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 |
|---|---|---|---|---|---|---|---|---|
38748
|
The use of UAVs by the FBI happened as a result of the 11 September terrorist attacks and remains classified top secret.
|
{
"evidence": [
"Use of UAVs in law enforcement_sentence_39"
],
"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 use of UAVs by the FBI happened as a result of the 11 September terrorist attacks and remains classified top secret.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Use of UAVs in law enforcement_sentence_39"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7084,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.329",
"value": "?search="
},
{
"operation": "Now on",
"time": "16.898",
"value": "?search="
},
{
"operation": "search",
"time": "24.405",
"value": "UAVs"
},
{
"operation": "Now on",
"time": "25.73",
"value": "Unmanned aerial vehicle"
},
{
"operation": "search",
"time": "64.022",
"value": "uavs fbi"
},
{
"operation": "Now on",
"time": "65.164",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "67.233",
"value": "Use of UAVs in law enforcement"
},
{
"operation": "Now on",
"time": "68.18",
"value": "Use of UAVs in law enforcement"
},
{
"operation": "Highlighting",
"time": "86.644",
"value": "Use of UAVs in law enforcement_sentence_39"
},
{
"operation": "finish",
"time": "137.583",
"value": "finish"
}
]
}
|
exfever_7095
|
Ty Cobb was an American major league baseball player who spent 22 seasons with the American professional baseball team based in Detroit.
|
{
"evidence": [
"Ty Cobb",
"Detroit Tigers"
],
"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:\nTy Cobb was an American major league baseball player who spent 22 seasons with the American professional baseball team based in Detroit.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ty Cobb",
"Detroit Tigers"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7085,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
53414
|
A professional Australian cricketer, Andy Delmont made his first-class debut against Tasmania at Bellerive Oval where he scored 56 in the second innings.
|
{
"evidence": [
"Andy Delmont_sentence_0",
"Andy Delmont_sentence_4",
"Andy Delmont_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:\nA professional Australian cricketer, Andy Delmont made his first-class debut against Tasmania at Bellerive Oval where he scored 56 in the second innings.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Andy Delmont_sentence_0",
"Andy Delmont_sentence_4",
"Andy Delmont_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7086,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.872",
"value": "?search="
},
{
"operation": "search",
"time": "13.834",
"value": "Andy Delmont"
},
{
"operation": "Now on",
"time": "14.721",
"value": "Andy Delmont"
},
{
"operation": "Highlighting",
"time": "22.27",
"value": "Andy Delmont_sentence_0"
},
{
"operation": "Highlighting",
"time": "26.089",
"value": "Andy Delmont_sentence_4"
},
{
"operation": "Highlighting",
"time": "30.536",
"value": "Andy Delmont_sentence_5"
},
{
"operation": "finish",
"time": "41.219",
"value": "finish"
}
]
}
|
exfever_4539
|
Stranger than Fiction is a 2006 British film directed by a German-Swiss film director best known for directing the feature films Monster's Ball.
|
{
"evidence": [
"Stranger than Fiction (2006 film)",
"Marc Forster"
],
"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:\nStranger than Fiction is a 2006 British film directed by a German-Swiss film director best known for directing the feature films Monster's Ball.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Stranger than Fiction (2006 film)",
"Marc Forster"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7087,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
36791
|
King of Dia Riata, Dunchad mac Conaing was co-ruler with the son of Eochaid Buide, Conall Crandomma until 654 when Conaing was killed by Talorcan.
|
{
"evidence": [
"Dúnchad mac Conaing_sentence_0",
"Dúnchad mac Conaing_sentence_1",
"Conall Crandomna_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:\nKing of Dia Riata, Dunchad mac Conaing was co-ruler with the son of Eochaid Buide, Conall Crandomma until 654 when Conaing was killed by Talorcan.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Dúnchad mac Conaing_sentence_0",
"Dúnchad mac Conaing_sentence_1",
"Conall Crandomna_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7088,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.356",
"value": "undefined"
},
{
"operation": "search",
"time": "8.76",
"value": "King of Dia Riata"
},
{
"operation": "Now on",
"time": "9.924",
"value": "undefined"
},
{
"operation": "search",
"time": "18.587",
"value": "Dunchad mac Conaing"
},
{
"operation": "Now on",
"time": "20.05",
"value": "Dúnchad mac Conaing"
},
{
"operation": "Highlighting",
"time": "23.694",
"value": "Dúnchad mac Conaing_sentence_0"
},
{
"operation": "Highlighting",
"time": "99.749",
"value": "Dúnchad mac Conaing_sentence_1"
},
{
"operation": "hyperlink",
"time": "150.438",
"value": "Conall Crandomna"
},
{
"operation": "Now on",
"time": "151.14",
"value": "Conall Crandomna"
},
{
"operation": "Page search",
"time": "169.145",
"value": "buide"
},
{
"operation": "Page search",
"time": "169.368",
"value": "buide"
},
{
"operation": "Highlighting",
"time": "173.553",
"value": "Conall Crandomna_sentence_1"
},
{
"operation": "finish",
"time": "188.254",
"value": "finish"
}
]
}
|
16323
|
Hereford Castle is a manor rebuilt by William FitzOsbern, Earl of Hereford and father of Roger FitzOsbern who was involved in an unsuccessful attempt to depose King William .
|
{
"evidence": [
"Hereford Castle_sentence_12",
"Hereford Castle_sentence_13",
"Hereford Castle_sentence_14"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHereford Castle is a manor rebuilt by William FitzOsbern, Earl of Hereford and father of Roger FitzOsbern who was involved in an unsuccessful attempt to depose King William .\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Hereford Castle_sentence_12",
"Hereford Castle_sentence_13",
"Hereford Castle_sentence_14"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7089,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.604",
"value": "?search="
},
{
"operation": "Now on",
"time": "10.859",
"value": "?search="
},
{
"operation": "search",
"time": "31.123",
"value": "Hereford Castle"
},
{
"operation": "Now on",
"time": "32.077",
"value": "Hereford Castle"
},
{
"operation": "Page search",
"time": "46.961",
"value": "William FitzOsbern"
},
{
"operation": "Highlighting",
"time": "71.808",
"value": "Hereford Castle_sentence_12"
},
{
"operation": "Highlighting",
"time": "72.742",
"value": "Hereford Castle_sentence_13"
},
{
"operation": "Highlighting",
"time": "73.669",
"value": "Hereford Castle_sentence_14"
},
{
"operation": "finish",
"time": "144.352",
"value": "finish"
}
]
}
|
47345
|
Diuris curta is a species of inedible orchid that flowers from August to early October and is endemic to eastern Australia, its labellum at 10–12 mm (0.39–0.47 in) long has three lobes, with the centre lobe being egg-shaped to wedge-shaped, and both side lobes being oblong to egg-shaped.
|
{
"evidence": [
"Diuris curta_sentence_0",
"Diuris curta_sentence_8",
"Diuris curta_sentence_9",
"Diuris curta_sentence_11"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDiuris curta is a species of inedible orchid that flowers from August to early October and is endemic to eastern Australia, its labellum at 10–12 mm (0.39–0.47 in) long has three lobes, with the centre lobe being egg-shaped to wedge-shaped, and both side lobes being oblong to egg-shaped.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Diuris curta_sentence_0",
"Diuris curta_sentence_8",
"Diuris curta_sentence_9",
"Diuris curta_sentence_11"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7090,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.845",
"value": "?search="
},
{
"operation": "search",
"time": "18.47",
"value": "Diuris curta"
},
{
"operation": "Now on",
"time": "18.998",
"value": "Diuris curta"
},
{
"operation": "Highlighting",
"time": "24.581",
"value": "Diuris curta_sentence_0"
},
{
"operation": "Highlighting",
"time": "35.054",
"value": "Diuris curta_sentence_8"
},
{
"operation": "Highlighting",
"time": "55.978",
"value": "Diuris curta_sentence_9"
},
{
"operation": "Highlighting deleted",
"time": "103.634",
"value": "Diuris curta_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "104.41",
"value": "Diuris curta_sentence_9"
},
{
"operation": "hyperlink",
"time": "125.729",
"value": "labellum"
},
{
"operation": "Now on",
"time": "126.514",
"value": "Labellum (botany)"
},
{
"operation": "back-button-clicked",
"time": "132.368",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "132.418",
"value": "Diuris curta"
},
{
"operation": "Highlighting",
"time": "136.349",
"value": "Diuris curta_sentence_8"
},
{
"operation": "Highlighting",
"time": "137.722",
"value": "Diuris curta_sentence_9"
},
{
"operation": "Highlighting",
"time": "170.302",
"value": "Diuris curta_sentence_11"
},
{
"operation": "finish",
"time": "175.364",
"value": "finish"
}
]
}
|
20350
|
Sinha has appeared as a guest on Midweek, Woman's Hour, Broadcasting House and as a political pundit on PM; then in March 2011, Sinha presented his own Radio 4 programme, Paul Sinha's Quiz Culture.
|
{
"evidence": [
"Paul Sinha_sentence_13",
"Paul Sinha_sentence_14"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSinha has appeared as a guest on Midweek, Woman's Hour, Broadcasting House and as a political pundit on PM; then in March 2011, Sinha presented his own Radio 4 programme, Paul Sinha's Quiz Culture.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Paul Sinha_sentence_13",
"Paul Sinha_sentence_14"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7091,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.6",
"value": "?search="
},
{
"operation": "search",
"time": "7.005",
"value": "Sinha"
},
{
"operation": "Now on",
"time": "7.709",
"value": "Sinha"
},
{
"operation": "search",
"time": "25.846",
"value": "Sinha"
},
{
"operation": "Now on",
"time": "26.517",
"value": "Sinha"
},
{
"operation": "Page search",
"time": "28.618",
"value": "Midweek"
},
{
"operation": "search",
"time": "51.039",
"value": "Paul Sinha"
},
{
"operation": "Now on",
"time": "51.725",
"value": "Paul Sinha"
},
{
"operation": "Page search",
"time": "57.309",
"value": "Midweek"
},
{
"operation": "Highlighting",
"time": "60.38",
"value": "Paul Sinha_sentence_13"
},
{
"operation": "Highlighting",
"time": "95.792",
"value": "Paul Sinha_sentence_14"
},
{
"operation": "finish",
"time": "329.134",
"value": "finish"
}
]
}
|
8410
|
Heldsberg, overlooking the Rhine (second-longest river in Central and Western Europe), was armed with four 75mm guns, mounted in individual concrete blockhouses, and supported by five machine gun blocks with a total of seven machine guns in between.
|
{
"evidence": [
"Rhine_sentence_4",
"Fort Heldsberg_sentence_9",
"Fort Heldsberg_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:\nHeldsberg, overlooking the Rhine (second-longest river in Central and Western Europe), was armed with four 75mm guns, mounted in individual concrete blockhouses, and supported by five machine gun blocks with a total of seven machine guns in between.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Rhine_sentence_4",
"Fort Heldsberg_sentence_9",
"Fort Heldsberg_sentence_10"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7092,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.758",
"value": "?search="
},
{
"operation": "search",
"time": "25.694",
"value": "Contains...heldsberg"
},
{
"operation": "Now on",
"time": "29.174",
"value": "undefined"
},
{
"operation": "search",
"time": "43.332",
"value": "Rhine"
},
{
"operation": "Now on",
"time": "44.44",
"value": "Rhine"
},
{
"operation": "search",
"time": "71.846",
"value": "heldsberg"
},
{
"operation": "Now on",
"time": "72.815",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "76.58",
"value": "healdsburg"
},
{
"operation": "Now on",
"time": "77.281",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "82.712",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "83.429",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "92.261",
"value": "List of fortifications in Switzerland"
},
{
"operation": "Now on",
"time": "93.364",
"value": "List of fortifications in Switzerland"
},
{
"operation": "hyperlink",
"time": "132.366",
"value": "Fort Heldsberg"
},
{
"operation": "Now on",
"time": "133.344",
"value": "Fort Heldsberg"
},
{
"operation": "Highlighting",
"time": "150.768",
"value": "Fort Heldsberg_sentence_0Fort Heldsberg_sentence_1"
},
{
"operation": "hyperlink",
"time": "155.056",
"value": "Rhine"
},
{
"operation": "Now on",
"time": "155.893",
"value": "Rhine"
},
{
"operation": "Highlighting",
"time": "175.639",
"value": "Rhine_sentence_4"
},
{
"operation": "back-button-clicked",
"time": "177.948",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "178.21",
"value": "Fort Heldsberg"
},
{
"operation": "Highlighting deleted",
"time": "187.488",
"value": "Fort Heldsberg_sentence_0Fort Heldsberg_sentence_1"
},
{
"operation": "Highlighting",
"time": "189.547",
"value": "Fort Heldsberg_sentence_9"
},
{
"operation": "Highlighting",
"time": "211.948",
"value": "Fort Heldsberg_sentence_10"
},
{
"operation": "finish",
"time": "259.728",
"value": "finish"
}
]
}
|
91359
|
I Knew It Was You is an American documentary film released on 16 January 2009,directed by Richard Shepard and produced by Brett Ratner, Stacey Reiss and Richard Shepard and distributed by HBO and Oscilloscope.
|
{
"evidence": [
"I Knew It Was You_sentence_4",
"2009 Sundance Film Festival_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:\nI Knew It Was You is an American documentary film released on 16 January 2009,directed by Richard Shepard and produced by Brett Ratner, Stacey Reiss and Richard Shepard and distributed by HBO and Oscilloscope.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"I Knew It Was You_sentence_4",
"2009 Sundance Film Festival_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7093,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.913",
"value": "?search="
},
{
"operation": "search",
"time": "23.015",
"value": "I Knew It Was You"
},
{
"operation": "Now on",
"time": "23.825",
"value": "I Knew It Was You"
},
{
"operation": "Highlighting",
"time": "45.292",
"value": "I Knew It Was You_sentence_0"
},
{
"operation": "Page search",
"time": "89.209",
"value": "january"
},
{
"operation": "Highlighting",
"time": "174.419",
"value": "I Knew It Was You_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "177.337",
"value": "I Knew It Was You_sentence_0"
},
{
"operation": "hyperlink",
"time": "202.984",
"value": "2009 Sundance Film Festival"
},
{
"operation": "Now on",
"time": "203.856",
"value": "2009 Sundance Film Festival"
},
{
"operation": "Highlighting",
"time": "251.826",
"value": "2009 Sundance Film Festival_sentence_0"
},
{
"operation": "finish",
"time": "275.719",
"value": "finish"
}
]
}
|
88114
|
On November 5, 1968 the United States presidential election in Oregon, Richard Nixon won the highest votes while George Wallace got the lowest votes.
|
{
"evidence": [
"1968 United States presidential election in Oregon_sentence_1",
"1968 United States presidential election in Oregon_sentence_3",
"1968 United States presidential election in Oregon_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:\nOn November 5, 1968 the United States presidential election in Oregon, Richard Nixon won the highest votes while George Wallace got the lowest votes.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"1968 United States presidential election in Oregon_sentence_1",
"1968 United States presidential election in Oregon_sentence_3",
"1968 United States presidential election in Oregon_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7094,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.955",
"value": "?search="
},
{
"operation": "search",
"time": "16.138",
"value": "1968 the United States presidential election in Oregon"
},
{
"operation": "Now on",
"time": "18.184",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "20.017",
"value": "1968 United States presidential election in Oregon"
},
{
"operation": "Now on",
"time": "21.272",
"value": "1968 United States presidential election in Oregon"
},
{
"operation": "Highlighting",
"time": "24.739",
"value": "1968 United States presidential election in Oregon_sentence_1"
},
{
"operation": "Highlighting",
"time": "46.386",
"value": "1968 United States presidential election in Oregon_sentence_3"
},
{
"operation": "Highlighting",
"time": "60.38",
"value": "1968 United States presidential election in Oregon_sentence_4"
},
{
"operation": "Highlighting",
"time": "75.075",
"value": "1968 United States presidential election in Oregon_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "75.079",
"value": "1968 United States presidential election in Oregon_section_0"
},
{
"operation": "Highlighting",
"time": "75.081",
"value": "1968 United States presidential election in Oregon_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "76.176",
"value": "1968 United States presidential election in Oregon_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "76.182",
"value": "1968 United States presidential election in Oregon_section_0"
},
{
"operation": "Highlighting deleted",
"time": "76.185",
"value": "1968 United States presidential election in Oregon_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "76.939",
"value": "1968 United States presidential election in Oregon_cell_0_1_2"
},
{
"operation": "Highlighting",
"time": "76.942",
"value": "1968 United States presidential election in Oregon_section_0"
},
{
"operation": "Highlighting",
"time": "76.943",
"value": "1968 United States presidential election in Oregon_header_cell_0_0_2"
},
{
"operation": "Highlighting",
"time": "78.02",
"value": "1968 United States presidential election in Oregon_cell_0_1_4"
},
{
"operation": "Highlighting",
"time": "78.024",
"value": "1968 United States presidential election in Oregon_header_cell_0_0_4"
},
{
"operation": "Highlighting",
"time": "79.214",
"value": "1968 United States presidential election in Oregon_cell_0_2_2"
},
{
"operation": "Highlighting",
"time": "80.083",
"value": "1968 United States presidential election in Oregon_cell_0_2_4"
},
{
"operation": "Highlighting",
"time": "81.174",
"value": "1968 United States presidential election in Oregon_cell_0_3_2"
},
{
"operation": "Highlighting",
"time": "81.959",
"value": "1968 United States presidential election in Oregon_cell_0_3_4"
},
{
"operation": "finish",
"time": "114.896",
"value": "finish"
}
]
}
|
exfever_883
|
Pillowtalk is the latest song by Zayn Malik.
|
{
"evidence": [
"Zayn Malik",
"RCA Records",
"Sony Corporation of America"
],
"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:\nPillowtalk is the latest song by Zayn Malik.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Zayn Malik",
"RCA Records",
"Sony Corporation of America"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7095,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
67935
|
In Amanush 2 Raghu's problems begin when his father commits suicide and he seeks revenge, when he is released from juvie he becomes a medical student.
|
{
"evidence": [
"Amanush 2_sentence_6",
"Amanush 2_sentence_7",
"Amanush 2_sentence_8",
"Amanush 2_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:\nIn Amanush 2 Raghu's problems begin when his father commits suicide and he seeks revenge, when he is released from juvie he becomes a medical student.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Amanush 2_sentence_6",
"Amanush 2_sentence_7",
"Amanush 2_sentence_8",
"Amanush 2_sentence_10"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7096,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.109",
"value": "?search="
},
{
"operation": "search",
"time": "11.706",
"value": "Amanush 2"
},
{
"operation": "Now on",
"time": "13.243",
"value": "Amanush 2"
},
{
"operation": "Highlighting",
"time": "23.079",
"value": "Amanush 2_sentence_6"
},
{
"operation": "Highlighting",
"time": "27.047",
"value": "Amanush 2_sentence_7"
},
{
"operation": "Highlighting",
"time": "35.801",
"value": "Amanush 2_sentence_8"
},
{
"operation": "Highlighting",
"time": "43.837",
"value": "Amanush 2_sentence_10"
},
{
"operation": "finish",
"time": "67.137",
"value": "finish"
}
]
}
|
7032
|
The etymology of Desert wheatear came from the Latin word "deserti," which means "dessert" and "wheat," referring to the prominent wheat found in some of its species.
|
{
"evidence": [
"Desert wheatear_sentence_15",
"Desert wheatear_sentence_14"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe etymology of Desert wheatear came from the Latin word \"deserti,\" which means \"dessert\" and \"wheat,\" referring to the prominent wheat found in some of its species.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Desert wheatear_sentence_15",
"Desert wheatear_sentence_14"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7097,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.404",
"value": "?search="
},
{
"operation": "search",
"time": "23.648",
"value": "Desert wheatear"
},
{
"operation": "Now on",
"time": "24.36",
"value": "Desert wheatear"
},
{
"operation": "Highlighting",
"time": "65.91",
"value": "Desert wheatear_sentence_15"
},
{
"operation": "Highlighting",
"time": "79.872",
"value": "Desert wheatear_sentence_14"
},
{
"operation": "finish",
"time": "104.476",
"value": "finish"
}
]
}
|
exfever_7836
|
The Oprah Winfrey Show was an American daytime syndicated talk show that aired nationally in the most populous city in the U.S. state of Illinois and the third-most populous city in the United States.
|
{
"evidence": [
"The Oprah Winfrey Show",
"Chicago"
],
"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 Oprah Winfrey Show was an American daytime syndicated talk show that aired nationally in the most populous city in the U.S. state of Illinois and the third-most populous city in the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Oprah Winfrey Show",
"Chicago"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7098,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
13422
|
The HSP90AB1 (which is a molecular chaperone) is a carbohydrate that in animals is encoded by the HSP90AB1 gene which are classified into two major groups.
|
{
"evidence": [
"HSP90AB1_sentence_0"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe HSP90AB1 (which is a molecular chaperone) is a carbohydrate that in animals is encoded by the HSP90AB1 gene which are classified into two major groups.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"HSP90AB1_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7099,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.002",
"value": "?search="
},
{
"operation": "search",
"time": "17.552",
"value": "HSP90AB1"
},
{
"operation": "Now on",
"time": "18.296",
"value": "HSP90AB1"
},
{
"operation": "Highlighting",
"time": "24.643",
"value": "HSP90AB1_sentence_1"
},
{
"operation": "Highlighting",
"time": "24.644",
"value": "HSP90AB1_section_0"
},
{
"operation": "Highlighting",
"time": "36.602",
"value": "HSP90AB1_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "38.831",
"value": "HSP90AB1_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "38.834",
"value": "HSP90AB1_section_0"
},
{
"operation": "Page search",
"time": "59.674",
"value": "carbohydrate"
},
{
"operation": "finish",
"time": "75.005",
"value": "finish"
}
]
}
|
exfever_1069
|
The Boys from Brazil features an English actor whose well-known roles include Marcus Brody in a 1982 American action-adventure film.
|
{
"evidence": [
"The Boys from Brazil (film)",
"Denholm Elliott",
"Raiders of the Lost Ark"
],
"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 Boys from Brazil features an English actor whose well-known roles include Marcus Brody in a 1982 American action-adventure film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Boys from Brazil (film)",
"Denholm Elliott",
"Raiders of the Lost Ark"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7100,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
72291
|
Astragalus applegatei is a rare species of mikvetch which is a large genus of over 3,000 species of herbs and small shrubs.
|
{
"evidence": [
"Astragalus applegatei_sentence_0",
"Astragalus_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:\nAstragalus applegatei is a rare species of mikvetch which is a large genus of over 3,000 species of herbs and small shrubs.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Astragalus applegatei_sentence_0",
"Astragalus_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7101,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.228",
"value": "?search="
},
{
"operation": "search",
"time": "9.054",
"value": "Astragalus applegatei"
},
{
"operation": "Now on",
"time": "9.916",
"value": "Astragalus applegatei"
},
{
"operation": "Highlighting",
"time": "22.869",
"value": "Astragalus applegatei_sentence_0"
},
{
"operation": "hyperlink",
"time": "24.714",
"value": "milkvetch"
},
{
"operation": "Now on",
"time": "25.739",
"value": "Astragalus"
},
{
"operation": "Highlighting",
"time": "31.666",
"value": "Astragalus_sentence_2"
},
{
"operation": "finish",
"time": "63.436",
"value": "finish"
}
]
}
|
exfever_3739
|
Edmund North was a screenwriter who shared an award for artistic and technical merit in the film industry where the statuette sketch by a British art director.
|
{
"evidence": [
"Edmund H. North",
"Academy Awards",
"Cedric Gibbons"
],
"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:\nEdmund North was a screenwriter who shared an award for artistic and technical merit in the film industry where the statuette sketch by a British art director.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Edmund H. North",
"Academy Awards",
"Cedric Gibbons"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7102,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3385
|
Warcraft is a 2016 American action fantasy film based on a video game series of a franchise of video games, novels, and other media created by Blizzard Entertainment.
|
{
"evidence": [
"Warcraft (film)",
"Warcraft"
],
"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:\nWarcraft is a 2016 American action fantasy film based on a video game series of a franchise of video games, novels, and other media created by Blizzard Entertainment.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Warcraft (film)",
"Warcraft"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7103,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7777
|
Stephen Robert Irwin was an Australian zookeeper, conservationist, television personality, wildlife expert, and environmentalist who co-hosted The Crocodile Hunter with his wife, an Indian-American naturalist, conservationist, TV personality, and author who owns Australia Zoo in Beerwah.
|
{
"evidence": [
"Steve Irwin",
"Terri Irwin"
],
"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:\nStephen Robert Irwin was an Australian zookeeper, conservationist, television personality, wildlife expert, and environmentalist who co-hosted The Crocodile Hunter with his wife, an Indian-American naturalist, conservationist, TV personality, and author who owns Australia Zoo in Beerwah.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Steve Irwin",
"Terri Irwin"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7104,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
38350
|
TuS Heeslingen merged with Sportverein Viktoria Heeslingen in before breaking into the Verbandsliga Ost Niedesachsen (V).
|
{
"evidence": [
"TuS Heeslingen_sentence_3",
"TuS Heeslingen_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:\nTuS Heeslingen merged with Sportverein Viktoria Heeslingen in before breaking into the Verbandsliga Ost Niedesachsen (V).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"TuS Heeslingen_sentence_3",
"TuS Heeslingen_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7105,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.091",
"value": "?search="
},
{
"operation": "search",
"time": "47.568",
"value": "TuS Heeslingen"
},
{
"operation": "Now on",
"time": "48.42",
"value": "TuS Heeslingen"
},
{
"operation": "Page search",
"time": "83.28",
"value": "Sportverein Viktoria Heeslingen"
},
{
"operation": "Highlighting",
"time": "85.611",
"value": "TuS Heeslingen_sentence_3"
},
{
"operation": "Page search",
"time": "106.398",
"value": "Verbandsliga Ost"
},
{
"operation": "Highlighting",
"time": "112.973",
"value": "TuS Heeslingen_sentence_4"
},
{
"operation": "finish",
"time": "122.002",
"value": "finish"
}
]
}
|
exfever_4782
|
Andrew Maxwell Dwyer is a fictional character that appears in the first season as an American actress slacker boyfriend.
|
{
"evidence": [
"Andy Dwyer",
"Rashida Jones"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAndrew Maxwell Dwyer is a fictional character that appears in the first season as an American actress slacker boyfriend.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Andy Dwyer",
"Rashida Jones"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7106,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1491
|
Ellen Bursty is an American actress who earned a nomination for an Academy Award for her appearance in a 1973 American supernatural horror film directed by James Cameron.
|
{
"evidence": [
"Ellen Burstyn",
"The Exorcist (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:\nEllen Bursty is an American actress who earned a nomination for an Academy Award for her appearance in a 1973 American supernatural horror film directed by James Cameron.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ellen Burstyn",
"The Exorcist (film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7107,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7124
|
The Hills Have Eyes is a 1991 American psychological horror film directed by Jonathan Demme and adapted from Thomas Harris's psychological horror novel featuring the cannibalistic serial killer Dr. Hannibal Lecter.
|
{
"evidence": [
"The Silence of the Lambs (film)",
"The Silence of the Lambs (novel)"
],
"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 Hills Have Eyes is a 1991 American psychological horror film directed by Jonathan Demme and adapted from Thomas Harris's psychological horror novel featuring the cannibalistic serial killer Dr. Hannibal Lecter.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Silence of the Lambs (film)",
"The Silence of the Lambs (novel)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7108,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1918
|
Buffy the Vampire Slayer is the highest-grossing television series of all time.
|
{
"evidence": [
"Sarah Michelle Gellar",
"Daytime Emmy Award for Outstanding Younger Actress in a Drama Series"
],
"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 is the highest-grossing television series of all time.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sarah Michelle Gellar",
"Daytime Emmy Award for Outstanding Younger Actress in a Drama Series"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7109,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2047
|
Lonesome Dove is an American epic Western adventure television miniseries based on a four-part adaptation of the 1981 novel by Larry McMurtry directed by an American novelist, essayist, bookseller, and screenwriter whose work was predominantly set in either the Old West or contemporary Texas.
|
{
"evidence": [
"Lonesome Dove (miniseries)",
"Larry McMurtry"
],
"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:\nLonesome Dove is an American epic Western adventure television miniseries based on a four-part adaptation of the 1981 novel by Larry McMurtry directed by an American novelist, essayist, bookseller, and screenwriter whose work was predominantly set in either the Old West or contemporary Texas.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lonesome Dove (miniseries)",
"Larry McMurtry"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7110,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
58800
|
A curb box, typically located between neither a building nor the district's water main lines, is a vertical cast iron sleeve, accessible from the public way, housing the shut-off valve for a property's water service line.
|
{
"evidence": [
"Curb box_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:\nA curb box, typically located between neither a building nor the district's water main lines, is a vertical cast iron sleeve, accessible from the public way, housing the shut-off valve for a property's water service line.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Curb box_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7111,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.31",
"value": "?search="
},
{
"operation": "search",
"time": "46.333",
"value": "curb box"
},
{
"operation": "Now on",
"time": "47.286",
"value": "Curb box"
},
{
"operation": "Highlighting",
"time": "104.719",
"value": "Curb box_sentence_1"
},
{
"operation": "finish",
"time": "112.281",
"value": "finish"
}
]
}
|
exfever_3709
|
Doctor Who is the 15th television programme broadcast by the BBC.
|
{
"evidence": [
"Doctor Who",
"Time travel"
],
"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:\nDoctor Who is the 15th television programme broadcast by the BBC.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Doctor Who",
"Time travel"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7112,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5721
|
The Incredible Hulk is a 2008 American superhero film based on the brand name and primary imprint of Marvel Worldwide Inc., which was acquired by the Walt Disney Company in 2019.
|
{
"evidence": [
"The Incredible Hulk (film)",
"Marvel Comics"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Incredible Hulk is a 2008 American superhero film based on the brand name and primary imprint of Marvel Worldwide Inc., which was acquired by the Walt Disney Company in 2019.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Incredible Hulk (film)",
"Marvel Comics"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7113,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
25635
|
Restinga is a forest with medium-sized trees found in eastern Brazil.
|
{
"evidence": [
"Restinga_sentence_1",
"Restinga_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:\nRestinga is a forest with medium-sized trees found in eastern Brazil.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Restinga_sentence_1",
"Restinga_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7114,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.631",
"value": "?search="
},
{
"operation": "search",
"time": "4.58",
"value": "Restinga"
},
{
"operation": "Now on",
"time": "5.418",
"value": "Restinga"
},
{
"operation": "Page search",
"time": "19.906",
"value": "forest"
},
{
"operation": "Highlighting",
"time": "30.68",
"value": "Restinga_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "30.738",
"value": "Restinga_sentence_1"
},
{
"operation": "Highlighting",
"time": "31.459",
"value": "Restinga_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "31.522",
"value": "Restinga_sentence_1"
},
{
"operation": "Highlighting",
"time": "32.04",
"value": "Restinga_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "32.111",
"value": "Restinga_sentence_1"
},
{
"operation": "Highlighting",
"time": "32.61",
"value": "Restinga_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "33.392",
"value": "Restinga_sentence_1"
},
{
"operation": "Highlighting",
"time": "33.813",
"value": "Restinga_sentence_1"
},
{
"operation": "hyperlink",
"time": "62.421",
"value": "Tropical and subtropical moist broadleaf forest"
},
{
"operation": "Now on",
"time": "63.178",
"value": "Tropical and subtropical moist broadleaf forests"
},
{
"operation": "Page search",
"time": "69.651",
"value": "size"
},
{
"operation": "back-button-clicked",
"time": "74.066",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "74.128",
"value": "Restinga"
},
{
"operation": "Page search",
"time": "77.169",
"value": "sizer"
},
{
"operation": "Page search",
"time": "77.457",
"value": "sizer"
},
{
"operation": "Page search",
"time": "77.636",
"value": "sizer"
},
{
"operation": "Page search",
"time": "77.918",
"value": "sizer"
},
{
"operation": "Page search",
"time": "78.433",
"value": "size"
},
{
"operation": "Page search",
"time": "78.637",
"value": "size"
},
{
"operation": "Highlighting",
"time": "81.277",
"value": "Restinga_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "81.349",
"value": "Restinga_sentence_2"
},
{
"operation": "Highlighting",
"time": "82.162",
"value": "Restinga_sentence_2"
},
{
"operation": "finish",
"time": "87.376",
"value": "finish"
}
]
}
|
exfever_3651
|
Michael Trent Reznor is the youngest member of the industrial rock band Nine Inch Nails.
|
{
"evidence": [
"Trent Reznor",
"Synth-pop"
],
"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:\nMichael Trent Reznor is the youngest member of the industrial rock band Nine Inch Nails.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Trent Reznor",
"Synth-pop"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7115,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
44560
|
At the 1909–10 Lancashire Cup, Leigh scored 22 points against Wigan's 5.
|
{
"evidence": [
"1909–10 Lancashire Cup_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:\nAt the 1909–10 Lancashire Cup, Leigh scored 22 points against Wigan's 5.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"1909–10 Lancashire Cup_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7116,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.771",
"value": "?search="
},
{
"operation": "search",
"time": "12.517",
"value": "1909–10 Lancashire Cup"
},
{
"operation": "Now on",
"time": "13.317",
"value": "1909–10 Lancashire Cup"
},
{
"operation": "Highlighting",
"time": "75.856",
"value": "1909–10 Lancashire Cup_sentence_0"
},
{
"operation": "finish",
"time": "89.717",
"value": "finish"
}
]
}
|
exfever_7760
|
Margaret Hamilton also produced the stage production of An Evening with the Bourgeoisie.
|
{
"evidence": [
"Saratoga (film)",
"Margaret Hamilton (actress)"
],
"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:\nMargaret Hamilton also produced the stage production of An Evening with the Bourgeoisie.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Saratoga (film)",
"Margaret Hamilton (actress)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7117,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3227
|
Sir Sean Connery was a Scottish actor who appeared in several films, including a 1964 American psychological thriller film directed by a Canadian filmmaker widely regarded as one of the most influential figures in the history of cinema.
|
{
"evidence": [
"Sean Connery",
"Marnie (film)",
"Alfred Hitchcock"
],
"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:\nSir Sean Connery was a Scottish actor who appeared in several films, including a 1964 American psychological thriller film directed by a Canadian filmmaker widely regarded as one of the most influential figures in the history of cinema.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sean Connery",
"Marnie (film)",
"Alfred Hitchcock"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7118,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
1196
|
Before the first Europeans arrived or copra companies leased it, Maupihaa was home to Inca's in ancient times.
|
{
"evidence": [
"Maupihaa_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:\nBefore the first Europeans arrived or copra companies leased it, Maupihaa was home to Inca's in ancient times.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Maupihaa_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7119,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "20.116",
"value": "Maupihaa"
},
{
"operation": "Highlighting",
"time": "93.936",
"value": "Maupihaa_sentence_6"
},
{
"operation": "hyperlink",
"time": "99.455",
"value": "Polynesians"
},
{
"operation": "back-button-clicked",
"time": "114.073",
"value": "back-button-clicked"
},
{
"operation": "finish",
"time": "165.168",
"value": "finish"
}
]
}
|
35329
|
The Queensland Tennis Centre was designed by internationally recognized stadium designers HOK Sport Venue Event and the main court was designed to incorporate a fiber fabric roof.
|
{
"evidence": [
"Queensland Tennis Centre_sentence_3",
"Queensland Tennis Centre_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 Queensland Tennis Centre was designed by internationally recognized stadium designers HOK Sport Venue Event and the main court was designed to incorporate a fiber fabric roof.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Queensland Tennis Centre_sentence_3",
"Queensland Tennis Centre_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7120,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.091",
"value": "?search="
},
{
"operation": "search",
"time": "576.653",
"value": "Queensland Tennis Centre"
},
{
"operation": "Now on",
"time": "577.643",
"value": "Queensland Tennis Centre"
},
{
"operation": "Highlighting",
"time": "595.964",
"value": "Queensland Tennis Centre_sentence_3"
},
{
"operation": "Highlighting",
"time": "599.433",
"value": "Queensland Tennis Centre_sentence_4"
},
{
"operation": "finish",
"time": "619.782",
"value": "finish"
}
]
}
|
exfever_1784
|
No Way Out was the best-selling album in 1997, selling over 561,000 copies.
|
{
"evidence": [
"No Way Out (Puff Daddy album)",
"Busta Rhymes"
],
"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:\nNo Way Out was the best-selling album in 1997, selling over 561,000 copies.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"No Way Out (Puff Daddy album)",
"Busta Rhymes"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7121,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
31583
|
Reed Dickens is an American Businessman, entrepreneur and former CEO of Marucci Sports who founded Outside Eyes in 2005.
|
{
"evidence": [
"Reed Dickens_sentence_0",
"Reed Dickens_sentence_2",
"Reed Dickens_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:\nReed Dickens is an American Businessman, entrepreneur and former CEO of Marucci Sports who founded Outside Eyes in 2005.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Reed Dickens_sentence_0",
"Reed Dickens_sentence_2",
"Reed Dickens_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7122,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.331",
"value": "?search="
},
{
"operation": "search",
"time": "97.246",
"value": "reed dickens"
},
{
"operation": "Now on",
"time": "98.313",
"value": "Reed Dickens"
},
{
"operation": "Highlighting",
"time": "101.284",
"value": "Reed Dickens_sentence_0"
},
{
"operation": "Highlighting",
"time": "124.994",
"value": "Reed Dickens_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "125.209",
"value": "Reed Dickens_sentence_2"
},
{
"operation": "Highlighting",
"time": "126.491",
"value": "Reed Dickens_sentence_2"
},
{
"operation": "Highlighting",
"time": "161.99",
"value": "Reed Dickens_sentence_7"
},
{
"operation": "Page search",
"time": "210.921",
"value": "entrep"
},
{
"operation": "Highlighting",
"time": "220.021",
"value": "Reed Dickens_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "221.044",
"value": "Reed Dickens_sentence_1"
},
{
"operation": "finish",
"time": "243.405",
"value": "finish"
}
]
}
|
67345
|
František Svoboda was a Czech football player who played club football for SK Slavia Praha and played in the 1934 FIFA World Cup.
|
{
"evidence": [
"František Svoboda_sentence_1",
"František Svoboda_sentence_2",
"František Svoboda_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:\nFrantišek Svoboda was a Czech football player who played club football for SK Slavia Praha and played in the 1934 FIFA World Cup.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"František Svoboda_sentence_1",
"František Svoboda_sentence_2",
"František Svoboda_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7123,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.483",
"value": "?search="
},
{
"operation": "search",
"time": "10.914",
"value": "František Svoboda"
},
{
"operation": "Now on",
"time": "12.156",
"value": "František Svoboda"
},
{
"operation": "Highlighting",
"time": "31.162",
"value": "František Svoboda_sentence_1"
},
{
"operation": "Highlighting",
"time": "35.039",
"value": "František Svoboda_sentence_2"
},
{
"operation": "Highlighting",
"time": "38.824",
"value": "František Svoboda_sentence_3"
},
{
"operation": "finish",
"time": "50.758",
"value": "finish"
}
]
}
|
28284
|
Tomas Abraham refused to play for Sokol Hrotovice, FC Slavia Trebic, FC Zbrojovka Brno, a professional football club based in the city of Brno, Moravia, Czech Republic, and FK Postorna.
|
{
"evidence": [
"Tomáš Abrahám_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:\nTomas Abraham refused to play for Sokol Hrotovice, FC Slavia Trebic, FC Zbrojovka Brno, a professional football club based in the city of Brno, Moravia, Czech Republic, and FK Postorna.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Tomáš Abrahám_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7124,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.749",
"value": "?search="
},
{
"operation": "search",
"time": "9.476",
"value": "Tomas Abraham"
},
{
"operation": "Now on",
"time": "10.289",
"value": "Tomáš Abrahám"
},
{
"operation": "Highlighting",
"time": "12.025",
"value": "Tomáš Abrahám_sentence_1"
},
{
"operation": "finish",
"time": "21.011",
"value": "finish"
}
]
}
|
39018
|
Tennis at the Pacific Games was held in Suva in 1936 then in Port Moresby in 1966.
|
{
"evidence": [
"Pacific Games_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:\nTennis at the Pacific Games was held in Suva in 1936 then in Port Moresby in 1966.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Pacific Games_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7125,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.588",
"value": "?search="
},
{
"operation": "search",
"time": "12.572",
"value": "Pacific Games"
},
{
"operation": "Now on",
"time": "13.632",
"value": "Pacific Games"
},
{
"operation": "hyperlink",
"time": "48.823",
"value": "Suva"
},
{
"operation": "Now on",
"time": "49.543",
"value": "Suva"
},
{
"operation": "search",
"time": "66.893",
"value": "Tennis Pacific games"
},
{
"operation": "Now on",
"time": "68.77",
"value": "undefined"
},
{
"operation": "search",
"time": "94.817",
"value": "Pacific games"
},
{
"operation": "Now on",
"time": "95.337",
"value": "Pacific Games"
},
{
"operation": "hyperlink",
"time": "98.355",
"value": "Pacific Ocean Games"
},
{
"operation": "Now on",
"time": "98.773",
"value": "Pacific Ocean Games"
},
{
"operation": "hyperlink",
"time": "111.707",
"value": "Pacific Games"
},
{
"operation": "Now on",
"time": "112.069",
"value": "Pacific Games"
},
{
"operation": "Highlighting",
"time": "121.852",
"value": "Pacific Games_sentence_1"
},
{
"operation": "search",
"time": "129.247",
"value": "South Pacific Games"
},
{
"operation": "Now on",
"time": "130.093",
"value": "Pacific Games"
},
{
"operation": "search",
"time": "178.207",
"value": "Port Moresby"
},
{
"operation": "Now on",
"time": "178.99",
"value": "Port Moresby"
},
{
"operation": "search",
"time": "249.395",
"value": "South Pacific Games"
},
{
"operation": "Now on",
"time": "250.023",
"value": "South Pacific Games"
},
{
"operation": "hyperlink",
"time": "255.5",
"value": "South Pacific Games"
},
{
"operation": "Now on",
"time": "256.081",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "257.533",
"value": "Pacific Games"
},
{
"operation": "Now on",
"time": "257.798",
"value": "Pacific Games"
},
{
"operation": "hyperlink",
"time": "269.847",
"value": "Pacific Ocean Games"
},
{
"operation": "Now on",
"time": "270.408",
"value": "Pacific Ocean Games"
},
{
"operation": "back-button-clicked",
"time": "278.434",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "278.568",
"value": "Pacific Games"
},
{
"operation": "Highlighting",
"time": "281.791",
"value": "Pacific Games_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "300.926",
"value": "Pacific Games_sentence_1"
},
{
"operation": "finish",
"time": "353.673",
"value": "finish"
}
]
}
|
19880
|
The House of Representatives is part of the executive branch of the federal government of the United States consisting of congressional district representatives serving six-year terms.
|
{
"evidence": [
"United States Congress_sentence_2",
"United States Congress_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:\nThe House of Representatives is part of the executive branch of the federal government of the United States consisting of congressional district representatives serving six-year terms.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"United States Congress_sentence_2",
"United States Congress_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7126,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.722",
"value": "undefined"
},
{
"operation": "search",
"time": "18.574",
"value": "Executive branch of the United States government"
},
{
"operation": "Now on",
"time": "19.883",
"value": "Federal government of the United States"
},
{
"operation": "hyperlink",
"time": "29.787",
"value": "legislative"
},
{
"operation": "Now on",
"time": "31.461",
"value": "United States Congress"
},
{
"operation": "Highlighting",
"time": "47.124",
"value": "United States Congress_sentence_2"
},
{
"operation": "Highlighting",
"time": "99.729",
"value": "United States Congress_sentence_15"
},
{
"operation": "Highlighting deleted",
"time": "149.039",
"value": "United States Congress_sentence_15"
},
{
"operation": "Highlighting",
"time": "169.795",
"value": "United States Congress_sentence_8"
},
{
"operation": "finish",
"time": "179.796",
"value": "finish"
}
]
}
|
exfever_5874
|
Cougar Town is an American television sitcom produced by the production company of an American screenwriter, producer, and director.
|
{
"evidence": [
"Cougar Town",
"Doozer (company)",
"Bill Lawrence (TV producer)"
],
"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:\nCougar Town is an American television sitcom produced by the production company of an American screenwriter, producer, and director.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Cougar Town",
"Doozer (company)",
"Bill Lawrence (TV producer)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7127,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
35691
|
The professional wrestling pay-per-view event put on by Juggalo Championship Wrestling took place in 2011 and featured nine different matches with wrestlers such as X-Pac, Corporal Robinson, Bill Martel and Officer Colt Cabana.
|
{
"evidence": [
"Juggalo Championship Wrestling_sentence_11"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe professional wrestling pay-per-view event put on by Juggalo Championship Wrestling took place in 2011 and featured nine different matches with wrestlers such as X-Pac, Corporal Robinson, Bill Martel and Officer Colt Cabana.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Juggalo Championship Wrestling_sentence_11"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7128,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.099",
"value": "?search="
},
{
"operation": "search",
"time": "40.252",
"value": "Juggalo Championship Wrestling"
},
{
"operation": "Now on",
"time": "41.121",
"value": "Juggalo Championship Wrestling"
},
{
"operation": "hyperlink",
"time": "121.534",
"value": "List of Juggalo Championship Wrestling employees"
},
{
"operation": "Now on",
"time": "122.21",
"value": "List of Juggalo Championship Wrestling personnel"
},
{
"operation": "back-button-clicked",
"time": "159.119",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "159.23",
"value": "Juggalo Championship Wrestling"
},
{
"operation": "Highlighting",
"time": "177.654",
"value": "Juggalo Championship Wrestling_sentence_11"
},
{
"operation": "hyperlink",
"time": "179.379",
"value": "internet"
},
{
"operation": "Now on",
"time": "180.465",
"value": "Internet"
},
{
"operation": "back-button-clicked",
"time": "182.42",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "182.534",
"value": "Juggalo Championship Wrestling"
},
{
"operation": "hyperlink",
"time": "183.538",
"value": "pay-per-view"
},
{
"operation": "Now on",
"time": "185.089",
"value": "Pay-per-view"
},
{
"operation": "back-button-clicked",
"time": "204.22",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "204.351",
"value": "Juggalo Championship Wrestling"
},
{
"operation": "search",
"time": "243.804",
"value": "X-Pac"
},
{
"operation": "Now on",
"time": "244.918",
"value": "Sean Waltman"
},
{
"operation": "finish",
"time": "267.703",
"value": "finish"
}
]
}
|
38407
|
According to the IEA, a policy advisor focusing on energy security, economic development, and environmental protection, energy security and renewable technology could make a direct contribution to heating.
|
{
"evidence": [
"International Energy Agency_sentence_2",
"International Energy Agency_sentence_3",
"International Energy Agency_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:\nAccording to the IEA, a policy advisor focusing on energy security, economic development, and environmental protection, energy security and renewable technology could make a direct contribution to heating.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"International Energy Agency_sentence_2",
"International Energy Agency_sentence_3",
"International Energy Agency_sentence_5"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7129,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.952",
"value": "?search="
},
{
"operation": "search",
"time": "30.782",
"value": "IEA"
},
{
"operation": "Now on",
"time": "31.45",
"value": "IEA"
},
{
"operation": "hyperlink",
"time": "51.829",
"value": "International Energy Agency"
},
{
"operation": "Now on",
"time": "52.612",
"value": "International Energy Agency"
},
{
"operation": "Highlighting",
"time": "56.868",
"value": "International Energy Agency_sentence_2"
},
{
"operation": "Highlighting",
"time": "63.349",
"value": "International Energy Agency_sentence_3"
},
{
"operation": "Highlighting",
"time": "76.243",
"value": "International Energy Agency_sentence_5"
},
{
"operation": "back-button-clicked",
"time": "135.691",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "135.907",
"value": "IEA"
},
{
"operation": "hyperlink",
"time": "153.87",
"value": "International Energy Agency"
},
{
"operation": "Now on",
"time": "154.398",
"value": "International Energy Agency"
},
{
"operation": "search",
"time": "189.899",
"value": "renewable technology"
},
{
"operation": "Now on",
"time": "191.678",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "196.459",
"value": "Renewable energy"
},
{
"operation": "Now on",
"time": "198.284",
"value": "Renewable energy"
},
{
"operation": "finish",
"time": "210.014",
"value": "finish"
}
]
}
|
34204
|
János Pap (1925-1994) was a French politician in the MSZMP party.
|
{
"evidence": [
"János Pap_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:\nJános Pap (1925-1994) was a French politician in the MSZMP party.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"János Pap_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7130,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.721",
"value": "?search="
},
{
"operation": "search",
"time": "9.226",
"value": "János Pap"
},
{
"operation": "Now on",
"time": "10.151",
"value": "János Pap"
},
{
"operation": "Highlighting",
"time": "22.787",
"value": "János Pap_sentence_2"
},
{
"operation": "finish",
"time": "38.954",
"value": "finish"
}
]
}
|
exfever_4812
|
William Clark Gable was a film actor who played a supporting role for an American media company in which Kirk Kerkorian purchased 40% of the company and diversified its products, forming a hotel and casino company based in a major metropolitan area in the southern part of the United States state of Nevada.
|
{
"evidence": [
"Clark Gable",
"Metro-Goldwyn-Mayer",
"Las Vegas Valley"
],
"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:\nWilliam Clark Gable was a film actor who played a supporting role for an American media company in which Kirk Kerkorian purchased 40% of the company and diversified its products, forming a hotel and casino company based in a major metropolitan area in the southern part of the United States state of Nevada.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Clark Gable",
"Metro-Goldwyn-Mayer",
"Las Vegas Valley"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7131,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8939
|
Felicity Rose Hadley Jones is an English actress who played the role of Emma Grundy in The Archers, which aired on the national broadcaster of the United Kingdom, that charges all British households, companies, and organizations using any type of equipment to receive or record catch-up on a video-on-demand service available on a wide range of devices.
|
{
"evidence": [
"Felicity Jones",
"BBC",
"BBC iPlayer"
],
"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:\nFelicity Rose Hadley Jones is an English actress who played the role of Emma Grundy in The Archers, which aired on the national broadcaster of the United Kingdom, that charges all British households, companies, and organizations using any type of equipment to receive or record catch-up on a video-on-demand service available on a wide range of devices.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Felicity Jones",
"BBC",
"BBC iPlayer"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7132,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
43184
|
State Correctional Institution – Pine Grove is especially built for prisons that are 18 year-old and under.
|
{
"evidence": [
"State Correctional Institution – Pine Grove_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:\nState Correctional Institution – Pine Grove is especially built for prisons that are 18 year-old and under.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"State Correctional Institution – Pine Grove_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7133,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.259",
"value": "?search="
},
{
"operation": "search",
"time": "25.067",
"value": "State Correctional Institution - Pine Grove"
},
{
"operation": "Now on",
"time": "26.735",
"value": "State Correctional Institution – Pine Grove"
},
{
"operation": "Highlighting",
"time": "63.176",
"value": "State Correctional Institution – Pine Grove_sentence_5"
},
{
"operation": "Highlighting",
"time": "65.215",
"value": "State Correctional Institution – Pine Grove_sentence_6"
},
{
"operation": "Highlighting",
"time": "66.24",
"value": "State Correctional Institution – Pine Grove_sentence_7"
},
{
"operation": "Highlighting",
"time": "67.145",
"value": "State Correctional Institution – Pine Grove_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "74.265",
"value": "State Correctional Institution – Pine Grove_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "76.457",
"value": "State Correctional Institution – Pine Grove_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "77.896",
"value": "State Correctional Institution – Pine Grove_sentence_5"
},
{
"operation": "finish",
"time": "102.763",
"value": "finish"
}
]
}
|
exfever_5017
|
KFC is a Canadian fast-food restaurant chain that is a subsidiary of an American fast-food corporation listed on the Fortune 1000 that operates Pizza Hut except in China, where an American Chinese Fortune 500 fast-food restaurant company based in Plano, Texas is operated by a separate company,
|
{
"evidence": [
"KFC",
"Yum! Brands",
"Yum China"
],
"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:\nKFC is a Canadian fast-food restaurant chain that is a subsidiary of an American fast-food corporation listed on the Fortune 1000 that operates Pizza Hut except in China, where an American Chinese Fortune 500 fast-food restaurant company based in Plano, Texas is operated by a separate company,\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"KFC",
"Yum! Brands",
"Yum China"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7134,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4989
|
Broadcast News is a romantic comedy-drama film that stars an American actor who became a well-known supporting actor in the 1980s.
|
{
"evidence": [
"Broadcast News (film)",
"Robert Prosky"
],
"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:\nBroadcast News is a romantic comedy-drama film that stars an American actor who became a well-known supporting actor in the 1980s.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Broadcast News (film)",
"Robert Prosky"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7135,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
8585
|
Due in part to the fact that there is no screening examination for early detection, ovarian cancer is the most malignant tumour of the female reproductive organs with 9,000 women newly affected each year.
|
{
"evidence": [
"Ovarian cancer_sentence_23",
"Ovarian cancer_sentence_603"
],
"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:\nDue in part to the fact that there is no screening examination for early detection, ovarian cancer is the most malignant tumour of the female reproductive organs with 9,000 women newly affected each year.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Ovarian cancer_sentence_23",
"Ovarian cancer_sentence_603"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7136,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.567",
"value": "?search="
},
{
"operation": "search",
"time": "21.174",
"value": "Ovarian cancer"
},
{
"operation": "Now on",
"time": "22.792",
"value": "Ovarian cancer"
},
{
"operation": "Highlighting",
"time": "89.496",
"value": "Ovarian cancer_sentence_23"
},
{
"operation": "Highlighting",
"time": "156.449",
"value": "Ovarian cancer_sentence_603"
},
{
"operation": "finish",
"time": "254.781",
"value": "finish"
}
]
}
|
exfever_7585
|
WeChat is the first Chinese instant messaging service developed by Tencent.
|
{
"evidence": [
"WeChat",
"Mobile payment"
],
"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:\nWeChat is the first Chinese instant messaging service developed by Tencent.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"WeChat",
"Mobile payment"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7137,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1965
|
Judith Quiney, William Shakespeare's and Anne Hathaway's younger daughter, is married to a tobacconist in Stratford-upon-Avon who held several municipal offices in the corporation of Stratford-upon-Avon.
|
{
"evidence": [
"Judith Quiney",
"Thomas Quiney"
],
"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:\nJudith Quiney, William Shakespeare's and Anne Hathaway's younger daughter, is married to a tobacconist in Stratford-upon-Avon who held several municipal offices in the corporation of Stratford-upon-Avon.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Judith Quiney",
"Thomas Quiney"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7138,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
23705
|
In March 1995, Mayhugh refuse to sell KLOA-AM-FM to Adelman Broadcasting for $500,000; at that time, KLOA-FM carried a country music format.
|
{
"evidence": [
"KEPD_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:\nIn March 1995, Mayhugh refuse to sell KLOA-AM-FM to Adelman Broadcasting for $500,000; at that time, KLOA-FM carried a country music format.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"KEPD_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7139,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.412",
"value": "undefined"
},
{
"operation": "search",
"time": "27.387",
"value": "Mayhugh"
},
{
"operation": "Now on",
"time": "29.071",
"value": "undefined"
},
{
"operation": "search",
"time": "65.341",
"value": "KLOA-AM-FM"
},
{
"operation": "Now on",
"time": "66.116",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "70.62",
"value": "KLOA (AM)"
},
{
"operation": "Now on",
"time": "71.269",
"value": "KLOA (AM)"
},
{
"operation": "hyperlink",
"time": "93.549",
"value": "KEPD"
},
{
"operation": "Now on",
"time": "94.016",
"value": "KEPD"
},
{
"operation": "Highlighting",
"time": "102.287",
"value": "KEPD_sentence_6"
},
{
"operation": "finish",
"time": "109.922",
"value": "finish"
}
]
}
|
35670
|
Bigfoot and Wildboy, a live action ABC children's show, was released on video by Embassy Video in the 1980s.
|
{
"evidence": [
"Bigfoot and Wildboy_sentence_0",
"Bigfoot and Wildboy_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:\nBigfoot and Wildboy, a live action ABC children's show, was released on video by Embassy Video in the 1980s.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bigfoot and Wildboy_sentence_0",
"Bigfoot and Wildboy_sentence_10"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7140,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.884",
"value": "?search="
},
{
"operation": "search",
"time": "8.499",
"value": "Bigfoot and Wildboy"
},
{
"operation": "Now on",
"time": "9.884",
"value": "Bigfoot and Wildboy"
},
{
"operation": "Highlighting",
"time": "11.717",
"value": "Bigfoot and Wildboy_sentence_0"
},
{
"operation": "Page search",
"time": "37.371",
"value": "video"
},
{
"operation": "Highlighting",
"time": "50.998",
"value": "Bigfoot and Wildboy_sentence_10"
},
{
"operation": "finish",
"time": "75.3",
"value": "finish"
}
]
}
|
91655
|
Grog Run lets out into the Little Miami River (a Class I tributary of the Ohio River) where residents have found many Native American artifacts.
|
{
"evidence": [
"Grog Run_sentence_4",
"Grog Run_sentence_5",
"Little Miami River_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:\nGrog Run lets out into the Little Miami River (a Class I tributary of the Ohio River) where residents have found many Native American artifacts.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Grog Run_sentence_4",
"Grog Run_sentence_5",
"Little Miami River_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7141,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.012",
"value": "?search="
},
{
"operation": "search",
"time": "8.061",
"value": "Grog Run"
},
{
"operation": "Now on",
"time": "9.189",
"value": "Grog Run"
},
{
"operation": "Highlighting",
"time": "12.25",
"value": "Grog Run_sentence_4"
},
{
"operation": "Highlighting",
"time": "23.705",
"value": "Grog Run_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "31.717",
"value": "Grog Run_sentence_3"
},
{
"operation": "Highlighting",
"time": "34.018",
"value": "Grog Run_sentence_0Grog Run_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "37.742",
"value": "Grog Run_sentence_0Grog Run_sentence_1"
},
{
"operation": "Highlighting",
"time": "48.539",
"value": "Grog Run_sentence_5"
},
{
"operation": "search",
"time": "62.45",
"value": "Little Miami River"
},
{
"operation": "Now on",
"time": "64.124",
"value": "Little Miami River"
},
{
"operation": "Highlighting",
"time": "71.046",
"value": "Little Miami River_sentence_1"
},
{
"operation": "finish",
"time": "95.75",
"value": "finish"
}
]
}
|
exfever_1676
|
A Nightmare on Elm Street is a 2010 American horror film with a genre involving a killer stalking and murdering a group of people like the Texas Chain Saw Massacre and a 1978 American independent film directed and scored by John Carpenter.
|
{
"evidence": [
"A Nightmare on Elm Street (2010 film)",
"Slasher film",
"Halloween (1978 film)"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nA Nightmare on Elm Street is a 2010 American horror film with a genre involving a killer stalking and murdering a group of people like the Texas Chain Saw Massacre and a 1978 American independent film directed and scored by John Carpenter.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"A Nightmare on Elm Street (2010 film)",
"Slasher film",
"Halloween (1978 film)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7142,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1261
|
Baby Driver is a 2017 drama film with a supporting role like an American actor and producer.
|
{
"evidence": [
"Baby Driver",
"Jon Hamm"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBaby Driver is a 2017 drama film with a supporting role like an American actor and producer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Baby Driver",
"Jon Hamm"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7143,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5273
|
The Republican Party is one of the two political parties in the United States that holds substantial influence in a country's politics, standing in contrast to a minor party.
|
{
"evidence": [
"Republican Party (United States)",
"Major party"
],
"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 Republican Party is one of the two political parties in the United States that holds substantial influence in a country's politics, standing in contrast to a minor party.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Republican Party (United States)",
"Major party"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7144,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1352
|
The Boys from Brazil is a 1978 British-American science fiction thriller film starring an American actor and one of the most popular film stars from the 1940s to the 1960s.
|
{
"evidence": [
"The Boys from Brazil (film)",
"Gregory Peck"
],
"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 Boys from Brazil is a 1978 British-American science fiction thriller film starring an American actor and one of the most popular film stars from the 1940s to the 1960s.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Boys from Brazil (film)",
"Gregory Peck"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7145,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
53723
|
Hassan Illiassou competed in the 100 m event, but only ranked fifth with a result of 10.73, while Boureima Kimba was the only one to advance to the quarterfinals.
|
{
"evidence": [
"Hassane Illiassou_sentence_1",
"Hassane Illiassou_sentence_0",
"Bouriema Kimba_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:\nHassan Illiassou competed in the 100 m event, but only ranked fifth with a result of 10.73, while Boureima Kimba was the only one to advance to the quarterfinals.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Hassane Illiassou_sentence_1",
"Hassane Illiassou_sentence_0",
"Bouriema Kimba_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7146,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.75",
"value": "?search="
},
{
"operation": "search",
"time": "11.944",
"value": "Hassan Illiassou"
},
{
"operation": "Now on",
"time": "12.754",
"value": "Hassane Illiassou"
},
{
"operation": "Highlighting",
"time": "19.572",
"value": "Hassane Illiassou_sentence_0"
},
{
"operation": "Highlighting",
"time": "33.767",
"value": "Hassane Illiassou_sentence_1"
},
{
"operation": "hyperlink",
"time": "46.005",
"value": "1992 Summer Olympics"
},
{
"operation": "Now on",
"time": "73.851",
"value": "1992 Summer Olympics"
},
{
"operation": "back-button-clicked",
"time": "79.001",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "79.441",
"value": "Hassane Illiassou"
},
{
"operation": "Highlighting deleted",
"time": "92.99",
"value": "Hassane Illiassou_sentence_0"
},
{
"operation": "Highlighting",
"time": "97.795",
"value": "Hassane Illiassou_sentence_0"
},
{
"operation": "search",
"time": "102.654",
"value": "Contains... men's 100m competition at the 1992 Summer Olympics"
},
{
"operation": "Now on",
"time": "106.606",
"value": "undefined"
},
{
"operation": "search",
"time": "138.155",
"value": "Boureima Kimba"
},
{
"operation": "Now on",
"time": "138.836",
"value": "Bouriema Kimba"
},
{
"operation": "Highlighting",
"time": "145.986",
"value": "Bouriema Kimba_sentence_0"
},
{
"operation": "hyperlink",
"time": "150.968",
"value": "1996 Summer Olympics"
},
{
"operation": "Now on",
"time": "164.958",
"value": "1996 Summer Olympics"
},
{
"operation": "finish",
"time": "185.386",
"value": "finish"
}
]
}
|
56998
|
Martin (one of fourteen women executed for witchcraft during the Salem witch trials of colonial Massachusetts) was baptized in Olney, Buckinghamshire, England on September 30, 1621; then on August 11, 1646 at Salisbury, she married a widower named George Martin.
|
{
"evidence": [
"Susannah Martin_sentence_0",
"Susannah Martin_sentence_4",
"Susannah Martin_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:\nMartin (one of fourteen women executed for witchcraft during the Salem witch trials of colonial Massachusetts) was baptized in Olney, Buckinghamshire, England on September 30, 1621; then on August 11, 1646 at Salisbury, she married a widower named George Martin.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Susannah Martin_sentence_0",
"Susannah Martin_sentence_4",
"Susannah Martin_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7147,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.819",
"value": "?search="
},
{
"operation": "search",
"time": "22.018",
"value": "George Martin salem"
},
{
"operation": "Now on",
"time": "24.166",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "44.817",
"value": "Salem witch trials"
},
{
"operation": "Now on",
"time": "47.853",
"value": "Salem witch trials"
},
{
"operation": "hyperlink",
"time": "74.658",
"value": "Susannah Martin"
},
{
"operation": "Now on",
"time": "75.172",
"value": "Susannah Martin"
},
{
"operation": "Highlighting",
"time": "77.075",
"value": "Susannah Martin_sentence_0"
},
{
"operation": "Highlighting",
"time": "83.69",
"value": "Susannah Martin_sentence_4"
},
{
"operation": "Highlighting",
"time": "92.793",
"value": "Susannah Martin_sentence_5"
},
{
"operation": "finish",
"time": "119.546",
"value": "finish"
}
]
}
|
51765
|
Mark Vermeulen was called up to the Zimbabwe squad for the One Day International Series, a form of limited overs cricket.
|
{
"evidence": [
"Mark Vermeulen_sentence_20",
"One Day International_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:\nMark Vermeulen was called up to the Zimbabwe squad for the One Day International Series, a form of limited overs cricket.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Mark Vermeulen_sentence_20",
"One Day International_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7148,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.702",
"value": "?search="
},
{
"operation": "search",
"time": "23.278",
"value": "Mark Vermelulen"
},
{
"operation": "Now on",
"time": "24.377",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "31.542",
"value": "Mark Vermeulen"
},
{
"operation": "Now on",
"time": "32.097",
"value": "Mark Vermeulen"
},
{
"operation": "Page search",
"time": "36.381",
"value": "called up"
},
{
"operation": "Page search",
"time": "54.197",
"value": "One Day"
},
{
"operation": "Highlighting",
"time": "62.093",
"value": "Mark Vermeulen_sentence_20"
},
{
"operation": "hyperlink",
"time": "66.748",
"value": "One Day International"
},
{
"operation": "Now on",
"time": "67.555",
"value": "One Day International"
},
{
"operation": "Highlighting",
"time": "72.916",
"value": "One Day International_sentence_2"
},
{
"operation": "finish",
"time": "79.379",
"value": "finish"
}
]
}
|
58778
|
Erna Baumbauer (8 January 1919 – 30 January 2010) was a casting agent and honorary awardee of the German Film Award from Germany who was called the "grand dame of German casting agents" by The Hollywood Reporter".
|
{
"evidence": [
"Erna Baumbauer_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:\nErna Baumbauer (8 January 1919 – 30 January 2010) was a casting agent and honorary awardee of the German Film Award from Germany who was called the \"grand dame of German casting agents\" by The Hollywood Reporter\".\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Erna Baumbauer_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7149,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.928",
"value": "?search="
},
{
"operation": "search",
"time": "11.996",
"value": "Erna Baumbauer"
},
{
"operation": "Now on",
"time": "13.021",
"value": "Erna Baumbauer"
},
{
"operation": "hyperlink",
"time": "45.977",
"value": "German"
},
{
"operation": "Now on",
"time": "47.906",
"value": "Erna Baumbauer"
},
{
"operation": "Highlighting",
"time": "49.051",
"value": "Erna Baumbauer_sentence_0"
},
{
"operation": "Page search",
"time": "93.765",
"value": "honorary"
},
{
"operation": "search",
"time": "135.071",
"value": "German Film Award"
},
{
"operation": "Now on",
"time": "139.416",
"value": "Deutscher Filmpreis"
},
{
"operation": "Page search",
"time": "181.617",
"value": "baum"
},
{
"operation": "back-button-clicked",
"time": "209.441",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "209.507",
"value": "Erna Baumbauer"
},
{
"operation": "Highlighting",
"time": "211.889",
"value": "Erna Baumbauer_sentence_1"
},
{
"operation": "Highlighting",
"time": "214.495",
"value": "Erna Baumbauer_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "215.001",
"value": "Erna Baumbauer_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "215.507",
"value": "Erna Baumbauer_sentence_1"
},
{
"operation": "search",
"time": "234.325",
"value": "honorary awardee of the German Film Award"
},
{
"operation": "Now on",
"time": "235.517",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "249.159",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "249.208",
"value": "Erna Baumbauer"
},
{
"operation": "hyperlink",
"time": "253.472",
"value": "The Hollywood Reporter"
},
{
"operation": "Now on",
"time": "254.467",
"value": "The Hollywood Reporter"
},
{
"operation": "back-button-clicked",
"time": "256.003",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "256.051",
"value": "Erna Baumbauer"
},
{
"operation": "Page search",
"time": "263.534",
"value": "German Film Award"
},
{
"operation": "hyperlink",
"time": "274.197",
"value": "lifetime achievement award"
},
{
"operation": "Now on",
"time": "274.618",
"value": "Lifetime achievement award"
},
{
"operation": "back-button-clicked",
"time": "284.365",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "284.418",
"value": "Erna Baumbauer"
},
{
"operation": "finish",
"time": "346.385",
"value": "finish"
}
]
}
|
50621
|
Sabie is a forestry town situated on the banks of the Sabie River in Mpumalanga, South Africa.
|
{
"evidence": [
"Sabie_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:\nSabie is a forestry town situated on the banks of the Sabie River in Mpumalanga, South Africa.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Sabie_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7150,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0",
"value": "?search="
},
{
"operation": "search",
"time": "0.023",
"value": "Sabie"
},
{
"operation": "Now on",
"time": "0.024",
"value": "Sabie"
},
{
"operation": "Highlighting",
"time": "0.029",
"value": "Sabie_sentence_1"
},
{
"operation": "finish",
"time": "0.043",
"value": "finish"
},
{
"operation": "Now on",
"time": "1621091012.268",
"value": "Sabie"
},
{
"operation": "finish",
"time": "1621091022.57",
"value": "finish"
}
]
}
|
21142
|
Rufus Granderson, an American football defensive tackle who played one season with the Dallas Texans of the American Football League, died in 2015 at the age of 79.
|
{
"evidence": [
"Rufus Granderson_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:\nRufus Granderson, an American football defensive tackle who played one season with the Dallas Texans of the American Football League, died in 2015 at the age of 79.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Rufus Granderson_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7151,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.962",
"value": "?search="
},
{
"operation": "search",
"time": "16.329",
"value": "Rufus Granderson"
},
{
"operation": "Now on",
"time": "17.528",
"value": "Rufus Granderson"
},
{
"operation": "Highlighting",
"time": "22.245",
"value": "Rufus Granderson_sentence_0"
},
{
"operation": "Highlighting",
"time": "38.226",
"value": "Rufus Granderson_header_cell_0_4_0"
},
{
"operation": "Highlighting",
"time": "39.404",
"value": "Rufus Granderson_cell_0_4_1"
},
{
"operation": "Highlighting",
"time": "39.412",
"value": "Rufus Granderson_header_cell_0_4_0"
},
{
"operation": "Highlighting",
"time": "39.412",
"value": "Rufus Granderson_header_cell_0_2_0"
},
{
"operation": "Highlighting",
"time": "40.557",
"value": "Rufus Granderson_table_caption_0"
},
{
"operation": "search",
"time": "121.414",
"value": "dallas texans Granderson"
},
{
"operation": "Now on",
"time": "122.906",
"value": "undefined"
},
{
"operation": "search",
"time": "132.37",
"value": "Rufus Granderson Dallas"
},
{
"operation": "Now on",
"time": "133.098",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "137.969",
"value": "Rufus Granderson"
},
{
"operation": "Now on",
"time": "138.325",
"value": "Rufus Granderson"
},
{
"operation": "Highlighting deleted",
"time": "208.917",
"value": "Rufus Granderson_table_caption_0"
},
{
"operation": "Highlighting deleted",
"time": "210.696",
"value": "Rufus Granderson_cell_0_4_1"
},
{
"operation": "Highlighting deleted",
"time": "210.702",
"value": "Rufus Granderson_header_cell_0_4_0"
},
{
"operation": "Highlighting deleted",
"time": "210.703",
"value": "Rufus Granderson_header_cell_0_2_0"
},
{
"operation": "finish",
"time": "250.939",
"value": "finish"
}
]
}
|
exfever_8022
|
G.I. Jane is a 1997 American war drama film directed by an English film director and producer.
|
{
"evidence": [
"G.I. Jane",
"Ridley Scott"
],
"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:\nG.I. Jane is a 1997 American war drama film directed by an English film director and producer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"G.I. Jane",
"Ridley Scott"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7152,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
92049
|
The 2004 Kentucky Amendment 1 made it unconstitutional for Kentucky to recognize same-sex marriages, and the referendum was rejected by 75% of voters.
|
{
"evidence": [
"2004 Kentucky Amendment 1_sentence_0",
"2004 Kentucky Amendment 1_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 2004 Kentucky Amendment 1 made it unconstitutional for Kentucky to recognize same-sex marriages, and the referendum was rejected by 75% of voters.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"2004 Kentucky Amendment 1_sentence_0",
"2004 Kentucky Amendment 1_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7153,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.742",
"value": "?search="
},
{
"operation": "search",
"time": "14.233",
"value": "2004 Kentucky Amendment 1"
},
{
"operation": "Now on",
"time": "15.199",
"value": "2004 Kentucky Amendment 1"
},
{
"operation": "Highlighting",
"time": "23.238",
"value": "2004 Kentucky Amendment 1_sentence_0"
},
{
"operation": "Highlighting",
"time": "31.233",
"value": "2004 Kentucky Amendment 1_sentence_1"
},
{
"operation": "finish",
"time": "41.301",
"value": "finish"
}
]
}
|
exfever_431
|
Constantine is a 2005 Canadian superhero horror film starring Keanu Reeves as an antihero who appears in an American publication that consists of comic art in the form of sequentially juxtaposed panels that represent individual scenes published by DC Comics.
|
{
"evidence": [
"Constantine (film)",
"John Constantine",
"Comic book"
],
"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:\nConstantine is a 2005 Canadian superhero horror film starring Keanu Reeves as an antihero who appears in an American publication that consists of comic art in the form of sequentially juxtaposed panels that represent individual scenes published by DC Comics.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Constantine (film)",
"John Constantine",
"Comic book"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7154,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3422
|
Thomas William Hiddleston is an English actor who appeared in a 2011 American superhero film produced by an American film and television production company that is a subsidiary of Walt Disney Studios, a division of the Walt Disney Company.
|
{
"evidence": [
"Tom Hiddleston",
"Thor (film)",
"Marvel Studios"
],
"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 William Hiddleston is an English actor who appeared in a 2011 American superhero film produced by an American film and television production company that is a subsidiary of Walt Disney Studios, a division of the Walt Disney Company.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tom Hiddleston",
"Thor (film)",
"Marvel Studios"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7155,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5394
|
Estonia is a country in Northern Europe with a population of just over 1.3 billion, making it one of the least-populous members of a political and economic union of 27 member states that are located primarily in Europe.
|
{
"evidence": [
"Estonia",
"European Union"
],
"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:\nEstonia is a country in Northern Europe with a population of just over 1.3 billion, making it one of the least-populous members of a political and economic union of 27 member states that are located primarily in Europe.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Estonia",
"European Union"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7156,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3617
|
Ray Milland's last film was The Lost Weekend.
|
{
"evidence": [
"Ray Milland",
"Billy Wilder"
],
"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:\nRay Milland's last film was The Lost Weekend.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ray Milland",
"Billy Wilder"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7157,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5966
|
Cinderella is Chris Weitz highest-grossing film.
|
{
"evidence": [
"Chris Weitz",
"The Golden Compass (film)",
"Dakota Blue Richards"
],
"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:\nCinderella is Chris Weitz highest-grossing film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Chris Weitz",
"The Golden Compass (film)",
"Dakota Blue Richards"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7158,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1494
|
Pablo Picasso was a Spanish theatre designer and painter who was the co-inventor of a technique of art creation, primarily used in art forms such as video, design, and dancing.
|
{
"evidence": [
"Pablo Picasso",
"Collage",
"Visual arts"
],
"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 Picasso was a Spanish theatre designer and painter who was the co-inventor of a technique of art creation, primarily used in art forms such as video, design, and dancing.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Pablo Picasso",
"Collage",
"Visual arts"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7159,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5111
|
Gemini is a Tamil-language crime action film written and directed by Saran that features 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 a Tamil-language crime action film written and directed by Saran that features 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": 7160,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5410
|
Mrs Henderson Presents is a 2001 biographical musical film written by American dramatist and screenwriter best known for his 20 stage plays, which have been produced in over 60 countries.
|
{
"evidence": [
"Mrs Henderson Presents",
"Martin Sherman"
],
"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:\nMrs Henderson Presents is a 2001 biographical musical film written by American dramatist and screenwriter best known for his 20 stage plays, which have been produced in over 60 countries.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mrs Henderson Presents",
"Martin Sherman"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7161,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5108
|
Taylor Swift was the youngest songwriter to sign with Sony/ATV Music Publishing.
|
{
"evidence": [
"Taylor Swift",
"Nashville, Tennessee"
],
"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:\nTaylor Swift was the youngest songwriter to sign with Sony/ATV Music Publishing.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Taylor Swift",
"Nashville, Tennessee"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7162,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
77380
|
Mike Iggulden, who enjoyed a successful debut season in the 2009-10 season, signed a lucrative one-year contract with the SCL Tigers, a professional ice hockey team from Langnau im Emmental, Switzerland, of the Swiss National League A (NLA) on July 16, 2010.
|
{
"evidence": [
"Mike Iggulden_sentence_11",
"Mike Iggulden_sentence_12",
"SCL Tigers_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:\nMike Iggulden, who enjoyed a successful debut season in the 2009-10 season, signed a lucrative one-year contract with the SCL Tigers, a professional ice hockey team from Langnau im Emmental, Switzerland, of the Swiss National League A (NLA) on July 16, 2010.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Mike Iggulden_sentence_11",
"Mike Iggulden_sentence_12",
"SCL Tigers_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7163,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.845",
"value": "?search="
},
{
"operation": "search",
"time": "9.579",
"value": "Mike Iggulden"
},
{
"operation": "Now on",
"time": "10.39",
"value": "Mike Iggulden"
},
{
"operation": "Highlighting",
"time": "158.495",
"value": "Mike Iggulden_sentence_11"
},
{
"operation": "Highlighting",
"time": "164.382",
"value": "Mike Iggulden_sentence_12"
},
{
"operation": "hyperlink",
"time": "169.607",
"value": "SCL Tigers"
},
{
"operation": "Now on",
"time": "170.415",
"value": "SCL Tigers"
},
{
"operation": "Highlighting",
"time": "173.136",
"value": "SCL Tigers_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "176.052",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "176.142",
"value": "Mike Iggulden"
},
{
"operation": "finish",
"time": "187.618",
"value": "finish"
}
]
}
|
90037
|
Ditrigona chama has a wingspan that is the same between males and females and its fore and hind wings are lustrous white.
|
{
"evidence": [
"Ditrigona chama_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:\nDitrigona chama has a wingspan that is the same between males and females and its fore and hind wings are lustrous white.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ditrigona chama_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7164,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.372",
"value": "?search="
},
{
"operation": "search",
"time": "10.03",
"value": "Ditrigona chama"
},
{
"operation": "Now on",
"time": "11.134",
"value": "Ditrigona chama"
},
{
"operation": "Highlighting",
"time": "19.439",
"value": "Ditrigona chama_sentence_3"
},
{
"operation": "finish",
"time": "34.18",
"value": "finish"
}
]
}
|
30734
|
Christian Democratic Party politician Odd Holten was elected to the Norwegian Parliament from Østfold in 1989.
|
{
"evidence": [
"Odd Holten_sentence_0",
"Odd Holten_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:\nChristian Democratic Party politician Odd Holten was elected to the Norwegian Parliament from Østfold in 1989.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Odd Holten_sentence_0",
"Odd Holten_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7165,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.226",
"value": "The Perils of Penelope Pitstop"
},
{
"operation": "Now on",
"time": "1.342",
"value": "?search="
},
{
"operation": "search",
"time": "8.056",
"value": "Odd Holten"
},
{
"operation": "Now on",
"time": "9.008",
"value": "Odd Holten"
},
{
"operation": "Highlighting",
"time": "10.972",
"value": "Odd Holten_sentence_0"
},
{
"operation": "Highlighting",
"time": "19.066",
"value": "Odd Holten_sentence_1"
},
{
"operation": "finish",
"time": "36.164",
"value": "finish"
}
]
}
|
exfever_8321
|
Perry Saturn is an American retired professional wrestler and a former champion in a professional wrestling world tag team championship in Extreme Championship Wrestling established under ECW in 1994.
|
{
"evidence": [
"Perry Saturn",
"ECW World Tag Team Championship"
],
"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:\nPerry Saturn is an American retired professional wrestler and a former champion in a professional wrestling world tag team championship in Extreme Championship Wrestling established under ECW in 1994.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Perry Saturn",
"ECW World Tag Team Championship"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7166,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3533
|
Gabrielle Monique Union's breakthrough role was in a 2000 American teen cheerleading comedy film.
|
{
"evidence": [
"Gabrielle Union",
"Bring It On (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:\nGabrielle Monique Union's breakthrough role was in a 2000 American teen cheerleading comedy film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Gabrielle Union",
"Bring It On (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7167,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4326
|
X-Men is a 2000 American superhero film starring an Australian actor, who has headlined films in various genres, including a 2006 mystery thriller film directed by Christopher Nolan.
|
{
"evidence": [
"X-Men (film)",
"Hugh Jackman",
"The Prestige (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:\nX-Men is a 2000 American superhero film starring an Australian actor, who has headlined films in various genres, including a 2006 mystery thriller film directed by Christopher Nolan.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"X-Men (film)",
"Hugh Jackman",
"The Prestige (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7168,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_328
|
Cary Fukunaga is an Academy Award-winning director.
|
{
"evidence": [
"Jane Eyre (2011 film)",
"Michael Fassbender"
],
"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:\nCary Fukunaga is an Academy Award-winning director.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jane Eyre (2011 film)",
"Michael Fassbender"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7169,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
18044
|
Special Scout Connie Ryan (born on February 27, 1920) was named to finish out the string with only a handful of games left in the season; then for 1976, GM Eddie Robinson promised to hire a "firebrand" to replace the scholarly King.
|
{
"evidence": [
"Connie Ryan_sentence_0",
"Connie Ryan_sentence_6",
"Clyde King_sentence_25"
],
"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:\nSpecial Scout Connie Ryan (born on February 27, 1920) was named to finish out the string with only a handful of games left in the season; then for 1976, GM Eddie Robinson promised to hire a \"firebrand\" to replace the scholarly King.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Connie Ryan_sentence_0",
"Connie Ryan_sentence_6",
"Clyde King_sentence_25"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7170,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.668",
"value": "?search="
},
{
"operation": "search",
"time": "14.719",
"value": "Connie Ryan"
},
{
"operation": "Now on",
"time": "15.66",
"value": "Connie Ryan"
},
{
"operation": "Highlighting",
"time": "44.272",
"value": "Connie Ryan_sentence_0"
},
{
"operation": "Page search",
"time": "84.279",
"value": "string"
},
{
"operation": "Highlighting",
"time": "121.08",
"value": "Connie Ryan_sentence_6"
},
{
"operation": "Page search",
"time": "188.231",
"value": "eddie robinson"
},
{
"operation": "search",
"time": "206.161",
"value": "Eddie Robinson (baseball)"
},
{
"operation": "Now on",
"time": "207.366",
"value": "Eddie Robinson (baseball)"
},
{
"operation": "back-button-clicked",
"time": "265.673",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "265.78",
"value": "Connie Ryan"
},
{
"operation": "hyperlink",
"time": "270.871",
"value": "Clyde King"
},
{
"operation": "Now on",
"time": "271.669",
"value": "Clyde King"
},
{
"operation": "Highlighting",
"time": "297.454",
"value": "Clyde King_sentence_25"
},
{
"operation": "finish",
"time": "322.726",
"value": "finish"
}
]
}
|
exfever_8075
|
Yung Rich Nation is the debut album by Migos, distributed by an American record label founded by Lyor Cohen.
|
{
"evidence": [
"Yung Rich Nation",
"300 Entertainment"
],
"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:\nYung Rich Nation is the debut album by Migos, distributed by an American record label founded by Lyor Cohen.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Yung Rich Nation",
"300 Entertainment"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7171,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_501
|
Tom Hanks is an American filmmaker who played the title character in a 1994 American comedy-drama series.
|
{
"evidence": [
"Tom Hanks",
"Forrest Gump"
],
"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:\nTom Hanks is an American filmmaker who played the title character in a 1994 American comedy-drama series.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tom Hanks",
"Forrest Gump"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7172,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
51634
|
Since the ions are charged, they cannot pass through the membrane via simple diffusion; but two different mechanisms can transport the ions across the membrane, namely, active or passive transport.
|
{
"evidence": [
"Electrochemical gradient_sentence_43",
"Electrochemical gradient_sentence_44"
],
"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:\nSince the ions are charged, they cannot pass through the membrane via simple diffusion; but two different mechanisms can transport the ions across the membrane, namely, active or passive transport.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Electrochemical gradient_sentence_43",
"Electrochemical gradient_sentence_44"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7173,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.489",
"value": "?search="
},
{
"operation": "search",
"time": "23.432",
"value": "charged ions"
},
{
"operation": "Now on",
"time": "24.465",
"value": "undefined"
},
{
"operation": "search",
"time": "40.913",
"value": "Active transport"
},
{
"operation": "Now on",
"time": "41.926",
"value": "Active transport"
},
{
"operation": "search",
"time": "108.55",
"value": "active and passive transport"
},
{
"operation": "Now on",
"time": "111.258",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "118.415",
"value": "Electrochemical gradient"
},
{
"operation": "Now on",
"time": "119.419",
"value": "Electrochemical gradient"
},
{
"operation": "Highlighting",
"time": "188.18",
"value": "Electrochemical gradient_sentence_43"
},
{
"operation": "Highlighting",
"time": "191.552",
"value": "Electrochemical gradient_sentence_44"
},
{
"operation": "finish",
"time": "239.208",
"value": "finish"
}
]
}
|
exfever_2287
|
Snowden is a thriller film directed by Oliver Stone based on the book The Snowden Files, which starred Joseph Gordon-Levitt as a former computer intelligence consultant who leaked highly classified information from the National Security Agency in 2013.
|
{
"evidence": [
"Snowden (film)",
"Edward Snowden"
],
"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:\nSnowden is a thriller film directed by Oliver Stone based on the book The Snowden Files, which starred Joseph Gordon-Levitt as a former computer intelligence consultant who leaked highly classified information from the National Security Agency in 2013.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Snowden (film)",
"Edward Snowden"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7174,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
36684
|
The 1.4 engine is 4cyl ( a four-cylinder piston engine where cylinders are arranged in a line along a common crankshaft) and runs on diesel.
|
{
"evidence": [
"Straight-four engine_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:\nThe 1.4 engine is 4cyl ( a four-cylinder piston engine where cylinders are arranged in a line along a common crankshaft) and runs on diesel.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Straight-four engine_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7175,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.345",
"value": "?search="
},
{
"operation": "search",
"time": "24.042",
"value": "1.4 engine"
},
{
"operation": "Now on",
"time": "28.956",
"value": "undefined"
},
{
"operation": "search",
"time": "98.446",
"value": "1.4 4cyl"
},
{
"operation": "Now on",
"time": "99.983",
"value": "undefined"
},
{
"operation": "search",
"time": "113.025",
"value": "Diesel engine"
},
{
"operation": "Now on",
"time": "115.812",
"value": "Diesel engine"
},
{
"operation": "search",
"time": "186.522",
"value": "Straight-four engine"
},
{
"operation": "Now on",
"time": "187.973",
"value": "Straight-four engine"
},
{
"operation": "Highlighting",
"time": "203.874",
"value": "Straight-four engine_sentence_0"
},
{
"operation": "finish",
"time": "306.739",
"value": "finish"
}
]
}
|
exfever_3576
|
A rubber mounted on a structure and intended to protect the structure from a lightning strike is called a lightning rod or hip-knob, which is an element marking the top or end of some object, often formed to be a decorative feature.
|
{
"evidence": [
"Lightning rod",
"Finial"
],
"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:\nA rubber mounted on a structure and intended to protect the structure from a lightning strike is called a lightning rod or hip-knob, which is an element marking the top or end of some object, often formed to be a decorative feature.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lightning rod",
"Finial"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7176,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4507
|
Nick Kroll has also acted in films, including a 2016 American biographical romantic drama film.
|
{
"evidence": [
"Nick Kroll",
"Loving (2016 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:\nNick Kroll has also acted in films, including a 2016 American biographical romantic drama film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Nick Kroll",
"Loving (2016 film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7177,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
48048
|
Abaco Air, a company in the consumer services industry and airlines sector, is founded in 1975 with headquarters in Marsh Harbour, Bahamas (a state of the Lucayan Archipelago).
|
{
"evidence": [
"Abaco Air_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:\nAbaco Air, a company in the consumer services industry and airlines sector, is founded in 1975 with headquarters in Marsh Harbour, Bahamas (a state of the Lucayan Archipelago).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Abaco Air_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7178,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.994",
"value": "?search="
},
{
"operation": "search",
"time": "5.577",
"value": "Abaco Air"
},
{
"operation": "Now on",
"time": "6.688",
"value": "Abaco Air"
},
{
"operation": "search",
"time": "28.404",
"value": "Abaco Air"
},
{
"operation": "Now on",
"time": "29.098",
"value": "Abaco Air"
},
{
"operation": "Highlighting",
"time": "48.182",
"value": "Abaco Air_sentence_0"
},
{
"operation": "hyperlink",
"time": "61.097",
"value": "Marsh Harbour"
},
{
"operation": "Now on",
"time": "61.653",
"value": "Marsh Harbour"
},
{
"operation": "back-button-clicked",
"time": "82.395",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "82.51",
"value": "Abaco Air"
},
{
"operation": "search",
"time": "109.291",
"value": "Abaco Airlines"
},
{
"operation": "Now on",
"time": "110.941",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "117.538",
"value": "List of defunct airlines of the Bahamas"
},
{
"operation": "Now on",
"time": "118.072",
"value": "List of defunct airlines of the Bahamas"
},
{
"operation": "finish",
"time": "138.51",
"value": "finish"
}
]
}
|
exfever_3649
|
Naruto is a comics originating from Korea written and illustrated by Masashi Kishimoto that tells the story of Naruto Uzumaki.
|
{
"evidence": [
"Naruto",
"Manga"
],
"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:\nNaruto is a comics originating from Korea written and illustrated by Masashi Kishimoto that tells the story of Naruto Uzumaki.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Naruto",
"Manga"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7179,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
27778
|
GM2A is a gene which encodes a small glycolipid transport protein and also stimulates the enzymatic processing of gangliosides, which are molecules devoid of any acids or chemicals.
|
{
"evidence": [
"Ganglioside_sentence_0",
"GM2A_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:\nGM2A is a gene which encodes a small glycolipid transport protein and also stimulates the enzymatic processing of gangliosides, which are molecules devoid of any acids or chemicals.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Ganglioside_sentence_0",
"GM2A_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7180,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.022",
"value": "?search="
},
{
"operation": "search",
"time": "36.417",
"value": "GM2A"
},
{
"operation": "Now on",
"time": "37.137",
"value": "GM2A"
},
{
"operation": "Highlighting",
"time": "39.953",
"value": "GM2A_sentence_0"
},
{
"operation": "Highlighting",
"time": "51.462",
"value": "GM2A_sentence_1"
},
{
"operation": "Highlighting",
"time": "61.872",
"value": "GM2A_sentence_3"
},
{
"operation": "hyperlink",
"time": "64.81",
"value": "gangliosides"
},
{
"operation": "Now on",
"time": "65.832",
"value": "Gangliosides"
},
{
"operation": "Page search",
"time": "78.197",
"value": "devoid"
},
{
"operation": "Highlighting",
"time": "190.445",
"value": "Ganglioside_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "191.281",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "191.473",
"value": "GM2A"
},
{
"operation": "Highlighting deleted",
"time": "192.65",
"value": "GM2A_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "193.129",
"value": "GM2A_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "194.185",
"value": "GM2A_sentence_0"
},
{
"operation": "Highlighting",
"time": "207.343",
"value": "GM2A_sentence_2"
},
{
"operation": "Highlighting",
"time": "209.529",
"value": "GM2A_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "209.929",
"value": "GM2A_sentence_2"
},
{
"operation": "finish",
"time": "212.239",
"value": "finish"
}
]
}
|
exfever_6441
|
Sid and Nancy is the first film where Courtney Michelle Love first appeared.
|
{
"evidence": [
"Courtney Love",
"Hole (band)"
],
"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:\nSid and Nancy is the first film where Courtney Michelle Love first appeared.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Courtney Love",
"Hole (band)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7181,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1192
|
Scandal is an American rock band that scored a hit in the United States with a song featuring Patty Smyth, from the album Warrior, written by Holly Knight and Nick Gilder.
|
{
"evidence": [
"Scandal (American band)",
"The Warrior (song)"
],
"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:\nScandal is an American rock band that scored a hit in the United States with a song featuring Patty Smyth, from the album Warrior, written by Holly Knight and Nick Gilder.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Scandal (American band)",
"The Warrior (song)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7182,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6176
|
Friends with Benefits is a 2011 film that starred a Canadian actor known for his role as Ben Epstein in How to Make It in America.
|
{
"evidence": [
"Friends with Benefits (film)",
"Bryan Greenberg"
],
"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:\nFriends with Benefits is a 2011 film that starred a Canadian actor known for his role as Ben Epstein in How to Make It in America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Friends with Benefits (film)",
"Bryan Greenberg"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7183,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.