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 |
|---|---|---|---|---|---|---|---|---|
39185
|
Suzette Charles was the first runner-up, Williams' reign as Miss America was not without its challenges and she was a target of death threats.
|
{
"evidence": [
"Miss America 1984_sentence_8",
"Miss America 1984_sentence_10",
"Miss America 1984_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:\nSuzette Charles was the first runner-up, Williams' reign as Miss America was not without its challenges and she was a target of death threats.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Miss America 1984_sentence_8",
"Miss America 1984_sentence_10",
"Miss America 1984_sentence_11"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1184,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.372",
"value": "undefined"
},
{
"operation": "search",
"time": "34.803",
"value": "Suzette Charles"
},
{
"operation": "Now on",
"time": "36.274",
"value": "Suzette Charles"
},
{
"operation": "hyperlink",
"time": "61.398",
"value": "Miss America 1984"
},
{
"operation": "Now on",
"time": "62.61",
"value": "Miss America 1984"
},
{
"operation": "Highlighting",
"time": "91.571",
"value": "Miss America 1984_sentence_8"
},
{
"operation": "Highlighting",
"time": "99.157",
"value": "Miss America 1984_sentence_10"
},
{
"operation": "Highlighting",
"time": "101.828",
"value": "Miss America 1984_sentence_11"
},
{
"operation": "finish",
"time": "138.73",
"value": "finish"
}
]
}
|
exfever_5638
|
Dilwale Dulhania Le Jayenge is a 1995 Indian Hindi-language musical romance film written and directed by an Indian filmmaker who works in Hindi cinema.
|
{
"evidence": [
"Dilwale Dulhania Le Jayenge",
"Aditya Chopra"
],
"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:\nDilwale Dulhania Le Jayenge is a 1995 Indian Hindi-language musical romance film written and directed by an Indian filmmaker who works in Hindi cinema.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dilwale Dulhania Le Jayenge",
"Aditya Chopra"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1185,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
56688
|
The role of Rebecca in I Think I Love My Wife, a 2007 American romantic comedy film starring Kerry Washington, Chris Rock and Gina Torres, was played by Kimberly Hébert Gregory.
|
{
"evidence": [
"I Think I Love My Wife_sentence_0",
"Kimberly Hébert Gregory_sentence_0",
"Kimberly Hébert Gregory_sentence_10"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe role of Rebecca in I Think I Love My Wife, a 2007 American romantic comedy film starring Kerry Washington, Chris Rock and Gina Torres, was played by Kimberly Hébert Gregory.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"I Think I Love My Wife_sentence_0",
"Kimberly Hébert Gregory_sentence_0",
"Kimberly Hébert Gregory_sentence_10"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1186,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.56",
"value": "?search="
},
{
"operation": "search",
"time": "19.04",
"value": "I Think I Love My Wife"
},
{
"operation": "Now on",
"time": "20.646",
"value": "I Think I Love My Wife"
},
{
"operation": "Highlighting",
"time": "24.91",
"value": "I Think I Love My Wife_sentence_0"
},
{
"operation": "search",
"time": "115.457",
"value": "Contains...Kimberly Hébert Gregory."
},
{
"operation": "Now on",
"time": "118.574",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "121.797",
"value": "Kimberly Hébert Gregory"
},
{
"operation": "Now on",
"time": "123.883",
"value": "Kimberly Hébert Gregory"
},
{
"operation": "Highlighting",
"time": "125.575",
"value": "Kimberly Hébert Gregory_sentence_0"
},
{
"operation": "Highlighting",
"time": "141.316",
"value": "Kimberly Hébert Gregory_sentence_10"
},
{
"operation": "hyperlink",
"time": "149.979",
"value": "I Think I Love My Wife"
},
{
"operation": "Now on",
"time": "151.033",
"value": "I Think I Love My Wife"
},
{
"operation": "finish",
"time": "209.072",
"value": "finish"
}
]
}
|
91036
|
In 2011, Sumlin coached Houston to a 12–0 start before losing the Conference USA Championship Game to the Southern Miss Golden Eagles; then on December 10, 2011, Sumlin told his players he was never leaving Houston.
|
{
"evidence": [
"Kevin Sumlin_sentence_16"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn 2011, Sumlin coached Houston to a 12–0 start before losing the Conference USA Championship Game to the Southern Miss Golden Eagles; then on December 10, 2011, Sumlin told his players he was never leaving Houston.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Kevin Sumlin_sentence_16"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1187,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.542",
"value": "?search="
},
{
"operation": "search",
"time": "59.961",
"value": "Sumlin"
},
{
"operation": "Now on",
"time": "60.976",
"value": "Sumlin"
},
{
"operation": "hyperlink",
"time": "63.781",
"value": "Kevin Sumlin"
},
{
"operation": "Now on",
"time": "64.233",
"value": "Kevin Sumlin"
},
{
"operation": "Page search",
"time": "75.485",
"value": "coached"
},
{
"operation": "Highlighting",
"time": "79.898",
"value": "Kevin Sumlin_sentence_15"
},
{
"operation": "Highlighting",
"time": "88.551",
"value": "Kevin Sumlin_sentence_16"
},
{
"operation": "Highlighting deleted",
"time": "91.209",
"value": "Kevin Sumlin_sentence_15"
},
{
"operation": "finish",
"time": "103.887",
"value": "finish"
}
]
}
|
84878
|
Tbilisi Aircraft Manufacturing (TAM), also known as JSC Tbilaviamsheni, is a Georgian aerospace development and manufacturing company, which was the only manufacturer of Sukhoi SU-25, which took its maiden voyage from the runway of Tbilisi State Association.
|
{
"evidence": [
"Tbilisi Aircraft Manufacturing_sentence_2",
"Tbilisi Aircraft Manufacturing_sentence_17",
"Tbilisi Aircraft Manufacturing_sentence_15"
],
"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:\nTbilisi Aircraft Manufacturing (TAM), also known as JSC Tbilaviamsheni, is a Georgian aerospace development and manufacturing company, which was the only manufacturer of Sukhoi SU-25, which took its maiden voyage from the runway of Tbilisi State Association.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Tbilisi Aircraft Manufacturing_sentence_2",
"Tbilisi Aircraft Manufacturing_sentence_17",
"Tbilisi Aircraft Manufacturing_sentence_15"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1188,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.379",
"value": "?search="
},
{
"operation": "search",
"time": "136.877",
"value": "Tbilisi Aircraft Manufacturing"
},
{
"operation": "Now on",
"time": "137.821",
"value": "Tbilisi Aircraft Manufacturing"
},
{
"operation": "Highlighting",
"time": "141.55",
"value": "Tbilisi Aircraft Manufacturing_sentence_2"
},
{
"operation": "Page search",
"time": "164.729",
"value": "Sukhoi"
},
{
"operation": "Page search",
"time": "181.232",
"value": "only"
},
{
"operation": "Highlighting",
"time": "184.497",
"value": "Tbilisi Aircraft Manufacturing_sentence_17"
},
{
"operation": "Page search",
"time": "212.544",
"value": "Tbilisi State Association"
},
{
"operation": "Highlighting",
"time": "219.797",
"value": "Tbilisi Aircraft Manufacturing_sentence_15"
},
{
"operation": "finish",
"time": "226.067",
"value": "finish"
}
]
}
|
exfever_5125
|
The Royal Tenenbaums is a 2005 American comedy-drama film directed by an American filmmaker.
|
{
"evidence": [
"The Royal Tenenbaums",
"Wes Anderson"
],
"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 Royal Tenenbaums is a 2005 American comedy-drama film directed by an American filmmaker.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Royal Tenenbaums",
"Wes Anderson"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1189,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6476
|
The Reader was the third film directed by Stephen Daldry.
|
{
"evidence": [
"The Reader (2008 film)",
"Kate Winslet",
"Grammy Award"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Reader was the third film directed by Stephen Daldry.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Reader (2008 film)",
"Kate Winslet",
"Grammy Award"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1190,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7985
|
Tall Story is a 1950 American romantic comedy film starring an American actor, director, and singer.
|
{
"evidence": [
"Tall Story",
"Anthony Perkins"
],
"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:\nTall Story is a 1950 American romantic comedy film starring an American actor, director, and singer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tall Story",
"Anthony Perkins"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1191,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
85022
|
The King of the Klondike comic hero Scrooge McDuck is a cartoon character created for Nickolodeon.
|
{
"evidence": [
"Scrooge McDuck_sentence_1"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe King of the Klondike comic hero Scrooge McDuck is a cartoon character created for Nickolodeon.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Scrooge McDuck_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1192,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.717",
"value": "?search="
},
{
"operation": "search",
"time": "14.655",
"value": "The King of the Klondike"
},
{
"operation": "Now on",
"time": "15.173",
"value": "The King of the Klondike"
},
{
"operation": "Highlighting",
"time": "27.885",
"value": "The King of the Klondike_cell_0_5_1"
},
{
"operation": "Highlighting",
"time": "27.886",
"value": "The King of the Klondike_title"
},
{
"operation": "Highlighting",
"time": "27.888",
"value": "The King of the Klondike_header_cell_0_5_0"
},
{
"operation": "Highlighting",
"time": "27.888",
"value": "The King of the Klondike_header_cell_0_0_0"
},
{
"operation": "hyperlink",
"time": "37.286",
"value": "Scrooge McDuck"
},
{
"operation": "Now on",
"time": "38.468",
"value": "Scrooge McDuck"
},
{
"operation": "Highlighting",
"time": "41.392",
"value": "Scrooge McDuck_sentence_1"
},
{
"operation": "finish",
"time": "161.156",
"value": "finish"
}
]
}
|
exfever_865
|
Temple of the Dog was an American rock supergroup that formed in Seattle, composed of talented musicians, including an American musician and songwriter that born on March 10, 1963.
|
{
"evidence": [
"Temple of the Dog",
"Jeff Ament"
],
"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:\nTemple of the Dog was an American rock supergroup that formed in Seattle, composed of talented musicians, including an American musician and songwriter that born on March 10, 1963.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Temple of the Dog",
"Jeff Ament"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1193,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5743
|
James Oliver Cromwell's debut film was "Babe".
|
{
"evidence": [
"James Cromwell",
"W. (film)",
"Biographical 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:\nJames Oliver Cromwell's debut film was \"Babe\".\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"James Cromwell",
"W. (film)",
"Biographical film"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1194,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4415
|
Kurt Steven Angle was named the greatest shoot wrestler ever and one of the top 15 college wrestlers of all time by the organization that currently governs freestyle wrestling and Greco-Roman wrestling in the United States.
|
{
"evidence": [
"Kurt Angle",
"USA Wrestling"
],
"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:\nKurt Steven Angle was named the greatest shoot wrestler ever and one of the top 15 college wrestlers of all time by the organization that currently governs freestyle wrestling and Greco-Roman wrestling in the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Kurt Angle",
"USA Wrestling"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1195,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
82265
|
The Mo was originally a house named after Morag, born 1881, daughter of Sir Thomas Digby Pigott a member of the Upcher family; and Richard Frederick Hayward, M.C., K.C., a maritime lawyer and farmer born on March 1879, was its last owner.
|
{
"evidence": [
"The Mo Sheringham Museum_sentence_5",
"The Mo Sheringham Museum_sentence_6"
],
"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 Mo was originally a house named after Morag, born 1881, daughter of Sir Thomas Digby Pigott a member of the Upcher family; and Richard Frederick Hayward, M.C., K.C., a maritime lawyer and farmer born on March 1879, was its last owner.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"The Mo Sheringham Museum_sentence_5",
"The Mo Sheringham Museum_sentence_6"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1196,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.636",
"value": "?search="
},
{
"operation": "search",
"time": "9.162",
"value": "The Mo"
},
{
"operation": "Now on",
"time": "9.624",
"value": "undefined"
},
{
"operation": "search",
"time": "28.369",
"value": "\"The Mo was originally a house named after Morag, born 1881\""
},
{
"operation": "Now on",
"time": "28.839",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "32.602",
"value": "The Mo Sheringham Museum"
},
{
"operation": "Now on",
"time": "33.004",
"value": "The Mo Sheringham Museum"
},
{
"operation": "Highlighting",
"time": "46.051",
"value": "The Mo Sheringham Museum_sentence_5"
},
{
"operation": "Highlighting",
"time": "88.749",
"value": "The Mo Sheringham Museum_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "118.215",
"value": "The Mo Sheringham Museum_sentence_6"
},
{
"operation": "Highlighting",
"time": "118.912",
"value": "The Mo Sheringham Museum_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "121.355",
"value": "The Mo Sheringham Museum_sentence_6"
},
{
"operation": "Highlighting",
"time": "122.813",
"value": "The Mo Sheringham Museum_sentence_6"
},
{
"operation": "search",
"time": "126.388",
"value": "Richard Frederick Hayward"
},
{
"operation": "Now on",
"time": "128.083",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "132.133",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "132.207",
"value": "The Mo Sheringham Museum"
},
{
"operation": "Highlighting",
"time": "183.415",
"value": "The Mo Sheringham Museum_sentence_4"
},
{
"operation": "Highlighting",
"time": "183.416",
"value": "The Mo Sheringham Museum_section_0"
},
{
"operation": "Highlighting deleted",
"time": "185.022",
"value": "The Mo Sheringham Museum_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "185.024",
"value": "The Mo Sheringham Museum_section_0"
},
{
"operation": "Highlighting",
"time": "185.357",
"value": "The Mo Sheringham Museum_sentence_4"
},
{
"operation": "Highlighting",
"time": "185.358",
"value": "The Mo Sheringham Museum_section_0"
},
{
"operation": "Highlighting deleted",
"time": "186.844",
"value": "The Mo Sheringham Museum_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "186.846",
"value": "The Mo Sheringham Museum_section_0"
},
{
"operation": "Highlighting deleted",
"time": "187.936",
"value": "The Mo Sheringham Museum_sentence_6"
},
{
"operation": "Highlighting",
"time": "191.063",
"value": "The Mo Sheringham Museum_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "192.64",
"value": "The Mo Sheringham Museum_sentence_6"
},
{
"operation": "Highlighting",
"time": "195.788",
"value": "The Mo Sheringham Museum_sentence_6"
},
{
"operation": "search",
"time": "279.798",
"value": "Richard Frederick Hayward, M.C., K.C., a maritime lawyer and farmer"
},
{
"operation": "Now on",
"time": "281.358",
"value": "undefined"
},
{
"operation": "search",
"time": "305.243",
"value": "Richard Frederick Hayward"
},
{
"operation": "Now on",
"time": "305.949",
"value": "undefined"
},
{
"operation": "finish",
"time": "379.596",
"value": "finish"
}
]
}
|
85491
|
In the leaf pouches of H. physophora lies the extrafloral nectaries, where ants can get the nectar that was created by the plant, it can be viewed as parasitism by the ants, but the plant benefits immensely from the presence of ants as well.
|
{
"evidence": [
"Allomerus decemarticulatus_sentence_24",
"Allomerus decemarticulatus_sentence_26",
"Allomerus decemarticulatus_sentence_27",
"Hirtella physophora_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:\nIn the leaf pouches of H. physophora lies the extrafloral nectaries, where ants can get the nectar that was created by the plant, it can be viewed as parasitism by the ants, but the plant benefits immensely from the presence of ants as well.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Allomerus decemarticulatus_sentence_24",
"Allomerus decemarticulatus_sentence_26",
"Allomerus decemarticulatus_sentence_27",
"Hirtella physophora_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1197,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "5.038",
"value": "undefined"
},
{
"operation": "search",
"time": "96.464",
"value": "H. physophora"
},
{
"operation": "Now on",
"time": "97.826",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "143.997",
"value": "Hirtella physophora"
},
{
"operation": "Now on",
"time": "145.425",
"value": "Hirtella physophora"
},
{
"operation": "hyperlink",
"time": "270.327",
"value": "Allomerus decemarticulatus"
},
{
"operation": "Now on",
"time": "271.131",
"value": "Allomerus decemarticulatus"
},
{
"operation": "Page search",
"time": "294.026",
"value": "nectar"
},
{
"operation": "Highlighting",
"time": "350.456",
"value": "Allomerus decemarticulatus_sentence_24"
},
{
"operation": "Page search",
"time": "553.507",
"value": "parasitism"
},
{
"operation": "Highlighting",
"time": "578.33",
"value": "Allomerus decemarticulatus_sentence_26"
},
{
"operation": "Highlighting",
"time": "597.353",
"value": "Allomerus decemarticulatus_sentence_27"
},
{
"operation": "back-button-clicked",
"time": "668.358",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "668.93",
"value": "Hirtella physophora"
},
{
"operation": "Highlighting",
"time": "759.28",
"value": "Hirtella physophora_sentence_0"
},
{
"operation": "Highlighting",
"time": "760.948",
"value": "Hirtella physophora_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "842.57",
"value": "Hirtella physophora_sentence_0"
},
{
"operation": "finish",
"time": "858.942",
"value": "finish"
}
]
}
|
41123
|
During the 2011 PGA Championship Steve Stricker set the early lead, Jerry Kelly finished with 65 and the pre-tournament favorite was Rory McIlroy.
|
{
"evidence": [
"2011 PGA Championship_sentence_45",
"2011 PGA Championship_sentence_46",
"2011 PGA Championship_sentence_47"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDuring the 2011 PGA Championship Steve Stricker set the early lead, Jerry Kelly finished with 65 and the pre-tournament favorite was Rory McIlroy.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"2011 PGA Championship_sentence_45",
"2011 PGA Championship_sentence_46",
"2011 PGA Championship_sentence_47"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1198,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.77",
"value": "?search="
},
{
"operation": "search",
"time": "11.22",
"value": "2011 PGA Championship"
},
{
"operation": "Now on",
"time": "12.59",
"value": "2011 PGA Championship"
},
{
"operation": "Highlighting",
"time": "126.061",
"value": "2011 PGA Championship_sentence_45"
},
{
"operation": "Highlighting",
"time": "138.036",
"value": "2011 PGA Championship_sentence_46"
},
{
"operation": "Highlighting",
"time": "138.956",
"value": "2011 PGA Championship_sentence_47"
},
{
"operation": "finish",
"time": "148.988",
"value": "finish"
}
]
}
|
exfever_8172
|
Tropic Thunder is a 2008 action-comedy film directed by an Australian actor, comedian, producer, director, and screenwriter with a 2013 American adventure comedy-drama film credited to him.
|
{
"evidence": [
"Tropic Thunder",
"Ben Stiller",
"The Secret Life of Walter Mitty (2013 film)"
],
"label": "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:\nTropic Thunder is a 2008 action-comedy film directed by an Australian actor, comedian, producer, director, and screenwriter with a 2013 American adventure comedy-drama film credited to him.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tropic Thunder",
"Ben Stiller",
"The Secret Life of Walter Mitty (2013 film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1199,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
5494
|
On Friday morning, February 23, 1945, four days after the Marines landed at Iwo Jima, Rosenthal hurried toward Suribachi, lugging along his bulky Speed Graphic camera when he heard that an American flag was being raised atop Mount Suribachi.
|
{
"evidence": [
"Joe Rosenthal_sentence_15",
"Joe Rosenthal_sentence_14"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nOn Friday morning, February 23, 1945, four days after the Marines landed at Iwo Jima, Rosenthal hurried toward Suribachi, lugging along his bulky Speed Graphic camera when he heard that an American flag was being raised atop Mount Suribachi.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Joe Rosenthal_sentence_15",
"Joe Rosenthal_sentence_14"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1200,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.401",
"value": "?search="
},
{
"operation": "search",
"time": "20.321",
"value": "Iwo Jima photograph"
},
{
"operation": "Now on",
"time": "21.307",
"value": "Raising the Flag on Iwo Jima"
},
{
"operation": "hyperlink",
"time": "89.191",
"value": "Joe Rosenthal"
},
{
"operation": "Now on",
"time": "89.688",
"value": "Joe Rosenthal"
},
{
"operation": "Highlighting",
"time": "103.602",
"value": "Joe Rosenthal_sentence_15"
},
{
"operation": "Highlighting",
"time": "106.628",
"value": "Joe Rosenthal_sentence_14"
},
{
"operation": "finish",
"time": "154.377",
"value": "finish"
}
]
}
|
exfever_7587
|
A drama school, stage school, or theatre school is an undergraduate and/or graduate school or department at a college or university where graduate students may take a discipline-based terminal doctoral degree that was originally conceived and designed to be an alternative to the traditional research-based Doctor of Philosophy.
|
{
"evidence": [
"Drama school",
"Doctor of Arts"
],
"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:\nA drama school, stage school, or theatre school is an undergraduate and/or graduate school or department at a college or university where graduate students may take a discipline-based terminal doctoral degree that was originally conceived and designed to be an alternative to the traditional research-based Doctor of Philosophy.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Drama school",
"Doctor of Arts"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1201,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
47103
|
The James Young High School is a secondary school named after James Young that was established in 1982.
|
{
"evidence": [
"The James Young High School_sentence_0",
"The James Young High School_sentence_1"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe James Young High School is a secondary school named after James Young that was established in 1982.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"The James Young High School_sentence_0",
"The James Young High School_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1202,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.809",
"value": "?search="
},
{
"operation": "search",
"time": "16.897",
"value": "James Young High School"
},
{
"operation": "Now on",
"time": "19.489",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "24.324",
"value": "The James Young High School"
},
{
"operation": "Now on",
"time": "25.053",
"value": "The James Young High School"
},
{
"operation": "Highlighting",
"time": "28.937",
"value": "The James Young High School_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "36.784",
"value": "The James Young High School_sentence_1"
},
{
"operation": "Highlighting",
"time": "37.234",
"value": "The James Young High School_sentence_0"
},
{
"operation": "Highlighting",
"time": "37.688",
"value": "The James Young High School_sentence_1"
},
{
"operation": "finish",
"time": "71.4",
"value": "finish"
}
]
}
|
16643
|
"Red Lights" was released as the lead single from A Town Called Paradise in 2013 and peaked at number six in Scotland; the second song released from the album was "Wasted" in 2014.
|
{
"evidence": [
"Red Lights (song)_sentence_32"
],
"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:\n\"Red Lights\" was released as the lead single from A Town Called Paradise in 2013 and peaked at number six in Scotland; the second song released from the album was \"Wasted\" in 2014.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Red Lights (song)_sentence_32"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1203,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.191",
"value": "?search="
},
{
"operation": "search",
"time": "18.821",
"value": "\"Red Lights\""
},
{
"operation": "Now on",
"time": "20.826",
"value": "Red light"
},
{
"operation": "search",
"time": "89.269",
"value": "A Town Called Paradise"
},
{
"operation": "Now on",
"time": "90.343",
"value": "A Town Called Paradise"
},
{
"operation": "Highlighting",
"time": "95.862",
"value": "A Town Called Paradise_sentence_1"
},
{
"operation": "Highlighting",
"time": "95.865",
"value": "A Town Called Paradise_title"
},
{
"operation": "hyperlink",
"time": "107.262",
"value": "Red Lights"
},
{
"operation": "Now on",
"time": "109.185",
"value": "Red Lights (song)"
},
{
"operation": "Highlighting",
"time": "123.147",
"value": "Red Lights (song)_sentence_2"
},
{
"operation": "Page search",
"time": "191.229",
"value": "peak"
},
{
"operation": "Page search",
"time": "229.488",
"value": "scotland"
},
{
"operation": "Highlighting",
"time": "237.376",
"value": "Red Lights (song)_sentence_5"
},
{
"operation": "Highlighting",
"time": "250.164",
"value": "Red Lights (song)_sentence_32"
},
{
"operation": "finish",
"time": "272.738",
"value": "finish"
}
]
}
|
exfever_4718
|
Myles Richard Bass, known professionally as Myles Kennedy, is an American dancer who was asked to join the American rock band from Orlando, formed in 2004 by an American musician and songwriter best known for his tenures with Creed.
|
{
"evidence": [
"Myles Kennedy",
"Mark Tremonti",
"Alter Bridge"
],
"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:\nMyles Richard Bass, known professionally as Myles Kennedy, is an American dancer who was asked to join the American rock band from Orlando, formed in 2004 by an American musician and songwriter best known for his tenures with Creed.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Myles Kennedy",
"Mark Tremonti",
"Alter Bridge"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1204,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
12315
|
Sweden's Jonas Sandqvist is the only player that made 26 appearances in Allsvenkan league, a Swedish professional league for men's association football clubs.
|
{
"evidence": [
"Jonas Sandqvist_sentence_0",
"Allsvenskan_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:\nSweden's Jonas Sandqvist is the only player that made 26 appearances in Allsvenkan league, a Swedish professional league for men's association football clubs.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Jonas Sandqvist_sentence_0",
"Allsvenskan_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1205,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.549",
"value": "?search="
},
{
"operation": "search",
"time": "18.073",
"value": "Contains...Allsvenkan league"
},
{
"operation": "Now on",
"time": "18.526",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "35.229",
"value": "allsvenskan league"
},
{
"operation": "Now on",
"time": "36.253",
"value": "undefined"
},
{
"operation": "search",
"time": "57.711",
"value": "Jonas Sandqvist"
},
{
"operation": "Now on",
"time": "58.401",
"value": "Jonas Sandqvist"
},
{
"operation": "Highlighting",
"time": "64.696",
"value": "Jonas Sandqvist_sentence_0"
},
{
"operation": "hyperlink",
"time": "97.682",
"value": "Allsvenskan"
},
{
"operation": "Now on",
"time": "98.552",
"value": "Allsvenskan"
},
{
"operation": "Highlighting",
"time": "112.451",
"value": "Allsvenskan_sentence_2"
},
{
"operation": "hyperlink",
"time": "156.752",
"value": "List of Allsvenskan players"
},
{
"operation": "Now on",
"time": "157.96",
"value": "List of Allsvenskan players"
},
{
"operation": "search",
"time": "195.321",
"value": "Contains...26 appearances"
},
{
"operation": "Now on",
"time": "196.956",
"value": "undefined"
},
{
"operation": "search",
"time": "224.933",
"value": "Sandqvist 26 appearances"
},
{
"operation": "Now on",
"time": "225.801",
"value": "undefined"
},
{
"operation": "finish",
"time": "252.834",
"value": "finish"
}
]
}
|
exfever_2954
|
10 Cloverfield Lane is a 2016 American science fiction film written by an American film director, producer, and screenwriter.
|
{
"evidence": [
"10 Cloverfield Lane",
"Damien Chazelle"
],
"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:\n10 Cloverfield Lane is a 2016 American science fiction film written by an American film director, producer, and screenwriter.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"10 Cloverfield Lane",
"Damien Chazelle"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1206,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
81601
|
The Show Grounds Greyhound Track was built to hold 20,000 people, with room to park 1,000 cars , and five prices levels tf rdifferent areas of the grandstand---much of the structure had already been in place from previous use ---and there was tram service to Ballintemple.
|
{
"evidence": [
"The Show Grounds Greyhound Track_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:\nThe Show Grounds Greyhound Track was built to hold 20,000 people, with room to park 1,000 cars , and five prices levels tf rdifferent areas of the grandstand---much of the structure had already been in place from previous use ---and there was tram service to Ballintemple.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"The Show Grounds Greyhound Track_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1207,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.416",
"value": "?search="
},
{
"operation": "search",
"time": "14.538",
"value": "Contains...Show Grounds Greyhound Track"
},
{
"operation": "Now on",
"time": "19.161",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "23.095",
"value": "The Show Grounds Greyhound Track"
},
{
"operation": "Now on",
"time": "23.949",
"value": "The Show Grounds Greyhound Track"
},
{
"operation": "Page search",
"time": "34.619",
"value": "Ballintemple"
},
{
"operation": "Highlighting",
"time": "50.275",
"value": "The Show Grounds Greyhound Track_sentence_5"
},
{
"operation": "Highlighting",
"time": "60.216",
"value": "The Show Grounds Greyhound Track_sentence_6"
},
{
"operation": "Highlighting",
"time": "106.409",
"value": "The Show Grounds Greyhound Track_sentence_0"
},
{
"operation": "Highlighting",
"time": "133.315",
"value": "The Show Grounds Greyhound Track_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "134.526",
"value": "The Show Grounds Greyhound Track_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "135.235",
"value": "The Show Grounds Greyhound Track_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "136.458",
"value": "The Show Grounds Greyhound Track_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "175.019",
"value": "The Show Grounds Greyhound Track_sentence_4"
},
{
"operation": "Highlighting",
"time": "176.587",
"value": "The Show Grounds Greyhound Track_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "185.492",
"value": "The Show Grounds Greyhound Track_sentence_4"
},
{
"operation": "Highlighting",
"time": "186.7",
"value": "The Show Grounds Greyhound Track_sentence_6"
},
{
"operation": "finish",
"time": "193.203",
"value": "finish"
}
]
}
|
4476
|
During the negation that was adopted in 2007, the Declaration on the Rights of Indigenous Peoples, Bill C-15 was introduced in 2020 which Samoa has since endorsed.
|
{
"evidence": [
"Declaration on the Rights of Indigenous Peoples_sentence_0",
"Declaration on the Rights of Indigenous Peoples_sentence_50",
"Declaration on the Rights of Indigenous Peoples_sentence_53"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDuring the negation that was adopted in 2007, the Declaration on the Rights of Indigenous Peoples, Bill C-15 was introduced in 2020 which Samoa has since endorsed.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Declaration on the Rights of Indigenous Peoples_sentence_0",
"Declaration on the Rights of Indigenous Peoples_sentence_50",
"Declaration on the Rights of Indigenous Peoples_sentence_53"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1208,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "18.985",
"value": "Declaration on the Rights of Indigenous Peoples"
},
{
"operation": "Highlighting",
"time": "103.312",
"value": "Declaration on the Rights of Indigenous Peoples_sentence_0"
},
{
"operation": "Page search",
"time": "119.011",
"value": "bill c-15"
},
{
"operation": "Highlighting",
"time": "129.085",
"value": "Declaration on the Rights of Indigenous Peoples_sentence_50"
},
{
"operation": "Highlighting",
"time": "140.775",
"value": "Declaration on the Rights of Indigenous Peoples_sentence_53"
},
{
"operation": "finish",
"time": "164.629",
"value": "finish"
}
]
}
|
83774
|
Like previous games, virtual avatars of famous solo singers, and the band No Doubt are presented in Band Hero, which received mixed reviews from journalists.
|
{
"evidence": [
"Band Hero_sentence_3",
"Band Hero_sentence_4"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLike previous games, virtual avatars of famous solo singers, and the band No Doubt are presented in Band Hero, which received mixed reviews from journalists.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Band Hero_sentence_3",
"Band Hero_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1209,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.112",
"value": "?search="
},
{
"operation": "search",
"time": "25.522",
"value": "Contains...No Doubt are presented in Band Hero"
},
{
"operation": "Now on",
"time": "28.125",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "30.625",
"value": "Band Hero"
},
{
"operation": "Now on",
"time": "32.393",
"value": "Band Hero"
},
{
"operation": "Page search",
"time": "35.552",
"value": "No Doubt are presented in Band Hero"
},
{
"operation": "Page search",
"time": "49.987",
"value": "No Doubt"
},
{
"operation": "Highlighting",
"time": "52.689",
"value": "Band Hero_sentence_3"
},
{
"operation": "Highlighting",
"time": "56.998",
"value": "Band Hero_sentence_4"
},
{
"operation": "finish",
"time": "62.865",
"value": "finish"
}
]
}
|
exfever_6594
|
Kick-Ass is a 2009 superhero film that tells the story of an ordinary teenager, which is a high school student and comic book fan whose dreams inspire him to become a real-life superhero dubbed "Kick-Ass".
|
{
"evidence": [
"Kick-Ass (film)",
"Kick-Ass (character)"
],
"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:\nKick-Ass is a 2009 superhero film that tells the story of an ordinary teenager, which is a high school student and comic book fan whose dreams inspire him to become a real-life superhero dubbed \"Kick-Ass\".\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Kick-Ass (film)",
"Kick-Ass (character)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1210,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4372
|
Grunge is an alternative rock genre and a group of people within a culture that differentiates itself from the parent culture to which it belongs.
|
{
"evidence": [
"Grunge",
"Subculture"
],
"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:\nGrunge is an alternative rock genre and a group of people within a culture that differentiates itself from the parent culture to which it belongs.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Grunge",
"Subculture"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1211,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
48308
|
The Šance Dam, constructed from 1964-1969 in the Moravian-Silesian Region of the Czech Republic, is built on the Elbe river and has a surface area of 3.37 km.
|
{
"evidence": [
"Å ance Dam_sentence_1"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Šance Dam, constructed from 1964-1969 in the Moravian-Silesian Region of the Czech Republic, is built on the Elbe river and has a surface area of 3.37 km.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Å ance Dam_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1212,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.705",
"value": "?search="
},
{
"operation": "search",
"time": "14.524",
"value": "Contains...The Šance Dam"
},
{
"operation": "Now on",
"time": "15.971",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "19.559",
"value": "Staré Hamry"
},
{
"operation": "Now on",
"time": "20.115",
"value": "Staré Hamry"
},
{
"operation": "hyperlink",
"time": "32.023",
"value": "Šance Dam"
},
{
"operation": "Now on",
"time": "32.892",
"value": "Å ance Dam"
},
{
"operation": "Highlighting",
"time": "39.677",
"value": "Å ance Dam_sentence_2"
},
{
"operation": "Highlighting",
"time": "47.063",
"value": "Å ance Dam_sentence_0"
},
{
"operation": "Highlighting",
"time": "50.258",
"value": "Å ance Dam_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "52.306",
"value": "Å ance Dam_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "53.059",
"value": "Å ance Dam_sentence_2"
},
{
"operation": "finish",
"time": "67.844",
"value": "finish"
}
]
}
|
72393
|
Nathan Freer is a rugby player known as the claw who played for the Sharks first then was promoted with York City Knights in 2000 the went to the Rovers and back to the Knights.
|
{
"evidence": [
"Nathan Freer_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:\nNathan Freer is a rugby player known as the claw who played for the Sharks first then was promoted with York City Knights in 2000 the went to the Rovers and back to the Knights.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Nathan Freer_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1213,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.328",
"value": "?search="
},
{
"operation": "search",
"time": "8.294",
"value": "Nathan Freer"
},
{
"operation": "Now on",
"time": "9.11",
"value": "Nathan Freer"
},
{
"operation": "Highlighting",
"time": "14.888",
"value": "Nathan Freer_sentence_0"
},
{
"operation": "Highlighting",
"time": "50.614",
"value": "Nathan Freer_sentence_1"
},
{
"operation": "Page search",
"time": "71.236",
"value": "2000"
},
{
"operation": "Page search",
"time": "100.078",
"value": "promoted"
},
{
"operation": "Highlighting deleted",
"time": "113.241",
"value": "Nathan Freer_sentence_1"
},
{
"operation": "Highlighting",
"time": "114.632",
"value": "Nathan Freer_sentence_1"
},
{
"operation": "Highlighting",
"time": "160.681",
"value": "Nathan Freer_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "161.761",
"value": "Nathan Freer_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "162.625",
"value": "Nathan Freer_sentence_0"
},
{
"operation": "finish",
"time": "180.946",
"value": "finish"
}
]
}
|
exfever_6006
|
The Hateful Eight is a 2015 film starring an Italian actor, often playing supporting villainous characters of unstable nature.
|
{
"evidence": [
"The Hateful Eight",
"Bruce Dern"
],
"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 Hateful Eight is a 2015 film starring an Italian actor, often playing supporting villainous characters of unstable nature.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Hateful Eight",
"Bruce Dern"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1214,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3646
|
True Detective is an American anthology crime drama television series starring an American actress best known for her starring roles in the film Kiss Kiss Bang Bang.
|
{
"evidence": [
"True Detective",
"Michelle Monaghan"
],
"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:\nTrue Detective is an American anthology crime drama television series starring an American actress best known for her starring roles in the film Kiss Kiss Bang Bang.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"True Detective",
"Michelle Monaghan"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1215,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5214
|
PageRank is an algorithm used by Google Search to rank web pages in their search engine results, which is named after its co-founder, an Australian computer scientist and internet entrepreneur.
|
{
"evidence": [
"PageRank",
"Larry Page"
],
"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:\nPageRank is an algorithm used by Google Search to rank web pages in their search engine results, which is named after its co-founder, an Australian computer scientist and internet entrepreneur.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"PageRank",
"Larry Page"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1216,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
82054
|
After Lionel Beaumont Thomas decided to retire on health grounds in 1933, Cartland was chosen to replace him in Herbert Austin's former constituency of King's Norton, Birmingham; then he won in the 1935 election, and became one of the youngest MPs in the Commons (neither the lower house nor the de facto primary chamber of the Parliament of the United Kingdom).
|
{
"evidence": [
"House of Commons of the United Kingdom_sentence_3"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAfter Lionel Beaumont Thomas decided to retire on health grounds in 1933, Cartland was chosen to replace him in Herbert Austin's former constituency of King's Norton, Birmingham; then he won in the 1935 election, and became one of the youngest MPs in the Commons (neither the lower house nor the de facto primary chamber of the Parliament of the United Kingdom).\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"House of Commons of the United Kingdom_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1217,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.672",
"value": "?search="
},
{
"operation": "search",
"time": "62.463",
"value": "Lionel Beaumont-Thomas"
},
{
"operation": "Now on",
"time": "64.404",
"value": "Lionel Beaumont-Thomas"
},
{
"operation": "hyperlink",
"time": "88.839",
"value": "Birmingham King's Norton"
},
{
"operation": "Now on",
"time": "92.125",
"value": "Birmingham King's Norton (UK Parliament constituency)"
},
{
"operation": "Highlighting",
"time": "130.5",
"value": "Birmingham King's Norton (UK Parliament constituency)_cell_1_5_2"
},
{
"operation": "Highlighting",
"time": "130.502",
"value": "Birmingham King's Norton (UK Parliament constituency)_title"
},
{
"operation": "Highlighting",
"time": "130.504",
"value": "Birmingham King's Norton (UK Parliament constituency)_section_1"
},
{
"operation": "Highlighting",
"time": "130.506",
"value": "Birmingham King's Norton (UK Parliament constituency)_header_cell_1_0_2"
},
{
"operation": "hyperlink",
"time": "134.039",
"value": "Ronald Cartland"
},
{
"operation": "Now on",
"time": "135.232",
"value": "Ronald Cartland"
},
{
"operation": "Highlighting",
"time": "176.832",
"value": "Ronald Cartland_sentence_1"
},
{
"operation": "Highlighting",
"time": "176.833",
"value": "Ronald Cartland_title"
},
{
"operation": "Highlighting",
"time": "301.063",
"value": "Ronald Cartland_sentence_14"
},
{
"operation": "Highlighting",
"time": "301.066",
"value": "Ronald Cartland_section_1"
},
{
"operation": "Highlighting",
"time": "327.96",
"value": "Ronald Cartland_sentence_16"
},
{
"operation": "Highlighting deleted",
"time": "348.848",
"value": "Ronald Cartland_sentence_1"
},
{
"operation": "hyperlink",
"time": "356.611",
"value": "Commons"
},
{
"operation": "Now on",
"time": "362.223",
"value": "House of Commons of the United Kingdom"
},
{
"operation": "Highlighting",
"time": "377.201",
"value": "House of Commons of the United Kingdom_sentence_3"
},
{
"operation": "back-button-clicked",
"time": "397.728",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "397.909",
"value": "Ronald Cartland"
},
{
"operation": "Highlighting deleted",
"time": "400.224",
"value": "Ronald Cartland_sentence_16"
},
{
"operation": "Highlighting deleted",
"time": "402.036",
"value": "Ronald Cartland_sentence_14"
},
{
"operation": "Highlighting deleted",
"time": "402.038",
"value": "Ronald Cartland_title"
},
{
"operation": "Highlighting deleted",
"time": "402.041",
"value": "Ronald Cartland_section_1"
},
{
"operation": "finish",
"time": "453.208",
"value": "finish"
}
]
}
|
exfever_3813
|
Orphan Black is a Canadian thriller series created by screenwriter Graeme Manson.
|
{
"evidence": [
"Tatiana Maslany",
"Critics' Choice Television Award for Best Actress in a Drama Series"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nOrphan Black is a Canadian thriller series created by screenwriter Graeme Manson.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tatiana Maslany",
"Critics' Choice Television Award for Best Actress in a Drama Series"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1218,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4825
|
Death Note is a series of movie programs that are a staple of television and are broadcast daily and produce a genre of film, television, video game, and theatre in Japan.
|
{
"evidence": [
"Death Note (2015 TV series)",
"Japanese television drama",
"Jidaigeki"
],
"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:\nDeath Note is a series of movie programs that are a staple of television and are broadcast daily and produce a genre of film, television, video game, and theatre in Japan.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Death Note (2015 TV series)",
"Japanese television drama",
"Jidaigeki"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1219,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
79520
|
The major highway in Bouza was completed in the late 1900s, while the major unpaved road in the area (RN16), runs through the town of Bouza from Madaoua to the south to Keita in the north.
|
{
"evidence": [
"Bouza_sentence_4",
"Bouza_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 major highway in Bouza was completed in the late 1900s, while the major unpaved road in the area (RN16), runs through the town of Bouza from Madaoua to the south to Keita in the north.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bouza_sentence_4",
"Bouza_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1220,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.053",
"value": "?search="
},
{
"operation": "search",
"time": "23.775",
"value": "Bouza"
},
{
"operation": "Now on",
"time": "25.206",
"value": "Bouza"
},
{
"operation": "Highlighting",
"time": "31.768",
"value": "Bouza_sentence_4"
},
{
"operation": "Highlighting",
"time": "38.439",
"value": "Bouza_sentence_5"
},
{
"operation": "finish",
"time": "69.268",
"value": "finish"
}
]
}
|
exfever_7427
|
Alicia Keys is the first artist in history to have sold over 131,000 albums worldwide.
|
{
"evidence": [
"Here (Alicia Keys album)",
"Jimmy Napes"
],
"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:\nAlicia Keys is the first artist in history to have sold over 131,000 albums worldwide.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Here (Alicia Keys album)",
"Jimmy Napes"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1221,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7429
|
Lil Wayne is a songwriter who began his career when he was signed by the co-founder of Virgin Records, who released his first album on November 26, 2002.
|
{
"evidence": [
"Lil Wayne",
"Birdman (rapper)",
"Birdman (album)"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLil Wayne is a songwriter who began his career when he was signed by the co-founder of Virgin Records, who released his first album on November 26, 2002.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lil Wayne",
"Birdman (rapper)",
"Birdman (album)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1222,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_867
|
Boston Legal is an American legal action-drama television series that is the direct spin-off and continuation of an American legal drama television series created by David E.
|
{
"evidence": [
"Boston Legal",
"The Practice"
],
"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:\nBoston Legal is an American legal action-drama television series that is the direct spin-off and continuation of an American legal drama television series created by David E.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Boston Legal",
"The Practice"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1223,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
71031
|
Little Onn is a hamlet in Staffordshire, England one mile north of Wheaton Aston, a commuter village.
|
{
"evidence": [
"Little Onn_sentence_0",
"Wheaton Aston_sentence_4"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLittle Onn is a hamlet in Staffordshire, England one mile north of Wheaton Aston, a commuter village.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Little Onn_sentence_0",
"Wheaton Aston_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1224,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.246",
"value": "?search="
},
{
"operation": "search",
"time": "4.123",
"value": "Little Onn"
},
{
"operation": "Now on",
"time": "5.126",
"value": "Little Onn"
},
{
"operation": "Highlighting",
"time": "6.52",
"value": "Little Onn_sentence_0"
},
{
"operation": "hyperlink",
"time": "12.061",
"value": "Wheaton Aston"
},
{
"operation": "Now on",
"time": "12.858",
"value": "Wheaton Aston"
},
{
"operation": "Highlighting",
"time": "17.264",
"value": "Wheaton Aston_sentence_4"
},
{
"operation": "finish",
"time": "27.848",
"value": "finish"
}
]
}
|
exfever_981
|
Billy Ray Cyrus was the recipient of the Diamond Award for his singles in 1992.
|
{
"evidence": [
"Billy Ray Cyrus",
"Achy Breaky Heart",
"The Marcy Brothers"
],
"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:\nBilly Ray Cyrus was the recipient of the Diamond Award for his singles in 1992.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Billy Ray Cyrus",
"Achy Breaky Heart",
"The Marcy Brothers"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1225,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
49632
|
Mini-chromosome maintenance proteins such as MCM3 are in involved in genomic DNA replication.
|
{
"evidence": [
"Minichromosome maintenance_sentence_0",
"MCM3_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:\nMini-chromosome maintenance proteins such as MCM3 are in involved in genomic DNA replication.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Minichromosome maintenance_sentence_0",
"MCM3_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1226,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.249",
"value": "?search="
},
{
"operation": "search",
"time": "36.002",
"value": "Mini-chromosome"
},
{
"operation": "Now on",
"time": "37.062",
"value": "undefined"
},
{
"operation": "search",
"time": "48.424",
"value": "Mini-chromosome maintenance proteins such as MCM3 are in involved in genomic DNA replication"
},
{
"operation": "Now on",
"time": "50.085",
"value": "undefined"
},
{
"operation": "search",
"time": "72.292",
"value": "MCM3"
},
{
"operation": "Now on",
"time": "73.298",
"value": "MCM3"
},
{
"operation": "hyperlink",
"time": "80.88",
"value": "Mini Chromosome Maintenance"
},
{
"operation": "Now on",
"time": "84.647",
"value": "Mini Chromosome Maintenance"
},
{
"operation": "Page search",
"time": "92.197",
"value": "mcm3"
},
{
"operation": "Highlighting",
"time": "156.809",
"value": "Minichromosome maintenance_sentence_0"
},
{
"operation": "hyperlink",
"time": "172.149",
"value": "MCM3"
},
{
"operation": "Now on",
"time": "172.578",
"value": "MCM3"
},
{
"operation": "Highlighting",
"time": "183.026",
"value": "MCM3_sentence_1"
},
{
"operation": "finish",
"time": "187.723",
"value": "finish"
}
]
}
|
5980
|
Merryl Lawton represented the electoral district of Digby (existed from 1867 to 1993) and was a member of the Progressive conservative Party.
|
{
"evidence": [
"Merryl Lawton_sentence_1",
"Merryl Lawton_sentence_2",
"Digby (provincial electoral district)_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:\nMerryl Lawton represented the electoral district of Digby (existed from 1867 to 1993) and was a member of the Progressive conservative Party.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Merryl Lawton_sentence_1",
"Merryl Lawton_sentence_2",
"Digby (provincial electoral district)_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1227,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.391",
"value": "?search="
},
{
"operation": "search",
"time": "13.953",
"value": "Merryl Lawton"
},
{
"operation": "Now on",
"time": "14.455",
"value": "Merryl Lawton"
},
{
"operation": "Highlighting",
"time": "27.597",
"value": "Merryl Lawton_sentence_1"
},
{
"operation": "Highlighting",
"time": "28.283",
"value": "Merryl Lawton_sentence_2"
},
{
"operation": "hyperlink",
"time": "36.068",
"value": "Digby"
},
{
"operation": "Now on",
"time": "36.476",
"value": "Digby (provincial electoral district)"
},
{
"operation": "Highlighting",
"time": "38.779",
"value": "Digby (provincial electoral district)_sentence_1"
},
{
"operation": "finish",
"time": "58.152",
"value": "finish"
}
]
}
|
28193
|
Yoshiaki Numata lost the title in his first defense to Hiroshi Kobayashi by KO in the 12th after being down once in the 6th and three times in the 12th round.
|
{
"evidence": [
"Yoshiaki Numata_sentence_2",
"Yoshiaki Numata_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:\nYoshiaki Numata lost the title in his first defense to Hiroshi Kobayashi by KO in the 12th after being down once in the 6th and three times in the 12th round.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Yoshiaki Numata_sentence_2",
"Yoshiaki Numata_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1228,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.748",
"value": "?search="
},
{
"operation": "search",
"time": "12.109",
"value": "Yoshiaki Numata"
},
{
"operation": "Now on",
"time": "13.099",
"value": "Yoshiaki Numata"
},
{
"operation": "Highlighting",
"time": "29.133",
"value": "Yoshiaki Numata_sentence_2"
},
{
"operation": "Highlighting",
"time": "43.55",
"value": "Yoshiaki Numata_sentence_3"
},
{
"operation": "finish",
"time": "62.471",
"value": "finish"
}
]
}
|
47431
|
The rear cover of the album shows "Portrait U.S.A.", a satellite image of the United States produced by their Beltsville Photo Engineering Lab and created in 1976 by the General Electric Co. (an American multinational conglomerate incorporated in New York City).
|
{
"evidence": [
"More Songs About Buildings and Food_sentence_5",
"More Songs About Buildings and Food_sentence_6",
"General Electric_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 rear cover of the album shows \"Portrait U.S.A.\", a satellite image of the United States produced by their Beltsville Photo Engineering Lab and created in 1976 by the General Electric Co. (an American multinational conglomerate incorporated in New York City).\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"More Songs About Buildings and Food_sentence_5",
"More Songs About Buildings and Food_sentence_6",
"General Electric_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1229,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.635",
"value": "?search="
},
{
"operation": "search",
"time": "16.745",
"value": "rear cover of the album shows \"Portrait U.S.A.\""
},
{
"operation": "Now on",
"time": "18.35",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "21.971",
"value": "The Girls Want to Be With the Girls"
},
{
"operation": "Now on",
"time": "22.536",
"value": "More Songs About Buildings and Food"
},
{
"operation": "Highlighting",
"time": "31.247",
"value": "More Songs About Buildings and Food_sentence_5"
},
{
"operation": "Highlighting",
"time": "40.666",
"value": "More Songs About Buildings and Food_sentence_6"
},
{
"operation": "hyperlink",
"time": "47.64",
"value": "General Electric Co."
},
{
"operation": "Now on",
"time": "49.627",
"value": "General Electric"
},
{
"operation": "Highlighting",
"time": "51.572",
"value": "General Electric_sentence_3"
},
{
"operation": "finish",
"time": "70.235",
"value": "finish"
}
]
}
|
63995
|
St Peter's College was the site of the St Mungo centre, Newlands remained there for four years and was renamed Merrylee House.
|
{
"evidence": [
"Scotus College_sentence_5",
"Scotus College_sentence_6",
"Scotus College_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:\nSt Peter's College was the site of the St Mungo centre, Newlands remained there for four years and was renamed Merrylee House.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Scotus College_sentence_5",
"Scotus College_sentence_6",
"Scotus College_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1230,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.408",
"value": "?search="
},
{
"operation": "search",
"time": "6.906",
"value": "St Peter's College"
},
{
"operation": "Now on",
"time": "8.912",
"value": "St Peter's College"
},
{
"operation": "search",
"time": "46.775",
"value": "St Peter's College mungo"
},
{
"operation": "Now on",
"time": "49.229",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "92.863",
"value": "Scotus College"
},
{
"operation": "Now on",
"time": "94.219",
"value": "Scotus College"
},
{
"operation": "Highlighting",
"time": "98.97",
"value": "Scotus College_sentence_5"
},
{
"operation": "Highlighting",
"time": "113.431",
"value": "Scotus College_sentence_6"
},
{
"operation": "Highlighting",
"time": "121.624",
"value": "Scotus College_sentence_7"
},
{
"operation": "finish",
"time": "132.624",
"value": "finish"
}
]
}
|
exfever_2761
|
The Ten Commandments are a set of biblical principles that appear in the third book of the Bible.
|
{
"evidence": [
"Ten Commandments",
"Book of Exodus"
],
"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 Ten Commandments are a set of biblical principles that appear in the third book of the Bible.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ten Commandments",
"Book of Exodus"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1231,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
38299
|
The Arunachal Scouts’ role are similar to that of the Sikkim and Ladakh scouts, an infantry regiment of the Indian Army that specialized in mountain warfare tasked to guard India’s borders.
|
{
"evidence": [
"Ladakh Scouts_sentence_0",
"Ladakh Scouts_sentence_1",
"Arunachal Scouts_sentence_6"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Arunachal Scouts’ role are similar to that of the Sikkim and Ladakh scouts, an infantry regiment of the Indian Army that specialized in mountain warfare tasked to guard India’s borders.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Ladakh Scouts_sentence_0",
"Ladakh Scouts_sentence_1",
"Arunachal Scouts_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1232,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.768",
"value": "?search="
},
{
"operation": "search",
"time": "8.177",
"value": "Arunachal Scouts"
},
{
"operation": "Now on",
"time": "9.257",
"value": "Arunachal Scouts"
},
{
"operation": "Page search",
"time": "12.822",
"value": "role"
},
{
"operation": "Highlighting",
"time": "20.206",
"value": "Arunachal Scouts_sentence_31"
},
{
"operation": "Highlighting",
"time": "20.208",
"value": "Arunachal Scouts_section_4"
},
{
"operation": "hyperlink",
"time": "34.258",
"value": "Ladakh Scouts"
},
{
"operation": "Now on",
"time": "35.197",
"value": "Ladakh Scouts"
},
{
"operation": "back-button-clicked",
"time": "39.129",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "39.249",
"value": "Arunachal Scouts"
},
{
"operation": "hyperlink",
"time": "151.501",
"value": "Ladakh Scouts"
},
{
"operation": "Now on",
"time": "152.674",
"value": "Ladakh Scouts"
},
{
"operation": "Highlighting",
"time": "154.615",
"value": "Ladakh Scouts_sentence_0"
},
{
"operation": "Highlighting",
"time": "159.677",
"value": "Ladakh Scouts_sentence_1"
},
{
"operation": "back-button-clicked",
"time": "164.5",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "164.617",
"value": "Arunachal Scouts"
},
{
"operation": "Page search",
"time": "168.294",
"value": "sikkin"
},
{
"operation": "Page search",
"time": "169.405",
"value": "sikkim"
},
{
"operation": "Highlighting",
"time": "186.427",
"value": "Arunachal Scouts_sentence_13"
},
{
"operation": "Highlighting",
"time": "186.429",
"value": "Arunachal Scouts_section_2"
},
{
"operation": "Highlighting deleted",
"time": "190.847",
"value": "Arunachal Scouts_sentence_13"
},
{
"operation": "Highlighting deleted",
"time": "190.852",
"value": "Arunachal Scouts_section_2"
},
{
"operation": "page-search-reset",
"time": "204.969",
"value": "page-search-reset"
},
{
"operation": "Highlighting deleted",
"time": "216.972",
"value": "Arunachal Scouts_sentence_31"
},
{
"operation": "Highlighting deleted",
"time": "216.977",
"value": "Arunachal Scouts_section_4"
},
{
"operation": "Highlighting",
"time": "221.667",
"value": "Arunachal Scouts_sentence_6"
},
{
"operation": "finish",
"time": "231.185",
"value": "finish"
}
]
}
|
exfever_1585
|
The Flash is an American tv series that aired in an American commercial broadcast network that distributes over the air program content.
|
{
"evidence": [
"The Flash (1990 TV series)",
"CBS",
"Television network"
],
"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 Flash is an American tv series that aired in an American commercial broadcast network that distributes over the air program content.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Flash (1990 TV series)",
"CBS",
"Television network"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1233,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
32206
|
In 1988 Montreal Alouettes season schedule, on June 18 the Toronto Argonauts (professional Canadian football) won with a score of 35-23 but lost on August 7 with a score of 16–12.
|
{
"evidence": [
"List of Montreal Alouettes seasons_sentence_4",
"List of Montreal Alouettes seasons_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:\nIn 1988 Montreal Alouettes season schedule, on June 18 the Toronto Argonauts (professional Canadian football) won with a score of 35-23 but lost on August 7 with a score of 16–12.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"List of Montreal Alouettes seasons_sentence_4",
"List of Montreal Alouettes seasons_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1234,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.047",
"value": "?search="
},
{
"operation": "search",
"time": "38.457",
"value": "List of Montreal Alouettes seasons"
},
{
"operation": "Now on",
"time": "39.18",
"value": "List of Montreal Alouettes seasons"
},
{
"operation": "Highlighting",
"time": "73.836",
"value": "List of Montreal Alouettes seasons_sentence_3"
},
{
"operation": "Highlighting",
"time": "74.541",
"value": "List of Montreal Alouettes seasons_sentence_4"
},
{
"operation": "Highlighting",
"time": "79.637",
"value": "List of Montreal Alouettes seasons_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "81.099",
"value": "List of Montreal Alouettes seasons_sentence_3"
},
{
"operation": "finish",
"time": "89.465",
"value": "finish"
}
]
}
|
27686
|
In the History of Zionism, the Bund was founded in 1897, one of the largest socialist movements in Europe and campaigned for Jewish autonomy and recognition national rights within Russia.
|
{
"evidence": [
"History of Zionism_sentence_199",
"History of Zionism_sentence_200"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn the History of Zionism, the Bund was founded in 1897, one of the largest socialist movements in Europe and campaigned for Jewish autonomy and recognition national rights within Russia.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"History of Zionism_sentence_199",
"History of Zionism_sentence_200"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1235,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.338",
"value": "?search="
},
{
"operation": "search",
"time": "5.303",
"value": "History of Zionism"
},
{
"operation": "Now on",
"time": "10.717",
"value": "History of Zionism"
},
{
"operation": "hyperlink",
"time": "18.765",
"value": "3.3.3 The Bund History of Zionism_section_22"
},
{
"operation": "Highlighting",
"time": "20.972",
"value": "History of Zionism_sentence_199"
},
{
"operation": "Highlighting",
"time": "30.087",
"value": "History of Zionism_sentence_200"
},
{
"operation": "finish",
"time": "41.581",
"value": "finish"
}
]
}
|
exfever_3146
|
Rogue is a fictional character appearing in American comic books commonly in association with the fictional team of superheroes that was initially canceled in 1980, following its 1975 revival and subsequent direction under a British-born American comic book writer and novelist, known for his 17-year stint on Uncanny X-Men.
|
{
"evidence": [
"Rogue (comics)",
"X-Men",
"Chris Claremont"
],
"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:\nRogue is a fictional character appearing in American comic books commonly in association with the fictional team of superheroes that was initially canceled in 1980, following its 1975 revival and subsequent direction under a British-born American comic book writer and novelist, known for his 17-year stint on Uncanny X-Men.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Rogue (comics)",
"X-Men",
"Chris Claremont"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1236,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
63905
|
Kaser, ranked No. 11 in the nation in 2014, kicked his longest punt of the season for 76 yards against Rice University (a 300-acre private research university in Houston, Texas) and would later be named special teams MVP at the Aggies annual banquet.
|
{
"evidence": [
"Drew Kaser_sentence_9",
"Drew Kaser_sentence_10",
"Drew Kaser_sentence_7",
"Rice University_sentence_1",
"Rice University_sentence_0",
"Drew Kaser_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:\nKaser, ranked No. 11 in the nation in 2014, kicked his longest punt of the season for 76 yards against Rice University (a 300-acre private research university in Houston, Texas) and would later be named special teams MVP at the Aggies annual banquet.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Drew Kaser_sentence_9",
"Drew Kaser_sentence_10",
"Drew Kaser_sentence_7",
"Rice University_sentence_1",
"Rice University_sentence_0",
"Drew Kaser_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1237,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.624",
"value": "?search="
},
{
"operation": "search",
"time": "64.029",
"value": "\"kicked his longest punt of the season for 76 yards against Rice University\""
},
{
"operation": "Now on",
"time": "65.949",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "67.027",
"value": "Drew Kaser"
},
{
"operation": "Now on",
"time": "68.685",
"value": "Drew Kaser"
},
{
"operation": "Page search",
"time": "69.589",
"value": "kicked his longest punt of the season for 76 yards against Rice University"
},
{
"operation": "Highlighting",
"time": "76.432",
"value": "Drew Kaser_sentence_9"
},
{
"operation": "Highlighting",
"time": "77.75",
"value": "Drew Kaser_sentence_10"
},
{
"operation": "Highlighting",
"time": "81.78",
"value": "Drew Kaser_sentence_7"
},
{
"operation": "Page search",
"time": "106.705",
"value": "300"
},
{
"operation": "Page search",
"time": "135.481",
"value": "acre"
},
{
"operation": "Highlighting deleted",
"time": "167.123",
"value": "Drew Kaser_sentence_7"
},
{
"operation": "Highlighting",
"time": "168.01",
"value": "Drew Kaser_sentence_7"
},
{
"operation": "hyperlink",
"time": "168.807",
"value": "Rice University"
},
{
"operation": "Now on",
"time": "176.829",
"value": "Rice University"
},
{
"operation": "Highlighting",
"time": "184.461",
"value": "Rice University_sentence_1"
},
{
"operation": "Highlighting",
"time": "193.42",
"value": "Rice University_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "198.584",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "198.683",
"value": "Drew Kaser"
},
{
"operation": "Highlighting",
"time": "206.643",
"value": "Drew Kaser_sentence_8"
},
{
"operation": "finish",
"time": "296.241",
"value": "finish"
}
]
}
|
exfever_6787
|
The Who are an English rock band that consists of a bass guitarist and a singer nicknamed The Ox.
|
{
"evidence": [
"The Who",
"John Entwistle"
],
"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 Who are an English rock band that consists of a bass guitarist and a singer nicknamed The Ox.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Who",
"John Entwistle"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1238,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4231
|
Michael Edward Love was an American singer and songwriter who co-founded One Direction with his cousin, best remembered as their drummer.
|
{
"evidence": [
"Mike Love",
"Dennis Wilson"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMichael Edward Love was an American singer and songwriter who co-founded One Direction with his cousin, best remembered as their drummer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mike Love",
"Dennis Wilson"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1239,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
74855
|
The Mandai Crematorium and Columbarium, located at Mandai Road in Mandai, Singapore, was completed in 1982, and in 1999, the government decided that all cremation services are to be consolidated at Mandai (a planning area located in the North Region of Singapore).
|
{
"evidence": [
"Mandai Crematorium and Columbarium_sentence_9"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Mandai Crematorium and Columbarium, located at Mandai Road in Mandai, Singapore, was completed in 1982, and in 1999, the government decided that all cremation services are to be consolidated at Mandai (a planning area located in the North Region of Singapore).\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Mandai Crematorium and Columbarium_sentence_9"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1240,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.998",
"value": "?search="
},
{
"operation": "search",
"time": "11.22",
"value": "Mandai Crematorium and Columbarium"
},
{
"operation": "Now on",
"time": "12.125",
"value": "Mandai Crematorium and Columbarium"
},
{
"operation": "Highlighting",
"time": "30.217",
"value": "Mandai Crematorium and Columbarium_cell_0_3_1"
},
{
"operation": "Highlighting",
"time": "30.225",
"value": "Mandai Crematorium and Columbarium_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "30.225",
"value": "Mandai Crematorium and Columbarium_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "30.226",
"value": "Mandai Crematorium and Columbarium_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "39.951",
"value": "Mandai Crematorium and Columbarium_cell_0_4_1"
},
{
"operation": "Highlighting",
"time": "39.956",
"value": "Mandai Crematorium and Columbarium_header_cell_0_4_0"
},
{
"operation": "Highlighting",
"time": "63.543",
"value": "Mandai Crematorium and Columbarium_sentence_9"
},
{
"operation": "Highlighting deleted",
"time": "69.359",
"value": "Mandai Crematorium and Columbarium_cell_0_4_1"
},
{
"operation": "Highlighting deleted",
"time": "69.366",
"value": "Mandai Crematorium and Columbarium_header_cell_0_4_0"
},
{
"operation": "Highlighting deleted",
"time": "70.527",
"value": "Mandai Crematorium and Columbarium_cell_0_3_1"
},
{
"operation": "Highlighting deleted",
"time": "70.533",
"value": "Mandai Crematorium and Columbarium_header_cell_0_3_0"
},
{
"operation": "Highlighting deleted",
"time": "70.536",
"value": "Mandai Crematorium and Columbarium_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "70.538",
"value": "Mandai Crematorium and Columbarium_header_cell_0_0_0"
},
{
"operation": "finish",
"time": "78.815",
"value": "finish"
}
]
}
|
9239
|
Irwin Haskett(born April 22, 1903) was a politician in Ontario who was one of the founders of the annual Tulip Festival in Ottawa and served as a president of the Ottawa Board of Trade and the Ontario Chamber of Commerce.
|
{
"evidence": [
"Irwin Haskett_sentence_0",
"Irwin Haskett_sentence_9",
"Irwin Haskett_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:\nIrwin Haskett(born April 22, 1903) was a politician in Ontario who was one of the founders of the annual Tulip Festival in Ottawa and served as a president of the Ottawa Board of Trade and the Ontario Chamber of Commerce.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Irwin Haskett_sentence_0",
"Irwin Haskett_sentence_9",
"Irwin Haskett_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1241,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.694",
"value": "?search="
},
{
"operation": "search",
"time": "20.172",
"value": "Irwin Haskett"
},
{
"operation": "Now on",
"time": "21.173",
"value": "Irwin Haskett"
},
{
"operation": "Highlighting",
"time": "25.272",
"value": "Irwin Haskett_sentence_0"
},
{
"operation": "Highlighting",
"time": "34.107",
"value": "Irwin Haskett_sentence_9"
},
{
"operation": "Highlighting",
"time": "42.272",
"value": "Irwin Haskett_sentence_8"
},
{
"operation": "finish",
"time": "60.712",
"value": "finish"
}
]
}
|
exfever_7453
|
The Boomerang is a 1992 American romantic comedy film featuring an American actor and comedian best known as Bernard, from an American television sitcom that ran for thirteen episodes.
|
{
"evidence": [
"Boomerang (1992 film)",
"David Alan Grier",
"Damon (TV series)"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Boomerang is a 1992 American romantic comedy film featuring an American actor and comedian best known as Bernard, from an American television sitcom that ran for thirteen episodes.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Boomerang (1992 film)",
"David Alan Grier",
"Damon (TV series)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1242,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
51850
|
The organization, which has local branches in Tromsø, and Østlandet, and is in-charge of improving the media coverage on Kven issues and establishing a Kven kindergarten, is Norwegian Kven Organization.
|
{
"evidence": [
"Norwegian Kven Organization_sentence_1",
"Norwegian Kven Organization_sentence_2",
"Norwegian Kven Organization_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 organization, which has local branches in Tromsø, and Østlandet, and is in-charge of improving the media coverage on Kven issues and establishing a Kven kindergarten, is Norwegian Kven Organization.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Norwegian Kven Organization_sentence_1",
"Norwegian Kven Organization_sentence_2",
"Norwegian Kven Organization_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1243,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.303",
"value": "?search="
},
{
"operation": "search",
"time": "15.859",
"value": "Tromsø"
},
{
"operation": "Now on",
"time": "18.23",
"value": "Tromsø"
},
{
"operation": "search",
"time": "31.039",
"value": "Norwegian Kven Organization"
},
{
"operation": "Now on",
"time": "31.566",
"value": "Norwegian Kven Organization"
},
{
"operation": "Highlighting",
"time": "62.006",
"value": "Norwegian Kven Organization_sentence_1"
},
{
"operation": "Highlighting",
"time": "78.435",
"value": "Norwegian Kven Organization_sentence_2"
},
{
"operation": "Highlighting",
"time": "151.337",
"value": "Norwegian Kven Organization_sentence_3"
},
{
"operation": "finish",
"time": "204.945",
"value": "finish"
}
]
}
|
20883
|
Found in Canada, Drepanidae chinoea has a wingspan of 16-20.5 mm for males, with lustrous white fore and hindwings.
|
{
"evidence": [
"Drepanidae_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:\nFound in Canada, Drepanidae chinoea has a wingspan of 16-20.5 mm for males, with lustrous white fore and hindwings.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Drepanidae_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1244,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.124",
"value": "?search="
},
{
"operation": "search",
"time": "36.108",
"value": "Drepanidae"
},
{
"operation": "Now on",
"time": "37.076",
"value": "Drepanidae"
},
{
"operation": "search",
"time": "57.835",
"value": "drepanidae chinoea"
},
{
"operation": "Now on",
"time": "58.666",
"value": "undefined"
},
{
"operation": "search",
"time": "67.851",
"value": "drepanidae"
},
{
"operation": "Now on",
"time": "68.427",
"value": "Drepanidae"
},
{
"operation": "Highlighting",
"time": "97.197",
"value": "Drepanidae_sentence_4"
},
{
"operation": "hyperlink",
"time": "97.962",
"value": "subfamilies"
},
{
"operation": "Now on",
"time": "98.526",
"value": "Subfamily"
},
{
"operation": "back-button-clicked",
"time": "104.749",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "104.845",
"value": "Drepanidae"
},
{
"operation": "Highlighting",
"time": "116.219",
"value": "Drepanidae_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "118.643",
"value": "Drepanidae_sentence_4"
},
{
"operation": "finish",
"time": "129.964",
"value": "finish"
}
]
}
|
67585
|
Willem Vrolik, a Dutch anatomist and pathologist who is a pioneer in the field of vertebrate teratology, published teratological works on cyclopia, the pathogenesis of congenital anomalies, and a treatise on conjoined twins as well as Handboek der ziektekundige ontleedkunde (Handbook of pathological anatomy); however, these were later on stopped in production.
|
{
"evidence": [
"Willem Vrolik_sentence_0",
"Willem Vrolik_sentence_1",
"Willem Vrolik_sentence_10",
"Willem Vrolik_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:\nWillem Vrolik, a Dutch anatomist and pathologist who is a pioneer in the field of vertebrate teratology, published teratological works on cyclopia, the pathogenesis of congenital anomalies, and a treatise on conjoined twins as well as Handboek der ziektekundige ontleedkunde (Handbook of pathological anatomy); however, these were later on stopped in production.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Willem Vrolik_sentence_0",
"Willem Vrolik_sentence_1",
"Willem Vrolik_sentence_10",
"Willem Vrolik_sentence_11"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1245,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.061",
"value": "?search="
},
{
"operation": "search",
"time": "8.913",
"value": "Willem Vrolik"
},
{
"operation": "Now on",
"time": "10.331",
"value": "Willem Vrolik"
},
{
"operation": "Highlighting",
"time": "723.861",
"value": "Willem Vrolik_sentence_0"
},
{
"operation": "Highlighting",
"time": "727.076",
"value": "Willem Vrolik_sentence_1"
},
{
"operation": "Highlighting",
"time": "734.464",
"value": "Willem Vrolik_sentence_10"
},
{
"operation": "Highlighting",
"time": "747.617",
"value": "Willem Vrolik_sentence_11"
},
{
"operation": "finish",
"time": "804.963",
"value": "finish"
}
]
}
|
47686
|
Wolff Schoemaker designed the Trade Fair Center, it has a cross-shaped floor plan and was designed in a modern style.
|
{
"evidence": [
"Charles Prosper Wolff Schoemaker_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:\nWolff Schoemaker designed the Trade Fair Center, it has a cross-shaped floor plan and was designed in a modern style.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Charles Prosper Wolff Schoemaker_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1246,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.329",
"value": "?search="
},
{
"operation": "search",
"time": "9.752",
"value": "Wolff Schoemaker"
},
{
"operation": "Now on",
"time": "11.416",
"value": "Charles Prosper Wolff Schoemaker"
},
{
"operation": "Highlighting",
"time": "25.697",
"value": "Charles Prosper Wolff Schoemaker_sentence_0"
},
{
"operation": "hyperlink",
"time": "29.656",
"value": "Wolff Schoemaker"
},
{
"operation": "Now on",
"time": "30.933",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "36.582",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "37.333",
"value": "Charles Prosper Wolff Schoemaker"
},
{
"operation": "Highlighting deleted",
"time": "43.118",
"value": "Charles Prosper Wolff Schoemaker_sentence_0"
},
{
"operation": "search",
"time": "48.504",
"value": "Wolff Schoemaker"
},
{
"operation": "Now on",
"time": "49.505",
"value": "Wolff Schoemaker"
},
{
"operation": "Highlighting",
"time": "56.719",
"value": "Charles Prosper Wolff Schoemaker_sentence_0"
},
{
"operation": "Page search",
"time": "69.399",
"value": "trade"
},
{
"operation": "Page search",
"time": "69.582",
"value": "trade"
},
{
"operation": "Page search",
"time": "69.741",
"value": "trade"
},
{
"operation": "search",
"time": "92.527",
"value": "trade fair center"
},
{
"operation": "Now on",
"time": "94.713",
"value": "undefined"
},
{
"operation": "search",
"time": "116.614",
"value": "Contains...trade fair center"
},
{
"operation": "Now on",
"time": "117.398",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "119.473",
"value": "Charlottenburg-Wilmersdorf"
},
{
"operation": "Now on",
"time": "120.457",
"value": "Charlottenburg-Wilmersdorf"
},
{
"operation": "Page search",
"time": "133.159",
"value": "trade"
},
{
"operation": "back-button-clicked",
"time": "151.648",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "152.595",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "153.507",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "153.54",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "154.421",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "155.191",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "155.221",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "155.738",
"value": "Charles Prosper Wolff Schoemaker"
},
{
"operation": "Highlighting",
"time": "160.065",
"value": "Charles Prosper Wolff Schoemaker_sentence_0"
},
{
"operation": "Page search",
"time": "176.034",
"value": "trade"
},
{
"operation": "Page search",
"time": "176.5",
"value": "trade"
},
{
"operation": "Page search",
"time": "176.704",
"value": "trade"
},
{
"operation": "Highlighting",
"time": "231.691",
"value": "Charles Prosper Wolff Schoemaker_sentence_13Charles Prosper Wolff Schoemaker_sentence_14"
},
{
"operation": "Highlighting",
"time": "231.696",
"value": "Charles Prosper Wolff Schoemaker_section_1"
},
{
"operation": "Highlighting deleted",
"time": "402.676",
"value": "Charles Prosper Wolff Schoemaker_sentence_13Charles Prosper Wolff Schoemaker_sentence_14"
},
{
"operation": "Highlighting deleted",
"time": "402.682",
"value": "Charles Prosper Wolff Schoemaker_section_1"
},
{
"operation": "finish",
"time": "408.597",
"value": "finish"
}
]
}
|
exfever_2519
|
Zhou Enlai served under Chairman Mao and helped the Communist Party rise to power and develop the Chinese economy, which founded the People's Republic of China.
|
{
"evidence": [
"Zhou Enlai",
"Mao Zedong"
],
"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:\nZhou Enlai served under Chairman Mao and helped the Communist Party rise to power and develop the Chinese economy, which founded the People's Republic of China.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Zhou Enlai",
"Mao Zedong"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1247,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
34162
|
Velous, a stage name based on a comment about his marvelous production skills, released his debut mixtape called Velocity The Mixtape in 2011 which gave him the opportunity to be signed to Swanky Music Group.
|
{
"evidence": [
"Velous_sentence_13",
"Velous_sentence_14",
"Velous_sentence_15"
],
"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:\nVelous, a stage name based on a comment about his marvelous production skills, released his debut mixtape called Velocity The Mixtape in 2011 which gave him the opportunity to be signed to Swanky Music Group.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Velous_sentence_13",
"Velous_sentence_14",
"Velous_sentence_15"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1248,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.852",
"value": "?search="
},
{
"operation": "search",
"time": "72.619",
"value": "Velous"
},
{
"operation": "Now on",
"time": "73.532",
"value": "Velous"
},
{
"operation": "Highlighting",
"time": "80.459",
"value": "Velous_sentence_0"
},
{
"operation": "Highlighting",
"time": "110.915",
"value": "Velous_sentence_13"
},
{
"operation": "Highlighting",
"time": "117.835",
"value": "Velous_sentence_14"
},
{
"operation": "Highlighting",
"time": "122.107",
"value": "Velous_sentence_15"
},
{
"operation": "Highlighting deleted",
"time": "132.611",
"value": "Velous_sentence_0"
},
{
"operation": "finish",
"time": "147.315",
"value": "finish"
}
]
}
|
61491
|
Dina Mangabeira was born in Bocaiúva and moved to Montes Claros two year later.
|
{
"evidence": [
"Dina Mangabeira_sentence_1",
"Dina Mangabeira_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:\nDina Mangabeira was born in Bocaiúva and moved to Montes Claros two year later.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Dina Mangabeira_sentence_1",
"Dina Mangabeira_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1249,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.953",
"value": "?search="
},
{
"operation": "search",
"time": "6.327",
"value": "Dina Mangabeira"
},
{
"operation": "Now on",
"time": "7.754",
"value": "Dina Mangabeira"
},
{
"operation": "Highlighting",
"time": "90.568",
"value": "Dina Mangabeira_sentence_1"
},
{
"operation": "Highlighting",
"time": "96.534",
"value": "Dina Mangabeira_sentence_3"
},
{
"operation": "finish",
"time": "105.743",
"value": "finish"
}
]
}
|
exfever_5741
|
Katherine Marie Heigl is an American actress, producer and former fashion model who appeared in a 1990 French-American comedy-drama directed by Steve Miner.
|
{
"evidence": [
"Katherine Heigl",
"My Father the Hero (1994 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:\nKatherine Marie Heigl is an American actress, producer and former fashion model who appeared in a 1990 French-American comedy-drama directed by Steve Miner.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Katherine Heigl",
"My Father the Hero (1994 film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1250,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
54814
|
Slovakia's national football team was tied in a friendly match with Estadio Nacional de Costa Rica on February 5, 1997, resulting in a T2-2 score.
|
{
"evidence": [
"Slovakia national football team_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:\nSlovakia's national football team was tied in a friendly match with Estadio Nacional de Costa Rica on February 5, 1997, resulting in a T2-2 score.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Slovakia national football team_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1251,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.321",
"value": "?search="
},
{
"operation": "search",
"time": "17.085",
"value": "Slovakia national football team"
},
{
"operation": "Now on",
"time": "71.45",
"value": "Slovakia national football team"
},
{
"operation": "search",
"time": "266.391",
"value": "Estadio Nacional de Costa Rica"
},
{
"operation": "Now on",
"time": "267.998",
"value": "Estadio Nacional de Costa Rica"
},
{
"operation": "hyperlink",
"time": "275.788",
"value": "Estadio Nacional de Costa Rica (1924)"
},
{
"operation": "Now on",
"time": "276.482",
"value": "Estadio Nacional de Costa Rica (1924)"
},
{
"operation": "back-button-clicked",
"time": "291.118",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "291.153",
"value": "Estadio Nacional de Costa Rica"
},
{
"operation": "back-button-clicked",
"time": "292.753",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "292.757",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "293.013",
"value": "Slovakia national football team"
},
{
"operation": "Page search",
"time": "300.106",
"value": "Estadio"
},
{
"operation": "Page search",
"time": "343.315",
"value": "friendly"
},
{
"operation": "search",
"time": "392.842",
"value": "Contains...February 5, 1997"
},
{
"operation": "Now on",
"time": "397.198",
"value": "undefined"
},
{
"operation": "search",
"time": "453.337",
"value": "\"slovakia\", \"February 5, 1997\""
},
{
"operation": "Now on",
"time": "454.5",
"value": "undefined"
},
{
"operation": "search",
"time": "470.123",
"value": "Estadio Nacional de Costa Rica (1924)"
},
{
"operation": "Now on",
"time": "471.675",
"value": "Estadio Nacional de Costa Rica (1924)"
},
{
"operation": "search",
"time": "491.327",
"value": "Estadio Nacional de Costa Rica (2011)"
},
{
"operation": "Now on",
"time": "496.428",
"value": "Estadio Nacional de Costa Rica (2011)"
},
{
"operation": "search",
"time": "522.249",
"value": "Slovakia national football team"
},
{
"operation": "Now on",
"time": "523.389",
"value": "Slovakia national football team"
},
{
"operation": "hyperlink",
"time": "526.514",
"value": "Slovakia women's national football team"
},
{
"operation": "Now on",
"time": "528.471",
"value": "Slovakia women's national football team"
},
{
"operation": "Page search",
"time": "584.154",
"value": "1997"
},
{
"operation": "search",
"time": "626.984",
"value": "Slovakia national football team"
},
{
"operation": "Now on",
"time": "628.139",
"value": "Slovakia national football team"
},
{
"operation": "Highlighting",
"time": "631.729",
"value": "Slovakia national football team_sentence_2"
},
{
"operation": "finish",
"time": "649.979",
"value": "finish"
}
]
}
|
exfever_6988
|
Dopamine is a neuromodulatory molecule that plays several important roles in cells an amine synthesized by removing a carboxyl group from a molecule of its amino acid that is made and used as part of the normal biology of some plants and animals and the precursor to the neurotransmitters neuromodulatory molecule that plays several important roles in cells.
|
{
"evidence": [
"Dopamine",
"L-DOPA",
"Dopamine"
],
"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:\nDopamine is a neuromodulatory molecule that plays several important roles in cells an amine synthesized by removing a carboxyl group from a molecule of its amino acid that is made and used as part of the normal biology of some plants and animals and the precursor to the neurotransmitters neuromodulatory molecule that plays several important roles in cells.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dopamine",
"L-DOPA",
"Dopamine"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1252,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
80366
|
Jovanovitch's formal schooling began two years later at Newtown Infants School, and he was an avid reader from an early age.
|
{
"evidence": [
"Danilo Jovanovitch_sentence_10",
"Danilo Jovanovitch_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:\nJovanovitch's formal schooling began two years later at Newtown Infants School, and he was an avid reader from an early age.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Danilo Jovanovitch_sentence_10",
"Danilo Jovanovitch_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1253,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.745",
"value": "?search="
},
{
"operation": "search",
"time": "11.665",
"value": "Jovanovitch"
},
{
"operation": "Now on",
"time": "12.222",
"value": "Jovanović"
},
{
"operation": "search",
"time": "36.352",
"value": "jovanovitch newtown"
},
{
"operation": "Now on",
"time": "37.032",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "38.681",
"value": "Danilo Jovanovitch"
},
{
"operation": "Now on",
"time": "39.09",
"value": "Danilo Jovanovitch"
},
{
"operation": "Highlighting",
"time": "51.367",
"value": "Danilo Jovanovitch_sentence_10"
},
{
"operation": "Highlighting",
"time": "60.111",
"value": "Danilo Jovanovitch_sentence_8"
},
{
"operation": "finish",
"time": "186.31",
"value": "finish"
}
]
}
|
49803
|
Yukinori Yanagi is a Japanese contemporary artist who earned a bachelor's and master's of fine arts, both from Musashino Art University.
|
{
"evidence": [
"Yukinori Yanagi_sentence_0",
"Yukinori Yanagi_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:\nYukinori Yanagi is a Japanese contemporary artist who earned a bachelor's and master's of fine arts, both from Musashino Art University.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Yukinori Yanagi_sentence_0",
"Yukinori Yanagi_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1254,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.41",
"value": "?search="
},
{
"operation": "search",
"time": "9.09",
"value": "Yukinori Yanagi"
},
{
"operation": "Now on",
"time": "10.081",
"value": "Yukinori Yanagi"
},
{
"operation": "Highlighting",
"time": "13.767",
"value": "Yukinori Yanagi_sentence_0"
},
{
"operation": "Highlighting",
"time": "17.429",
"value": "Yukinori Yanagi_sentence_1"
},
{
"operation": "finish",
"time": "49.742",
"value": "finish"
}
]
}
|
exfever_8000
|
Rebecca Ferguson came to international prominence with her portrayal of the former Queen of England in the British television miniseries The White Queen.
|
{
"evidence": [
"Rebecca Ferguson",
"Elizabeth Woodville"
],
"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:\nRebecca Ferguson came to international prominence with her portrayal of the former Queen of England in the British television miniseries The White Queen.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Rebecca Ferguson",
"Elizabeth Woodville"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1255,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2252
|
Telangana is a state in India situated on the south-central stretch of the physiographical region in Southern Asia.
|
{
"evidence": [
"Telangana",
"Indian subcontinent"
],
"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:\nTelangana is a state in India situated on the south-central stretch of the physiographical region in Southern Asia.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Telangana",
"Indian subcontinent"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1256,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2135
|
Tessanne Chin is best known for winning Season 5 of NBC's The Voice, as part of an American singer and musician's team.
|
{
"evidence": [
"Tessanne Chin",
"Adam Levine"
],
"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:\nTessanne Chin is best known for winning Season 5 of NBC's The Voice, as part of an American singer and musician's team.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tessanne Chin",
"Adam Levine"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1257,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1647
|
Valerian and the City of a Thousand Planets is the most expensive movie worldwide.
|
{
"evidence": [
"Valerian and the City of a Thousand Planets",
"Luc Besson"
],
"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:\nValerian and the City of a Thousand Planets is the most expensive movie worldwide.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Valerian and the City of a Thousand Planets",
"Luc Besson"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1258,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
48513
|
Christian Sands studied piano for years to accomplish the goal of being in an accomplished band.
|
{
"evidence": [
"Christian Sands_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:\nChristian Sands studied piano for years to accomplish the goal of being in an accomplished band.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Christian Sands_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1259,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.024",
"value": "?search="
},
{
"operation": "search",
"time": "9.425",
"value": "Christian Sands"
},
{
"operation": "Now on",
"time": "10.381",
"value": "Christian Sands"
},
{
"operation": "Page search",
"time": "16.897",
"value": "piano"
},
{
"operation": "Highlighting",
"time": "20.207",
"value": "Christian Sands_sentence_3"
},
{
"operation": "Page search",
"time": "46.954",
"value": "goal"
},
{
"operation": "Page search",
"time": "71.533",
"value": "accomplished"
},
{
"operation": "Highlighting deleted",
"time": "78.862",
"value": "Christian Sands_sentence_3"
},
{
"operation": "Highlighting",
"time": "80.166",
"value": "Christian Sands_sentence_3"
},
{
"operation": "finish",
"time": "138.355",
"value": "finish"
}
]
}
|
63820
|
While at Repton, Fisher met Rosamond Chevallier Forman, and they were married on 12 April 1917 and based on Webster, a person who wrote Fisher's biography, that marriage was chaotic and only lasted for five years.
|
{
"evidence": [
"Geoffrey Fisher_sentence_41",
"Geoffrey Fisher_sentence_43",
"Geoffrey Fisher_sentence_42"
],
"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 at Repton, Fisher met Rosamond Chevallier Forman, and they were married on 12 April 1917 and based on Webster, a person who wrote Fisher's biography, that marriage was chaotic and only lasted for five years.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Geoffrey Fisher_sentence_41",
"Geoffrey Fisher_sentence_43",
"Geoffrey Fisher_sentence_42"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1260,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.622",
"value": "?search="
},
{
"operation": "search",
"time": "30.43",
"value": "Rosamond Chevallier Forman"
},
{
"operation": "Now on",
"time": "31.15",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "34.333",
"value": "Geoffrey Fisher"
},
{
"operation": "Now on",
"time": "35.093",
"value": "Geoffrey Fisher"
},
{
"operation": "Page search",
"time": "46.205",
"value": "rosamond"
},
{
"operation": "Highlighting",
"time": "54.996",
"value": "Geoffrey Fisher_sentence_41"
},
{
"operation": "Highlighting",
"time": "58.756",
"value": "Geoffrey Fisher_sentence_42"
},
{
"operation": "Highlighting",
"time": "61.764",
"value": "Geoffrey Fisher_sentence_43"
},
{
"operation": "Highlighting deleted",
"time": "77.067",
"value": "Geoffrey Fisher_sentence_42"
},
{
"operation": "Highlighting",
"time": "101.204",
"value": "Geoffrey Fisher_sentence_42"
},
{
"operation": "finish",
"time": "104.075",
"value": "finish"
}
]
}
|
exfever_853
|
Fantastic Beasts and Where to Find Them is a 2017 fantasy film starring a British-American actress who debuted in Michael Clayton.
|
{
"evidence": [
"Fantastic Beasts and Where to Find Them (film)",
"Katherine Waterston"
],
"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:\nFantastic Beasts and Where to Find Them is a 2017 fantasy film starring a British-American actress who debuted in Michael Clayton.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Fantastic Beasts and Where to Find Them (film)",
"Katherine Waterston"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1261,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
72876
|
Two of Douglas Gordon's best-known art works are The 24 Hour Psycho (1993), which slows down Alfred Hitchcock's film Psycho so that it lasts twenty four hours, and Feature Film, a projection of Gordon's own film of James Conlon (born on March 18, 1950) conducting Bernard Herrmann's score to Vertigo.
|
{
"evidence": [
"Douglas Gordon_sentence_10",
"Douglas Gordon_sentence_13",
"James Conlon_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:\nTwo of Douglas Gordon's best-known art works are The 24 Hour Psycho (1993), which slows down Alfred Hitchcock's film Psycho so that it lasts twenty four hours, and Feature Film, a projection of Gordon's own film of James Conlon (born on March 18, 1950) conducting Bernard Herrmann's score to Vertigo.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Douglas Gordon_sentence_10",
"Douglas Gordon_sentence_13",
"James Conlon_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1262,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.855",
"value": "?search="
},
{
"operation": "search",
"time": "17.441",
"value": "Douglas Gordon"
},
{
"operation": "Now on",
"time": "18.26",
"value": "Douglas Gordon"
},
{
"operation": "Highlighting",
"time": "55.308",
"value": "Douglas Gordon_sentence_10"
},
{
"operation": "Highlighting",
"time": "86.021",
"value": "Douglas Gordon_sentence_13"
},
{
"operation": "hyperlink",
"time": "97.19",
"value": "James Conlon"
},
{
"operation": "Now on",
"time": "98.103",
"value": "James Conlon"
},
{
"operation": "Highlighting",
"time": "99.318",
"value": "James Conlon_sentence_1"
},
{
"operation": "finish",
"time": "104.301",
"value": "finish"
}
]
}
|
38269
|
Bernhard von Beskow born in Stockholm was a Swedish dramatist and entered the civil service in 1814.
|
{
"evidence": [
"Bernhard von Beskow_sentence_0",
"Bernhard von Beskow_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:\nBernhard von Beskow born in Stockholm was a Swedish dramatist and entered the civil service in 1814.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bernhard von Beskow_sentence_0",
"Bernhard von Beskow_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1263,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.806",
"value": "?search="
},
{
"operation": "search",
"time": "7.619",
"value": "Bernhard von Beskow"
},
{
"operation": "Now on",
"time": "8.557",
"value": "Bernhard von Beskow"
},
{
"operation": "Highlighting",
"time": "12.079",
"value": "Bernhard von Beskow_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "12.083",
"value": "Bernhard von Beskow_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "12.083",
"value": "Bernhard von Beskow_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "14.068",
"value": "Bernhard von Beskow_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "15.748",
"value": "Bernhard von Beskow_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "15.753",
"value": "Bernhard von Beskow_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "15.754",
"value": "Bernhard von Beskow_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "21.284",
"value": "Bernhard von Beskow_sentence_2"
},
{
"operation": "finish",
"time": "26.425",
"value": "finish"
}
]
}
|
exfever_4449
|
Animal Farm is a satirical allegorical novella by George Orwell that is classified as a literary genre that is anthropomorphized, and that illustrates or leads to a particular moral lesson.
|
{
"evidence": [
"Animal Farm",
"Fable"
],
"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:\nAnimal Farm is a satirical allegorical novella by George Orwell that is classified as a literary genre that is anthropomorphized, and that illustrates or leads to a particular moral lesson.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Animal Farm",
"Fable"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1264,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
3848
|
Voting rights in Singapore is impacted by Paines view, and the Supreme Court says that representative democracy does not appear in the Constitution of Singapore.
|
{
"evidence": [
"Voting rights in Singapore_sentence_0",
"Voting rights in Singapore_sentence_6",
"Voting rights in Singapore_sentence_20",
"Voting rights in Singapore_sentence_22",
"Voting rights in Singapore_sentence_24",
"Voting rights in Singapore_sentence_25",
"Voting rights in Singapore_sentence_30",
"Voting rights in Singapore_sentence_31"
],
"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:\nVoting rights in Singapore is impacted by Paines view, and the Supreme Court says that representative democracy does not appear in the Constitution of Singapore.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Voting rights in Singapore_sentence_0",
"Voting rights in Singapore_sentence_6",
"Voting rights in Singapore_sentence_20",
"Voting rights in Singapore_sentence_22",
"Voting rights in Singapore_sentence_24",
"Voting rights in Singapore_sentence_25",
"Voting rights in Singapore_sentence_30",
"Voting rights in Singapore_sentence_31"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1265,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "17.248",
"value": "Voting rights in Singapore"
},
{
"operation": "Highlighting",
"time": "35.167",
"value": "Voting rights in Singapore_sentence_0"
},
{
"operation": "Highlighting",
"time": "85.024",
"value": "Voting rights in Singapore_sentence_6"
},
{
"operation": "Page search",
"time": "112.098",
"value": "paine"
},
{
"operation": "Highlighting",
"time": "180.651",
"value": "Voting rights in Singapore_sentence_20"
},
{
"operation": "Highlighting",
"time": "181.82",
"value": "Voting rights in Singapore_sentence_22"
},
{
"operation": "Highlighting",
"time": "198.501",
"value": "Voting rights in Singapore_sentence_24"
},
{
"operation": "Highlighting",
"time": "203.748",
"value": "Voting rights in Singapore_sentence_25"
},
{
"operation": "Highlighting",
"time": "304.049",
"value": "Voting rights in Singapore_sentence_1Voting rights in Singapore_sentence_2"
},
{
"operation": "Highlighting",
"time": "421.617",
"value": "Voting rights in Singapore_sentence_30"
},
{
"operation": "Highlighting",
"time": "430.559",
"value": "Voting rights in Singapore_sentence_31"
},
{
"operation": "finish",
"time": "559.606",
"value": "finish"
}
]
}
|
exfever_4964
|
Murder on the Orient Express is a 2017 mystery film based on the 1934 novel that is a work of detective fiction by English writer Agatha Christie that retailed for seven shillings and a sixpenny bit in the UK edition.
|
{
"evidence": [
"Murder on the Orient Express (2017 film)",
"Murder on the Orient Express",
"Sixpence (British coin)"
],
"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:\nMurder on the Orient Express is a 2017 mystery film based on the 1934 novel that is a work of detective fiction by English writer Agatha Christie that retailed for seven shillings and a sixpenny bit in the UK edition.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Murder on the Orient Express (2017 film)",
"Murder on the Orient Express",
"Sixpence (British coin)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1266,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
18731
|
Franjo Maixner served as a rector of the University of Zagreb and taught Latin language, he never wrote works on grammar.
|
{
"evidence": [
"Franjo Maixner_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:\nFranjo Maixner served as a rector of the University of Zagreb and taught Latin language, he never wrote works on grammar.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Franjo Maixner_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1267,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.536",
"value": "?search="
},
{
"operation": "search",
"time": "20.807",
"value": "Franjo Maixner"
},
{
"operation": "Now on",
"time": "22.123",
"value": "Franjo Maixner"
},
{
"operation": "Highlighting",
"time": "30.591",
"value": "Franjo Maixner_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "46.588",
"value": "Franjo Maixner_sentence_0"
},
{
"operation": "Highlighting",
"time": "49.485",
"value": "Franjo Maixner_sentence_6"
},
{
"operation": "finish",
"time": "67.914",
"value": "finish"
}
]
}
|
exfever_3084
|
The Oranges is the first film directed by Julian Farino.
|
{
"evidence": [
"Leighton Meester",
"The Oranges (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:\nThe Oranges is the first film directed by Julian Farino.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Leighton Meester",
"The Oranges (film)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1268,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
82130
|
NADH dehydrogenase 1 alpha subcomplex subunit 3 is a protein in humans that is encoded by the NDUFA3 gene, which is gene codes for a subunit of Complex I of the respiratory chain that acts as an accessory subunit of the complex and is believed not to be involved in catalysis.
|
{
"evidence": [
"NDUFA3_sentence_0",
"NDUFA3_sentence_9",
"NDUFA3_sentence_10"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nNADH dehydrogenase 1 alpha subcomplex subunit 3 is a protein in humans that is encoded by the NDUFA3 gene, which is gene codes for a subunit of Complex I of the respiratory chain that acts as an accessory subunit of the complex and is believed not to be involved in catalysis.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"NDUFA3_sentence_0",
"NDUFA3_sentence_9",
"NDUFA3_sentence_10"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1269,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.416",
"value": "?search="
},
{
"operation": "search",
"time": "27.206",
"value": "Contains...NADH dehydrogenase 1 alpha subcomplex subunit 3"
},
{
"operation": "Now on",
"time": "29.908",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "40.036",
"value": "NDUFA3"
},
{
"operation": "Now on",
"time": "41.084",
"value": "NDUFA3"
},
{
"operation": "Highlighting",
"time": "119.294",
"value": "NDUFA3_sentence_0"
},
{
"operation": "Highlighting",
"time": "244.494",
"value": "NDUFA3_sentence_9"
},
{
"operation": "Highlighting",
"time": "273.689",
"value": "NDUFA3_sentence_10"
},
{
"operation": "finish",
"time": "342.54",
"value": "finish"
}
]
}
|
exfever_3568
|
Willie Nelson is an American musician, actor, and activist who has been involved in activism for the use of a fuel that is produced over a short time span from biomass.
|
{
"evidence": [
"Willie Nelson",
"Biofuel"
],
"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:\nWillie Nelson is an American musician, actor, and activist who has been involved in activism for the use of a fuel that is produced over a short time span from biomass.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Willie Nelson",
"Biofuel"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1270,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
77421
|
Blues Forever received reviews from music critics such as Allmusic, The Penguin Guide to Jazz, and The Rolling Stone Jazz Record Guide, previously known as The Rolling Stone Record Guide.
|
{
"evidence": [
"Blues Forever_sentence_1",
"Blues Forever_sentence_3",
"Blues Forever_sentence_2",
"The Rolling Stone Album Guide_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:\nBlues Forever received reviews from music critics such as Allmusic, The Penguin Guide to Jazz, and The Rolling Stone Jazz Record Guide, previously known as The Rolling Stone Record Guide.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Blues Forever_sentence_1",
"Blues Forever_sentence_3",
"Blues Forever_sentence_2",
"The Rolling Stone Album Guide_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1271,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.477",
"value": "undefined"
},
{
"operation": "search",
"time": "9.464",
"value": "Blues Forever"
},
{
"operation": "Now on",
"time": "10.111",
"value": "Blues Forever"
},
{
"operation": "Highlighting",
"time": "43.88",
"value": "Blues Forever_sentence_1"
},
{
"operation": "Highlighting",
"time": "48.956",
"value": "Blues Forever_sentence_3"
},
{
"operation": "Highlighting",
"time": "51.743",
"value": "Blues Forever_sentence_2"
},
{
"operation": "hyperlink",
"time": "57.441",
"value": "The Rolling Stone Jazz Record Guide"
},
{
"operation": "Now on",
"time": "59.45",
"value": "The Rolling Stone Album Guide"
},
{
"operation": "Highlighting",
"time": "60.497",
"value": "The Rolling Stone Album Guide_sentence_0"
},
{
"operation": "finish",
"time": "101.399",
"value": "finish"
}
]
}
|
93467
|
The IAS machine, or "von Neumann" machine, was built in 1945-51, and was the first electronic robot to be built at the Institute for Advanced Study in Roswell, New Mexico.
|
{
"evidence": [
"IAS machine_sentence_0"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe IAS machine, or \"von Neumann\" machine, was built in 1945-51, and was the first electronic robot to be built at the Institute for Advanced Study in Roswell, New Mexico.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"IAS machine_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1272,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.919",
"value": "?search="
},
{
"operation": "search",
"time": "10.2",
"value": "IAS machine"
},
{
"operation": "Now on",
"time": "11.246",
"value": "IAS machine"
},
{
"operation": "Highlighting",
"time": "14.386",
"value": "IAS machine_sentence_1"
},
{
"operation": "Highlighting",
"time": "17.402",
"value": "IAS machine_sentence_2"
},
{
"operation": "Highlighting",
"time": "28.367",
"value": "IAS machine_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "30.667",
"value": "IAS machine_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "31.723",
"value": "IAS machine_sentence_2"
},
{
"operation": "finish",
"time": "44.585",
"value": "finish"
}
]
}
|
4190
|
According to Bostrom, a computer program that faithfully emulates a human brain but with processing speeds faster than a human when strictly comparing speeds as Biological neurons operate at about 200 Hz, while a modern microprocessor operates at 2,000,000,000 Hz; same with comparing the human axons that travel at around 120m/s while a computer signal travels at the speed of sound.
|
{
"evidence": [
"AI takeover_sentence_39",
"AI takeover_sentence_40",
"AI takeover_sentence_41"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAccording to Bostrom, a computer program that faithfully emulates a human brain but with processing speeds faster than a human when strictly comparing speeds as Biological neurons operate at about 200 Hz, while a modern microprocessor operates at 2,000,000,000 Hz; same with comparing the human axons that travel at around 120m/s while a computer signal travels at the speed of sound.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"AI takeover_sentence_39",
"AI takeover_sentence_40",
"AI takeover_sentence_41"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1273,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "36.899",
"value": "bostrom"
},
{
"operation": "search",
"time": "55.026",
"value": "AI takeover"
},
{
"operation": "Page search",
"time": "59.65",
"value": "bostrom"
},
{
"operation": "Highlighting",
"time": "78.721",
"value": "AI takeover_sentence_39"
},
{
"operation": "Highlighting",
"time": "95.738",
"value": "AI takeover_sentence_40"
},
{
"operation": "Highlighting",
"time": "102.665",
"value": "AI takeover_sentence_41"
},
{
"operation": "finish",
"time": "120.09",
"value": "finish"
}
]
}
|
86354
|
32532 Thereus (named after thēreios bia 'beastly strength') is a 80 kilometers in diameter centaur from the outer Solar System that was discovered on 9 August 2001 by astronomers in NEAT in the the Palomar Observatory.
|
{
"evidence": [
"32532 Thereus_sentence_0",
"32532 Thereus_sentence_2",
"32532 Thereus_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:\n32532 Thereus (named after thēreios bia 'beastly strength') is a 80 kilometers in diameter centaur from the outer Solar System that was discovered on 9 August 2001 by astronomers in NEAT in the the Palomar Observatory.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"32532 Thereus_sentence_0",
"32532 Thereus_sentence_2",
"32532 Thereus_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1274,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.756",
"value": "?search="
},
{
"operation": "search",
"time": "7.057",
"value": "32532 Thereus"
},
{
"operation": "Now on",
"time": "8.18",
"value": "32532 Thereus"
},
{
"operation": "Highlighting",
"time": "37.139",
"value": "32532 Thereus_sentence_0"
},
{
"operation": "Highlighting",
"time": "50.566",
"value": "32532 Thereus_sentence_2"
},
{
"operation": "Highlighting",
"time": "70.3",
"value": "32532 Thereus_sentence_1"
},
{
"operation": "Highlighting",
"time": "94.309",
"value": "32532 Thereus_cell_0_7_1"
},
{
"operation": "Highlighting",
"time": "94.332",
"value": "32532 Thereus_header_cell_0_7_0"
},
{
"operation": "Highlighting",
"time": "94.333",
"value": "32532 Thereus_header_cell_0_4_0"
},
{
"operation": "Highlighting",
"time": "96.417",
"value": "32532 Thereus_cell_0_9_1"
},
{
"operation": "Highlighting",
"time": "96.439",
"value": "32532 Thereus_header_cell_0_9_0"
},
{
"operation": "Highlighting",
"time": "121.736",
"value": "32532 Thereus_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "121.753",
"value": "32532 Thereus_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "121.753",
"value": "32532 Thereus_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "122.646",
"value": "32532 Thereus_cell_0_2_1"
},
{
"operation": "Highlighting",
"time": "122.662",
"value": "32532 Thereus_header_cell_0_2_0"
},
{
"operation": "Highlighting",
"time": "123.383",
"value": "32532 Thereus_cell_0_3_1"
},
{
"operation": "Highlighting",
"time": "123.4",
"value": "32532 Thereus_header_cell_0_3_0"
},
{
"operation": "Highlighting deleted",
"time": "131.22",
"value": "32532 Thereus_cell_0_9_1"
},
{
"operation": "Highlighting deleted",
"time": "131.239",
"value": "32532 Thereus_header_cell_0_9_0"
},
{
"operation": "Highlighting deleted",
"time": "131.877",
"value": "32532 Thereus_cell_0_7_1"
},
{
"operation": "Highlighting deleted",
"time": "131.894",
"value": "32532 Thereus_header_cell_0_7_0"
},
{
"operation": "Highlighting deleted",
"time": "131.895",
"value": "32532 Thereus_header_cell_0_4_0"
},
{
"operation": "Highlighting deleted",
"time": "132.639",
"value": "32532 Thereus_cell_0_3_1"
},
{
"operation": "Highlighting deleted",
"time": "132.656",
"value": "32532 Thereus_header_cell_0_3_0"
},
{
"operation": "Highlighting deleted",
"time": "133.232",
"value": "32532 Thereus_cell_0_2_1"
},
{
"operation": "Highlighting deleted",
"time": "133.248",
"value": "32532 Thereus_header_cell_0_2_0"
},
{
"operation": "Highlighting deleted",
"time": "133.761",
"value": "32532 Thereus_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "133.777",
"value": "32532 Thereus_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "133.778",
"value": "32532 Thereus_header_cell_0_0_0"
},
{
"operation": "finish",
"time": "150.06",
"value": "finish"
}
]
}
|
exfever_4063
|
Queen Victoria was the only legitimate grandchild of George III.
|
{
"evidence": [
"Queen Victoria",
"British Empire"
],
"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:\nQueen Victoria was the only legitimate grandchild of George III.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Queen Victoria",
"British Empire"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1275,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
20896
|
Deep Down Happy is the debut studio album by English alternative rock band Sports Team, On 12 June 2019, Sports Team released the first single "Here It Comes,Again", and On 12 June 2020, the album ranked at No. 2 on the Official Charts Company's UK Albums Chart.
|
{
"evidence": [
"Deep Down Happy_sentence_0",
"Deep Down Happy_sentence_3",
"Deep Down Happy_sentence_7",
"Deep Down Happy_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:\nDeep Down Happy is the debut studio album by English alternative rock band Sports Team, On 12 June 2019, Sports Team released the first single \"Here It Comes,Again\", and On 12 June 2020, the album ranked at No. 2 on the Official Charts Company's UK Albums Chart.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Deep Down Happy_sentence_0",
"Deep Down Happy_sentence_3",
"Deep Down Happy_sentence_7",
"Deep Down Happy_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1276,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.904",
"value": "?search="
},
{
"operation": "search",
"time": "11.721",
"value": "Deep Down Happy"
},
{
"operation": "Now on",
"time": "12.586",
"value": "Deep Down Happy"
},
{
"operation": "Highlighting",
"time": "27.377",
"value": "Deep Down Happy_sentence_0"
},
{
"operation": "Highlighting",
"time": "257.032",
"value": "Deep Down Happy_sentence_3"
},
{
"operation": "Highlighting",
"time": "270.179",
"value": "Deep Down Happy_sentence_7"
},
{
"operation": "Highlighting",
"time": "273.792",
"value": "Deep Down Happy_sentence_8"
},
{
"operation": "finish",
"time": "283.017",
"value": "finish"
}
]
}
|
exfever_5452
|
NSYNC consisted of members that included an American singer who writes and produces music for acts, including an American pop singer.
|
{
"evidence": [
"NSYNC",
"JC Chasez",
"David Archuleta"
],
"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:\nNSYNC consisted of members that included an American singer who writes and produces music for acts, including an American pop singer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"NSYNC",
"JC Chasez",
"David Archuleta"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1277,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
4831
|
Huang Xiaowei began work in 1998 and became the Vice Minister of Supervision after 14 years.
|
{
"evidence": [
"Huang Xiaowei_sentence_4",
"Huang Xiaowei_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:\nHuang Xiaowei began work in 1998 and became the Vice Minister of Supervision after 14 years.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Huang Xiaowei_sentence_4",
"Huang Xiaowei_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1278,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "8.415",
"value": "Huang Xiaowei"
},
{
"operation": "Highlighting",
"time": "15.751",
"value": "Huang Xiaowei_sentence_4"
},
{
"operation": "Highlighting",
"time": "17.974",
"value": "Huang Xiaowei_sentence_7"
},
{
"operation": "finish",
"time": "35.978",
"value": "finish"
}
]
}
|
exfever_3384
|
Travis Landon Barker is a musician who serves as the drummer for an American rock band that has a musical style that blends catchy pop melodies with the fast-paced music genre that emerged in the mid-1970s.
|
{
"evidence": [
"Travis Barker",
"Blink-182",
"Punk rock"
],
"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:\nTravis Landon Barker is a musician who serves as the drummer for an American rock band that has a musical style that blends catchy pop melodies with the fast-paced music genre that emerged in the mid-1970s.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Travis Barker",
"Blink-182",
"Punk rock"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1279,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8694
|
The Democratic Party is one of the two major contemporary political parties in the United States that promotes political ideology and reform movements from the early 20th century, including the introduction of toxins into the natural environment that creates undesirable change.
|
{
"evidence": [
"Democratic Party (United States)",
"Progressivism in the United States",
"Pollution"
],
"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 Democratic Party is one of the two major contemporary political parties in the United States that promotes political ideology and reform movements from the early 20th century, including the introduction of toxins into the natural environment that creates undesirable change.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Democratic Party (United States)",
"Progressivism in the United States",
"Pollution"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1280,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8761
|
Jacqueline Ruth Weaver is the most prominent Australian actress in the film The Disaster Artist.
|
{
"evidence": [
"Jacki Weaver",
"Picnic at Hanging Rock (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:\nJacqueline Ruth Weaver is the most prominent Australian actress in the film The Disaster Artist.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jacki Weaver",
"Picnic at Hanging Rock (film)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1281,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5964
|
Argo is an American historical drama-thriller film based on a Wired article by a producer who contributed to the documentary, The Guilty.
|
{
"evidence": [
"Argo (2012 film)",
"Joshuah Bearman"
],
"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:\nArgo is an American historical drama-thriller film based on a Wired article by a producer who contributed to the documentary, The Guilty.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Argo (2012 film)",
"Joshuah Bearman"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1282,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
89354
|
Tom Campbell Black was the Chief Pilot of Wilson Airways until March 1939 with the outbreak of the Second World War.
|
{
"evidence": [
"Tom Campbell Black_sentence_23"
],
"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:\nTom Campbell Black was the Chief Pilot of Wilson Airways until March 1939 with the outbreak of the Second World War.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Tom Campbell Black_sentence_23"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1283,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.929",
"value": "?search="
},
{
"operation": "search",
"time": "13.71",
"value": "Contains...Tom Campbell Black"
},
{
"operation": "Now on",
"time": "15.554",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "17.795",
"value": "Tom Campbell Black"
},
{
"operation": "Now on",
"time": "18.272",
"value": "Tom Campbell Black"
},
{
"operation": "Page search",
"time": "75.626",
"value": "wilson"
},
{
"operation": "Page search",
"time": "89.794",
"value": "chief"
},
{
"operation": "Highlighting",
"time": "98.802",
"value": "Tom Campbell Black_sentence_23"
},
{
"operation": "finish",
"time": "115.491",
"value": "finish"
}
]
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.