
"time": "252.068",
"value": "finish"
}
]
}
|
56790
|
Lady Flash first recorded appearance was in 1975 later appeared with Reparata and performed on his first European tour
|
{
"evidence": [
"Lady Flash_sentence_7",
"Lady Flash_sentence_5"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLady Flash first recorded appearance was in 1975 later appeared with Reparata and performed on his first European tour\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Lady Flash_sentence_7",
"Lady Flash_sentence_5"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 423,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.068",
"value": "?search="
},
{
"operation": "search",
"time": "11.971",
"value": "Lady Flash"
},
{
"operation": "Now on",
"time": "12.758",
"value": "Lady Flash"
},
{
"operation": "Page search",
"time": "21.768",
"value": "appear"
},
{
"operation": "Highlighting",
"time": "25.661",
"value": "Lady Flash_sentence_5"
},
{
"operation": "Page search",
"time": "49.278",
"value": "repara"
},
{
"operation": "Page search",
"time": "74.522",
"value": "reparata"
},
{
"operation": "Page search",
"time": "150.366",
"value": "europe"
},
{
"operation": "Highlighting",
"time": "157.834",
"value": "Lady Flash_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "160.524",
"value": "Lady Flash_sentence_5"
},
{
"operation": "Highlighting",
"time": "161.453",
"value": "Lady Flash_sentence_5"
},
{
"operation": "finish",
"time": "316.581",
"value": "finish"
}
]
}
|
90973
|
Hubbard married Ann E. Chamberlin of Louisiana in late 1860s, then he was a clerk of the West Virginia House of Delegates from 1866 to 1870.
|
{
"evidence": [
"William Pallister Hubbard_sentence_8",
"William Pallister Hubbard_sentence_9"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHubbard married Ann E. Chamberlin of Louisiana in late 1860s, then he was a clerk of the West Virginia House of Delegates from 1866 to 1870.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"William Pallister Hubbard_sentence_8",
"William Pallister Hubbard_sentence_9"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 424,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.271",
"value": "?search="
},
{
"operation": "search",
"time": "20.782",
"value": "\"Ann E. Chamberlin\""
},
{
"operation": "Now on",
"time": "21.863",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "23.598",
"value": "William Pallister Hubbard"
},
{
"operation": "Now on",
"time": "24.189",
"value": "William Pallister Hubbard"
},
{
"operation": "Page search",
"time": "27.853",
"value": "Ann E. Chamberlin"
},
{
"operation": "Highlighting",
"time": "29.903",
"value": "William Pallister Hubbard_sentence_8"
},
{
"operation": "Highlighting",
"time": "34.425",
"value": "William Pallister Hubbard_sentence_9"
},
{
"operation": "finish",
"time": "51.828",
"value": "finish"
}
]
}
|
exfever_8733
|
Leslie Uggams is an American actress who is recognized for portraying a role in a television series that tells a story in a predetermined, limited number of episodes.
|
{
"evidence": [
"Leslie Uggams",
"Miniseries"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLeslie Uggams is an American actress who is recognized for portraying a role in a television series that tells a story in a predetermined, limited number of episodes.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Leslie Uggams",
"Miniseries"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 425,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_1302
|
Philip Seymour Hoffman was an actor, director, and producer who gained recognition for his supporting work in a drama film adapted by a screenwriter and playwright who received three Academy Awards.
|
{
"evidence": [
"Philip Seymour Hoffman",
"Scent of a Woman (1992 film)",
"Bo Goldman"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPhilip Seymour Hoffman was an actor, director, and producer who gained recognition for his supporting work in a drama film adapted by a screenwriter and playwright who received three Academy Awards.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Philip Seymour Hoffman",
"Scent of a Woman (1992 film)",
"Bo Goldman"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 426,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
57340
|
There were 40 animated films released in 1992 including Aladdin which ranked at number one and Porco Rosso which ranked at number two.
|
{
"evidence": [
"Aladdin (1992 Disney film)_sentence_2"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThere were 40 animated films released in 1992 including Aladdin which ranked at number one and Porco Rosso which ranked at number two.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Aladdin (1992 Disney film)_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 427,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.479",
"value": "?search="
},
{
"operation": "search",
"time": "17.027",
"value": "Aladdin"
},
{
"operation": "Now on",
"time": "17.727",
"value": "Aladdin"
},
{
"operation": "hyperlink",
"time": "25.246",
"value": "Aladdin (disambiguation)"
},
{
"operation": "Now on",
"time": "26.33",
"value": "Aladdin (disambiguation)"
},
{
"operation": "hyperlink",
"time": "33.638",
"value": "1.1 Disney franchise Aladdin (disambiguation)_section_1"
},
{
"operation": "hyperlink",
"time": "39.802",
"value": "Aladdin (1992 Disney film)"
},
{
"operation": "Now on",
"time": "41.05",
"value": "Aladdin (1992 Disney film)"
},
{
"operation": "Highlighting",
"time": "81.635",
"value": "Aladdin (1992 Disney film)_sentence_2"
},
{
"operation": "search",
"time": "128.645",
"value": "animated films released in 1992"
},
{
"operation": "Now on",
"time": "132.675",
"value": "undefined"
},
{
"operation": "search",
"time": "163.962",
"value": "animated films"
},
{
"operation": "Now on",
"time": "167.618",
"value": "Animation"
},
{
"operation": "search",
"time": "198.236",
"value": "Porco Rosso"
},
{
"operation": "Now on",
"time": "215.556",
"value": "Porco Rosso"
},
{
"operation": "Now on",
"time": "325.23",
"value": "Porco Rosso"
},
{
"operation": "finish",
"time": "333.91",
"value": "finish"
}
]
}
|
91976
|
Jordan Peterson is a Canadian clinical psychologist which is an integration of science, theory and clinical knowledge for the purpose of understanding, preventing, and relieving psychologically-based distress or dysfunction.
|
{
"evidence": [
"Jordan Peterson_sentence_0",
"Clinical psychology_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJordan Peterson is a Canadian clinical psychologist which is an integration of science, theory and clinical knowledge for the purpose of understanding, preventing, and relieving psychologically-based distress or dysfunction.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Jordan Peterson_sentence_0",
"Clinical psychology_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 428,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.801",
"value": "undefined"
},
{
"operation": "search",
"time": "23.098",
"value": "Jordan Peterson"
},
{
"operation": "Now on",
"time": "24.321",
"value": "Jordan Peterson"
},
{
"operation": "Highlighting",
"time": "30.968",
"value": "Jordan Peterson_sentence_0"
},
{
"operation": "hyperlink",
"time": "43.534",
"value": "clinical psychologist"
},
{
"operation": "Now on",
"time": "45.631",
"value": "Clinical psychologist"
},
{
"operation": "Highlighting",
"time": "48.117",
"value": "Clinical psychology_sentence_0"
},
{
"operation": "finish",
"time": "77.266",
"value": "finish"
}
]
}
|
exfever_4335
|
The Crown is a historical television series about Queen Elizabeth II that was produced by an American film and television production company that was formed in 2007,?the first British media company to receive investment from BBC Worldwide.
|
{
"evidence": [
"The Crown (TV series)",
"Left Bank Pictures"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Crown is a historical television series about Queen Elizabeth II that was produced by an American film and television production company that was formed in 2007,?the first British media company to receive investment from BBC Worldwide.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Crown (TV series)",
"Left Bank Pictures"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 429,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
68449
|
Katherine Glass played as Allison MacKenzie (neither a fictional character nor one of the protagonists in the novel Peyton Place) in the daytime soap opera Return to Peyton Place from 1972 to 1973; then as Kim Jordan in The Best of Everything in 1970.
|
{
"evidence": [
"Allison MacKenzie_sentence_0"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nKatherine Glass played as Allison MacKenzie (neither a fictional character nor one of the protagonists in the novel Peyton Place) in the daytime soap opera Return to Peyton Place from 1972 to 1973; then as Kim Jordan in The Best of Everything in 1970.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Allison MacKenzie_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 430,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.324",
"value": "?search="
},
{
"operation": "search",
"time": "9.002",
"value": "Katherine Glass"
},
{
"operation": "Now on",
"time": "10.634",
"value": "Katherine Glass"
},
{
"operation": "Highlighting",
"time": "28.153",
"value": "Katherine Glass_cell_1_3_2"
},
{
"operation": "Highlighting",
"time": "28.154",
"value": "Katherine Glass_title"
},
{
"operation": "Highlighting",
"time": "28.155",
"value": "Katherine Glass_section_2"
},
{
"operation": "Highlighting",
"time": "28.156",
"value": "Katherine Glass_header_cell_1_1_2"
},
{
"operation": "Highlighting",
"time": "28.157",
"value": "Katherine Glass_header_cell_1_0_0"
},
{
"operation": "Highlighting",
"time": "32.664",
"value": "Katherine Glass_cell_1_3_1"
},
{
"operation": "Highlighting",
"time": "32.668",
"value": "Katherine Glass_header_cell_1_1_1"
},
{
"operation": "Highlighting",
"time": "35.152",
"value": "Katherine Glass_cell_1_3_0"
},
{
"operation": "Highlighting",
"time": "35.154",
"value": "Katherine Glass_header_cell_1_1_0"
},
{
"operation": "Highlighting",
"time": "39.648",
"value": "Katherine Glass_cell_1_2_2"
},
{
"operation": "Highlighting",
"time": "40.376",
"value": "Katherine Glass_cell_1_2_1"
},
{
"operation": "Highlighting",
"time": "41.128",
"value": "Katherine Glass_cell_1_2_0"
},
{
"operation": "hyperlink",
"time": "44.671",
"value": "Allison MacKenzie"
},
{
"operation": "Now on",
"time": "45.398",
"value": "Allison MacKenzie"
},
{
"operation": "Highlighting",
"time": "48.56",
"value": "Allison MacKenzie_sentence_0"
},
{
"operation": "finish",
"time": "62.653",
"value": "finish"
}
]
}
|
exfever_6585
|
Heikegani is a species of crab believed that these are the Heike warriors' new life in the same physical form or body after biological death.
|
{
"evidence": [
"Heikegani",
"Reincarnation"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHeikegani is a species of crab believed that these are the Heike warriors' new life in the same physical form or body after biological death.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Heikegani",
"Reincarnation"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 431,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
29769
|
There are some clubs from western Bavaria playing in the Württemberg football league system and the SpVgg Au/Iller played in the Verbandsliga during the 2009-10 season.
|
{
"evidence": [
"Verbandsliga Württemberg_sentence_9",
"Verbandsliga Württemberg_sentence_10"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThere are some clubs from western Bavaria playing in the Württemberg football league system and the SpVgg Au/Iller played in the Verbandsliga during the 2009-10 season.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Verbandsliga Württemberg_sentence_9",
"Verbandsliga Württemberg_sentence_10"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 432,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.677",
"value": "?search="
},
{
"operation": "search",
"time": "75.682",
"value": "2009-10 VERBANDSLIGA"
},
{
"operation": "Now on",
"time": "77.295",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "87.529",
"value": "Verbandsliga Hessen-Süd"
},
{
"operation": "Now on",
"time": "88.12",
"value": "Verbandsliga Hessen-Süd"
},
{
"operation": "search",
"time": "122.047",
"value": "Contains...2009-10 SpVgg Au/Iller"
},
{
"operation": "Now on",
"time": "122.955",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "133.364",
"value": "Swabian football league system"
},
{
"operation": "Now on",
"time": "133.905",
"value": "Swabian football league system"
},
{
"operation": "search",
"time": "162.086",
"value": "Contains... Württemberg football league"
},
{
"operation": "Now on",
"time": "164.119",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "201.401",
"value": "Verbandsliga Württemberg"
},
{
"operation": "Now on",
"time": "203.132",
"value": "Verbandsliga Württemberg"
},
{
"operation": "Highlighting",
"time": "252.129",
"value": "Verbandsliga Württemberg_sentence_9"
},
{
"operation": "Highlighting",
"time": "261.648",
"value": "Verbandsliga Württemberg_sentence_10"
},
{
"operation": "finish",
"time": "293.146",
"value": "finish"
}
]
}
|
exfever_5883
|
Jazz music was originated in San Francisco.
|
{
"evidence": [
"Rock music",
"Rhythm and blues"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJazz music was originated in San Francisco.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Rock music",
"Rhythm and blues"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 433,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
78070
|
Writing a total of nine scripts from 1954-1984, Michel Foucault wrote the Archaeology of Knowledge in 1969.
|
{
"evidence": [
"Michel Foucault_sentence_9"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWriting a total of nine scripts from 1954-1984, Michel Foucault wrote the Archaeology of Knowledge in 1969.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Michel Foucault_sentence_9"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 434,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.712",
"value": "?search="
},
{
"operation": "search",
"time": "13.992",
"value": "Michel Foucault"
},
{
"operation": "Now on",
"time": "14.964",
"value": "Michel Foucault"
},
{
"operation": "Page search",
"time": "35.101",
"value": "scripts"
},
{
"operation": "Page search",
"time": "60.7",
"value": "archaeology"
},
{
"operation": "Highlighting",
"time": "83.996",
"value": "Michel Foucault_sentence_9"
},
{
"operation": "Highlighting",
"time": "131.714",
"value": "Michel Foucault_sentence_12"
},
{
"operation": "Highlighting deleted",
"time": "139.584",
"value": "Michel Foucault_sentence_12"
},
{
"operation": "Page search",
"time": "209.515",
"value": "script"
},
{
"operation": "search",
"time": "290.187",
"value": "Script"
},
{
"operation": "Now on",
"time": "290.921",
"value": "Script"
},
{
"operation": "hyperlink",
"time": "318.216",
"value": "Manuscript"
},
{
"operation": "Now on",
"time": "318.926",
"value": "Manuscript"
},
{
"operation": "finish",
"time": "325.27",
"value": "finish"
}
]
}
|
40935
|
Adama Diomande has the most goals scored in the record of football as he was able to score a total of 105 goals from all of his 281 appearances in games.
|
{
"evidence": [
"List of footballers with 500 or more goals_sentence_0"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAdama Diomande has the most goals scored in the record of football as he was able to score a total of 105 goals from all of his 281 appearances in games.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"List of footballers with 500 or more goals_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 435,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.174",
"value": "?search="
},
{
"operation": "search",
"time": "9.704",
"value": "Adama Diomande"
},
{
"operation": "Now on",
"time": "10.907",
"value": "Adama Diomande"
},
{
"operation": "search",
"time": "83.841",
"value": "most goals football"
},
{
"operation": "Now on",
"time": "85.757",
"value": "undefined"
},
{
"operation": "search",
"time": "130.447",
"value": "footballer goals"
},
{
"operation": "Now on",
"time": "131.256",
"value": "undefined"
},
{
"operation": "search",
"time": "160.504",
"value": "footballers most goals"
},
{
"operation": "Now on",
"time": "161.454",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "164.872",
"value": "List of footballers with 500 or more goals"
},
{
"operation": "Now on",
"time": "166.047",
"value": "List of footballers with 500 or more goals"
},
{
"operation": "Highlighting",
"time": "191.642",
"value": "List of footballers with 500 or more goals_sentence_0"
},
{
"operation": "finish",
"time": "207.704",
"value": "finish"
}
]
}
|
18168
|
Roman Romulo, who authored (and later disavowed) two education bills, has been endorsed by the man who removed Ferdinand Marcos from power in the Philippines.
|
{
"evidence": [
"Roman Romulo_sentence_25",
"Roman Romulo_sentence_22",
"Jovito Salonga_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRoman Romulo, who authored (and later disavowed) two education bills, has been endorsed by the man who removed Ferdinand Marcos from power in the Philippines.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Roman Romulo_sentence_25",
"Roman Romulo_sentence_22",
"Jovito Salonga_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 436,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.623",
"value": "?search="
},
{
"operation": "search",
"time": "10.187",
"value": "Roman Romulo"
},
{
"operation": "Now on",
"time": "11.041",
"value": "Roman Romulo"
},
{
"operation": "Page search",
"time": "329.827",
"value": "authored"
},
{
"operation": "Highlighting",
"time": "343.377",
"value": "Roman Romulo_sentence_25"
},
{
"operation": "Page search",
"time": "378.307",
"value": "disavowed"
},
{
"operation": "Page search",
"time": "406.409",
"value": "education bill"
},
{
"operation": "Page search",
"time": "426.433",
"value": "ferdinand marcos"
},
{
"operation": "search",
"time": "452.4",
"value": "ferdinand marcos"
},
{
"operation": "Now on",
"time": "453.651",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "459.713",
"value": "1984 Philippine parliamentary election"
},
{
"operation": "Now on",
"time": "460.538",
"value": "1984 Philippine parliamentary election"
},
{
"operation": "Page search",
"time": "470.857",
"value": "ferdinand marcos"
},
{
"operation": "back-button-clicked",
"time": "488.612",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "489.24",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "505.907",
"value": "History of the Philippines under Ferdinand Marcos"
},
{
"operation": "Now on",
"time": "506.781",
"value": "History of the Philippines under Ferdinand Marcos"
},
{
"operation": "Page search",
"time": "513.984",
"value": "roman romulo"
},
{
"operation": "back-button-clicked",
"time": "598.817",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "599.466",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "599.804",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "599.807",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "599.811",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "599.905",
"value": "Roman Romulo"
},
{
"operation": "Page search",
"time": "626.537",
"value": "endorse"
},
{
"operation": "Highlighting",
"time": "637.506",
"value": "Roman Romulo_sentence_22"
},
{
"operation": "hyperlink",
"time": "641.048",
"value": "Jovito Salonga"
},
{
"operation": "Now on",
"time": "641.853",
"value": "Jovito Salonga"
},
{
"operation": "Page search",
"time": "650.53",
"value": "ferdinand marcos"
},
{
"operation": "Highlighting",
"time": "655.213",
"value": "Jovito Salonga_sentence_1"
},
{
"operation": "back-button-clicked",
"time": "658.774",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "658.867",
"value": "Roman Romulo"
},
{
"operation": "finish",
"time": "670.939",
"value": "finish"
}
]
}
|
65283
|
Max Stewart (14 March 1935 – 19 March 1977), an Australian racing driver, is the fastest driver in the history of Australian Grand Prix.
|
{
"evidence": [
"Max Stewart_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMax Stewart (14 March 1935 – 19 March 1977), an Australian racing driver, is the fastest driver in the history of Australian Grand Prix.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Max Stewart_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 437,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.698",
"value": "?search="
},
{
"operation": "search",
"time": "6.899",
"value": "Max Stewart"
},
{
"operation": "Now on",
"time": "11.478",
"value": "Max Stewart"
},
{
"operation": "Highlighting",
"time": "73.413",
"value": "Max Stewart_sentence_0"
},
{
"operation": "Highlighting",
"time": "94.2",
"value": "Max Stewart_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "100.386",
"value": "Max Stewart_sentence_1"
},
{
"operation": "search",
"time": "193.165",
"value": "Australian Grand Prix"
},
{
"operation": "Now on",
"time": "196.233",
"value": "undefined"
},
{
"operation": "search",
"time": "219.126",
"value": "Australian Grand Prix (disambiguation)"
},
{
"operation": "Now on",
"time": "220.093",
"value": "Australian Grand Prix (disambiguation)"
},
{
"operation": "back-button-clicked",
"time": "229.599",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "230.566",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "232.282",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "232.285",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "232.354",
"value": "Max Stewart"
},
{
"operation": "search",
"time": "248.457",
"value": "Australian Grand Prix"
},
{
"operation": "Now on",
"time": "249.337",
"value": "undefined"
},
{
"operation": "search",
"time": "264.618",
"value": "Australian Grand Prix (disambiguation)"
},
{
"operation": "Now on",
"time": "265.156",
"value": "Australian Grand Prix (disambiguation)"
},
{
"operation": "hyperlink",
"time": "266.57",
"value": "Australian motorcycle Grand Prix"
},
{
"operation": "Now on",
"time": "281.78",
"value": "Australian motorcycle Grand Prix"
},
{
"operation": "Page search",
"time": "299.596",
"value": "max"
},
{
"operation": "back-button-clicked",
"time": "546.63",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "546.683",
"value": "Australian Grand Prix (disambiguation)"
},
{
"operation": "back-button-clicked",
"time": "547.818",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "547.821",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "548.69",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "578.704",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "578.707",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "578.776",
"value": "Max Stewart"
},
{
"operation": "search",
"time": "636.208",
"value": "fastest driver in the history of Australian Grand Prix"
},
{
"operation": "Now on",
"time": "638.892",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "645.214",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "645.275",
"value": "Max Stewart"
},
{
"operation": "search",
"time": "655.601",
"value": "history of Australian Grand Prix"
},
{
"operation": "Now on",
"time": "656.813",
"value": "undefined"
},
{
"operation": "finish",
"time": "665.819",
"value": "finish"
}
]
}
|
exfever_2902
|
Cate Blanchett is the recipient of numerous accolades, including two Oscars, which are given annually by the AMPAS, a professional honorary organization with the stated goal of advancing the arts and sciences of motion pictures.
|
{
"evidence": [
"Cate Blanchett",
"Academy Awards",
"Academy of Motion Picture Arts and Sciences"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nCate Blanchett is the recipient of numerous accolades, including two Oscars, which are given annually by the AMPAS, a professional honorary organization with the stated goal of advancing the arts and sciences of motion pictures.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Cate Blanchett",
"Academy Awards",
"Academy of Motion Picture Arts and Sciences"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 438,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_2653
|
Prince was an American singer-songwriter who produced his albums himself, pioneering the subgenre of funk-rock with elements of synth-pop.
|
{
"evidence": [
"Prince (musician)",
"Minneapolis sound"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPrince was an American singer-songwriter who produced his albums himself, pioneering the subgenre of funk-rock with elements of synth-pop.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Prince (musician)",
"Minneapolis sound"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 439,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
54047
|
Including 23,810 species, Lamiales holds erythranthe moschata.
|
{
"evidence": [
"Lamiales_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIncluding 23,810 species, Lamiales holds erythranthe moschata.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Lamiales_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 440,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.49",
"value": "?search="
},
{
"operation": "search",
"time": "9.526",
"value": "Lamiales"
},
{
"operation": "Now on",
"time": "10.866",
"value": "Lamiales"
},
{
"operation": "Highlighting",
"time": "13.017",
"value": "Lamiales_sentence_1"
},
{
"operation": "Page search",
"time": "43.636",
"value": "erythranthe moschata"
},
{
"operation": "Page search",
"time": "75.388",
"value": "moschata"
},
{
"operation": "Page search",
"time": "89.489",
"value": "erythranthe"
},
{
"operation": "Page search",
"time": "98.623",
"value": "holds"
},
{
"operation": "hyperlink",
"time": "104.684",
"value": "families"
},
{
"operation": "Now on",
"time": "105.362",
"value": "Family (biology)"
},
{
"operation": "back-button-clicked",
"time": "108.065",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "108.17",
"value": "Lamiales"
},
{
"operation": "finish",
"time": "141.758",
"value": "finish"
}
]
}
|
43075
|
The 1932 Dayton Flyers football team, under American football player, coach of football and basketball, and college athletics administrator Harry Baujan, was scheduled to play eleven matches and has won nine of these matches.
|
{
"evidence": [
"1932 Dayton Flyers football team_sentence_1",
"Harry Baujan_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe 1932 Dayton Flyers football team, under American football player, coach of football and basketball, and college athletics administrator Harry Baujan, was scheduled to play eleven matches and has won nine of these matches.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"1932 Dayton Flyers football team_sentence_1",
"Harry Baujan_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 441,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.701",
"value": "?search="
},
{
"operation": "search",
"time": "13.084",
"value": "1932 Dayton Flyers football team"
},
{
"operation": "Now on",
"time": "13.776",
"value": "1932 Dayton Flyers football team"
},
{
"operation": "Highlighting",
"time": "65.381",
"value": "1932 Dayton Flyers football team_cell_1_1_3"
},
{
"operation": "Highlighting",
"time": "65.382",
"value": "1932 Dayton Flyers football team_section_0"
},
{
"operation": "Highlighting",
"time": "65.384",
"value": "1932 Dayton Flyers football team_header_cell_1_0_3"
},
{
"operation": "Highlighting",
"time": "66.164",
"value": "1932 Dayton Flyers football team_cell_1_2_3"
},
{
"operation": "Highlighting",
"time": "66.984",
"value": "1932 Dayton Flyers football team_cell_1_3_3"
},
{
"operation": "Highlighting",
"time": "67.732",
"value": "1932 Dayton Flyers football team_cell_1_4_3"
},
{
"operation": "Highlighting",
"time": "69.165",
"value": "1932 Dayton Flyers football team_cell_1_5_3"
},
{
"operation": "Highlighting",
"time": "70.126",
"value": "1932 Dayton Flyers football team_cell_1_6_3"
},
{
"operation": "Highlighting",
"time": "71.057",
"value": "1932 Dayton Flyers football team_cell_1_7_3"
},
{
"operation": "Highlighting",
"time": "72.251",
"value": "1932 Dayton Flyers football team_cell_1_9_3"
},
{
"operation": "Highlighting",
"time": "73.325",
"value": "1932 Dayton Flyers football team_cell_1_11_3"
},
{
"operation": "Highlighting",
"time": "75.005",
"value": "1932 Dayton Flyers football team_cell_1_8_3"
},
{
"operation": "Highlighting",
"time": "75.888",
"value": "1932 Dayton Flyers football team_cell_1_10_3"
},
{
"operation": "Highlighting",
"time": "90.249",
"value": "1932 Dayton Flyers football team_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "107.846",
"value": "1932 Dayton Flyers football team_cell_1_11_3"
},
{
"operation": "Highlighting deleted",
"time": "108.614",
"value": "1932 Dayton Flyers football team_cell_1_10_3"
},
{
"operation": "Highlighting",
"time": "109.159",
"value": "1932 Dayton Flyers football team_cell_1_10_3"
},
{
"operation": "Highlighting deleted",
"time": "110.571",
"value": "1932 Dayton Flyers football team_cell_1_10_3"
},
{
"operation": "Highlighting deleted",
"time": "111.27",
"value": "1932 Dayton Flyers football team_cell_1_9_3"
},
{
"operation": "Highlighting deleted",
"time": "111.871",
"value": "1932 Dayton Flyers football team_cell_1_8_3"
},
{
"operation": "Highlighting deleted",
"time": "112.649",
"value": "1932 Dayton Flyers football team_cell_1_7_3"
},
{
"operation": "Highlighting deleted",
"time": "113.707",
"value": "1932 Dayton Flyers football team_cell_1_6_3"
},
{
"operation": "Highlighting deleted",
"time": "114.442",
"value": "1932 Dayton Flyers football team_cell_1_5_3"
},
{
"operation": "Highlighting deleted",
"time": "115.34",
"value": "1932 Dayton Flyers football team_cell_1_4_3"
},
{
"operation": "Highlighting deleted",
"time": "116.207",
"value": "1932 Dayton Flyers football team_cell_1_3_3"
},
{
"operation": "Highlighting deleted",
"time": "117.196",
"value": "1932 Dayton Flyers football team_cell_1_2_3"
},
{
"operation": "Highlighting deleted",
"time": "118.454",
"value": "1932 Dayton Flyers football team_cell_1_1_3"
},
{
"operation": "Highlighting deleted",
"time": "118.456",
"value": "1932 Dayton Flyers football team_section_0"
},
{
"operation": "Highlighting deleted",
"time": "118.459",
"value": "1932 Dayton Flyers football team_header_cell_1_0_3"
},
{
"operation": "hyperlink",
"time": "121.771",
"value": "Harry Baujan"
},
{
"operation": "Now on",
"time": "122.285",
"value": "Harry Baujan"
},
{
"operation": "Highlighting",
"time": "129.952",
"value": "Harry Baujan_sentence_0"
},
{
"operation": "finish",
"time": "141.138",
"value": "finish"
}
]
}
|
45010
|
In 1999, with a 17th place finish, Engström Motorsport moved up to race in the Swedish Touring Car Championship as the Chrysler works team under the Chrysler Racing Team Sweden name, running a 1998 spec Stratus.
|
{
"evidence": [
"Engström Motorsport_sentence_4",
"Engström Motorsport_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn 1999, with a 17th place finish, Engström Motorsport moved up to race in the Swedish Touring Car Championship as the Chrysler works team under the Chrysler Racing Team Sweden name, running a 1998 spec Stratus.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Engström Motorsport_sentence_4",
"Engström Motorsport_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 442,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.751",
"value": "?search="
},
{
"operation": "search",
"time": "79.509",
"value": "1999 Swedish Touring Car Championship"
},
{
"operation": "Now on",
"time": "80.901",
"value": "1999 Swedish Touring Car Championship"
},
{
"operation": "Page search",
"time": "114.485",
"value": "Engström Motorsport"
},
{
"operation": "search",
"time": "130.687",
"value": "Engström Motorsport"
},
{
"operation": "Now on",
"time": "132.338",
"value": "Engström Motorsport"
},
{
"operation": "Highlighting",
"time": "145.583",
"value": "Engström Motorsport_sentence_4"
},
{
"operation": "Highlighting",
"time": "150.767",
"value": "Engström Motorsport_sentence_5"
},
{
"operation": "finish",
"time": "167.137",
"value": "finish"
}
]
}
|
87434
|
Rubeho forest partridge, formerly considered to be a well-marked subspecies of the Udzungwa forest partridge, inhabits and is endemic to forests of the Rubeho Mountains, a mountain range in central Tanzania.
|
{
"evidence": [
"Rubeho forest partridge_sentence_4",
"Rubeho forest partridge_sentence_3",
"Rubeho Mountains_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRubeho forest partridge, formerly considered to be a well-marked subspecies of the Udzungwa forest partridge, inhabits and is endemic to forests of the Rubeho Mountains, a mountain range in central Tanzania.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Rubeho forest partridge_sentence_4",
"Rubeho forest partridge_sentence_3",
"Rubeho Mountains_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 443,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.539",
"value": "?search="
},
{
"operation": "search",
"time": "10.47",
"value": "Rubeho forest partridge"
},
{
"operation": "Now on",
"time": "11.412",
"value": "Rubeho forest partridge"
},
{
"operation": "Highlighting",
"time": "17.172",
"value": "Rubeho forest partridge_sentence_4"
},
{
"operation": "Highlighting",
"time": "26.609",
"value": "Rubeho forest partridge_sentence_3"
},
{
"operation": "hyperlink",
"time": "31.174",
"value": "Rubeho Mountains"
},
{
"operation": "Now on",
"time": "31.98",
"value": "Rubeho Mountains"
},
{
"operation": "Highlighting",
"time": "37.255",
"value": "Rubeho Mountains_sentence_0"
},
{
"operation": "finish",
"time": "174.182",
"value": "finish"
}
]
}
|
exfever_6356
|
Neighbors is a 2014 American comedy film starring an American actor who performed the role of Chris D'Amico in a 2010 superhero film directed by Matthew Vaughn.
|
{
"evidence": [
"Neighbors (2014 film)",
"Christopher Mintz-Plasse",
"Kick-Ass (film)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nNeighbors is a 2014 American comedy film starring an American actor who performed the role of Chris D'Amico in a 2010 superhero film directed by Matthew Vaughn.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Neighbors (2014 film)",
"Christopher Mintz-Plasse",
"Kick-Ass (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 444,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
76592
|
The 2009 Saint Francis Cougars football team had no coach.
|
{
"evidence": [
"2009 Saint Francis Cougars football team_sentence_1"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe 2009 Saint Francis Cougars football team had no coach.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"2009 Saint Francis Cougars football team_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 445,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.303",
"value": "?search="
},
{
"operation": "search",
"time": "12.931",
"value": "2009 Saint Francis Cougars football team"
},
{
"operation": "Now on",
"time": "13.895",
"value": "2009 Saint Francis Cougars football team"
},
{
"operation": "Highlighting",
"time": "22.594",
"value": "2009 Saint Francis Cougars football team_sentence_1"
},
{
"operation": "finish",
"time": "39.746",
"value": "finish"
}
]
}
|
exfever_6939
|
21 Jump Street is a 2012 Canadian buddy cop action comedy film directed by Phil Lord that was adapted from the television series of the same name by an American television producer and writer.
|
{
"evidence": [
"21 Jump Street (film)",
"Patrick Hasburgh"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n21 Jump Street is a 2012 Canadian buddy cop action comedy film directed by Phil Lord that was adapted from the television series of the same name by an American television producer and writer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"21 Jump Street (film)",
"Patrick Hasburgh"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 446,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
13196
|
Confessions of a Matchmaker is an American reality television series that debuted on June 16, 2007, which shows Patti Novak, a real-life matchmaker pairing new clients with her pre-existing clients.
|
{
"evidence": [
"Confessions of a Matchmaker_sentence_1",
"Confessions of a Matchmaker_sentence_0",
"Confessions of a Matchmaker_sentence_4",
"Confessions of a Matchmaker_sentence_7"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nConfessions of a Matchmaker is an American reality television series that debuted on June 16, 2007, which shows Patti Novak, a real-life matchmaker pairing new clients with her pre-existing clients.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Confessions of a Matchmaker_sentence_1",
"Confessions of a Matchmaker_sentence_0",
"Confessions of a Matchmaker_sentence_4",
"Confessions of a Matchmaker_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 447,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.974",
"value": "?search="
},
{
"operation": "Now on",
"time": "20.53",
"value": "?search="
},
{
"operation": "search",
"time": "103.106",
"value": "Confessions of a Matchmaker"
},
{
"operation": "Now on",
"time": "103.993",
"value": "Confessions of a Matchmaker"
},
{
"operation": "Highlighting",
"time": "184.953",
"value": "Confessions of a Matchmaker_sentence_1"
},
{
"operation": "Highlighting",
"time": "189.829",
"value": "Confessions of a Matchmaker_sentence_0"
},
{
"operation": "Highlighting",
"time": "198.626",
"value": "Confessions of a Matchmaker_sentence_4"
},
{
"operation": "Highlighting",
"time": "209.014",
"value": "Confessions of a Matchmaker_sentence_7"
},
{
"operation": "finish",
"time": "233.522",
"value": "finish"
}
]
}
|
exfever_2303
|
The Inhumans are a fictional group of superheroes introduced in the Filipino media franchise and shared universe produced by an American film and television production company that is a subsidiary of Walt Disney Studios.
|
{
"evidence": [
"Inhumans",
"Marvel Cinematic Universe",
"Marvel Studios"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Inhumans are a fictional group of superheroes introduced in the Filipino media franchise and shared universe produced by an American film and television production company that is a subsidiary of Walt Disney Studios.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Inhumans",
"Marvel Cinematic Universe",
"Marvel Studios"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 448,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_7505
|
Gone with the Wind was the most popular among American readers from the outset.
|
{
"evidence": [
"Gone with the Wind (novel)",
"Margaret Mitchell"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nGone with the Wind was the most popular among American readers from the outset.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Gone with the Wind (novel)",
"Margaret Mitchell"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 449,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
87352
|
The 1993 IAAF World Half Marathon Championships was held on October 3, 1993 in Brussels, Belgium where 133 athletes competed from 49 countries.
|
{
"evidence": [
"1993 IAAF World Half Marathon Championships_sentence_1"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe 1993 IAAF World Half Marathon Championships was held on October 3, 1993 in Brussels, Belgium where 133 athletes competed from 49 countries.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"1993 IAAF World Half Marathon Championships_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 450,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.23",
"value": "?search="
},
{
"operation": "search",
"time": "5.629",
"value": "1993 IAAF World Half Marathon Championships"
},
{
"operation": "Now on",
"time": "10.845",
"value": "1993 IAAF World Half Marathon Championships"
},
{
"operation": "Highlighting",
"time": "44.099",
"value": "1993 IAAF World Half Marathon Championships_sentence_0"
},
{
"operation": "Highlighting",
"time": "49.676",
"value": "1993 IAAF World Half Marathon Championships_sentence_1"
},
{
"operation": "finish",
"time": "59.153",
"value": "finish"
}
]
}
|
45892
|
Reaching up to 3.7 metres (12.1 ft) from claw to claw, the Japanese spider crab can weigh up to 19 kilograms (42 lb).
|
{
"evidence": [
"Japanese spider crab_sentence_8",
"Japanese spider crab_sentence_9"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nReaching up to 3.7 metres (12.1 ft) from claw to claw, the Japanese spider crab can weigh up to 19 kilograms (42 lb).\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Japanese spider crab_sentence_8",
"Japanese spider crab_sentence_9"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 451,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.827",
"value": "?search="
},
{
"operation": "search",
"time": "9.381",
"value": "Japanese spider crab"
},
{
"operation": "Now on",
"time": "10.415",
"value": "Japanese spider crab"
},
{
"operation": "Highlighting",
"time": "62.97",
"value": "Japanese spider crab_sentence_8"
},
{
"operation": "Highlighting",
"time": "69.394",
"value": "Japanese spider crab_sentence_9"
},
{
"operation": "finish",
"time": "94.971",
"value": "finish"
}
]
}
|
13295
|
The BMW F650CS, a standard motorcycle made by BMW Motorrad, the motorcycle brand of the German company BMW, from 2001 to 2005, was known for its offbeat styling intended to attract new motorcyclists.
|
{
"evidence": [
"BMW F650CS_sentence_2",
"BMW F650CS_sentence_5",
"BMW Motorrad_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe BMW F650CS, a standard motorcycle made by BMW Motorrad, the motorcycle brand of the German company BMW, from 2001 to 2005, was known for its offbeat styling intended to attract new motorcyclists.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"BMW F650CS_sentence_2",
"BMW F650CS_sentence_5",
"BMW Motorrad_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 452,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.874",
"value": "undefined"
},
{
"operation": "search",
"time": "33.872",
"value": "BMW F650CS"
},
{
"operation": "Now on",
"time": "35.411",
"value": "BMW F650CS"
},
{
"operation": "Highlighting",
"time": "42.268",
"value": "BMW F650CS_sentence_2"
},
{
"operation": "Highlighting",
"time": "48.457",
"value": "BMW F650CS_sentence_5"
},
{
"operation": "hyperlink",
"time": "54.336",
"value": "BMW Motorrad"
},
{
"operation": "Now on",
"time": "56.429",
"value": "BMW Motorrad"
},
{
"operation": "Highlighting",
"time": "57.692",
"value": "BMW Motorrad_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "62",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "62.169",
"value": "BMW F650CS"
},
{
"operation": "finish",
"time": "82.359",
"value": "finish"
}
]
}
|
exfever_5914
|
Emily Ratajkowski is an Indonesian model who appeared in a music video for American singer Robin Thicke’s song from the sixth studio album, first released in Germany on July 12, 2013.
|
{
"evidence": [
"Emily Ratajkowski",
"Blurred Lines",
"Blurred Lines (album)"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nEmily Ratajkowski is an Indonesian model who appeared in a music video for American singer Robin Thicke’s song from the sixth studio album, first released in Germany on July 12, 2013.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Emily Ratajkowski",
"Blurred Lines",
"Blurred Lines (album)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 453,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
17837
|
The Namibian Special Forces training includes the 5-week Paratroopers Pre Selection phase (that tests a candidates skills in Musketry, Navigation, and Field Craft), and then the Parachute Selection Phase.
|
{
"evidence": [
"Namibian Special Forces_sentence_14",
"Namibian Special Forces_sentence_15"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Namibian Special Forces training includes the 5-week Paratroopers Pre Selection phase (that tests a candidates skills in Musketry, Navigation, and Field Craft), and then the Parachute Selection Phase.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Namibian Special Forces_sentence_14",
"Namibian Special Forces_sentence_15"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 454,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.969",
"value": "?search="
},
{
"operation": "search",
"time": "11.785",
"value": "Namibian Special Forces"
},
{
"operation": "Now on",
"time": "12.635",
"value": "Namibian Special Forces"
},
{
"operation": "Highlighting",
"time": "51.129",
"value": "Namibian Special Forces_sentence_14"
},
{
"operation": "Highlighting",
"time": "60.295",
"value": "Namibian Special Forces_sentence_15"
},
{
"operation": "finish",
"time": "74.183",
"value": "finish"
}
]
}
|
exfever_6398
|
Thomas Jeffrey Hanks is an American actor who made his breakthrough with a leading role in a 1984 American sex comedy film directed by Neal Israel.
|
{
"evidence": [
"Tom Hanks",
"Bachelor Party (1984 film)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThomas Jeffrey Hanks is an American actor who made his breakthrough with a leading role in a 1984 American sex comedy film directed by Neal Israel.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tom Hanks",
"Bachelor Party (1984 film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 455,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_1415
|
Mankatha is a 2011 Indonesian Tamil-language action thriller film with an ensemble cast including an Indian producer who works predominantly in Tamil, Kannada, and Telugu language films.
|
{
"evidence": [
"Mankatha",
"Arjun Sarja"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMankatha is a 2011 Indonesian Tamil-language action thriller film with an ensemble cast including an Indian producer who works predominantly in Tamil, Kannada, and Telugu language films.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mankatha",
"Arjun Sarja"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 456,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_1742
|
Scarlett Johansson is also a singer and released an album titled Anywhere I Lay My Head.
|
{
"evidence": [
"Scarlett Johansson",
"Off-Broadway"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nScarlett Johansson is also a singer and released an album titled Anywhere I Lay My Head.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Scarlett Johansson",
"Off-Broadway"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 457,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_2217
|
Pablo Fenjves is an Australian screenwriter based in Los Angeles, California, who ghostwrote a book by O. J. Simpson in which Simpson puts forth a hypothetical description of the murders of Nicole Brown Simpson and Ron Goldman.
|
{
"evidence": [
"Pablo Fenjves",
"If I Did It"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPablo Fenjves is an Australian screenwriter based in Los Angeles, California, who ghostwrote a book by O. J. Simpson in which Simpson puts forth a hypothetical description of the murders of Nicole Brown Simpson and Ron Goldman.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Pablo Fenjves",
"If I Did It"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 458,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
5694
|
Komodo IDE's editor component is implemented using the Netscape Plugin Application Programming Interface and has an open-source counterpart called Komodo Edit.
|
{
"evidence": [
"Komodo IDE_sentence_4",
"Komodo IDE_sentence_5"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nKomodo IDE's editor component is implemented using the Netscape Plugin Application Programming Interface and has an open-source counterpart called Komodo Edit.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Komodo IDE_sentence_4",
"Komodo IDE_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 459,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.557",
"value": "?search="
},
{
"operation": "search",
"time": "15.06",
"value": "Komodo IDE"
},
{
"operation": "Now on",
"time": "15.794",
"value": "Komodo IDE"
},
{
"operation": "Page search",
"time": "124.46",
"value": "netscape plugin"
},
{
"operation": "Highlighting",
"time": "130.535",
"value": "Komodo IDE_sentence_4"
},
{
"operation": "Highlighting",
"time": "142.345",
"value": "Komodo IDE_sentence_5"
},
{
"operation": "finish",
"time": "151.195",
"value": "finish"
}
]
}
|
exfever_3700
|
The Wolf of Wall Street is a 2013 American biographical crime black comedy film written by an American writer and producer of television and films.
|
{
"evidence": [
"The Wolf of Wall Street (2013 film)",
"Terence Winter"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Wolf of Wall Street is a 2013 American biographical crime black comedy film written by an American writer and producer of television and films.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Wolf of Wall Street (2013 film)",
"Terence Winter"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 460,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_2039
|
Seth Woodbury MacFarlane is a French actor and the creator of the television series inspired primarily by the original Star Trek and its Next Generation successor, which it heavily parodies and pays homage to.
|
{
"evidence": [
"Seth MacFarlane",
"The Orville"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSeth Woodbury MacFarlane is a French actor and the creator of the television series inspired primarily by the original Star Trek and its Next Generation successor, which it heavily parodies and pays homage to.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Seth MacFarlane",
"The Orville"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 461,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_6408
|
Shakira is the highest-paid Colombian singer?at the age of 13.
|
{
"evidence": [
"Shakira",
"Pies Descalzos"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nShakira is the highest-paid Colombian singer?at the age of 13.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Shakira",
"Pies Descalzos"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 462,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
90692
|
McWilliams Arroyo sewed the flag of Puerto Rico for the 2008 Olympic games where he joined along with his sister.
|
{
"evidence": [
"McWilliams Arroyo_sentence_46",
"McWilliams Arroyo_sentence_55"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMcWilliams Arroyo sewed the flag of Puerto Rico for the 2008 Olympic games where he joined along with his sister.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"McWilliams Arroyo_sentence_46",
"McWilliams Arroyo_sentence_55"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 463,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "3.155",
"value": "?search="
},
{
"operation": "search",
"time": "12.935",
"value": "McWilliams Arroyo"
},
{
"operation": "Now on",
"time": "15.138",
"value": "McWilliams Arroyo"
},
{
"operation": "Page search",
"time": "30.045",
"value": "flag"
},
{
"operation": "Highlighting",
"time": "38.587",
"value": "McWilliams Arroyo_sentence_46"
},
{
"operation": "Highlighting",
"time": "81.082",
"value": "McWilliams Arroyo_sentence_55"
},
{
"operation": "finish",
"time": "91.094",
"value": "finish"
}
]
}
|
exfever_1574
|
Buffy the Vampire Slayer was the last film directed by Fran Rubel Kuzui.
|
{
"evidence": [
"Hilary Swank",
"Buffy the Vampire Slayer (film)",
"Luke Perry"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBuffy the Vampire Slayer was the last film directed by Fran Rubel Kuzui.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hilary Swank",
"Buffy the Vampire Slayer (film)",
"Luke Perry"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 464,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_5239
|
Nepal is a country that does not have territory connected to an ocean or whose coastlines lie in endorheic basins in South Asia.
|
{
"evidence": [
"Nepal",
"Landlocked country"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nNepal is a country that does not have territory connected to an ocean or whose coastlines lie in endorheic basins in South Asia.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Nepal",
"Landlocked country"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 465,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
75240
|
Andawala is a remote village in Central Province, Sri Lanka with two Buddhist temples and where most of the 400 families are working as government officials.
|
{
"evidence": [
"Andawala_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAndawala is a remote village in Central Province, Sri Lanka with two Buddhist temples and where most of the 400 families are working as government officials.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Andawala_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 466,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.034",
"value": "?search="
},
{
"operation": "search",
"time": "16.812",
"value": "Andawala"
},
{
"operation": "Now on",
"time": "17.469",
"value": "Andawala"
},
{
"operation": "Highlighting",
"time": "31.394",
"value": "Andawala_sentence_2"
},
{
"operation": "Highlighting",
"time": "46.18",
"value": "Andawala_sentence_0"
},
{
"operation": "Highlighting",
"time": "90.255",
"value": "Andawala_sentence_8"
},
{
"operation": "Highlighting",
"time": "112.973",
"value": "Andawala_sentence_9"
},
{
"operation": "Highlighting",
"time": "116.673",
"value": "Andawala_sentence_5"
},
{
"operation": "hyperlink",
"time": "190.598",
"value": "List of towns in Central Province, Sri Lanka"
},
{
"operation": "Now on",
"time": "192.222",
"value": "List of settlements in Central Province (Sri Lanka)"
},
{
"operation": "back-button-clicked",
"time": "232.746",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "232.806",
"value": "Andawala"
},
{
"operation": "Highlighting",
"time": "259.664",
"value": "Andawala_sentence_5"
},
{
"operation": "finish",
"time": "270.966",
"value": "finish"
}
]
}
|
exfever_12
|
The Block has also been certified gold in the United States.
|
{
"evidence": [
"The Block (album)",
"Akon"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Block has also been certified gold in the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Block (album)",
"Akon"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 467,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_4047
|
JAG is a legal drama television series produced in association with an American commercial broadcast television located in the most populous city in the United States.
|
{
"evidence": [
"JAG (TV series)",
"CBS",
"New York City"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJAG is a legal drama television series produced in association with an American commercial broadcast television located in the most populous city in the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"JAG (TV series)",
"CBS",
"New York City"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 468,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
40723
|
WNVM is a radio station licensed to serve Cidra, it is owned by New Life Broadcasting and broadcasts country music.
|
{
"evidence": [
"WNVM_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWNVM is a radio station licensed to serve Cidra, it is owned by New Life Broadcasting and broadcasts country music.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"WNVM_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 469,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.532",
"value": "?search="
},
{
"operation": "search",
"time": "14.144",
"value": "WNVM"
},
{
"operation": "Now on",
"time": "14.902",
"value": "WNVM"
},
{
"operation": "Highlighting",
"time": "23.687",
"value": "WNVM_sentence_2"
},
{
"operation": "finish",
"time": "33.39",
"value": "finish"
}
]
}
|
48141
|
Nicolai Abraham Holten established his own business as a bankier, he resisted against the government and was the personal advisor on financial matters of Frederick VII who was the King of Denmark.
|
{
"evidence": [
"Nicolai Abraham Holten_sentence_6"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nNicolai Abraham Holten established his own business as a bankier, he resisted against the government and was the personal advisor on financial matters of Frederick VII who was the King of Denmark.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Nicolai Abraham Holten_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 470,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.666",
"value": "?search="
},
{
"operation": "search",
"time": "11.113",
"value": "Nicolai Abraham Holten"
},
{
"operation": "Now on",
"time": "11.979",
"value": "Nicolai Abraham Holten"
},
{
"operation": "Highlighting",
"time": "23.035",
"value": "Nicolai Abraham Holten_sentence_5"
},
{
"operation": "Highlighting",
"time": "29.751",
"value": "Nicolai Abraham Holten_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "50.761",
"value": "Nicolai Abraham Holten_sentence_5"
},
{
"operation": "finish",
"time": "56.706",
"value": "finish"
}
]
}
|
67254
|
Black American Sign Language the first of its kind was founded by Platt Skinner in Niagara Falls, New York.
|
{
"evidence": [
"Black American Sign Language_sentence_12"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBlack American Sign Language the first of its kind was founded by Platt Skinner in Niagara Falls, New York.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Black American Sign Language_sentence_12"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 471,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "84.244",
"value": "1979 Italian Grand Prix"
},
{
"operation": "search",
"time": "87.763",
"value": "Black American Sign Language"
},
{
"operation": "Now on",
"time": "104.737",
"value": "1979 Italian Grand Prix"
},
{
"operation": "search",
"time": "137.433",
"value": "Black American Sign Language"
},
{
"operation": "Now on",
"time": "148.521",
"value": "Black American Sign Language"
},
{
"operation": "Highlighting",
"time": "274.488",
"value": "Black American Sign Language_sentence_11"
},
{
"operation": "Highlighting",
"time": "494.317",
"value": "Black American Sign Language_sentence_12"
},
{
"operation": "Highlighting deleted",
"time": "502.797",
"value": "Black American Sign Language_sentence_11"
},
{
"operation": "finish",
"time": "593.085",
"value": "finish"
}
]
}
|
47160
|
Farnborough Town won in the Premiere Division during the 1993–94 Southern Football League season, which was the league's 91st.
|
{
"evidence": [
"1993–94 Southern Football League_sentence_0",
"1993–94 Southern Football League_sentence_1"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nFarnborough Town won in the Premiere Division during the 1993–94 Southern Football League season, which was the league's 91st.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"1993–94 Southern Football League_sentence_0",
"1993–94 Southern Football League_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 472,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.727",
"value": "?search="
},
{
"operation": "search",
"time": "73.508",
"value": "Contains... 1993–94 Southern Football League season"
},
{
"operation": "Now on",
"time": "76.875",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "79.896",
"value": "1993–94 Southern Football League"
},
{
"operation": "Now on",
"time": "81.016",
"value": "1993–94 Southern Football League"
},
{
"operation": "Highlighting",
"time": "87.145",
"value": "1993–94 Southern Football League_sentence_0"
},
{
"operation": "Highlighting",
"time": "96.241",
"value": "1993–94 Southern Football League_sentence_1"
},
{
"operation": "finish",
"time": "114.953",
"value": "finish"
}
]
}
|
exfever_8461
|
There may have no symptoms of a specific infection for very young, old, and people with Immunodeficiency, a state in which the immune system's ability to fight infectious diseases and cancer is compromised or entirely absent.
|
{
"evidence": [
"Sepsis",
"Immunodeficiency"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThere may have no symptoms of a specific infection for very young, old, and people with Immunodeficiency, a state in which the immune system's ability to fight infectious diseases and cancer is compromised or entirely absent.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sepsis",
"Immunodeficiency"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 473,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_7542
|
Honey to the B is the debut album by the pop singer Billie Piper that includes the title track, released on March 22, 1999, that debuted at number three on the US singles chart.
|
{
"evidence": [
"Honey to the B",
"Honey to the Bee"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHoney to the B is the debut album by the pop singer Billie Piper that includes the title track, released on March 22, 1999, that debuted at number three on the US singles chart.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Honey to the B",
"Honey to the Bee"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 474,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
78191
|
Jeff Berding grew up in the Price Hill neighborhood of Cincinnati, Ohio and became interested in sports at an early age.
|
{
"evidence": [
"Jeff Berding_sentence_4"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJeff Berding grew up in the Price Hill neighborhood of Cincinnati, Ohio and became interested in sports at an early age.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Jeff Berding_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 475,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.097",
"value": "?search="
},
{
"operation": "search",
"time": "36.459",
"value": "Jeff Berding"
},
{
"operation": "Now on",
"time": "37.287",
"value": "Jeff Berding"
},
{
"operation": "Highlighting",
"time": "59.268",
"value": "Jeff Berding_sentence_4"
},
{
"operation": "finish",
"time": "76.901",
"value": "finish"
}
]
}
|
exfever_2735
|
7 Things is a song by American singer Miley Cyrus that was released by an American record label with a current roster of artists, including an American pop band.
|
{
"evidence": [
"7 Things",
"Hollywood Records",
"Ocean Park Standoff"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n7 Things is a song by American singer Miley Cyrus that was released by an American record label with a current roster of artists, including an American pop band.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"7 Things",
"Hollywood Records",
"Ocean Park Standoff"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 476,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_2196
|
In Poona Pact, eight electoral seats are reserved for Punjab.
|
{
"evidence": [
"Poona Pact",
"Legislatures of British India"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn Poona Pact, eight electoral seats are reserved for Punjab.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Poona Pact",
"Legislatures of British India"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 477,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_2340
|
For Habib Faisal, nostalgia is like the warmth of a cotton 'razaai' in the winter of pre-liberalisation Delhi.
|
{
"evidence": [
"Arjun Kapoor",
"Ishaqzaade"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nFor Habib Faisal, nostalgia is like the warmth of a cotton 'razaai' in the winter of pre-liberalisation Delhi.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Arjun Kapoor",
"Ishaqzaade"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 478,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
85162
|
Wallace, born in Sweden to neither Swedish nor Danish parents, studied at Uppsala University and trained on archaeological sites in Sweden and Finland.
|
{
"evidence": [
"Birgitta Wallace_sentence_3"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWallace, born in Sweden to neither Swedish nor Danish parents, studied at Uppsala University and trained on archaeological sites in Sweden and Finland.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Birgitta Wallace_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 479,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.742",
"value": "?search="
},
{
"operation": "search",
"time": "13.646",
"value": "Contains...Wallace, born in Sweden"
},
{
"operation": "Now on",
"time": "15.984",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "23.49",
"value": "Birgitta Wallace"
},
{
"operation": "Now on",
"time": "24.145",
"value": "Birgitta Wallace"
},
{
"operation": "Highlighting",
"time": "28.45",
"value": "Birgitta Wallace_sentence_3"
},
{
"operation": "finish",
"time": "44.285",
"value": "finish"
}
]
}
|
exfever_8528
|
Avatar is a 2009 epic science fiction film starring an American actress who began her career playing in a 2000 American sports drama film written and directed by Karyn Kusama.
|
{
"evidence": [
"Avatar (2009 film)",
"Michelle Rodriguez",
"Girlfight"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAvatar is a 2009 epic science fiction film starring an American actress who began her career playing in a 2000 American sports drama film written and directed by Karyn Kusama.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Avatar (2009 film)",
"Michelle Rodriguez",
"Girlfight"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 480,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_8778
|
The Homesman is a 2014 Western historical drama film starring an American actress.
|
{
"evidence": [
"The Homesman",
"Grace Gummer"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Homesman is a 2014 Western historical drama film starring an American actress.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Homesman",
"Grace Gummer"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 481,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_7809
|
David Judah Simon is best known for his work on an American comedy drama television series.
|
{
"evidence": [
"David Simon",
"The Wire"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDavid Judah Simon is best known for his work on an American comedy drama television series.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"David Simon",
"The Wire"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 482,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
10210
|
Atractaspidinae snakes can easily envenomate a person once bitten.
|
{
"evidence": [
"Atractaspidinae_sentence_7"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAtractaspidinae snakes can easily envenomate a person once bitten.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Atractaspidinae_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 483,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.539",
"value": "?search="
},
{
"operation": "search",
"time": "13.808",
"value": "Contains...Atractaspidinae snakes"
},
{
"operation": "Now on",
"time": "14.257",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "26.84",
"value": "Atractaspidinae"
},
{
"operation": "Now on",
"time": "27.35",
"value": "Atractaspidinae"
},
{
"operation": "Highlighting",
"time": "49.375",
"value": "Atractaspidinae_sentence_7"
},
{
"operation": "finish",
"time": "137.497",
"value": "finish"
}
]
}
|
exfever_6335
|
The director and co-writer of the action film produced by Mac Cappuccino also directed numerous episodes of the parody television news series Onion News Network.
|
{
"evidence": [
"Jon Watts",
"Clown (film)",
"Jon Watts"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe director and co-writer of the action film produced by Mac Cappuccino also directed numerous episodes of the parody television news series Onion News Network.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jon Watts",
"Clown (film)",
"Jon Watts"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 484,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
exfever_7975
|
Batman Begins is a supervillain film directed by a British-American film director and producer.
|
{
"evidence": [
"Batman Begins",
"Christopher Nolan"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBatman Begins is a supervillain film directed by a British-American film director and producer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Batman Begins",
"Christopher Nolan"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 485,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
71316
|
The Feldberg Lake District Nature Park region sits on the Havel river, which provides an important link in waterway connections between eastern and western Germany.
|
{
"evidence": [
"Feldberg Lake District Nature Park_sentence_17"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Feldberg Lake District Nature Park region sits on the Havel river, which provides an important link in waterway connections between eastern and western Germany.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Feldberg Lake District Nature Park_sentence_17"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 486,
"split": "test_on"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.427",
"value": "?search="
},
{
"operation": "search",
"time": "8.149",
"value": "Feldberg Lake District Nature Park"
},
{
"operation": "Now on",
"time": "10.124",
"value": "Feldberg Lake District Nature Park"
},
{
"operation": "Page search",
"time": "68.236",
"value": "river"
},
{
"operation": "Highlighting",
"time": "127.472",
"value": "Feldberg Lake District Nature Park_sentence_17"
},
{
"operation": "search",
"time": "164.644",
"value": "havel river"
},
{
"operation": "Now on",
"time": "166.254",
"value": "Havel"
},
{
"operation": "Page search",
"time": "178.37",
"value": "Feldberg"
},
{
"operation": "Page search",
"time": "191.949",
"value": "lake district"
},
{
"operation": "finish",
"time": "208.461",
"value": "finish"
}
]
}
|
exfever_6768
|
Daniel Craig is an English actor who began his career on stage after training at the youth theatre and registered charity in London, founded in 1954.
|
{
"evidence": [
"Daniel Craig",
"National Youth Theatre"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDaniel Craig is an English actor who began his career on stage after training at the youth theatre and registered charity in London, founded in 1954.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Daniel Craig",
"National Youth Theatre"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 487,
"split": "test_on"
}
|
{
"annotator_operations": []
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.