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 |
|---|---|---|---|---|---|---|---|---|
17971
|
Kesen is spoken in the Kesen district, Harutsugu Yamaura developed a writing system for Kesen has been described as a variety of the Tōhoku dialect which is a group of the Japanese dialects.
|
{
"evidence": [
"Kesen dialect_sentence_3",
"Kesen dialect_sentence_6",
"Kesen dialect_sentence_4",
"Tōhoku dialect_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:\nKesen is spoken in the Kesen district, Harutsugu Yamaura developed a writing system for Kesen has been described as a variety of the Tōhoku dialect which is a group of the Japanese dialects.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Kesen dialect_sentence_3",
"Kesen dialect_sentence_6",
"Kesen dialect_sentence_4",
"Tōhoku dialect_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1484,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.806",
"value": "?search="
},
{
"operation": "search",
"time": "17.194",
"value": "kesen"
},
{
"operation": "Now on",
"time": "17.663",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "32.649",
"value": "Rikuzen Province"
},
{
"operation": "Now on",
"time": "33.232",
"value": "Rikuzen Province"
},
{
"operation": "hyperlink",
"time": "36.729",
"value": "Iwate Prefecture"
},
{
"operation": "Now on",
"time": "37.28",
"value": "Iwate Prefecture"
},
{
"operation": "hyperlink",
"time": "49.321",
"value": "district"
},
{
"operation": "Now on",
"time": "49.771",
"value": "Districts of Japan"
},
{
"operation": "search",
"time": "109.33",
"value": "Harutsugu Yamaura"
},
{
"operation": "Now on",
"time": "109.929",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "114.608",
"value": "Harutsugu Yamaura Kesen dialect_section_1"
},
{
"operation": "Now on",
"time": "115.041",
"value": "Kesen dialect#Harutsugu Yamaura Kesen dialect section 1"
},
{
"operation": "Highlighting",
"time": "173.489",
"value": "Kesen dialect_sentence_6"
},
{
"operation": "Highlighting",
"time": "195.192",
"value": "Kesen dialect_sentence_9"
},
{
"operation": "Highlighting",
"time": "195.192",
"value": "Kesen dialect_section_1"
},
{
"operation": "hyperlink",
"time": "196.624",
"value": "Japanese dialect"
},
{
"operation": "Now on",
"time": "197.082",
"value": "Japanese dialects"
},
{
"operation": "back-button-clicked",
"time": "209.976",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "210.025",
"value": "Kesen dialect#Harutsugu Yamaura Kesen dialect section 1"
},
{
"operation": "Highlighting",
"time": "245.463",
"value": "Kesen dialect_sentence_3"
},
{
"operation": "Highlighting",
"time": "256.103",
"value": "Kesen dialect_sentence_6"
},
{
"operation": "Highlighting",
"time": "264.647",
"value": "Kesen dialect_sentence_4"
},
{
"operation": "hyperlink",
"time": "265.559",
"value": "Tōhoku dialect"
},
{
"operation": "Now on",
"time": "266.007",
"value": "Tōhoku dialect"
},
{
"operation": "Highlighting",
"time": "270.527",
"value": "Tōhoku dialect_sentence_0"
},
{
"operation": "finish",
"time": "276.999",
"value": "finish"
}
]
}
|
exfever_4769
|
The Czech Republic also called Czechia, is a landlocked country in the Central part of a continent that shares the continental landmass of Afro-Eurasia with both Africa and Earth's largest and most populous continent, located primarily in the Eastern and Northern Hemispheres.
|
{
"evidence": [
"Czech Republic",
"Europe",
"Asia"
],
"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 Czech Republic also called Czechia, is a landlocked country in the Central part of a continent that shares the continental landmass of Afro-Eurasia with both Africa and Earth's largest and most populous continent, located primarily in the Eastern and Northern Hemispheres.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Czech Republic",
"Europe",
"Asia"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1485,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5806
|
Turkey is a transcontinental country located primarily on a large peninsula in Western Europe and the westernmost protrusion of the Asian continent that constitutes the major part of modern-day Turkey.
|
{
"evidence": [
"Turkey",
"Anatolia"
],
"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:\nTurkey is a transcontinental country located primarily on a large peninsula in Western Europe and the westernmost protrusion of the Asian continent that constitutes the major part of modern-day Turkey.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Turkey",
"Anatolia"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1486,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2019
|
The Fifth Element was written by two screenwriters, including the creator of The Karate Kid franchise, who collaborated with a Belgian filmmaker associated with the Cinéma du Look film movement.
|
{
"evidence": [
"The Fifth Element",
"Robert Mark Kamen",
"Luc Besson"
],
"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 Fifth Element was written by two screenwriters, including the creator of The Karate Kid franchise, who collaborated with a Belgian filmmaker associated with the Cinéma du Look film movement.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Fifth Element",
"Robert Mark Kamen",
"Luc Besson"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1487,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
35556
|
With the United States' entry into World War II, it brought a relocation of the game to Milan Puskar Stadium in Morgantown, West Virginia in spite of the fact that Mountaineer Field was favored by two touchdowns, the Duke pulled off a major upset, winning 20–10.
|
{
"evidence": [
"Milan Puskar Stadium_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWith the United States' entry into World War II, it brought a relocation of the game to Milan Puskar Stadium in Morgantown, West Virginia in spite of the fact that Mountaineer Field was favored by two touchdowns, the Duke pulled off a major upset, winning 20–10.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Milan Puskar Stadium_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1488,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.644",
"value": "undefined"
},
{
"operation": "search",
"time": "15.53",
"value": "Milan Puskar Stadium"
},
{
"operation": "Now on",
"time": "16.247",
"value": "Milan Puskar Stadium"
},
{
"operation": "Highlighting",
"time": "27.741",
"value": "Milan Puskar Stadium_sentence_1"
},
{
"operation": "hyperlink",
"time": "54.514",
"value": "West Virginia Mountaineers football"
},
{
"operation": "Now on",
"time": "55.335",
"value": "West Virginia Mountaineers football"
},
{
"operation": "hyperlink",
"time": "62.993",
"value": "West Virginia University"
},
{
"operation": "Now on",
"time": "64.472",
"value": "West Virginia University"
},
{
"operation": "hyperlink",
"time": "157.155",
"value": "West Virginia Mountaineers football"
},
{
"operation": "Now on",
"time": "157.964",
"value": "West Virginia Mountaineers football"
},
{
"operation": "Page search",
"time": "188.047",
"value": "duke"
},
{
"operation": "finish",
"time": "249.861",
"value": "finish"
}
]
}
|
exfever_3769
|
Dawn of the Planet of the Apes is a 2014 American science fiction action film directed by an American film director who first gained recognition for the WB drama series Felicity which he co-created with J. J. Abrams.
|
{
"evidence": [
"Dawn of the Planet of the Apes",
"Matt Reeves"
],
"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:\nDawn of the Planet of the Apes is a 2014 American science fiction action film directed by an American film director who first gained recognition for the WB drama series Felicity which he co-created with J. J. Abrams.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dawn of the Planet of the Apes",
"Matt Reeves"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1489,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
35180
|
The United Nations Security Council Resolution 431 on Nambia's situation was adopted by 13 votes to none on 27 July 1978.; Czechoslovakia and the Soviet Union abstained from voting.
|
{
"evidence": [
"United Nations Security Council Resolution 431_sentence_0",
"United Nations Security Council Resolution 431_sentence_2"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe United Nations Security Council Resolution 431 on Nambia's situation was adopted by 13 votes to none on 27 July 1978.; Czechoslovakia and the Soviet Union abstained from voting.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"United Nations Security Council Resolution 431_sentence_0",
"United Nations Security Council Resolution 431_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1490,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.205",
"value": "?search="
},
{
"operation": "search",
"time": "31.213",
"value": "United Nations Security Council Resolution 431"
},
{
"operation": "Now on",
"time": "32.007",
"value": "United Nations Security Council Resolution 431"
},
{
"operation": "Highlighting",
"time": "46.651",
"value": "United Nations Security Council Resolution 431_sentence_0"
},
{
"operation": "Highlighting",
"time": "52.959",
"value": "United Nations Security Council Resolution 431_sentence_2"
},
{
"operation": "finish",
"time": "70.116",
"value": "finish"
}
]
}
|
18489
|
Levitation is a high-wing monoplane, powered by a 180 hp (134 kW) Lycoming O-360 and has a fixed conventional landing gear.
|
{
"evidence": [
"Tapanee Levitation 4_sentence_2",
"Tapanee Levitation 4_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:\nLevitation is a high-wing monoplane, powered by a 180 hp (134 kW) Lycoming O-360 and has a fixed conventional landing gear.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Tapanee Levitation 4_sentence_2",
"Tapanee Levitation 4_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1491,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.483",
"value": "?search="
},
{
"operation": "search",
"time": "20.51",
"value": "Levitation (disambiguation)"
},
{
"operation": "Now on",
"time": "22.255",
"value": "Levitation (disambiguation)"
},
{
"operation": "hyperlink",
"time": "72.948",
"value": "4 See also Levitation (disambiguation)_section_5"
},
{
"operation": "Highlighting",
"time": "75.016",
"value": "Levitation (disambiguation)_item_5_20"
},
{
"operation": "Highlighting",
"time": "75.018",
"value": "Levitation (disambiguation)_title"
},
{
"operation": "Highlighting",
"time": "75.019",
"value": "Levitation (disambiguation)_section_5"
},
{
"operation": "Highlighting deleted",
"time": "75.875",
"value": "Levitation (disambiguation)_item_5_20"
},
{
"operation": "Highlighting deleted",
"time": "75.877",
"value": "Levitation (disambiguation)_title"
},
{
"operation": "Highlighting deleted",
"time": "75.879",
"value": "Levitation (disambiguation)_section_5"
},
{
"operation": "hyperlink",
"time": "79.974",
"value": "Levitation"
},
{
"operation": "Now on",
"time": "81.281",
"value": "Levitation"
},
{
"operation": "search",
"time": "103.721",
"value": "Levitation monoplane"
},
{
"operation": "Now on",
"time": "104.618",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "129.518",
"value": "Tapanee Pegazair-100"
},
{
"operation": "Now on",
"time": "130.541",
"value": "Tapanee Pegazair-100"
},
{
"operation": "hyperlink",
"time": "153.011",
"value": "Tapanee Levitation 4"
},
{
"operation": "Now on",
"time": "154.04",
"value": "Tapanee Levitation 4"
},
{
"operation": "Highlighting",
"time": "170.704",
"value": "Tapanee Levitation 4_sentence_0"
},
{
"operation": "Highlighting",
"time": "174.591",
"value": "Tapanee Levitation 4_sentence_2"
},
{
"operation": "Highlighting",
"time": "188.834",
"value": "Tapanee Levitation 4_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "222.403",
"value": "Tapanee Levitation 4_sentence_0"
},
{
"operation": "finish",
"time": "247.05",
"value": "finish"
}
]
}
|
exfever_2186
|
Penny Dreadful is a horror drama television series co-produced by a British pay television channel owned by Sky Group Limited broadcast.
|
{
"evidence": [
"Penny Dreadful (TV series)",
"Sky Atlantic"
],
"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:\nPenny Dreadful is a horror drama television series co-produced by a British pay television channel owned by Sky Group Limited broadcast.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Penny Dreadful (TV series)",
"Sky Atlantic"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1492,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
78795
|
Wendy Monk was married to J.C. Trewin, and were described as "an inseparable couple, whose shared interests also bore fruit in literary collaboration, they had two sons": Ion and Mark Antony, and since 1990, an award has been given by the Critics' Circle for the best Shakespearian performance of the Year: "The John And Wendy Trewin Award For Best Shakespearian Performance".
|
{
"evidence": [
"Wendy Monk_sentence_10"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWendy Monk was married to J.C. Trewin, and were described as \"an inseparable couple, whose shared interests also bore fruit in literary collaboration, they had two sons\": Ion and Mark Antony, and since 1990, an award has been given by the Critics' Circle for the best Shakespearian performance of the Year: \"The John And Wendy Trewin Award For Best Shakespearian Performance\".\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Wendy Monk_sentence_10"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1493,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "3.965",
"value": "?search="
},
{
"operation": "search",
"time": "44.174",
"value": "Wendy Monk"
},
{
"operation": "Now on",
"time": "46.242",
"value": "Wendy Monk"
},
{
"operation": "Highlighting",
"time": "58.886",
"value": "Wendy Monk_sentence_7Wendy Monk_sentence_8"
},
{
"operation": "Highlighting",
"time": "63.839",
"value": "Wendy Monk_sentence_9"
},
{
"operation": "Highlighting",
"time": "75.118",
"value": "Wendy Monk_sentence_10"
},
{
"operation": "Highlighting deleted",
"time": "75.866",
"value": "Wendy Monk_sentence_9"
},
{
"operation": "Highlighting deleted",
"time": "76.276",
"value": "Wendy Monk_sentence_7Wendy Monk_sentence_8"
},
{
"operation": "finish",
"time": "89.212",
"value": "finish"
}
]
}
|
35203
|
Pavel Belyayev was awarded the Distinguished Combat service medal (made of gold plated brass and lacquer) in 1953; then in 1954, he was promoted to captain.
|
{
"evidence": [
"Pavel Belyayev_sentence_26",
"Pavel Belyayev_sentence_27"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPavel Belyayev was awarded the Distinguished Combat service medal (made of gold plated brass and lacquer) in 1953; then in 1954, he was promoted to captain.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Pavel Belyayev_sentence_26",
"Pavel Belyayev_sentence_27"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1494,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.118",
"value": "?search="
},
{
"operation": "search",
"time": "9.77",
"value": "Pavel Belyayev"
},
{
"operation": "Now on",
"time": "10.889",
"value": "Pavel Belyayev"
},
{
"operation": "hyperlink",
"time": "26.713",
"value": "6 Honors and awards Pavel Belyayev_section_6"
},
{
"operation": "Page search",
"time": "46.137",
"value": "Distinguished Combat"
},
{
"operation": "Highlighting",
"time": "50.944",
"value": "Pavel Belyayev_sentence_26"
},
{
"operation": "Highlighting",
"time": "59.437",
"value": "Pavel Belyayev_sentence_27"
},
{
"operation": "search",
"time": "73.045",
"value": "Distinguished Combat service medal"
},
{
"operation": "Now on",
"time": "75.831",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "86.159",
"value": "Air Force Distinguished Service Medal"
},
{
"operation": "Now on",
"time": "87.582",
"value": "Air Force Distinguished Service Medal"
},
{
"operation": "back-button-clicked",
"time": "113.317",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "114.838",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "120.752",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "120.757",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "120.889",
"value": "Pavel Belyayev#Honors and awards Pavel Belyayev section 6"
},
{
"operation": "Page search",
"time": "155.086",
"value": "made of"
},
{
"operation": "search",
"time": "170.733",
"value": "Soviet Air Forces"
},
{
"operation": "Now on",
"time": "172.956",
"value": "Soviet Air Forces"
},
{
"operation": "Page search",
"time": "180.074",
"value": "service medal"
},
{
"operation": "back-button-clicked",
"time": "191.035",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "191.184",
"value": "Pavel Belyayev#Honors and awards Pavel Belyayev section 6"
},
{
"operation": "finish",
"time": "199.988",
"value": "finish"
}
]
}
|
exfever_8391
|
South America is a continent in the half of the third planet from the Sun that is south of the Equator.
|
{
"evidence": [
"South America",
"Southern Hemisphere",
"Earth"
],
"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:\nSouth America is a continent in the half of the third planet from the Sun that is south of the Equator.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"South America",
"Southern Hemisphere",
"Earth"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1495,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5360
|
"What a Girl Wants" is a song-recorded version that was released following the lead single by Christina Aguilera from her self-titled debut studio album.
|
{
"evidence": [
"What a Girl Wants (Christina Aguilera song)",
"Genie in a Bottle"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n\"What a Girl Wants\" is a song-recorded version that was released following the lead single by Christina Aguilera from her self-titled debut studio album.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"What a Girl Wants (Christina Aguilera song)",
"Genie in a Bottle"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1496,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
64074
|
Casino Moose Jaw, which is located in downtown Paradise, Nevada, U.S., was opened on October 6, 2000.
|
{
"evidence": [
"Casino Moose Jaw_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:\nCasino Moose Jaw, which is located in downtown Paradise, Nevada, U.S., was opened on October 6, 2000.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Casino Moose Jaw_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1497,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.288",
"value": "?search="
},
{
"operation": "search",
"time": "114.723",
"value": "Casino Moose Jaw,"
},
{
"operation": "Now on",
"time": "123.099",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "319.95",
"value": "Casino Moose Jaw"
},
{
"operation": "Now on",
"time": "321.919",
"value": "Casino Moose Jaw"
},
{
"operation": "Highlighting",
"time": "331.299",
"value": "Casino Moose Jaw_sentence_0"
},
{
"operation": "finish",
"time": "341.307",
"value": "finish"
}
]
}
|
78887
|
Thiago Rocha da Cunha is falsely believed to have been born November 22, 1984 in Rio de Janeiro, Brazil and played the Defensive midfielder position.
|
{
"evidence": [
"Thiago Rocha da Cunha_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:\nThiago Rocha da Cunha is falsely believed to have been born November 22, 1984 in Rio de Janeiro, Brazil and played the Defensive midfielder position.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Thiago Rocha da Cunha_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1498,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.752",
"value": "?search="
},
{
"operation": "search",
"time": "22.618",
"value": "Thiago Rocha da Cunha"
},
{
"operation": "Now on",
"time": "23.326",
"value": "Thiago Rocha da Cunha"
},
{
"operation": "Highlighting",
"time": "41.851",
"value": "Thiago Rocha da Cunha_sentence_0"
},
{
"operation": "finish",
"time": "48.923",
"value": "finish"
}
]
}
|
43787
|
On 29 November 2011, Syamsudin Noor Airport's radar detected Dassault Falcon 900 EX that carried Papua New Guinean Deputy Prime Minister H.O.N. meanwhile, the airport's ATC and Kohanudnas tries to contact the aircraft as it is an unscheduled flight, but no response.
|
{
"evidence": [
"Indonesian National Air Defense Forces Command_sentence_20",
"Indonesian National Air Defense Forces Command_sentence_21"
],
"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 29 November 2011, Syamsudin Noor Airport's radar detected Dassault Falcon 900 EX that carried Papua New Guinean Deputy Prime Minister H.O.N. meanwhile, the airport's ATC and Kohanudnas tries to contact the aircraft as it is an unscheduled flight, but no response.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Indonesian National Air Defense Forces Command_sentence_20",
"Indonesian National Air Defense Forces Command_sentence_21"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1499,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.115",
"value": "?search="
},
{
"operation": "search",
"time": "36.29",
"value": "Dassault Falcon 900 EX"
},
{
"operation": "Now on",
"time": "37.862",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "44.642",
"value": "Dassault Falcon 900"
},
{
"operation": "Now on",
"time": "45.451",
"value": "Dassault Falcon 900"
},
{
"operation": "Page search",
"time": "48.731",
"value": "noor"
},
{
"operation": "search",
"time": "67.346",
"value": "Syamsudin Noor Airport's radar"
},
{
"operation": "Now on",
"time": "68.555",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "75.19",
"value": "Indonesian National Air Defense Forces Command"
},
{
"operation": "Now on",
"time": "75.798",
"value": "Indonesian National Air Defense Forces Command"
},
{
"operation": "Page search",
"time": "91.712",
"value": "Syamsudin Noor"
},
{
"operation": "Highlighting",
"time": "133.438",
"value": "Indonesian National Air Defense Forces Command_sentence_20"
},
{
"operation": "Highlighting",
"time": "147.223",
"value": "Indonesian National Air Defense Forces Command_sentence_21"
},
{
"operation": "finish",
"time": "255.678",
"value": "finish"
}
]
}
|
24229
|
ThousandEyes received a $500K National Science Foundation grant in 2011 and also received a $5.5M investment from Sequoia Capital in the same year.
|
{
"evidence": [
"ThousandEyes_sentence_4",
"ThousandEyes_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:\nThousandEyes received a $500K National Science Foundation grant in 2011 and also received a $5.5M investment from Sequoia Capital in the same year.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"ThousandEyes_sentence_4",
"ThousandEyes_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1500,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.622",
"value": "?search="
},
{
"operation": "search",
"time": "8.466",
"value": "ThousandEyes"
},
{
"operation": "Now on",
"time": "9.16",
"value": "ThousandEyes"
},
{
"operation": "Highlighting",
"time": "35.971",
"value": "ThousandEyes_sentence_4"
},
{
"operation": "Highlighting",
"time": "44.557",
"value": "ThousandEyes_sentence_5"
},
{
"operation": "finish",
"time": "51.871",
"value": "finish"
}
]
}
|
exfever_7223
|
Doctor Strange is a fictional character appearing in a thin book, averaging 45 pages and containing a publication that consists of comic art in the form of individual scenes.
|
{
"evidence": [
"Doctor Strange",
"American comic book",
"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:\nDoctor Strange is a fictional character appearing in a thin book, averaging 45 pages and containing a publication that consists of comic art in the form of individual scenes.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Doctor Strange",
"American comic book",
"Comic book"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1501,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
49132
|
Nadia Comăneci is a Romanian gymnast who has won a total of nine Olympic medals, the most of any athlete Romania has ever fielded to the Olympics.
|
{
"evidence": [
"Nadia Comăneci_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:\nNadia Comăneci is a Romanian gymnast who has won a total of nine Olympic medals, the most of any athlete Romania has ever fielded to the Olympics.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Nadia Comăneci_sentence_5"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1502,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.134",
"value": "?search="
},
{
"operation": "search",
"time": "18.162",
"value": "Nadia Comăneci"
},
{
"operation": "Now on",
"time": "19.071",
"value": "Nadia Comăneci"
},
{
"operation": "Page search",
"time": "35.915",
"value": "Olympic medals"
},
{
"operation": "Highlighting",
"time": "40.082",
"value": "Nadia Comăneci_sentence_5"
},
{
"operation": "Page search",
"time": "75.307",
"value": "most"
},
{
"operation": "Page search",
"time": "118.36",
"value": "Olympic"
},
{
"operation": "Page search",
"time": "133.722",
"value": "Romania"
},
{
"operation": "finish",
"time": "326.552",
"value": "finish"
}
]
}
|
36872
|
Tower of Glass is a Nebula Award-nominated science fiction novel by American writer Robert Westall, published in 1970 that involves a 24th-century entrepreneur-tycoon-scientist, Simeon Krug.
|
{
"evidence": [
"Tower of Glass_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:\nTower of Glass is a Nebula Award-nominated science fiction novel by American writer Robert Westall, published in 1970 that involves a 24th-century entrepreneur-tycoon-scientist, Simeon Krug.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Tower of Glass_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1503,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.699",
"value": "?search="
},
{
"operation": "search",
"time": "11.58",
"value": "Tower of Glass"
},
{
"operation": "Now on",
"time": "12.317",
"value": "Tower of Glass"
},
{
"operation": "Highlighting",
"time": "18.637",
"value": "Tower of Glass_sentence_0"
},
{
"operation": "Highlighting",
"time": "24.269",
"value": "Tower of Glass_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "32.676",
"value": "Tower of Glass_sentence_1"
},
{
"operation": "finish",
"time": "41.597",
"value": "finish"
}
]
}
|
52825
|
Brian Keith Jackson has written for New York Magazine, Paper Magazine, The London Observer, Nylon Magazine and various publications about art and contemporary culture, but the plays her wrote were never performed.
|
{
"evidence": [
"Brian Keith Jackson_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:\nBrian Keith Jackson has written for New York Magazine, Paper Magazine, The London Observer, Nylon Magazine and various publications about art and contemporary culture, but the plays her wrote were never performed.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Brian Keith Jackson_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1504,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.622",
"value": "undefined"
},
{
"operation": "search",
"time": "28.503",
"value": "Brian Keith Jackson"
},
{
"operation": "Now on",
"time": "29.321",
"value": "Brian Keith Jackson"
},
{
"operation": "Highlighting",
"time": "45.791",
"value": "Brian Keith Jackson_sentence_9"
},
{
"operation": "Highlighting",
"time": "81.445",
"value": "Brian Keith Jackson_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "82.125",
"value": "Brian Keith Jackson_sentence_9"
},
{
"operation": "finish",
"time": "110.228",
"value": "finish"
}
]
}
|
48006
|
Sherluck placed in the Pimlico Futurity and won the Blue Grass Stakes.
|
{
"evidence": [
"Sherluck_sentence_7",
"Sherluck_sentence_8"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSherluck placed in the Pimlico Futurity and won the Blue Grass Stakes.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sherluck_sentence_7",
"Sherluck_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1505,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.741",
"value": "?search="
},
{
"operation": "search",
"time": "16.842",
"value": "Sherluck"
},
{
"operation": "Now on",
"time": "17.525",
"value": "Sherluck"
},
{
"operation": "Highlighting",
"time": "33.897",
"value": "Sherluck_sentence_7"
},
{
"operation": "Highlighting",
"time": "40.374",
"value": "Sherluck_sentence_8"
},
{
"operation": "finish",
"time": "62.748",
"value": "finish"
}
]
}
|
76267
|
Vindicius discovered papers exposing Tarquin's plot to regain power, he gave these papers to Publius Valerius Publicola and he was a Roman slave.
|
{
"evidence": [
"Vindicius_sentence_0",
"Vindicius_sentence_1",
"Vindicius_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:\nVindicius discovered papers exposing Tarquin's plot to regain power, he gave these papers to Publius Valerius Publicola and he was a Roman slave.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Vindicius_sentence_0",
"Vindicius_sentence_1",
"Vindicius_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1506,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.113",
"value": "?search="
},
{
"operation": "search",
"time": "9.861",
"value": "Vindicius"
},
{
"operation": "Now on",
"time": "10.926",
"value": "Vindicius"
},
{
"operation": "Highlighting",
"time": "12.621",
"value": "Vindicius_sentence_0"
},
{
"operation": "Highlighting",
"time": "16.609",
"value": "Vindicius_sentence_1"
},
{
"operation": "Highlighting",
"time": "41.371",
"value": "Vindicius_sentence_3"
},
{
"operation": "finish",
"time": "117.883",
"value": "finish"
}
]
}
|
79095
|
The Clarence Bolton was born and raised in Wallingford, Connecticut which is centrally located between New Haven and Hartford.
|
{
"evidence": [
"Wallingford, Connecticut_sentence_0",
"Clarence Bolton_sentence_2"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Clarence Bolton was born and raised in Wallingford, Connecticut which is centrally located between New Haven and Hartford.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Wallingford, Connecticut_sentence_0",
"Clarence Bolton_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1507,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.773",
"value": "?search="
},
{
"operation": "search",
"time": "19.131",
"value": "Clarence Bolton"
},
{
"operation": "Now on",
"time": "19.922",
"value": "Clarence Bolton"
},
{
"operation": "Highlighting",
"time": "40.826",
"value": "Clarence Bolton_sentence_2"
},
{
"operation": "hyperlink",
"time": "65.199",
"value": "Wallingford, Connecticut"
},
{
"operation": "Now on",
"time": "65.89",
"value": "Wallingford, Connecticut"
},
{
"operation": "Highlighting",
"time": "67.734",
"value": "Wallingford, Connecticut_sentence_0"
},
{
"operation": "Highlighting",
"time": "96.805",
"value": "Wallingford, Connecticut_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "157.719",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "157.782",
"value": "Clarence Bolton"
},
{
"operation": "Highlighting",
"time": "160.291",
"value": "Clarence Bolton_sentence_2"
},
{
"operation": "finish",
"time": "177.305",
"value": "finish"
}
]
}
|
exfever_5246
|
Mark Alan Ruffalo is an American actor who starred in the 2007 American mystery thriller film, starring an American actor and the son of director Stephen Gyllenhaal.
|
{
"evidence": [
"Mark Ruffalo",
"Zodiac (film)",
"Jake Gyllenhaal"
],
"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:\nMark Alan Ruffalo is an American actor who starred in the 2007 American mystery thriller film, starring an American actor and the son of director Stephen Gyllenhaal.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mark Ruffalo",
"Zodiac (film)",
"Jake Gyllenhaal"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1508,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7852
|
Subhash Ghai formed a successful collaboration with Dilip Kumar.
|
{
"evidence": [
"Taal (film)",
"Subhash Ghai"
],
"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:\nSubhash Ghai formed a successful collaboration with Dilip Kumar.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Taal (film)",
"Subhash Ghai"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1509,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5467
|
Faheem Rasheed Najm is an Italian rapper, singer, songwriter, and record producer who was signed to Konvict Muzik by a Senegalese American singer who rose to prominence in 2004 following the release of "Locked Up" the first single from his debut album Trouble.
|
{
"evidence": [
"T-Pain",
"Akon"
],
"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:\nFaheem Rasheed Najm is an Italian rapper, singer, songwriter, and record producer who was signed to Konvict Muzik by a Senegalese American singer who rose to prominence in 2004 following the release of \"Locked Up\" the first single from his debut album Trouble.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"T-Pain",
"Akon"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1510,
"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": 1511,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8040
|
Oscar Palmer Robertson is an American former professional basketball player who became the first player in NBA history to average a triple-double in the 16th season of the National Basketball Association that ended with the Boston Celtics winning their 4th straight NBA title, beating the Los Angeles Lakers.
|
{
"evidence": [
"Oscar Robertson",
"1961–62 NBA season"
],
"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:\nOscar Palmer Robertson is an American former professional basketball player who became the first player in NBA history to average a triple-double in the 16th season of the National Basketball Association that ended with the Boston Celtics winning their 4th straight NBA title, beating the Los Angeles Lakers.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Oscar Robertson",
"1961–62 NBA season"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1512,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
10209
|
The Polish Criminal police was analogous to the German "Kriminalpolizei" and was estimated to have between 1,790 to 2,800 ethnic Poles in the Polish Kripo units.
|
{
"evidence": [
"Polish Criminal Police_sentence_4",
"Polish Criminal Police_sentence_5"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Polish Criminal police was analogous to the German \"Kriminalpolizei\" and was estimated to have between 1,790 to 2,800 ethnic Poles in the Polish Kripo units.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Polish Criminal Police_sentence_4",
"Polish Criminal Police_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1513,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.027",
"value": "?search="
},
{
"operation": "search",
"time": "11.13",
"value": "the polish criminal police"
},
{
"operation": "Now on",
"time": "13.023",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "14.845",
"value": "Polish Criminal Police"
},
{
"operation": "Now on",
"time": "15.557",
"value": "Polish Criminal Police"
},
{
"operation": "Highlighting",
"time": "44.759",
"value": "Polish Criminal Police_sentence_1"
},
{
"operation": "Highlighting",
"time": "50.123",
"value": "Polish Criminal Police_sentence_4"
},
{
"operation": "search",
"time": "74.105",
"value": "Kriminalpolizei (Nazi Germany)"
},
{
"operation": "Now on",
"time": "76.054",
"value": "Kriminalpolizei (Nazi Germany)"
},
{
"operation": "Highlighting",
"time": "80.036",
"value": "Kriminalpolizei (Nazi Germany)_sentence_2"
},
{
"operation": "Highlighting",
"time": "80.038",
"value": "Kriminalpolizei (Nazi Germany)_title"
},
{
"operation": "back-button-clicked",
"time": "82.682",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "82.912",
"value": "Polish Criminal Police"
},
{
"operation": "Highlighting deleted",
"time": "91.623",
"value": "Polish Criminal Police_sentence_1"
},
{
"operation": "Highlighting",
"time": "93.029",
"value": "Polish Criminal Police_sentence_5"
},
{
"operation": "finish",
"time": "110.392",
"value": "finish"
}
]
}
|
exfever_4239
|
The Temple Mount is a hill in the Old City of Jerusalem that has been venerated as a holy site in an Abrahamic, monotheistic religion based on the teachings and existence of Jesus of Nazareth, which is primarily outlined in the four canonical gospels based on the New Testament.
|
{
"evidence": [
"Temple Mount",
"Christianity",
"Life of Jesus in the New Testament"
],
"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 Temple Mount is a hill in the Old City of Jerusalem that has been venerated as a holy site in an Abrahamic, monotheistic religion based on the teachings and existence of Jesus of Nazareth, which is primarily outlined in the four canonical gospels based on the New Testament.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Temple Mount",
"Christianity",
"Life of Jesus in the New Testament"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1514,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7026
|
Robert Francis Kennedy was an American lawyer and politician who was born into a wealthy political family in a town in Norfolk County, Massachusetts, in the United Kingdom.
|
{
"evidence": [
"Robert F. Kennedy",
"Brookline, Massachusetts"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRobert Francis Kennedy was an American lawyer and politician who was born into a wealthy political family in a town in Norfolk County, Massachusetts, in the United Kingdom.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Robert F. Kennedy",
"Brookline, Massachusetts"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1515,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
5845
|
Rajo Jack pseudonym Jack DeSoto called his wife, she thought that he meant to get ready for the ride but saw him backing up their truck to the garage and they wheeled the car onto the truck.
|
{
"evidence": [
"Rajo Jack_sentence_0",
"Rajo Jack_sentence_25",
"Rajo Jack_sentence_26",
"Rajo Jack_sentence_27",
"Rajo Jack_sentence_28"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRajo Jack pseudonym Jack DeSoto called his wife, she thought that he meant to get ready for the ride but saw him backing up their truck to the garage and they wheeled the car onto the truck.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Rajo Jack_sentence_0",
"Rajo Jack_sentence_25",
"Rajo Jack_sentence_26",
"Rajo Jack_sentence_27",
"Rajo Jack_sentence_28"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1516,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.938",
"value": "?search="
},
{
"operation": "search",
"time": "14.591",
"value": "Rajo Jack"
},
{
"operation": "Now on",
"time": "15.548",
"value": "Rajo Jack"
},
{
"operation": "Highlighting",
"time": "19.767",
"value": "Rajo Jack_sentence_0"
},
{
"operation": "Highlighting",
"time": "33.652",
"value": "Rajo Jack_sentence_25"
},
{
"operation": "Highlighting",
"time": "36.472",
"value": "Rajo Jack_sentence_26"
},
{
"operation": "Highlighting",
"time": "42.876",
"value": "Rajo Jack_sentence_27"
},
{
"operation": "Highlighting",
"time": "45.515",
"value": "Rajo Jack_sentence_28"
},
{
"operation": "finish",
"time": "75.365",
"value": "finish"
}
]
}
|
35079
|
The brucea javanica's is from Latin, flower anthers are typically red, flowers in June and July and measures up to 0.5 cm.
|
{
"evidence": [
"Brucea javanica_sentence_1",
"Brucea javanica_sentence_6",
"Brucea javanica_sentence_7",
"Brucea javanica_sentence_8"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe brucea javanica's is from Latin, flower anthers are typically red, flowers in June and July and measures up to 0.5 cm.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Brucea javanica_sentence_1",
"Brucea javanica_sentence_6",
"Brucea javanica_sentence_7",
"Brucea javanica_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1517,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.671",
"value": "?search="
},
{
"operation": "search",
"time": "5.882",
"value": "brucea javanica"
},
{
"operation": "Now on",
"time": "6.6",
"value": "Brucea javanica"
},
{
"operation": "Highlighting",
"time": "9.163",
"value": "Brucea javanica_sentence_1"
},
{
"operation": "Highlighting",
"time": "15.883",
"value": "Brucea javanica_sentence_6"
},
{
"operation": "Highlighting",
"time": "18.626",
"value": "Brucea javanica_sentence_7"
},
{
"operation": "Highlighting",
"time": "21.377",
"value": "Brucea javanica_sentence_8"
},
{
"operation": "finish",
"time": "27.941",
"value": "finish"
}
]
}
|
77677
|
Gerrit van der Veen's remains were reburied in 1945, and on March 9, 2002, he was made Righteous Among the Nations (an honorific used by the State of Israel to describe non-Jews who risked their lives during the Holocaust to save Jews from extermination by the Nazis for altruistic reason) by Yad Vashem.
|
{
"evidence": [
"Gerrit van der Veen_sentence_5",
"Gerrit van der Veen_sentence_6",
"Righteous Among the Nations_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:\nGerrit van der Veen's remains were reburied in 1945, and on March 9, 2002, he was made Righteous Among the Nations (an honorific used by the State of Israel to describe non-Jews who risked their lives during the Holocaust to save Jews from extermination by the Nazis for altruistic reason) by Yad Vashem.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Gerrit van der Veen_sentence_5",
"Gerrit van der Veen_sentence_6",
"Righteous Among the Nations_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1518,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.825",
"value": "?search="
},
{
"operation": "search",
"time": "9.697",
"value": "Gerrit van der Veen"
},
{
"operation": "Now on",
"time": "10.511",
"value": "Gerrit van der Veen"
},
{
"operation": "Highlighting",
"time": "28.492",
"value": "Gerrit van der Veen_sentence_5"
},
{
"operation": "Highlighting",
"time": "30.951",
"value": "Gerrit van der Veen_sentence_6"
},
{
"operation": "hyperlink",
"time": "35.928",
"value": "Righteous Among the Nations"
},
{
"operation": "Now on",
"time": "36.938",
"value": "Righteous Among the Nations"
},
{
"operation": "Highlighting",
"time": "39.373",
"value": "Righteous Among the Nations_sentence_3"
},
{
"operation": "finish",
"time": "53.72",
"value": "finish"
}
]
}
|
exfever_5089
|
Jared Joseph Leto is an actor who started his career in the American teen drama television series created by Winnie Holzman.
|
{
"evidence": [
"Jared Leto",
"My So-Called Life"
],
"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:\nJared Joseph Leto is an actor who started his career in the American teen drama television series created by Winnie Holzman.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jared Leto",
"My So-Called Life"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1519,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_541
|
Mithun Chakraborty is an Indian actor who was a former member of the upper house of the bicameral Parliament of India.
|
{
"evidence": [
"Mithun Chakraborty",
"Rajya Sabha"
],
"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:\nMithun Chakraborty is an Indian actor who was a former member of the upper house of the bicameral Parliament of India.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mithun Chakraborty",
"Rajya Sabha"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1520,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
85455
|
"Fall in Line" is centered around lyrical themes of female empowerment, the song received positive reviews from music critics and a lyric video was uploaded on YouTube which was founded in 2005.
|
{
"evidence": [
"Fall in Line_sentence_3",
"Fall in Line_sentence_4",
"Fall in Line_sentence_6",
"YouTube_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:\n\"Fall in Line\" is centered around lyrical themes of female empowerment, the song received positive reviews from music critics and a lyric video was uploaded on YouTube which was founded in 2005.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Fall in Line_sentence_3",
"Fall in Line_sentence_4",
"Fall in Line_sentence_6",
"YouTube_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1521,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.82",
"value": "?search="
},
{
"operation": "search",
"time": "13.616",
"value": "Fall in Line"
},
{
"operation": "Now on",
"time": "14.536",
"value": "Fall in Line"
},
{
"operation": "Highlighting",
"time": "87.934",
"value": "Fall in Line_sentence_3"
},
{
"operation": "Highlighting",
"time": "93.12",
"value": "Fall in Line_sentence_4"
},
{
"operation": "Highlighting",
"time": "103.134",
"value": "Fall in Line_sentence_6"
},
{
"operation": "hyperlink",
"time": "110.019",
"value": "YouTube"
},
{
"operation": "Now on",
"time": "112.176",
"value": "YouTube"
},
{
"operation": "Highlighting",
"time": "112.739",
"value": "YouTube_sentence_2"
},
{
"operation": "finish",
"time": "118.185",
"value": "finish"
}
]
}
|
85839
|
The display of Xerox 820 is a 24-line CRT, it has a bulky 96-character keyboard and its communication ports are two 25-pin RS-232 which was introduced in 1960.
|
{
"evidence": [
"Xerox 820_sentence_11",
"Xerox 820_sentence_10",
"RS-232_sentence_4",
"Xerox 820_sentence_9"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe display of Xerox 820 is a 24-line CRT, it has a bulky 96-character keyboard and its communication ports are two 25-pin RS-232 which was introduced in 1960.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Xerox 820_sentence_11",
"Xerox 820_sentence_10",
"RS-232_sentence_4",
"Xerox 820_sentence_9"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1522,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.732",
"value": "?search="
},
{
"operation": "search",
"time": "16.466",
"value": "Xerox 820"
},
{
"operation": "Now on",
"time": "17.121",
"value": "Xerox 820"
},
{
"operation": "Highlighting",
"time": "146.151",
"value": "Xerox 820_sentence_11"
},
{
"operation": "Highlighting",
"time": "157.711",
"value": "Xerox 820_sentence_10"
},
{
"operation": "hyperlink",
"time": "170.035",
"value": "RS-232"
},
{
"operation": "Now on",
"time": "172.236",
"value": "RS-232"
},
{
"operation": "Highlighting",
"time": "174.204",
"value": "RS-232_sentence_4"
},
{
"operation": "back-button-clicked",
"time": "176.632",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "176.752",
"value": "Xerox 820"
},
{
"operation": "Highlighting",
"time": "184.748",
"value": "Xerox 820_sentence_9"
},
{
"operation": "finish",
"time": "207.335",
"value": "finish"
}
]
}
|
exfever_5991
|
Alkaline Trio is a British rock band from Chicago consisted of an American singer and musician and other band member.
|
{
"evidence": [
"Alkaline Trio",
"Dan Andriano"
],
"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:\nAlkaline Trio is a British rock band from Chicago consisted of an American singer and musician and other band member.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Alkaline Trio",
"Dan Andriano"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1523,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
72016
|
While Montrose was in castle, Lord Sinclair (title in the Peerage of Scotland) sacked Mugdock however until 1634 when Montrose began his Royalist revolt, becoming the Prince's commander in Greenland.
|
{
"evidence": [
"Mugdock Castle_sentence_10"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWhile Montrose was in castle, Lord Sinclair (title in the Peerage of Scotland) sacked Mugdock however until 1634 when Montrose began his Royalist revolt, becoming the Prince's commander in Greenland.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Mugdock Castle_sentence_10"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1524,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.714",
"value": "?search="
},
{
"operation": "search",
"time": "43.883",
"value": "While Montrose was in castle, Lord Sinclair (title in the Peerage of Scotland) sacked Mugdock however until 1634 when Montrose began his Royalist revolt, becoming the Prince's commander in Greenland"
},
{
"operation": "Now on",
"time": "46.561",
"value": "undefined"
},
{
"operation": "search",
"time": "59.441",
"value": "Mugdock Castle"
},
{
"operation": "Now on",
"time": "60.543",
"value": "Mugdock Castle"
},
{
"operation": "Highlighting",
"time": "75.528",
"value": "Mugdock Castle_sentence_10"
},
{
"operation": "finish",
"time": "95.333",
"value": "finish"
}
]
}
|
10253
|
The title Baron de Longueuil was granted to Charles Le Moyne in 1657, and when he died in Montreal in 1729, the barony was passed to his son.
|
{
"evidence": [
"Charles le Moyne de Longueuil, Baron de Longueuil_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 title Baron de Longueuil was granted to Charles Le Moyne in 1657, and when he died in Montreal in 1729, the barony was passed to his son.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Charles le Moyne de Longueuil, Baron de Longueuil_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1525,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.079",
"value": "?search="
},
{
"operation": "search",
"time": "9.644",
"value": "charles le moyne"
},
{
"operation": "Now on",
"time": "10.405",
"value": "Charles Le Moyne"
},
{
"operation": "hyperlink",
"time": "19.7",
"value": "Charles le Moyne de Longueuil, Baron de Longueuil"
},
{
"operation": "Now on",
"time": "20.282",
"value": "Charles le Moyne de Longueuil, Baron de Longueuil"
},
{
"operation": "Highlighting",
"time": "25.908",
"value": "Charles le Moyne de Longueuil, Baron de Longueuil_sentence_0"
},
{
"operation": "search",
"time": "78.084",
"value": "baron du longueuil"
},
{
"operation": "Now on",
"time": "80.399",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "98.139",
"value": "Beloeil, Quebec"
},
{
"operation": "Now on",
"time": "99.06",
"value": "Beloeil, Quebec"
},
{
"operation": "back-button-clicked",
"time": "127.748",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "128.507",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "136.676",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "136.681",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "136.892",
"value": "Charles le Moyne de Longueuil, Baron de Longueuil"
},
{
"operation": "finish",
"time": "185.804",
"value": "finish"
}
]
}
|
91363
|
Karl Toft(30 June 1936-28 April 2018) was a Canadian convicted sex predator and pederast regarded as one of Canada's worst sex offenders, who was fired in 1968 and imprisoned the sexual and physical abuse of his foster child.
|
{
"evidence": [
"Karl Toft_sentence_0",
"Karl Toft_sentence_1",
"Karl Toft_sentence_6",
"Karl Toft_sentence_8",
"Karl Toft_sentence_9"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nKarl Toft(30 June 1936-28 April 2018) was a Canadian convicted sex predator and pederast regarded as one of Canada's worst sex offenders, who was fired in 1968 and imprisoned the sexual and physical abuse of his foster child.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Karl Toft_sentence_0",
"Karl Toft_sentence_1",
"Karl Toft_sentence_6",
"Karl Toft_sentence_8",
"Karl Toft_sentence_9"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1526,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.959",
"value": "?search="
},
{
"operation": "search",
"time": "13.578",
"value": "Karl Toft"
},
{
"operation": "Now on",
"time": "14.305",
"value": "Karl Toft"
},
{
"operation": "Highlighting",
"time": "32.132",
"value": "Karl Toft_sentence_0"
},
{
"operation": "Highlighting",
"time": "34.125",
"value": "Karl Toft_sentence_1"
},
{
"operation": "Highlighting",
"time": "48.015",
"value": "Karl Toft_sentence_6"
},
{
"operation": "Highlighting",
"time": "155.942",
"value": "Karl Toft_sentence_8"
},
{
"operation": "Highlighting",
"time": "170.731",
"value": "Karl Toft_sentence_9"
},
{
"operation": "finish",
"time": "273.756",
"value": "finish"
}
]
}
|
43298
|
Ismaila Sané, born on January 4, 1956 in Coubalan, Casamance, Senegal, is a Senegalese born percussionist, singer, solo dancer of African ballet, choreographer and pedagogue.
|
{
"evidence": [
"Ismaila Sané_sentence_0",
"Ismaila Sané_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:\nIsmaila Sané, born on January 4, 1956 in Coubalan, Casamance, Senegal, is a Senegalese born percussionist, singer, solo dancer of African ballet, choreographer and pedagogue.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Ismaila Sané_sentence_0",
"Ismaila Sané_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1527,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.385",
"value": "?search="
},
{
"operation": "search",
"time": "30.657",
"value": "Ismaila Sané"
},
{
"operation": "Now on",
"time": "32.059",
"value": "Ismaila Sané"
},
{
"operation": "Highlighting",
"time": "35.173",
"value": "Ismaila Sané_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "42.261",
"value": "Ismaila Sané_sentence_1"
},
{
"operation": "Highlighting",
"time": "43.576",
"value": "Ismaila Sané_sentence_0"
},
{
"operation": "Highlighting",
"time": "45.189",
"value": "Ismaila Sané_sentence_1"
},
{
"operation": "finish",
"time": "121.745",
"value": "finish"
}
]
}
|
48092
|
Michael Bean is a television, film and theater actor who is the founder of Biz Studio.
|
{
"evidence": [
"Michael Bean_sentence_1",
"Michael Bean_sentence_6",
"Michael Bean_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:\nMichael Bean is a television, film and theater actor who is the founder of Biz Studio.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Michael Bean_sentence_1",
"Michael Bean_sentence_6",
"Michael Bean_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1528,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0",
"value": "?search="
},
{
"operation": "search",
"time": "0.011",
"value": "Michael Bean"
},
{
"operation": "Now on",
"time": "0.012",
"value": "Michael Bean"
},
{
"operation": "Highlighting",
"time": "0.029",
"value": "Michael Bean_sentence_1"
},
{
"operation": "Highlighting",
"time": "0.041",
"value": "Michael Bean_sentence_6"
},
{
"operation": "Highlighting",
"time": "0.045",
"value": "Michael Bean_sentence_7"
},
{
"operation": "finish",
"time": "0.055",
"value": "finish"
},
{
"operation": "Now on",
"time": "1620942637.813",
"value": "Michael Bean"
},
{
"operation": "finish",
"time": "1620942645.463",
"value": "finish"
}
]
}
|
13730
|
The Glen Canyon National Recreation Area, which was established in 1972, was created to increase tourism and was developed to allow access to Lake Powell via five marinas.
|
{
"evidence": [
"Glen Canyon National Recreation Area_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:\nThe Glen Canyon National Recreation Area, which was established in 1972, was created to increase tourism and was developed to allow access to Lake Powell via five marinas.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Glen Canyon National Recreation Area_sentence_5"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1529,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.881",
"value": "?search="
},
{
"operation": "search",
"time": "9.958",
"value": "Glen Canyon National Recreation Area"
},
{
"operation": "Now on",
"time": "11.001",
"value": "Glen Canyon National Recreation Area"
},
{
"operation": "Highlighting",
"time": "18.816",
"value": "Glen Canyon National Recreation Area_sentence_5"
},
{
"operation": "Page search",
"time": "68.763",
"value": "tourism"
},
{
"operation": "finish",
"time": "147.779",
"value": "finish"
}
]
}
|
92675
|
Ludwikowice Kłodzkie was part of Germany between 1871 and 1945 wherein in 1942, the Germans established a forced labor camp for Jews in the village before it was attacked by the enemies.
|
{
"evidence": [
"Ludwikowice Kłodzkie_sentence_4",
"Ludwikowice Kłodzkie_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:\nLudwikowice Kłodzkie was part of Germany between 1871 and 1945 wherein in 1942, the Germans established a forced labor camp for Jews in the village before it was attacked by the enemies.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ludwikowice Kłodzkie_sentence_4",
"Ludwikowice Kłodzkie_sentence_5"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1530,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.618",
"value": "?search="
},
{
"operation": "search",
"time": "9.328",
"value": "Ludwikowice Kłodzkie"
},
{
"operation": "Now on",
"time": "10.678",
"value": "Ludwikowice Kłodzkie"
},
{
"operation": "Highlighting",
"time": "16.454",
"value": "Ludwikowice Kłodzkie_sentence_4"
},
{
"operation": "Highlighting",
"time": "21.57",
"value": "Ludwikowice Kłodzkie_sentence_5"
},
{
"operation": "Page search",
"time": "56.853",
"value": "attacked"
},
{
"operation": "hyperlink",
"time": "71.548",
"value": "Gross-Rosen concentration camp"
},
{
"operation": "Now on",
"time": "72.38",
"value": "Gross-Rosen concentration camp"
},
{
"operation": "Page search",
"time": "78.526",
"value": "attacked"
},
{
"operation": "Page search",
"time": "119.348",
"value": "enemies"
},
{
"operation": "Page search",
"time": "135.642",
"value": "Kłodzkie"
},
{
"operation": "finish",
"time": "179.895",
"value": "finish"
}
]
}
|
39658
|
The highest recorded Mean Absolute Error, a measure of errors between paired observations expressing the same phenomenon, in the Kalawa Polling Division was in 1988 and 1982.
|
{
"evidence": [
"Mean absolute error_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe highest recorded Mean Absolute Error, a measure of errors between paired observations expressing the same phenomenon, in the Kalawa Polling Division was in 1988 and 1982.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Mean absolute error_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1531,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.678",
"value": "?search="
},
{
"operation": "search",
"time": "11.511",
"value": "Mean absolute error"
},
{
"operation": "Now on",
"time": "12.483",
"value": "Mean absolute error"
},
{
"operation": "Page search",
"time": "17.854",
"value": "highest"
},
{
"operation": "Highlighting",
"time": "41.18",
"value": "Mean absolute error_sentence_1"
},
{
"operation": "search",
"time": "61.686",
"value": "Contains...Kalawa polling divison"
},
{
"operation": "Now on",
"time": "62.528",
"value": "undefined"
},
{
"operation": "search",
"time": "108.868",
"value": "Mbooni Constituency"
},
{
"operation": "Now on",
"time": "109.593",
"value": "Mbooni Constituency"
},
{
"operation": "Page search",
"time": "119.303",
"value": "Kalawa"
},
{
"operation": "Highlighting",
"time": "124.055",
"value": "Mbooni Constituency_cell_1_6_1"
},
{
"operation": "Highlighting",
"time": "124.057",
"value": "Mbooni Constituency_title"
},
{
"operation": "Highlighting",
"time": "124.058",
"value": "Mbooni Constituency_section_1"
},
{
"operation": "Highlighting",
"time": "124.059",
"value": "Mbooni Constituency_header_cell_1_2_1"
},
{
"operation": "Highlighting deleted",
"time": "124.251",
"value": "Mbooni Constituency_cell_1_6_1"
},
{
"operation": "Highlighting deleted",
"time": "124.253",
"value": "Mbooni Constituency_title"
},
{
"operation": "Highlighting deleted",
"time": "124.255",
"value": "Mbooni Constituency_section_1"
},
{
"operation": "Highlighting deleted",
"time": "124.257",
"value": "Mbooni Constituency_header_cell_1_2_1"
},
{
"operation": "Highlighting",
"time": "125.56",
"value": "Mbooni Constituency_cell_1_5_0"
},
{
"operation": "Highlighting",
"time": "125.56",
"value": "Mbooni Constituency_title"
},
{
"operation": "Highlighting",
"time": "125.56",
"value": "Mbooni Constituency_section_1"
},
{
"operation": "Highlighting",
"time": "125.561",
"value": "Mbooni Constituency_header_cell_1_2_0"
},
{
"operation": "Highlighting deleted",
"time": "126.191",
"value": "Mbooni Constituency_cell_1_5_0"
},
{
"operation": "Highlighting deleted",
"time": "126.193",
"value": "Mbooni Constituency_title"
},
{
"operation": "Highlighting deleted",
"time": "126.194",
"value": "Mbooni Constituency_section_1"
},
{
"operation": "Highlighting deleted",
"time": "126.195",
"value": "Mbooni Constituency_header_cell_1_2_0"
},
{
"operation": "search",
"time": "132.458",
"value": "Kalawa"
},
{
"operation": "Now on",
"time": "133.312",
"value": "Kalawa"
},
{
"operation": "search",
"time": "155.565",
"value": "Kalawa Jazmee Records"
},
{
"operation": "Now on",
"time": "156.247",
"value": "Kalawa Jazmee Records"
},
{
"operation": "Page search",
"time": "162.307",
"value": "division"
},
{
"operation": "search",
"time": "184.358",
"value": "Contains...Kalawa Polling Division"
},
{
"operation": "Now on",
"time": "185.22",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "190.766",
"value": "kagawa polling division"
},
{
"operation": "Now on",
"time": "191.33",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "193.304",
"value": "2020 in Japan"
},
{
"operation": "Now on",
"time": "194.559",
"value": "2020 in Japan"
},
{
"operation": "Page search",
"time": "198.811",
"value": "kalawa"
},
{
"operation": "finish",
"time": "279.274",
"value": "finish"
}
]
}
|
exfever_3350
|
Jerry Lewis was an American comedian, actor, and singer who professionally debuted as part of Martin and Lewis with an American singer, actor, and comedian of Italian descent.
|
{
"evidence": [
"Jerry Lewis",
"Dean Martin"
],
"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:\nJerry Lewis was an American comedian, actor, and singer who professionally debuted as part of Martin and Lewis with an American singer, actor, and comedian of Italian descent.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jerry Lewis",
"Dean Martin"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1532,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
33596
|
Vicious Cycle Software had produced 24 products suitable for PS2, GameCube, Xbox, and other consoles wherein these games includes series of games for Robotech, Ben 10, Pacman as well as games like Spy vs. Spy—a video game based on the MAD magazine’s titular comic strip which has up to five players for multiplayer mode.
|
{
"evidence": [
"Spy vs. Spy (2005 video game)_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:\nVicious Cycle Software had produced 24 products suitable for PS2, GameCube, Xbox, and other consoles wherein these games includes series of games for Robotech, Ben 10, Pacman as well as games like Spy vs. Spy—a video game based on the MAD magazine’s titular comic strip which has up to five players for multiplayer mode.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Spy vs. Spy (2005 video game)_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1533,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.056",
"value": "?search="
},
{
"operation": "search",
"time": "11.892",
"value": "Vicious Cycle Software"
},
{
"operation": "Now on",
"time": "12.889",
"value": "Vicious Cycle Software"
},
{
"operation": "Page search",
"time": "34.4",
"value": "24"
},
{
"operation": "Highlighting",
"time": "63.385",
"value": "Vicious Cycle Software_cell_1_1_0"
},
{
"operation": "Highlighting",
"time": "63.386",
"value": "Vicious Cycle Software_title"
},
{
"operation": "Highlighting",
"time": "63.386",
"value": "Vicious Cycle Software_section_1"
},
{
"operation": "Highlighting",
"time": "63.387",
"value": "Vicious Cycle Software_header_cell_1_0_0"
},
{
"operation": "Highlighting",
"time": "64.804",
"value": "Vicious Cycle Software_cell_1_1_2"
},
{
"operation": "Highlighting",
"time": "64.806",
"value": "Vicious Cycle Software_header_cell_1_0_2"
},
{
"operation": "Highlighting",
"time": "66.189",
"value": "Vicious Cycle Software_cell_1_2_0"
},
{
"operation": "Highlighting",
"time": "67.693",
"value": "Vicious Cycle Software_cell_1_2_2"
},
{
"operation": "Highlighting",
"time": "68.764",
"value": "Vicious Cycle Software_cell_1_3_0"
},
{
"operation": "Highlighting",
"time": "69.828",
"value": "Vicious Cycle Software_cell_1_3_2"
},
{
"operation": "Highlighting",
"time": "71.145",
"value": "Vicious Cycle Software_cell_1_4_0"
},
{
"operation": "Highlighting",
"time": "72.37",
"value": "Vicious Cycle Software_cell_1_4_2"
},
{
"operation": "Highlighting",
"time": "73.68",
"value": "Vicious Cycle Software_cell_1_5_0"
},
{
"operation": "Highlighting",
"time": "75.146",
"value": "Vicious Cycle Software_cell_1_5_2"
},
{
"operation": "Highlighting",
"time": "76.285",
"value": "Vicious Cycle Software_cell_1_6_0"
},
{
"operation": "Highlighting",
"time": "77.718",
"value": "Vicious Cycle Software_cell_1_6_2"
},
{
"operation": "Highlighting",
"time": "79.067",
"value": "Vicious Cycle Software_cell_1_7_0"
},
{
"operation": "Highlighting",
"time": "80.268",
"value": "Vicious Cycle Software_cell_1_7_2"
},
{
"operation": "Highlighting deleted",
"time": "80.929",
"value": "Vicious Cycle Software_cell_1_7_2"
},
{
"operation": "Highlighting",
"time": "82.147",
"value": "Vicious Cycle Software_cell_1_7_2"
},
{
"operation": "Highlighting deleted",
"time": "82.556",
"value": "Vicious Cycle Software_cell_1_7_2"
},
{
"operation": "Highlighting",
"time": "83.895",
"value": "Vicious Cycle Software_cell_1_8_0"
},
{
"operation": "Highlighting",
"time": "84.744",
"value": "Vicious Cycle Software_cell_1_8_2"
},
{
"operation": "Highlighting",
"time": "85.898",
"value": "Vicious Cycle Software_cell_1_9_0"
},
{
"operation": "Highlighting",
"time": "86.902",
"value": "Vicious Cycle Software_cell_1_9_2"
},
{
"operation": "Highlighting",
"time": "87.866",
"value": "Vicious Cycle Software_cell_1_10_0"
},
{
"operation": "Highlighting",
"time": "88.813",
"value": "Vicious Cycle Software_cell_1_10_2"
},
{
"operation": "Highlighting",
"time": "90.246",
"value": "Vicious Cycle Software_cell_1_11_0"
},
{
"operation": "Highlighting",
"time": "92.32",
"value": "Vicious Cycle Software_cell_1_11_2"
},
{
"operation": "Highlighting",
"time": "95.276",
"value": "Vicious Cycle Software_cell_1_12_0"
},
{
"operation": "Highlighting",
"time": "96.296",
"value": "Vicious Cycle Software_cell_1_12_2"
},
{
"operation": "Highlighting",
"time": "98.8",
"value": "Vicious Cycle Software_cell_1_7_2"
},
{
"operation": "Highlighting",
"time": "104.661",
"value": "Vicious Cycle Software_cell_1_13_0"
},
{
"operation": "Highlighting",
"time": "105.995",
"value": "Vicious Cycle Software_cell_1_13_2"
},
{
"operation": "Highlighting",
"time": "107.514",
"value": "Vicious Cycle Software_cell_1_14_0"
},
{
"operation": "Highlighting",
"time": "108.454",
"value": "Vicious Cycle Software_cell_1_14_2"
},
{
"operation": "Highlighting",
"time": "109.556",
"value": "Vicious Cycle Software_cell_1_15_0"
},
{
"operation": "Highlighting",
"time": "111.978",
"value": "Vicious Cycle Software_cell_1_15_2"
},
{
"operation": "Highlighting",
"time": "113.21",
"value": "Vicious Cycle Software_cell_1_16_0"
},
{
"operation": "Highlighting",
"time": "114.366",
"value": "Vicious Cycle Software_cell_1_16_2"
},
{
"operation": "Highlighting",
"time": "115.691",
"value": "Vicious Cycle Software_cell_1_17_0"
},
{
"operation": "Highlighting",
"time": "117.031",
"value": "Vicious Cycle Software_cell_1_17_2"
},
{
"operation": "Highlighting",
"time": "118.417",
"value": "Vicious Cycle Software_cell_1_18_0"
},
{
"operation": "Highlighting",
"time": "119.617",
"value": "Vicious Cycle Software_cell_1_18_2"
},
{
"operation": "Highlighting",
"time": "120.648",
"value": "Vicious Cycle Software_cell_1_19_0"
},
{
"operation": "Highlighting",
"time": "121.851",
"value": "Vicious Cycle Software_cell_1_19_2"
},
{
"operation": "Highlighting",
"time": "123.173",
"value": "Vicious Cycle Software_cell_1_20_0"
},
{
"operation": "Highlighting",
"time": "124.076",
"value": "Vicious Cycle Software_cell_1_20_2"
},
{
"operation": "Highlighting",
"time": "125.561",
"value": "Vicious Cycle Software_cell_1_21_0"
},
{
"operation": "Highlighting",
"time": "126.856",
"value": "Vicious Cycle Software_cell_1_21_2"
},
{
"operation": "Highlighting",
"time": "128.109",
"value": "Vicious Cycle Software_cell_1_22_0"
},
{
"operation": "Highlighting",
"time": "129.048",
"value": "Vicious Cycle Software_cell_1_22_2"
},
{
"operation": "Highlighting",
"time": "130.298",
"value": "Vicious Cycle Software_cell_1_23_0"
},
{
"operation": "Highlighting",
"time": "131.583",
"value": "Vicious Cycle Software_cell_1_23_2"
},
{
"operation": "Highlighting",
"time": "133.001",
"value": "Vicious Cycle Software_cell_1_24_0"
},
{
"operation": "Highlighting",
"time": "134.259",
"value": "Vicious Cycle Software_cell_1_24_2"
},
{
"operation": "Page search",
"time": "153.64",
"value": "robotech"
},
{
"operation": "Highlighting deleted",
"time": "171.236",
"value": "Vicious Cycle Software_cell_1_24_0"
},
{
"operation": "Highlighting",
"time": "172.657",
"value": "Vicious Cycle Software_cell_1_24_0"
},
{
"operation": "Page search",
"time": "196.345",
"value": "pacman"
},
{
"operation": "Highlighting deleted",
"time": "211.514",
"value": "Vicious Cycle Software_cell_1_1_0"
},
{
"operation": "Highlighting",
"time": "213.282",
"value": "Vicious Cycle Software_cell_1_1_0"
},
{
"operation": "hyperlink",
"time": "227.239",
"value": "Spy vs. Spy"
},
{
"operation": "Now on",
"time": "228.042",
"value": "Spy vs. Spy (2005 video game)"
},
{
"operation": "Highlighting",
"time": "242.897",
"value": "Spy vs. Spy (2005 video game)_sentence_0"
},
{
"operation": "Page search",
"time": "253.446",
"value": "player"
},
{
"operation": "Highlighting",
"time": "268.433",
"value": "Spy vs. Spy (2005 video game)_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "270.75",
"value": "Spy vs. Spy (2005 video game)_sentence_0"
},
{
"operation": "Highlighting",
"time": "294.576",
"value": "Spy vs. Spy (2005 video game)_sentence_1"
},
{
"operation": "finish",
"time": "344.992",
"value": "finish"
}
]
}
|
33446
|
Natioro language, a Niger–Congo language of Burkina Faso spoken by a caste of blacksmiths,
|
{
"evidence": [
"Natioro language_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:\nNatioro language, a Niger–Congo language of Burkina Faso spoken by a caste of blacksmiths,\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Natioro language_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1534,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.54",
"value": "?search="
},
{
"operation": "search",
"time": "6.886",
"value": "Natioro language"
},
{
"operation": "Now on",
"time": "8.024",
"value": "Natioro language"
},
{
"operation": "Highlighting",
"time": "11.889",
"value": "Natioro language_sentence_0"
},
{
"operation": "finish",
"time": "22.625",
"value": "finish"
}
]
}
|
exfever_1792
|
Detroit is the largest city in the U.S. state in the Great Lakes region.
|
{
"evidence": [
"Detroit",
"Michigan"
],
"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:\nDetroit is the largest city in the U.S. state in the Great Lakes region.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Detroit",
"Michigan"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1535,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
53167
|
Caroline de Maupéou who married Count René de Maupeou was born in Mulhouse and was trained by Léon Bonnat and later showed her work in Paris.
|
{
"evidence": [
"Caroline de Maupéou_sentence_2",
"Caroline de Maupéou_sentence_1",
"Caroline de Maupéou_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:\nCaroline de Maupéou who married Count René de Maupeou was born in Mulhouse and was trained by Léon Bonnat and later showed her work in Paris.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Caroline de Maupéou_sentence_2",
"Caroline de Maupéou_sentence_1",
"Caroline de Maupéou_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1536,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.145",
"value": "?search="
},
{
"operation": "search",
"time": "3.788",
"value": "Caroline de Maupéou"
},
{
"operation": "Now on",
"time": "6.921",
"value": "Caroline de Maupéou"
},
{
"operation": "Highlighting",
"time": "11.414",
"value": "Caroline de Maupéou_sentence_2"
},
{
"operation": "Highlighting",
"time": "17.962",
"value": "Caroline de Maupéou_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "17.964",
"value": "Caroline de Maupéou_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "17.965",
"value": "Caroline de Maupéou_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "25.458",
"value": "Caroline de Maupéou_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "26.654",
"value": "Caroline de Maupéou_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "26.658",
"value": "Caroline de Maupéou_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "26.66",
"value": "Caroline de Maupéou_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "29.618",
"value": "Caroline de Maupéou_sentence_3"
},
{
"operation": "finish",
"time": "34.453",
"value": "finish"
}
]
}
|
exfever_1854
|
L.A. Law is an American legal drama television series centered on the partners, associates, and staff of a law firm in the smallest city in California that is adjacent to the Pacific Ocean, extending into an urbanized valley known for its iconic film studios.
|
{
"evidence": [
"L.A. Law",
"Los Angeles",
"San Fernando Valley"
],
"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:\nL.A. Law is an American legal drama television series centered on the partners, associates, and staff of a law firm in the smallest city in California that is adjacent to the Pacific Ocean, extending into an urbanized valley known for its iconic film studios.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"L.A. Law",
"Los Angeles",
"San Fernando Valley"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1537,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
37972
|
Antonovo has a population of 1,453 (in 2009) located in Targvishte, a province in northeastern Bulgaria, named after its main city - Targovishte in Brunei.
|
{
"evidence": [
"Antonovo_sentence_0",
"Targovishte Province_sentence_0",
"Brunei_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:\nAntonovo has a population of 1,453 (in 2009) located in Targvishte, a province in northeastern Bulgaria, named after its main city - Targovishte in Brunei.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Antonovo_sentence_0",
"Targovishte Province_sentence_0",
"Brunei_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1538,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.596",
"value": "?search="
},
{
"operation": "search",
"time": "12.526",
"value": "Antonovo"
},
{
"operation": "Now on",
"time": "13.758",
"value": "Antonovo"
},
{
"operation": "Highlighting",
"time": "19.348",
"value": "Antonovo_sentence_0"
},
{
"operation": "Highlighting",
"time": "33.387",
"value": "Antonovo_sentence_3"
},
{
"operation": "hyperlink",
"time": "76.283",
"value": "Targovishte"
},
{
"operation": "Now on",
"time": "77.485",
"value": "Targovishte Province"
},
{
"operation": "Highlighting",
"time": "80.862",
"value": "Targovishte Province_sentence_0"
},
{
"operation": "hyperlink",
"time": "95.508",
"value": "Targovishte"
},
{
"operation": "Now on",
"time": "97.412",
"value": "Targovishte"
},
{
"operation": "search",
"time": "153.034",
"value": "Brunei"
},
{
"operation": "Now on",
"time": "160.65",
"value": "Brunei"
},
{
"operation": "Highlighting",
"time": "166.419",
"value": "Brunei_sentence_0"
},
{
"operation": "finish",
"time": "206.176",
"value": "finish"
}
]
}
|
52665
|
Neshaminy Creek, named Nishambanach in 1671, has two springs, one was known at the time as the 'Giant Spring'.
|
{
"evidence": [
"Neshaminy Creek_sentence_13",
"Neshaminy Creek_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:\nNeshaminy Creek, named Nishambanach in 1671, has two springs, one was known at the time as the 'Giant Spring'.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Neshaminy Creek_sentence_13",
"Neshaminy Creek_sentence_14"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1539,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.934",
"value": "?search="
},
{
"operation": "Now on",
"time": "19.779",
"value": "?search="
},
{
"operation": "search",
"time": "96.741",
"value": "Neshaminy Creek"
},
{
"operation": "Now on",
"time": "98.311",
"value": "Neshaminy Creek"
},
{
"operation": "hyperlink",
"time": "109.151",
"value": "2 History Neshaminy Creek_section_1"
},
{
"operation": "Highlighting",
"time": "120.729",
"value": "Neshaminy Creek_sentence_12"
},
{
"operation": "Highlighting deleted",
"time": "211.926",
"value": "Neshaminy Creek_sentence_12"
},
{
"operation": "Highlighting",
"time": "221.111",
"value": "Neshaminy Creek_sentence_13"
},
{
"operation": "Highlighting",
"time": "317.734",
"value": "Neshaminy Creek_sentence_14"
},
{
"operation": "finish",
"time": "583.342",
"value": "finish"
}
]
}
|
79722
|
In the first ever season of a top football league in South Korea, two players received awards for being the best midfielders.
|
{
"evidence": [
"Cho Kwang-rae_sentence_3"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn the first ever season of a top football league in South Korea, two players received awards for being the best midfielders.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Cho Kwang-rae_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1540,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.376",
"value": "?search="
},
{
"operation": "search",
"time": "33.837",
"value": "Contains...first ever season of a top football league in South Korea"
},
{
"operation": "Now on",
"time": "44.778",
"value": "undefined"
},
{
"operation": "search",
"time": "67.854",
"value": "Contains...top football league in South Korea"
},
{
"operation": "Now on",
"time": "69.392",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "130.355",
"value": "Korean League Cup"
},
{
"operation": "Now on",
"time": "132.032",
"value": "Korean League Cup"
},
{
"operation": "search",
"time": "264.922",
"value": "South Korea best midfielder"
},
{
"operation": "Now on",
"time": "266.261",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "289.019",
"value": "Yoo Sang-chul"
},
{
"operation": "Now on",
"time": "290.622",
"value": "Yoo Sang-chul"
},
{
"operation": "hyperlink",
"time": "397.766",
"value": "2004 Summer Olympics"
},
{
"operation": "Now on",
"time": "399.659",
"value": "Football at the 2004 Summer Olympics – Men's tournament"
},
{
"operation": "search",
"time": "429.836",
"value": "\"top football leagues in South Korea\""
},
{
"operation": "Now on",
"time": "430.898",
"value": "undefined"
},
{
"operation": "search",
"time": "469.639",
"value": "\"two players received awards for being the best midfielders\""
},
{
"operation": "Now on",
"time": "470.629",
"value": "undefined"
},
{
"operation": "search",
"time": "489.255",
"value": "Contains...two players received awards"
},
{
"operation": "Now on",
"time": "491.332",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "529.42",
"value": "next 20"
},
{
"operation": "Now on",
"time": "530.522",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "543.915",
"value": "Awards FIFA World Cup_section_20"
},
{
"operation": "Now on",
"time": "546.367",
"value": "FIFA World Cup#Awards FIFA World Cup section 20"
},
{
"operation": "search",
"time": "726.751",
"value": "FIFA World Cup 2002"
},
{
"operation": "Now on",
"time": "729.349",
"value": "2002 FIFA World Cup"
},
{
"operation": "search",
"time": "918.632",
"value": "\"in the first ever season\""
},
{
"operation": "Now on",
"time": "924.361",
"value": "undefined"
},
{
"operation": "search",
"time": "1065.251",
"value": "South Korean football league"
},
{
"operation": "Now on",
"time": "1066.679",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "1104.968",
"value": "Football in South Korea"
},
{
"operation": "Now on",
"time": "1106.02",
"value": "Football in South Korea"
},
{
"operation": "search",
"time": "1211.83",
"value": "South Korea Best Midfielder list"
},
{
"operation": "Now on",
"time": "1212.867",
"value": "undefined"
},
{
"operation": "search",
"time": "1355.816",
"value": "List of South Korea midfielders"
},
{
"operation": "Now on",
"time": "1357.636",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "1444.947",
"value": "Cho Kwang-rae"
},
{
"operation": "Now on",
"time": "1446.11",
"value": "Cho Kwang-rae"
},
{
"operation": "Highlighting",
"time": "1460.423",
"value": "Cho Kwang-rae_sentence_3"
},
{
"operation": "finish",
"time": "1507.981",
"value": "finish"
}
]
}
|
exfever_3094
|
Dennis William Quaid is an American actor with notable credits, including an American science fiction comedy film starring an American actress who began her acting career in 1969 in minor roles before joining the cast of the CBS soap opera As the World Turns in 1982.
|
{
"evidence": [
"Dennis Quaid",
"Innerspace",
"Meg Ryan"
],
"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:\nDennis William Quaid is an American actor with notable credits, including an American science fiction comedy film starring an American actress who began her acting career in 1969 in minor roles before joining the cast of the CBS soap opera As the World Turns in 1982.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dennis Quaid",
"Innerspace",
"Meg Ryan"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1541,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
9467
|
The Soul Cages is the third full-length studio album released by Sting (or Gordon Matthew Thomas Sumner) in 1991, and the first to feature longtime guitarist Dominic Miller.
|
{
"evidence": [
"The Soul Cages_sentence_4",
"The Soul Cages_sentence_5",
"Sting (musician)_sentence_2"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Soul Cages is the third full-length studio album released by Sting (or Gordon Matthew Thomas Sumner) in 1991, and the first to feature longtime guitarist Dominic Miller.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"The Soul Cages_sentence_4",
"The Soul Cages_sentence_5",
"Sting (musician)_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1542,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.408",
"value": "?search="
},
{
"operation": "search",
"time": "15.347",
"value": "Soul Cages"
},
{
"operation": "Now on",
"time": "16.038",
"value": "The Soul Cages"
},
{
"operation": "Highlighting",
"time": "23.899",
"value": "The Soul Cages_sentence_4"
},
{
"operation": "Highlighting",
"time": "30.938",
"value": "The Soul Cages_sentence_5"
},
{
"operation": "hyperlink",
"time": "34.162",
"value": "Sting"
},
{
"operation": "Now on",
"time": "34.989",
"value": "Sting (musician)"
},
{
"operation": "hyperlink",
"time": "39.338",
"value": "CBE"
},
{
"operation": "Now on",
"time": "39.858",
"value": "Order of the British Empire"
},
{
"operation": "back-button-clicked",
"time": "47.33",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "47.504",
"value": "Sting (musician)"
},
{
"operation": "Highlighting",
"time": "50.65",
"value": "Sting (musician)_sentence_2"
},
{
"operation": "back-button-clicked",
"time": "56.586",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "56.588",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "56.993",
"value": "The Soul Cages"
},
{
"operation": "finish",
"time": "78.441",
"value": "finish"
}
]
}
|
exfever_7654
|
The Boeing 707 makes its first flight from Renton Field, southeast of Seattle.
|
{
"evidence": [
"Boeing 707",
"Narrow-body aircraft"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Boeing 707 makes its first flight from Renton Field, southeast of Seattle.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Boeing 707",
"Narrow-body aircraft"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1543,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
3315
|
Tolarno Stations is on the traditional land of the Barkindji people (an Aboriginal tribal group of the Darling River in Australia), wherein a school was developed to educate the children station workers, and those on neighbouring stations.
|
{
"evidence": [
"Tolarno Station_sentence_3",
"Tolarno Station_sentence_5",
"Paakantyi_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:\nTolarno Stations is on the traditional land of the Barkindji people (an Aboriginal tribal group of the Darling River in Australia), wherein a school was developed to educate the children station workers, and those on neighbouring stations.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Tolarno Station_sentence_3",
"Tolarno Station_sentence_5",
"Paakantyi_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1544,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "22.4",
"value": "Contains...Tolarno Stations"
},
{
"operation": "hyperlink",
"time": "27.089",
"value": "Tolarno Station"
},
{
"operation": "Highlighting",
"time": "36.532",
"value": "Tolarno Station_sentence_3"
},
{
"operation": "Highlighting",
"time": "46.018",
"value": "Tolarno Station_sentence_5"
},
{
"operation": "hyperlink",
"time": "49.088",
"value": "Barkindji"
},
{
"operation": "Highlighting",
"time": "53.255",
"value": "Paakantyi_sentence_0"
},
{
"operation": "finish",
"time": "98.957",
"value": "finish"
}
]
}
|
exfever_8384
|
Match Point was also nominated for the Golden Eagle Award.
|
{
"evidence": [
"Match Point",
"Penelope Wilton",
"The Borrowers (miniseries)"
],
"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:\nMatch Point was also nominated for the Golden Eagle Award.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Match Point",
"Penelope Wilton",
"The Borrowers (miniseries)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1545,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5799
|
Rajon Pierre Rondo is an American professional baseball player who was drafted 21st overall by the Phoenix Suns in the NBA draft that was held on June 28, 2006, at the Theatre at Madison Square Garden in New York City.
|
{
"evidence": [
"Rajon Rondo",
"2006 NBA draft"
],
"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:\nRajon Pierre Rondo is an American professional baseball player who was drafted 21st overall by the Phoenix Suns in the NBA draft that was held on June 28, 2006, at the Theatre at Madison Square Garden in New York City.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Rajon Rondo",
"2006 NBA draft"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1546,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
72902
|
Sepia adami is a species of cuttlefish that lives at a depth of up to 99 m, females reach a mantle length of at least 59 mm but there has never been a deposit of the type specimen.
|
{
"evidence": [
"Sepia adami_sentence_4"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSepia adami is a species of cuttlefish that lives at a depth of up to 99 m, females reach a mantle length of at least 59 mm but there has never been a deposit of the type specimen.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sepia adami_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1547,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.655",
"value": "?search="
},
{
"operation": "search",
"time": "16.253",
"value": "Sepia adami"
},
{
"operation": "Now on",
"time": "17.37",
"value": "Sepia adami"
},
{
"operation": "Highlighting",
"time": "34.374",
"value": "Sepia adami_sentence_0"
},
{
"operation": "Highlighting",
"time": "38.879",
"value": "Sepia adami_sentence_2"
},
{
"operation": "Highlighting",
"time": "41.236",
"value": "Sepia adami_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "47.329",
"value": "Sepia adami_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "47.773",
"value": "Sepia adami_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "48.37",
"value": "Sepia adami_sentence_0"
},
{
"operation": "Highlighting",
"time": "49.475",
"value": "Sepia adami_sentence_4"
},
{
"operation": "finish",
"time": "56.827",
"value": "finish"
}
]
}
|
exfever_6310
|
Berlin is Germany's municipality holding primary status in a country, state, province, department, or other sub-national entity, usually as its seat of the government.
|
{
"evidence": [
"Berlin",
"Capital city"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBerlin is Germany's municipality holding primary status in a country, state, province, department, or other sub-national entity, usually as its seat of the government.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Berlin",
"Capital city"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1548,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
55262
|
Rolfe graduated from Harvard University in 1881 and from Cornell University (Ph.D.) in 1885; then he taught at Cornell (1882–1885), and at Harvard (1889–1890).
|
{
"evidence": [
"John Carew Rolfe_sentence_1",
"John Carew Rolfe_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:\nRolfe graduated from Harvard University in 1881 and from Cornell University (Ph.D.) in 1885; then he taught at Cornell (1882–1885), and at Harvard (1889–1890).\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"John Carew Rolfe_sentence_1",
"John Carew Rolfe_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1549,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.397",
"value": "?search="
},
{
"operation": "search",
"time": "12.164",
"value": "Contains...Rolfe graduated from Harvard University in 1881"
},
{
"operation": "Now on",
"time": "13.58",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "16.859",
"value": "John Carew Rolfe"
},
{
"operation": "Now on",
"time": "17.388",
"value": "John Carew Rolfe"
},
{
"operation": "Highlighting",
"time": "22.122",
"value": "John Carew Rolfe_sentence_1"
},
{
"operation": "hyperlink",
"time": "29.162",
"value": "Cornell"
},
{
"operation": "Now on",
"time": "33.921",
"value": "John Carew Rolfe"
},
{
"operation": "Highlighting",
"time": "35.659",
"value": "John Carew Rolfe_sentence_2"
},
{
"operation": "finish",
"time": "53.376",
"value": "finish"
}
]
}
|
exfever_7684
|
Evan Goldberg is an English director, and one of his films is the 2013 American apocalyptic film, whose genre belongs to a category of film that emphasizes humor.
|
{
"evidence": [
"Evan Goldberg",
"This Is the End",
"Comedy 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:\nEvan Goldberg is an English director, and one of his films is the 2013 American apocalyptic film, whose genre belongs to a category of film that emphasizes humor.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Evan Goldberg",
"This Is the End",
"Comedy film"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1550,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2858
|
Jared Joseph Leto is the youngest member of the rock band Thirty Seconds to Mars.
|
{
"evidence": [
"Jared Leto",
"Prefontaine (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:\nJared Joseph Leto is the youngest member of the rock band Thirty Seconds to Mars.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jared Leto",
"Prefontaine (film)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1551,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3641
|
The Texas Longhorns men's basketball team represents The University of Texas at Austin in the highest level of intercollegiate athletics sanctioned by the nonprofit organization that regulates student-athletes from up to 1,268 North American institutions and conferences.
|
{
"evidence": [
"Texas Longhorns men's basketball",
"NCAA Division I",
"National Collegiate Athletic Association"
],
"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 Texas Longhorns men's basketball team represents The University of Texas at Austin in the highest level of intercollegiate athletics sanctioned by the nonprofit organization that regulates student-athletes from up to 1,268 North American institutions and conferences.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Texas Longhorns men's basketball",
"NCAA Division I",
"National Collegiate Athletic Association"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1552,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
18423
|
The Tamarixia radiata is an hymenoptrean wasp, used to control D. Citri and around the world, and a vector for citrus greening disease; it develops from nymph to host in about eleven days.
|
{
"evidence": [
"Tamarixia radiata_sentence_0",
"Tamarixia radiata_sentence_35",
"Tamarixia radiata_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 Tamarixia radiata is an hymenoptrean wasp, used to control D. Citri and around the world, and a vector for citrus greening disease; it develops from nymph to host in about eleven days.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Tamarixia radiata_sentence_0",
"Tamarixia radiata_sentence_35",
"Tamarixia radiata_sentence_11"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1553,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.525",
"value": "?search="
},
{
"operation": "search",
"time": "24.927",
"value": "Tamarixia radiata"
},
{
"operation": "Now on",
"time": "26.266",
"value": "Tamarixia radiata"
},
{
"operation": "Highlighting",
"time": "133.894",
"value": "Tamarixia radiata_sentence_0"
},
{
"operation": "hyperlink",
"time": "153.234",
"value": "4 Use as a Biological Control Tamarixia radiata_section_3"
},
{
"operation": "Highlighting",
"time": "179.599",
"value": "Tamarixia radiata_sentence_35"
},
{
"operation": "Page search",
"time": "245.326",
"value": "world"
},
{
"operation": "Highlighting",
"time": "286.538",
"value": "Tamarixia radiata_sentence_11"
},
{
"operation": "Page search",
"time": "345.272",
"value": "eleven"
},
{
"operation": "Page search",
"time": "358.346",
"value": "11"
},
{
"operation": "finish",
"time": "504.382",
"value": "finish"
}
]
}
|
exfever_6550
|
The Ten Commandments is a 1956 American epic religious drama film produced, directed, and narrated by an American film director, producer, and actor regarded as a founding father of American cinema and the most commercially successful producer-director in film history.
|
{
"evidence": [
"The Ten Commandments (1956 film)",
"Cecil B. DeMille"
],
"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 Ten Commandments is a 1956 American epic religious drama film produced, directed, and narrated by an American film director, producer, and actor regarded as a founding father of American cinema and the most commercially successful producer-director in film history.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Ten Commandments (1956 film)",
"Cecil B. DeMille"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1554,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8361
|
Mr. Nobody is a 2009 drama film whose genre uses speculative, fictional science-based depictions of phenomena that are not fully accepted by mainstream science, including machines, especially ones programmable by a computer.
|
{
"evidence": [
"Mr. Nobody (film)",
"Science fiction film",
"Robot"
],
"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:\nMr. Nobody is a 2009 drama film whose genre uses speculative, fictional science-based depictions of phenomena that are not fully accepted by mainstream science, including machines, especially ones programmable by a computer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mr. Nobody (film)",
"Science fiction film",
"Robot"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1555,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
79212
|
As his name indicated, Ustad Ahmad originally hailed from Lahore and in writings by Lahori's son, Lutfullah Muhandis, one architects are mentioned by name; Ustad Ahmad Lahori.
|
{
"evidence": [
"Ustad Ahmad Lahori_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:\nAs his name indicated, Ustad Ahmad originally hailed from Lahore and in writings by Lahori's son, Lutfullah Muhandis, one architects are mentioned by name; Ustad Ahmad Lahori.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Ustad Ahmad Lahori_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1556,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.826",
"value": "?search="
},
{
"operation": "search",
"time": "9.834",
"value": "Ustad Ahmad"
},
{
"operation": "Now on",
"time": "11.417",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "15.12",
"value": "Ustad Ahmad Lahori"
},
{
"operation": "Now on",
"time": "15.766",
"value": "Ustad Ahmad Lahori"
},
{
"operation": "Highlighting",
"time": "24.313",
"value": "Ustad Ahmad Lahori_sentence_3"
},
{
"operation": "Highlighting",
"time": "34.648",
"value": "Ustad Ahmad Lahori_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "36.754",
"value": "Ustad Ahmad Lahori_sentence_3"
},
{
"operation": "finish",
"time": "44.283",
"value": "finish"
}
]
}
|
65148
|
Daniel Caspary, born on May 4, 1976 in Karlsruhe, Baden-Württemberg, Germany, was a member of the Christian Democratic Union and European People's Party.
|
{
"evidence": [
"Daniel Caspary_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:\nDaniel Caspary, born on May 4, 1976 in Karlsruhe, Baden-Württemberg, Germany, was a member of the Christian Democratic Union and European People's Party.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Daniel Caspary_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1557,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.456",
"value": "?search="
},
{
"operation": "search",
"time": "16.038",
"value": "Daniel Caspary"
},
{
"operation": "Now on",
"time": "25.354",
"value": "Daniel Caspary"
},
{
"operation": "Highlighting",
"time": "40.87",
"value": "Daniel Caspary_cell_0_4_1"
},
{
"operation": "Highlighting",
"time": "40.877",
"value": "Daniel Caspary_header_cell_0_4_0"
},
{
"operation": "Highlighting",
"time": "40.877",
"value": "Daniel Caspary_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "44.919",
"value": "Daniel Caspary_sentence_0"
},
{
"operation": "Highlighting",
"time": "59.137",
"value": "Daniel Caspary_cell_0_5_1"
},
{
"operation": "Highlighting",
"time": "59.142",
"value": "Daniel Caspary_header_cell_0_5_0"
},
{
"operation": "Highlighting deleted",
"time": "65.546",
"value": "Daniel Caspary_cell_0_4_1"
},
{
"operation": "Highlighting deleted",
"time": "65.552",
"value": "Daniel Caspary_header_cell_0_4_0"
},
{
"operation": "Highlighting",
"time": "66.222",
"value": "Daniel Caspary_cell_0_4_1"
},
{
"operation": "Highlighting",
"time": "66.227",
"value": "Daniel Caspary_header_cell_0_4_0"
},
{
"operation": "Highlighting deleted",
"time": "68.697",
"value": "Daniel Caspary_sentence_0"
},
{
"operation": "Highlighting",
"time": "74.2",
"value": "Daniel Caspary_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "77.834",
"value": "Daniel Caspary_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "97.983",
"value": "Daniel Caspary_cell_0_5_1"
},
{
"operation": "Highlighting deleted",
"time": "97.989",
"value": "Daniel Caspary_header_cell_0_5_0"
},
{
"operation": "Highlighting deleted",
"time": "98.471",
"value": "Daniel Caspary_cell_0_4_1"
},
{
"operation": "Highlighting deleted",
"time": "98.477",
"value": "Daniel Caspary_header_cell_0_4_0"
},
{
"operation": "Highlighting deleted",
"time": "98.478",
"value": "Daniel Caspary_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "101.577",
"value": "Daniel Caspary_sentence_0"
},
{
"operation": "finish",
"time": "107.044",
"value": "finish"
}
]
}
|
7796
|
Raymonde Delaunois was a female singer, performed outside Paris, was the first to sing the Songs of Bilitis in Brussels and made tours in Continental Europe.
|
{
"evidence": [
"Raymonde Delaunois_sentence_0",
"Raymonde Delaunois_sentence_10",
"Raymonde Delaunois_sentence_9",
"Raymonde Delaunois_sentence_11"
],
"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:\nRaymonde Delaunois was a female singer, performed outside Paris, was the first to sing the Songs of Bilitis in Brussels and made tours in Continental Europe.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Raymonde Delaunois_sentence_0",
"Raymonde Delaunois_sentence_10",
"Raymonde Delaunois_sentence_9",
"Raymonde Delaunois_sentence_11"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1558,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.713",
"value": "?search="
},
{
"operation": "search",
"time": "15.279",
"value": "Raymonde Delaunois"
},
{
"operation": "Now on",
"time": "16.631",
"value": "Raymonde Delaunois"
},
{
"operation": "Highlighting",
"time": "32.389",
"value": "Raymonde Delaunois_sentence_0"
},
{
"operation": "Highlighting",
"time": "43.644",
"value": "Raymonde Delaunois_sentence_10"
},
{
"operation": "Highlighting",
"time": "45.008",
"value": "Raymonde Delaunois_sentence_9"
},
{
"operation": "Highlighting",
"time": "51.388",
"value": "Raymonde Delaunois_sentence_11"
},
{
"operation": "finish",
"time": "59.771",
"value": "finish"
}
]
}
|
exfever_5859
|
Sikhism originated in the Panjāb, a geopolitical, cultural, and historical region in South Asia.
|
{
"evidence": [
"Sikhism",
"Punjab"
],
"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:\nSikhism originated in the Panjāb, a geopolitical, cultural, and historical region in South Asia.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sikhism",
"Punjab"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1559,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8425
|
Dinosaurs' dominance continued throughout the Jurassic and a geological period that lasted from about 145 to 66 million years ago (Mya), with its name derived from a soft, white, porous, sedimentary carbonate rock.
|
{
"evidence": [
"Dinosaur",
"Cretaceous",
"Chalk"
],
"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:\nDinosaurs' dominance continued throughout the Jurassic and a geological period that lasted from about 145 to 66 million years ago (Mya), with its name derived from a soft, white, porous, sedimentary carbonate rock.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dinosaur",
"Cretaceous",
"Chalk"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1560,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1006
|
Keanu Charles Reeves is a Canadian actor whose feature film debut is in a 1980 American drama sports film directed, co-produced, and co-written by Peter Markle.
|
{
"evidence": [
"Keanu Reeves",
"Youngblood (1986 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:\nKeanu Charles Reeves is a Canadian actor whose feature film debut is in a 1980 American drama sports film directed, co-produced, and co-written by Peter Markle.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Keanu Reeves",
"Youngblood (1986 film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1561,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3974
|
Atlantic City is a 1980 romantic crime film directed by a French film director, screenwriter, and producer.
|
{
"evidence": [
"Atlantic City (1980 film)",
"Louis Malle"
],
"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:\nAtlantic City is a 1980 romantic crime film directed by a French film director, screenwriter, and producer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Atlantic City (1980 film)",
"Louis Malle"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1562,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3104
|
The Lonesome Dove miniseries stars an ensemble cast headed by an American actor and filmmaker.
|
{
"evidence": [
"Lonesome Dove (miniseries)",
"Robert Duvall"
],
"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 Lonesome Dove miniseries stars an ensemble cast headed by an American actor and filmmaker.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lonesome Dove (miniseries)",
"Robert Duvall"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1563,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8802
|
"Little Boy" was the codename for the type of atomic bomb dropped by the Boeing B-29 piloted by the commander of a unit of the United States Army Air Forces created during World War II.
|
{
"evidence": [
"Little Boy",
"509th Composite Group"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n\"Little Boy\" was the codename for the type of atomic bomb dropped by the Boeing B-29 piloted by the commander of a unit of the United States Army Air Forces created during World War II.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Little Boy",
"509th Composite Group"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1564,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
80134
|
Daisuke Matsuzaka, who was nicknamed "The Monster", set a still-standing record of 60 strikeouts in National High School Baseball Invitational Tournament in the spring of 1973.
|
{
"evidence": [
"Daisuke Matsuzaka_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:\nDaisuke Matsuzaka, who was nicknamed \"The Monster\", set a still-standing record of 60 strikeouts in National High School Baseball Invitational Tournament in the spring of 1973.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Daisuke Matsuzaka_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1565,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.007",
"value": "?search="
},
{
"operation": "search",
"time": "27.841",
"value": "National High School Baseball Invitational Tournament"
},
{
"operation": "Now on",
"time": "29.164",
"value": "National High School Baseball Invitational Tournament"
},
{
"operation": "search",
"time": "80.087",
"value": "Daisuke Matsuzaka"
},
{
"operation": "Now on",
"time": "82.367",
"value": "Daisuke Matsuzaka"
},
{
"operation": "Highlighting",
"time": "248.243",
"value": "Daisuke Matsuzaka_sentence_0"
},
{
"operation": "finish",
"time": "286.958",
"value": "finish"
}
]
}
|
exfever_5314
|
Zayn Malik is an Indian filmmaker and television personality who is the son of an Indian film producer and the founder of Dharma Productions.
|
{
"evidence": [
"Karan Johar",
"Yash Johar"
],
"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:\nZayn Malik is an Indian filmmaker and television personality who is the son of an Indian film producer and the founder of Dharma Productions.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Karan Johar",
"Yash Johar"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1566,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2599
|
Leslie Howard was an English director and producer who had a role in an American film directed by Archie Mayo.
|
{
"evidence": [
"Leslie Howard (actor)",
"The Petrified Forest"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLeslie Howard was an English director and producer who had a role in an American film directed by Archie Mayo.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Leslie Howard (actor)",
"The Petrified Forest"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1567,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1608
|
The Road to Escondido is the last recorded album by John Weldon "J. J." Cale.
|
{
"evidence": [
"J. J. Cale",
"Eric Clapton"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Road to Escondido is the last recorded album by John Weldon \"J. J.\" Cale.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"J. J. Cale",
"Eric Clapton"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1568,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2666
|
Realism is a work of art that portrays an image or appearance that looks real to the naked eye.
|
{
"evidence": [
"Walt Whitman",
"Transcendentalism"
],
"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:\nRealism is a work of art that portrays an image or appearance that looks real to the naked eye.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Walt Whitman",
"Transcendentalism"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1569,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
95627
|
Stefan Pfannmöller who is now 48 years old, won 3 medals while on the K1 team in three European canoeing championships and came in first place in 2003 on the World Cup individual podiums.
|
{
"evidence": [
"Stefan Pfannmöller_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:\nStefan Pfannmöller who is now 48 years old, won 3 medals while on the K1 team in three European canoeing championships and came in first place in 2003 on the World Cup individual podiums.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Stefan Pfannmöller_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1570,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.87",
"value": "?search="
},
{
"operation": "search",
"time": "11.514",
"value": "Stefan Pfannmöller"
},
{
"operation": "Now on",
"time": "12.42",
"value": "Stefan Pfannmöller"
},
{
"operation": "Highlighting",
"time": "28.442",
"value": "Stefan Pfannmöller_sentence_0"
},
{
"operation": "finish",
"time": "162.546",
"value": "finish"
}
]
}
|
exfever_815
|
Logan is a 2017 American superhero film that takes inspiration from the "Old Man Logan" comic book co-storyline by a Canadian comic book artist who first gained prominence on CrossGen's Meridian.
|
{
"evidence": [
"Logan (film)",
"Steve McNiven"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLogan is a 2017 American superhero film that takes inspiration from the \"Old Man Logan\" comic book co-storyline by a Canadian comic book artist who first gained prominence on CrossGen's Meridian.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Logan (film)",
"Steve McNiven"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1571,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8130
|
I Believe is the first song written by Ervin Drake.
|
{
"evidence": [
"I Believe (Frankie Laine song)",
"Popular music"
],
"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:\nI Believe is the first song written by Ervin Drake.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"I Believe (Frankie Laine song)",
"Popular music"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1572,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_367
|
Louis T. Leonowens was the son of Anna Leonowens, famous as the English teacher hired by the king in the film The King and I, based on the 1946 drama film directed by John Cromwell.
|
{
"evidence": [
"Louis T. Leonowens",
"Mongkut",
"Anna and the King of Siam (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:\nLouis T. Leonowens was the son of Anna Leonowens, famous as the English teacher hired by the king in the film The King and I, based on the 1946 drama film directed by John Cromwell.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Louis T. Leonowens",
"Mongkut",
"Anna and the King of Siam (film)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1573,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4486
|
Doctor Who was based on the real-life brother of Freema Agyeman.
|
{
"evidence": [
"Donna Noble",
"Catherine Tate"
],
"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 was based on the real-life brother of Freema Agyeman.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Donna Noble",
"Catherine Tate"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1574,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
32466
|
Walter Calverley murdered Winona and Winnie, his two daughters, then refused to plead and was therefore pressed to death.
|
{
"evidence": [
"Walter Calverley_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:\nWalter Calverley murdered Winona and Winnie, his two daughters, then refused to plead and was therefore pressed to death.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Walter Calverley_sentence_14"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1575,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.768",
"value": "?search="
},
{
"operation": "search",
"time": "12.915",
"value": "Walter Calverley"
},
{
"operation": "Now on",
"time": "14.029",
"value": "Walter Calverley"
},
{
"operation": "Highlighting",
"time": "17.394",
"value": "Walter Calverley_sentence_1"
},
{
"operation": "Highlighting",
"time": "40.147",
"value": "Walter Calverley_sentence_14"
},
{
"operation": "Highlighting deleted",
"time": "93.795",
"value": "Walter Calverley_sentence_1"
},
{
"operation": "finish",
"time": "111.916",
"value": "finish"
}
]
}
|
43380
|
The Boca Museum of Art, which offers free admission once a month, offers educational programs, artist lectures, films, classes for children, and events and sees more than 200,000 patrons annually, making it a major cultural institution in Boca Raton and the surrounding area.
|
{
"evidence": [
"Boca Raton Museum of Art_sentence_7",
"Boca Raton Museum of Art_sentence_4",
"Boca Raton Museum of Art_sentence_5"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Boca Museum of Art, which offers free admission once a month, offers educational programs, artist lectures, films, classes for children, and events and sees more than 200,000 patrons annually, making it a major cultural institution in Boca Raton and the surrounding area.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Boca Raton Museum of Art_sentence_7",
"Boca Raton Museum of Art_sentence_4",
"Boca Raton Museum of Art_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1576,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.131",
"value": "?search="
},
{
"operation": "search",
"time": "20.686",
"value": "Contains... Boca Museum of Art"
},
{
"operation": "Now on",
"time": "22.096",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "75.333",
"value": "Boca Raton Museum of Art"
},
{
"operation": "Now on",
"time": "76.051",
"value": "Boca Raton Museum of Art"
},
{
"operation": "Page search",
"time": "90.835",
"value": "free"
},
{
"operation": "Highlighting",
"time": "95.231",
"value": "Boca Raton Museum of Art_sentence_7"
},
{
"operation": "Page search",
"time": "104.723",
"value": "children"
},
{
"operation": "Highlighting",
"time": "109.709",
"value": "Boca Raton Museum of Art_sentence_4"
},
{
"operation": "Highlighting",
"time": "120.686",
"value": "Boca Raton Museum of Art_sentence_5"
},
{
"operation": "finish",
"time": "130.986",
"value": "finish"
}
]
}
|
exfever_6773
|
Denis Colin Leary is an American actor who began taking roles in films starting in the early 1990s, including a 1993 American action crime thriller film.
|
{
"evidence": [
"Denis Leary",
"Judgment Night (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:\nDenis Colin Leary is an American actor who began taking roles in films starting in the early 1990s, including a 1993 American action crime thriller film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Denis Leary",
"Judgment Night (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1577,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
37216
|
Paul Troast, a wealthy building contractor from Passaic, New Jersey won the republican primary for the 1953 New Jersey gubernatorial with Malcolm Forbes coming in behind him.
|
{
"evidence": [
"Paul L. Troast_sentence_0",
"Paul L. Troast_sentence_20",
"Paul L. Troast_sentence_25"
],
"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:\nPaul Troast, a wealthy building contractor from Passaic, New Jersey won the republican primary for the 1953 New Jersey gubernatorial with Malcolm Forbes coming in behind him.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Paul L. Troast_sentence_0",
"Paul L. Troast_sentence_20",
"Paul L. Troast_sentence_25"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1578,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.105",
"value": "?search="
},
{
"operation": "search",
"time": "11.728",
"value": "Paul Troast"
},
{
"operation": "Now on",
"time": "12.427",
"value": "Paul L. Troast"
},
{
"operation": "Highlighting",
"time": "16.3",
"value": "Paul L. Troast_sentence_0"
},
{
"operation": "Highlighting",
"time": "59.809",
"value": "Paul L. Troast_sentence_20"
},
{
"operation": "hyperlink",
"time": "74.264",
"value": "Governor of New Jersey"
},
{
"operation": "Now on",
"time": "74.92",
"value": "Governor of New Jersey"
},
{
"operation": "back-button-clicked",
"time": "79.404",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "79.914",
"value": "Paul L. Troast"
},
{
"operation": "Highlighting",
"time": "100.873",
"value": "Paul L. Troast_sentence_25"
},
{
"operation": "finish",
"time": "115.556",
"value": "finish"
}
]
}
|
30410
|
Pratt was born in Boston, taught music and drama at Virginia and directed the Virginia Glee Club which was founded in 1871.
|
{
"evidence": [
"Harry Rogers Pratt_sentence_2",
"Harry Rogers Pratt_sentence_3",
"Harry Rogers Pratt_sentence_4",
"Virginia Glee Club_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:\nPratt was born in Boston, taught music and drama at Virginia and directed the Virginia Glee Club which was founded in 1871.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Harry Rogers Pratt_sentence_2",
"Harry Rogers Pratt_sentence_3",
"Harry Rogers Pratt_sentence_4",
"Virginia Glee Club_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1579,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.625",
"value": "?search="
},
{
"operation": "search",
"time": "20.167",
"value": "Contains...Pratt was born in Boston, taught music and drama at Virginia"
},
{
"operation": "Now on",
"time": "22.012",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "26.312",
"value": "Harry Rogers Pratt"
},
{
"operation": "Now on",
"time": "26.807",
"value": "Harry Rogers Pratt"
},
{
"operation": "Highlighting",
"time": "34.86",
"value": "Harry Rogers Pratt_sentence_2"
},
{
"operation": "Highlighting",
"time": "35.836",
"value": "Harry Rogers Pratt_sentence_3"
},
{
"operation": "Highlighting",
"time": "41.732",
"value": "Harry Rogers Pratt_sentence_4"
},
{
"operation": "hyperlink",
"time": "50.297",
"value": "Virginia Glee Club"
},
{
"operation": "Now on",
"time": "50.943",
"value": "Virginia Glee Club"
},
{
"operation": "Highlighting",
"time": "52.368",
"value": "Virginia Glee Club_sentence_2"
},
{
"operation": "finish",
"time": "91.982",
"value": "finish"
}
]
}
|
39218
|
Beijing has more municipal (public university administered by the municipality) than any other type of universities and colleges.
|
{
"evidence": [
"Peking University_sentence_2"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBeijing has more municipal (public university administered by the municipality) than any other type of universities and colleges.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Peking University_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1580,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.866",
"value": "?search="
},
{
"operation": "search",
"time": "16.29",
"value": "Contains...beijing universities"
},
{
"operation": "Now on",
"time": "18.839",
"value": "undefined"
},
{
"operation": "search",
"time": "61.115",
"value": "beijing"
},
{
"operation": "Now on",
"time": "66.339",
"value": "Beijing"
},
{
"operation": "hyperlink",
"time": "71.514",
"value": "Beijing (disambiguation)"
},
{
"operation": "Now on",
"time": "73.031",
"value": "Beijing (disambiguation)"
},
{
"operation": "Page search",
"time": "78.556",
"value": "universi"
},
{
"operation": "Page search",
"time": "78.722",
"value": "universi"
},
{
"operation": "search",
"time": "140.07",
"value": "Beijing has more municipal than any other type of universities"
},
{
"operation": "Now on",
"time": "142.204",
"value": "undefined"
},
{
"operation": "search",
"time": "161.303",
"value": "Beijing University"
},
{
"operation": "Now on",
"time": "163.051",
"value": "Beijing University"
},
{
"operation": "hyperlink",
"time": "168.157",
"value": "Beida (disambiguation)"
},
{
"operation": "Now on",
"time": "169.243",
"value": "Beida (disambiguation)"
},
{
"operation": "hyperlink",
"time": "176.355",
"value": "Peking University"
},
{
"operation": "Now on",
"time": "177.366",
"value": "Peking University"
},
{
"operation": "Page search",
"time": "194.179",
"value": "municipal"
},
{
"operation": "Page search",
"time": "194.36",
"value": "municipal"
},
{
"operation": "Page search",
"time": "194.502",
"value": "municipal"
},
{
"operation": "search",
"time": "247.016",
"value": "Contains...beijing municipal"
},
{
"operation": "Now on",
"time": "248.367",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "259.606",
"value": "Beijing College Student Film Festival"
},
{
"operation": "Now on",
"time": "260.722",
"value": "Beijing College Student Film Festival"
},
{
"operation": "search",
"time": "281.765",
"value": "beijing municipal colleges"
},
{
"operation": "Now on",
"time": "282.906",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "293.43",
"value": "Peking University"
},
{
"operation": "Now on",
"time": "294.35",
"value": "Peking University"
},
{
"operation": "Highlighting",
"time": "304.92",
"value": "Peking University_sentence_2"
},
{
"operation": "finish",
"time": "310.63",
"value": "finish"
}
]
}
|
exfever_3862
|
Hilda Roderick Ellis Davidson was the 34th English scholar at the University of Cambridge.
|
{
"evidence": [
"Hilda Ellis Davidson",
"Anglo-Saxon paganism"
],
"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:\nHilda Roderick Ellis Davidson was the 34th English scholar at the University of Cambridge.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hilda Ellis Davidson",
"Anglo-Saxon paganism"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1581,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
61498
|
The acoustic receivers pick up signals from the tagged animals and also allow for the estimation of parameters such as swimming speed and mortality.
|
{
"evidence": [
"Pacific Ocean Shelf Tracking Project_sentence_2",
"Pacific Ocean Shelf Tracking Project_sentence_3"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe acoustic receivers pick up signals from the tagged animals and also allow for the estimation of parameters such as swimming speed and mortality.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Pacific Ocean Shelf Tracking Project_sentence_2",
"Pacific Ocean Shelf Tracking Project_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1582,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.201",
"value": "?search="
},
{
"operation": "search",
"time": "83.335",
"value": "Pacific Ocean Shelf Tracking Project"
},
{
"operation": "Now on",
"time": "84.245",
"value": "Pacific Ocean Shelf Tracking Project"
},
{
"operation": "Page search",
"time": "89.422",
"value": "track"
},
{
"operation": "Page search",
"time": "94.732",
"value": "tagged"
},
{
"operation": "Highlighting",
"time": "98.546",
"value": "Pacific Ocean Shelf Tracking Project_sentence_2"
},
{
"operation": "Highlighting",
"time": "100.912",
"value": "Pacific Ocean Shelf Tracking Project_sentence_3"
},
{
"operation": "finish",
"time": "105.644",
"value": "finish"
}
]
}
|
exfever_3277
|
Susan Abigail Sarandon is a British actress and activist who is the recipient of numerous accolades, including Oscars, an award given annually by a professional honorary organization with the stated goal of advancing the arts and sciences of motion pictures.
|
{
"evidence": [
"Susan Sarandon",
"Academy Awards",
"Academy of Motion Picture Arts and Sciences"
],
"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:\nSusan Abigail Sarandon is a British actress and activist who is the recipient of numerous accolades, including Oscars, an award given annually by a professional honorary organization with the stated goal of advancing the arts and sciences of motion pictures.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Susan Sarandon",
"Academy Awards",
"Academy of Motion Picture Arts and Sciences"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1583,
"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.