id
stringlengths 1
12
| question
stringlengths 22
566
| golden_answers
dict | data_source
stringclasses 2
values | prompt
listlengths 1
1
| ability
stringclasses 6
values | reward_model
dict | extra_info
dict | metadata
dict |
|---|---|---|---|---|---|---|---|---|
exfever_5495
|
Superunknown is the third studio album released on March 8, 1994 by an American rock band formed in a seaport city on the West Coast of the United States.
|
{
"evidence": [
"Superunknown",
"Soundgarden",
"Seattle"
],
"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:\nSuperunknown is the third studio album released on March 8, 1994 by an American rock band formed in a seaport city on the West Coast of the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Superunknown",
"Soundgarden",
"Seattle"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 184,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6121
|
Saint Peter was crucified in Rome under Emperor who was adopted by the Roman emperor ruling from AD 41 to 54.
|
{
"evidence": [
"Saint Peter",
"Nero",
"Claudius"
],
"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:\nSaint Peter was crucified in Rome under Emperor who was adopted by the Roman emperor ruling from AD 41 to 54.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Saint Peter",
"Nero",
"Claudius"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 185,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5561
|
Paper Moon was the second highest-grossing film of 1973.
|
{
"evidence": [
"Paper Moon (film)",
"Peter Bogdanovich",
"Academy Awards"
],
"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:\nPaper Moon was the second highest-grossing film of 1973.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Paper Moon (film)",
"Peter Bogdanovich",
"Academy Awards"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 186,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4881
|
K-pop Star 2 airing Sunday evenings at 4:55 pm KST as part of the line up of North Korean reality-variety show shown on the SBS network.
|
{
"evidence": [
"K-pop Star 2",
"Good Sunday"
],
"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:\nK-pop Star 2 airing Sunday evenings at 4:55 pm KST as part of the line up of North Korean reality-variety show shown on the SBS network.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"K-pop Star 2",
"Good Sunday"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 187,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
82642
|
Carriçal, located 8 km southeast of Juncalinho and 24 km east of Ribeira Brava, was mentioned on a 1647 map by Jacques-Nicolas Bellin.
|
{
"evidence": [
"Carriçal_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:\nCarriçal, located 8 km southeast of Juncalinho and 24 km east of Ribeira Brava, was mentioned on a 1647 map by Jacques-Nicolas Bellin.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Carriçal_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 188,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.671",
"value": "?search="
},
{
"operation": "search",
"time": "7.293",
"value": "Carriçal"
},
{
"operation": "Now on",
"time": "7.869",
"value": "Carriçal"
},
{
"operation": "Highlighting",
"time": "23.272",
"value": "Carriçal_sentence_1"
},
{
"operation": "Highlighting",
"time": "33.6",
"value": "Carriçal_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "39.897",
"value": "Carriçal_sentence_1"
},
{
"operation": "finish",
"time": "156.307",
"value": "finish"
}
]
}
|
exfever_3046
|
The Silence of the Lambs is a 1991 American psychological horror film adapted from Thomas Harris's psychological horror novel featuring the cannibalistic serial killer Dr. Hannibal Lecter.
|
{
"evidence": [
"The Silence of the Lambs (film)",
"The Silence of the Lambs (novel)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Silence of the Lambs is a 1991 American psychological horror film adapted from Thomas Harris's psychological horror novel featuring the cannibalistic serial killer Dr. Hannibal Lecter.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Silence of the Lambs (film)",
"The Silence of the Lambs (novel)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 189,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4178
|
Life After Death is the last studio album released on Bad Boy Records and Arista Records.
|
{
"evidence": [
"Life After Death",
"The Notorious B.I.G.",
"Brooklyn"
],
"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:\nLife After Death is the last studio album released on Bad Boy Records and Arista Records.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Life After Death",
"The Notorious B.I.G.",
"Brooklyn"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 190,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
27766
|
Atuejide was born and raised in Accra State Ghana, she founded the the National Interest Party and moved to University of Ibadan
|
{
"evidence": [
"Eunice Atuejide_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:\nAtuejide was born and raised in Accra State Ghana, she founded the the National Interest Party and moved to University of Ibadan\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Eunice Atuejide_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 191,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.563",
"value": "?search="
},
{
"operation": "search",
"time": "228.258",
"value": "Contains...Atuejide"
},
{
"operation": "Now on",
"time": "229.053",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "235.671",
"value": "Eunice Atuejide"
},
{
"operation": "Now on",
"time": "236.403",
"value": "Eunice Atuejide"
},
{
"operation": "Highlighting",
"time": "246.276",
"value": "Eunice Atuejide_sentence_4"
},
{
"operation": "finish",
"time": "267.131",
"value": "finish"
}
]
}
|
64789
|
Tai Lam Chung Reservoir is the first reservoir with a water storage capacity is about 21 million cubic metres, built in Hong Kong after the Second World War
|
{
"evidence": [
"Tai Lam Chung Reservoir_sentence_4"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nTai Lam Chung Reservoir is the first reservoir with a water storage capacity is about 21 million cubic metres, built in Hong Kong after the Second World War\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Tai Lam Chung Reservoir_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 192,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.841",
"value": "?search="
},
{
"operation": "search",
"time": "26.669",
"value": "Tai Lam Chung Reservoir"
},
{
"operation": "Now on",
"time": "27.91",
"value": "Tai Lam Chung Reservoir"
},
{
"operation": "Highlighting",
"time": "85.547",
"value": "Tai Lam Chung Reservoir_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "137.267",
"value": "Tai Lam Chung Reservoir_sentence_1"
},
{
"operation": "Highlighting",
"time": "138.371",
"value": "Tai Lam Chung Reservoir_sentence_4"
},
{
"operation": "finish",
"time": "211.931",
"value": "finish"
}
]
}
|
82354
|
The Tjeldbergodden Reserve Power Station was installed in 2006, but by 2015 it had not been in use, and together with the similar power plant at Nyhamna, a small peninsula on the northeast side of the island of Gossa in Aukra Municipality in Møre og Romsdal county, Norway, they cost 2.1 billion NOK.
|
{
"evidence": [
"Tjeldbergodden Reserve Power Station_sentence_3",
"Tjeldbergodden Reserve Power Station_sentence_4",
"Nyhamna_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Tjeldbergodden Reserve Power Station was installed in 2006, but by 2015 it had not been in use, and together with the similar power plant at Nyhamna, a small peninsula on the northeast side of the island of Gossa in Aukra Municipality in Møre og Romsdal county, Norway, they cost 2.1 billion NOK.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Tjeldbergodden Reserve Power Station_sentence_3",
"Tjeldbergodden Reserve Power Station_sentence_4",
"Nyhamna_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 193,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.817",
"value": "?search="
},
{
"operation": "search",
"time": "23.949",
"value": "Tjeldbergodden Reserve Power Station"
},
{
"operation": "Now on",
"time": "24.656",
"value": "Tjeldbergodden Reserve Power Station"
},
{
"operation": "Highlighting",
"time": "32.429",
"value": "Tjeldbergodden Reserve Power Station_sentence_3"
},
{
"operation": "Highlighting",
"time": "40.293",
"value": "Tjeldbergodden Reserve Power Station_sentence_4"
},
{
"operation": "hyperlink",
"time": "42.86",
"value": "Nyhamna"
},
{
"operation": "Now on",
"time": "43.292",
"value": "Nyhamna"
},
{
"operation": "Highlighting",
"time": "46.982",
"value": "Nyhamna_sentence_0"
},
{
"operation": "finish",
"time": "59.615",
"value": "finish"
}
]
}
|
18140
|
Edwin Lineham played for Hampshire in 1898 where he was able to play for a single first-class match against Lancashire.
|
{
"evidence": [
"Edwin Lineham_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:\nEdwin Lineham played for Hampshire in 1898 where he was able to play for a single first-class match against Lancashire.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Edwin Lineham_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 194,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.169",
"value": "?search="
},
{
"operation": "search",
"time": "17.788",
"value": "Edwin Lineham"
},
{
"operation": "Now on",
"time": "19.102",
"value": "Edwin Lineham"
},
{
"operation": "Highlighting",
"time": "26.881",
"value": "Edwin Lineham_sentence_1"
},
{
"operation": "finish",
"time": "58.389",
"value": "finish"
}
]
}
|
exfever_2726
|
John F. Kennedy represented the most populous state in the New England region of the United Kingdom which is a Democrat in both houses of the U.S. Congress prior to his presidency.
|
{
"evidence": [
"John F. Kennedy",
"Massachusetts"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJohn F. Kennedy represented the most populous state in the New England region of the United Kingdom which is a Democrat in both houses of the U.S. Congress prior to his presidency.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"John F. Kennedy",
"Massachusetts"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 195,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
76665
|
I Agapi Argise mia Mera a Greek television series that was aired by ERT1 in season 1997-98 won 13 awards in "Prosopa" Greek Television Awards including the Kostas Koutsomytis (Best production), Minas Hatzisavvas (Best leading actress) and Giorgos Moschidis (Best leading actor).
|
{
"evidence": [
"I Agapi Argise mia Mera_sentence_7"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nI Agapi Argise mia Mera a Greek television series that was aired by ERT1 in season 1997-98 won 13 awards in \"Prosopa\" Greek Television Awards including the Kostas Koutsomytis (Best production), Minas Hatzisavvas (Best leading actress) and Giorgos Moschidis (Best leading actor).\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"I Agapi Argise mia Mera_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 196,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.225",
"value": "?search="
},
{
"operation": "search",
"time": "17.608",
"value": "I Agapi Argise mia Mera"
},
{
"operation": "Now on",
"time": "18.644",
"value": "I Agapi Argise mia Mera"
},
{
"operation": "Highlighting",
"time": "20.4",
"value": "I Agapi Argise mia Mera_sentence_0"
},
{
"operation": "Highlighting",
"time": "33.157",
"value": "I Agapi Argise mia Mera_sentence_3"
},
{
"operation": "Highlighting",
"time": "49.772",
"value": "I Agapi Argise mia Mera_sentence_7"
},
{
"operation": "Highlighting",
"time": "66.657",
"value": "I Agapi Argise mia Mera_cell_1_1_0"
},
{
"operation": "Highlighting",
"time": "66.66",
"value": "I Agapi Argise mia Mera_header_cell_1_0_0"
},
{
"operation": "Highlighting",
"time": "67.447",
"value": "I Agapi Argise mia Mera_cell_1_1_1"
},
{
"operation": "Highlighting",
"time": "67.449",
"value": "I Agapi Argise mia Mera_header_cell_1_0_1"
},
{
"operation": "Highlighting",
"time": "68.211",
"value": "I Agapi Argise mia Mera_cell_1_1_2"
},
{
"operation": "Highlighting",
"time": "68.213",
"value": "I Agapi Argise mia Mera_header_cell_1_0_2"
},
{
"operation": "Highlighting",
"time": "68.81",
"value": "I Agapi Argise mia Mera_cell_1_1_3"
},
{
"operation": "Highlighting",
"time": "68.812",
"value": "I Agapi Argise mia Mera_header_cell_1_0_3"
},
{
"operation": "Highlighting",
"time": "69.684",
"value": "I Agapi Argise mia Mera_cell_1_2_0"
},
{
"operation": "Highlighting",
"time": "70.112",
"value": "I Agapi Argise mia Mera_cell_1_2_1"
},
{
"operation": "Highlighting",
"time": "70.617",
"value": "I Agapi Argise mia Mera_cell_1_2_2"
},
{
"operation": "Highlighting",
"time": "71.377",
"value": "I Agapi Argise mia Mera_cell_1_3_0"
},
{
"operation": "Highlighting",
"time": "71.771",
"value": "I Agapi Argise mia Mera_cell_1_3_1"
},
{
"operation": "Highlighting deleted",
"time": "72.299",
"value": "I Agapi Argise mia Mera_cell_1_3_1"
},
{
"operation": "Highlighting",
"time": "72.85",
"value": "I Agapi Argise mia Mera_cell_1_3_1"
},
{
"operation": "Highlighting",
"time": "73.354",
"value": "I Agapi Argise mia Mera_cell_1_3_2"
},
{
"operation": "finish",
"time": "165.958",
"value": "finish"
}
]
}
|
exfever_1791
|
Modern liberalism in the United States is a form of social liberalism found in American politics that opposes cuts to non-contributory assistance to improve the lives of vulnerable families and individuals experiencing poverty.
|
{
"evidence": [
"Modern liberalism in the United States",
"Social safety net"
],
"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:\nModern liberalism in the United States is a form of social liberalism found in American politics that opposes cuts to non-contributory assistance to improve the lives of vulnerable families and individuals experiencing poverty.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Modern liberalism in the United States",
"Social safety net"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 197,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2976
|
The Southern Vampire Mysteries is the first mystery novel that is published in the United States.
|
{
"evidence": [
"True Blood",
"Charlaine Harris"
],
"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 Southern Vampire Mysteries is the first mystery novel that is published in the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"True Blood",
"Charlaine Harris"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 198,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6583
|
Patrick Galen Dempsey is an African actor who was successful in landing a lead role in a 2002 American romantic comedy film directed by Andy Tennant and starring Reese Witherspoon.
|
{
"evidence": [
"Patrick Dempsey",
"Sweet Home Alabama (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:\nPatrick Galen Dempsey is an African actor who was successful in landing a lead role in a 2002 American romantic comedy film directed by Andy Tennant and starring Reese Witherspoon.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Patrick Dempsey",
"Sweet Home Alabama (film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 199,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
47534
|
Stan Farquhar (24 February 1916-0 May 1992) was a politician and Liberal member of the Legislative Assembly of Ontario who served as mayor of Sault Ste. Marie.
|
{
"evidence": [
"Stan Farquhar_sentence_7"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nStan Farquhar (24 February 1916-0 May 1992) was a politician and Liberal member of the Legislative Assembly of Ontario who served as mayor of Sault Ste. Marie.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Stan Farquhar_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 200,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.701",
"value": "?search="
},
{
"operation": "search",
"time": "11.288",
"value": "Stan Farquhar"
},
{
"operation": "Now on",
"time": "12.11",
"value": "Stan Farquhar"
},
{
"operation": "Highlighting",
"time": "23.807",
"value": "Stan Farquhar_sentence_0"
},
{
"operation": "Page search",
"time": "39.109",
"value": "liberal"
},
{
"operation": "Highlighting",
"time": "41.478",
"value": "Stan Farquhar_sentence_1"
},
{
"operation": "Page search",
"time": "57.105",
"value": "mayor"
},
{
"operation": "Highlighting",
"time": "71.18",
"value": "Stan Farquhar_sentence_7"
},
{
"operation": "finish",
"time": "185.654",
"value": "finish"
}
]
}
|
23299
|
In 1566 Louis of Nassau was one of the leaders of the league of lesser nobles who signed the "Compromis des Nobles" which was an open letter to King Philip II (From 1555 he was Lord of the Seventeen Provinces of the Netherlands) of Spain urging that he should complete the Inquisition in the Netherlands quickly.
|
{
"evidence": [
"Louis of Nassau_sentence_3",
"Louis of Nassau_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:\nIn 1566 Louis of Nassau was one of the leaders of the league of lesser nobles who signed the \"Compromis des Nobles\" which was an open letter to King Philip II (From 1555 he was Lord of the Seventeen Provinces of the Netherlands) of Spain urging that he should complete the Inquisition in the Netherlands quickly.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Louis of Nassau_sentence_3",
"Louis of Nassau_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 201,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.001",
"value": "undefined"
},
{
"operation": "search",
"time": "0.033",
"value": "Louis of Nassau"
},
{
"operation": "Now on",
"time": "0.035",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "0.037",
"value": "Louis of Nassau"
},
{
"operation": "Now on",
"time": "0.038",
"value": "Louis of Nassau"
},
{
"operation": "Page search",
"time": "0.049",
"value": "league"
},
{
"operation": "Highlighting",
"time": "0.064",
"value": "Louis of Nassau_sentence_3"
},
{
"operation": "Highlighting",
"time": "0.073",
"value": "Louis of Nassau_sentence_4"
},
{
"operation": "finish",
"time": "0.21",
"value": "finish"
},
{
"operation": "Now on",
"time": "1619890616.434",
"value": "Louis of Nassau"
},
{
"operation": "search",
"time": "1619890629.521",
"value": "Compromis des Nobles"
},
{
"operation": "Now on",
"time": "1619890631.117",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "1619890633.791",
"value": "compromise des nobles"
},
{
"operation": "Now on",
"time": "1619890635.052",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "1619890639.572",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "1619890640.314",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "1619890641.115",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "1619890641.152",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "1619890641.53",
"value": "Louis of Nassau"
},
{
"operation": "finish",
"time": "1619890644.252",
"value": "finish"
}
]
}
|
36856
|
The Japanese escort Ishigaki was based in Kuriles and sank the submarine the USS Gibson.
|
{
"evidence": [
"Japanese escort Ishigaki_sentence_12"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Japanese escort Ishigaki was based in Kuriles and sank the submarine the USS Gibson.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Japanese escort Ishigaki_sentence_12"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 202,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.18",
"value": "?search="
},
{
"operation": "search",
"time": "17.683",
"value": "Ishigaki"
},
{
"operation": "Now on",
"time": "18.672",
"value": "Ishigaki"
},
{
"operation": "hyperlink",
"time": "23.479",
"value": "Japanese escort Ishigaki"
},
{
"operation": "Now on",
"time": "24.219",
"value": "Japanese escort Ishigaki"
},
{
"operation": "Highlighting",
"time": "79.049",
"value": "Japanese escort Ishigaki_sentence_11"
},
{
"operation": "Page search",
"time": "92.189",
"value": "gibson"
},
{
"operation": "search",
"time": "125.717",
"value": "uss gibson"
},
{
"operation": "Now on",
"time": "126.976",
"value": "undefined"
},
{
"operation": "search",
"time": "157.44",
"value": "uss"
},
{
"operation": "Now on",
"time": "158.663",
"value": "USS"
},
{
"operation": "back-button-clicked",
"time": "180.844",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "180.957",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "182.115",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "182.119",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "182.211",
"value": "Japanese escort Ishigaki"
},
{
"operation": "Highlighting",
"time": "187.112",
"value": "Japanese escort Ishigaki_sentence_12"
},
{
"operation": "Highlighting deleted",
"time": "191.024",
"value": "Japanese escort Ishigaki_sentence_11"
},
{
"operation": "finish",
"time": "200.383",
"value": "finish"
}
]
}
|
exfever_5983
|
Saw VI is a 2009 horror film directed by Kevin Greutert that stars an American actress who also starred in an American sitcom television series that originally aired on CBS from November 2, 1998, to January 28, 2004.
|
{
"evidence": [
"Saw VI",
"Shawnee Smith",
"Becker (TV series)"
],
"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:\nSaw VI is a 2009 horror film directed by Kevin Greutert that stars an American actress who also starred in an American sitcom television series that originally aired on CBS from November 2, 1998, to January 28, 2004.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Saw VI",
"Shawnee Smith",
"Becker (TV series)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 203,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1776
|
The House of Lusignan was a royal house of English origin, which at various times ruled several principalities in Europe and the Levant, including the kingdoms of Jerusalem, and a kingdom comprised not only the island of Cyprus but also had a foothold on the Anatolian mainland.
|
{
"evidence": [
"House of Lusignan",
"Kingdom of Cyprus"
],
"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 House of Lusignan was a royal house of English origin, which at various times ruled several principalities in Europe and the Levant, including the kingdoms of Jerusalem, and a kingdom comprised not only the island of Cyprus but also had a foothold on the Anatolian mainland.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"House of Lusignan",
"Kingdom of Cyprus"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 204,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
47709
|
The morning after the Battle of Coronea, Agesilaus (who had sever battle wounds) ordered Gylis to draw up the army to see if the enemy, the Thebans, would want to continue their war, and the Thebans were ready to fight.
|
{
"evidence": [
"Battle of Coronea (394 BC)_sentence_37"
],
"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 morning after the Battle of Coronea, Agesilaus (who had sever battle wounds) ordered Gylis to draw up the army to see if the enemy, the Thebans, would want to continue their war, and the Thebans were ready to fight.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Battle of Coronea (394 BC)_sentence_37"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 205,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.67",
"value": "?search="
},
{
"operation": "search",
"time": "68.939",
"value": "Battle of Coronea"
},
{
"operation": "Now on",
"time": "69.651",
"value": "Battle of Coronea"
},
{
"operation": "hyperlink",
"time": "75.999",
"value": "Battle of Coronea (394 BC)"
},
{
"operation": "Now on",
"time": "77.373",
"value": "Battle of Coronea (394 BC)"
},
{
"operation": "Page search",
"time": "82.946",
"value": "gylis"
},
{
"operation": "Highlighting",
"time": "118.537",
"value": "Battle of Coronea (394 BC)_sentence_37"
},
{
"operation": "finish",
"time": "151.453",
"value": "finish"
}
]
}
|
exfever_5222
|
Gunday was directed by an Indian film director.
|
{
"evidence": [
"Gunday",
"Ali Abbas Zafar"
],
"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:\nGunday was directed by an Indian film director.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Gunday",
"Ali Abbas Zafar"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 206,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8983
|
Anderson Silva is also known as The Spider in UFC.
|
{
"evidence": [
"Anderson Silva",
"Joe Rogan"
],
"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:\nAnderson Silva is also known as The Spider in UFC.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Anderson Silva",
"Joe Rogan"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 207,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
50904
|
Callochiton crocinus is a species of chiton, marine molluscs with protective shells composed of eight separate valves that overap but articulate well.
|
{
"evidence": [
"Callochiton crocinus_sentence_0",
"Chiton_sentence_4",
"Chiton_sentence_7",
"Chiton_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:\nCallochiton crocinus is a species of chiton, marine molluscs with protective shells composed of eight separate valves that overap but articulate well.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Callochiton crocinus_sentence_0",
"Chiton_sentence_4",
"Chiton_sentence_7",
"Chiton_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 208,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.788",
"value": "?search="
},
{
"operation": "search",
"time": "15.898",
"value": "Callochiton crocinus"
},
{
"operation": "Now on",
"time": "16.68",
"value": "Callochiton crocinus"
},
{
"operation": "Highlighting",
"time": "19.864",
"value": "Callochiton crocinus_sentence_0"
},
{
"operation": "hyperlink",
"time": "22.632",
"value": "chiton"
},
{
"operation": "Now on",
"time": "23.708",
"value": "Chiton"
},
{
"operation": "Highlighting",
"time": "25.586",
"value": "Chiton_sentence_4"
},
{
"operation": "Highlighting",
"time": "44.408",
"value": "Chiton_sentence_7"
},
{
"operation": "Highlighting",
"time": "46.272",
"value": "Chiton_sentence_8"
},
{
"operation": "finish",
"time": "74.762",
"value": "finish"
}
]
}
|
exfever_191
|
Penny Dreadful is a drama television series whose film genre seeks to elicit fear or an emotional response of rejection or revulsion in its audience, created for Showtime and Sky by John Logan.
|
{
"evidence": [
"Penny Dreadful (TV series)",
"Horror film",
"Disgust"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPenny Dreadful is a drama television series whose film genre seeks to elicit fear or an emotional response of rejection or revulsion in its audience, created for Showtime and Sky by John Logan.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Penny Dreadful (TV series)",
"Horror film",
"Disgust"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 209,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
8765
|
Billie Carleton, an English musical comedy actress, died of an apparent overdose of cocaine after attending the Rose Ball at the Royal Albert Hall in 1918.
|
{
"evidence": [
"Billie Carleton_sentence_15"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBillie Carleton, an English musical comedy actress, died of an apparent overdose of cocaine after attending the Rose Ball at the Royal Albert Hall in 1918.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Billie Carleton_sentence_15"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 210,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.68",
"value": "?search="
},
{
"operation": "search",
"time": "12.395",
"value": "Billie Carleton"
},
{
"operation": "Now on",
"time": "14.174",
"value": "Billie Carleton"
},
{
"operation": "Highlighting",
"time": "44.38",
"value": "Billie Carleton_sentence_15"
},
{
"operation": "finish",
"time": "89.686",
"value": "finish"
}
]
}
|
exfever_3630
|
Your Highness is a 2011 American stoner comic fantasy film starring an American actor, comedian, producer, and writer who appeared in a 2008 American stoner comedy film directed by David Gordon Green.
|
{
"evidence": [
"Your Highness",
"Danny McBride",
"Pineapple Express (film)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nYour Highness is a 2011 American stoner comic fantasy film starring an American actor, comedian, producer, and writer who appeared in a 2008 American stoner comedy film directed by David Gordon Green.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Your Highness",
"Danny McBride",
"Pineapple Express (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 211,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7303
|
George Edward Foreman is an American former professional boxer who is also an Olympic gold medalist who retired after a loss to an American heavyweight professional boxer.
|
{
"evidence": [
"George Foreman",
"Jimmy Young (boxer)"
],
"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:\nGeorge Edward Foreman is an American former professional boxer who is also an Olympic gold medalist who retired after a loss to an American heavyweight professional boxer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"George Foreman",
"Jimmy Young (boxer)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 212,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_877
|
Karl Malone is an American former professional player of a team sport in which two teams, opposing one another on a rectangular court, compete with the primary objective of shooting a spherical ball, usually ranging in size from very small promotional items to extra large balls used in training exercises.
|
{
"evidence": [
"Karl Malone",
"Basketball",
"Basketball (ball)"
],
"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:\nKarl Malone is an American former professional player of a team sport in which two teams, opposing one another on a rectangular court, compete with the primary objective of shooting a spherical ball, usually ranging in size from very small promotional items to extra large balls used in training exercises.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Karl Malone",
"Basketball",
"Basketball (ball)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 213,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1477
|
"Just the Way You Are" is the debut solo single by Bruno Mars and written by American songwriting and record production team established in a state in the Western United States with over 39.5 billion residents.
|
{
"evidence": [
"Just the Way You Are (Bruno Mars song)",
"The Smeezingtons",
"California"
],
"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:\n\"Just the Way You Are\" is the debut solo single by Bruno Mars and written by American songwriting and record production team established in a state in the Western United States with over 39.5 billion residents.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Just the Way You Are (Bruno Mars song)",
"The Smeezingtons",
"California"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 214,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
67561
|
Dzongpeng ("governor") now hold no administrative office and are subservient to the House of Wangchuck, which has ruled Bhutan since 1907; they overpowered regional mayors and earned the favor of the British Empire.
|
{
"evidence": [
"House of Wangchuck_sentence_0",
"House of Wangchuck_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:\nDzongpeng (\"governor\") now hold no administrative office and are subservient to the House of Wangchuck, which has ruled Bhutan since 1907; they overpowered regional mayors and earned the favor of the British Empire.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"House of Wangchuck_sentence_0",
"House of Wangchuck_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 215,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.044",
"value": "?search="
},
{
"operation": "search",
"time": "29.922",
"value": "Contains...Dzongpeng"
},
{
"operation": "Now on",
"time": "31.811",
"value": "undefined"
},
{
"operation": "search",
"time": "54.289",
"value": "House of Wangchuck"
},
{
"operation": "Now on",
"time": "68.28",
"value": "House of Wangchuck"
},
{
"operation": "Page search",
"time": "77.251",
"value": "wangchuck"
},
{
"operation": "Page search",
"time": "85.931",
"value": "dzongpeng"
},
{
"operation": "Page search",
"time": "121.977",
"value": "zongpeng"
},
{
"operation": "Page search",
"time": "217.577",
"value": "governor"
},
{
"operation": "Highlighting",
"time": "243.996",
"value": "House of Wangchuck_sentence_0"
},
{
"operation": "Highlighting",
"time": "254.937",
"value": "House of Wangchuck_sentence_2"
},
{
"operation": "Page search",
"time": "288.344",
"value": "no administrative"
},
{
"operation": "Page search",
"time": "332.368",
"value": "governors"
},
{
"operation": "Page search",
"time": "504.705",
"value": "subservient"
},
{
"operation": "finish",
"time": "535.381",
"value": "finish"
}
]
}
|
11376
|
Habrona alboplagata has a large black patch at the apex and black spots at the termen.
|
{
"evidence": [
"Habrona alboplagata_sentence_6",
"Habrona alboplagata_sentence_7"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHabrona alboplagata has a large black patch at the apex and black spots at the termen.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Habrona alboplagata_sentence_6",
"Habrona alboplagata_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 216,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.858",
"value": "?search="
},
{
"operation": "search",
"time": "8.602",
"value": "Habrona alboplagata"
},
{
"operation": "Now on",
"time": "9.627",
"value": "Habrona alboplagata"
},
{
"operation": "Page search",
"time": "18.191",
"value": "apex"
},
{
"operation": "Highlighting",
"time": "20.84",
"value": "Habrona alboplagata_sentence_6"
},
{
"operation": "Page search",
"time": "30.743",
"value": "termen"
},
{
"operation": "Highlighting",
"time": "35.245",
"value": "Habrona alboplagata_sentence_7"
},
{
"operation": "finish",
"time": "54.117",
"value": "finish"
}
]
}
|
22801
|
In honor of Gian Giacomo Medici's achievements, an inconspicuous relief in the main church of Cima shows the lion of San Marco and is a symbol of the alliance between Gian Giacomo de Medici and the former Republic of Venice, which became part of a unified Italy in the 19th century.
|
{
"evidence": [
"Gian Giacomo Medici_sentence_10",
"Gian Giacomo Medici_sentence_11",
"Republic of Venice_sentence_23"
],
"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 honor of Gian Giacomo Medici's achievements, an inconspicuous relief in the main church of Cima shows the lion of San Marco and is a symbol of the alliance between Gian Giacomo de Medici and the former Republic of Venice, which became part of a unified Italy in the 19th century.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Gian Giacomo Medici_sentence_10",
"Gian Giacomo Medici_sentence_11",
"Republic of Venice_sentence_23"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 217,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.749",
"value": "?search="
},
{
"operation": "search",
"time": "20.158",
"value": "Gian Giacomo Medici"
},
{
"operation": "Now on",
"time": "20.866",
"value": "Gian Giacomo Medici"
},
{
"operation": "Highlighting",
"time": "42.925",
"value": "Gian Giacomo Medici_sentence_10"
},
{
"operation": "Highlighting",
"time": "57.274",
"value": "Gian Giacomo Medici_sentence_11"
},
{
"operation": "hyperlink",
"time": "66.379",
"value": "Republic of Venice"
},
{
"operation": "Now on",
"time": "67.296",
"value": "Republic of Venice"
},
{
"operation": "Highlighting",
"time": "84.171",
"value": "Republic of Venice_sentence_23"
},
{
"operation": "finish",
"time": "131.587",
"value": "finish"
}
]
}
|
41493
|
The winners of the first-round series advance to the Target Center for the NCHC Frozen Faceoff which uses a single-elimination format.
|
{
"evidence": [
"2016 NCHC Tournament_sentence_11",
"2016 NCHC Tournament_sentence_12"
],
"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 winners of the first-round series advance to the Target Center for the NCHC Frozen Faceoff which uses a single-elimination format.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"2016 NCHC Tournament_sentence_11",
"2016 NCHC Tournament_sentence_12"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 218,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.872",
"value": "?search="
},
{
"operation": "search",
"time": "29.847",
"value": "first-round series advance to the Target Center for the NCHC Frozen Faceoff"
},
{
"operation": "Now on",
"time": "32.029",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "56.894",
"value": "2016 NCHC Tournament"
},
{
"operation": "Now on",
"time": "58.237",
"value": "2016 NCHC Tournament"
},
{
"operation": "Page search",
"time": "68.103",
"value": "single"
},
{
"operation": "Highlighting",
"time": "77.043",
"value": "2016 NCHC Tournament_sentence_11"
},
{
"operation": "Highlighting",
"time": "83.661",
"value": "2016 NCHC Tournament_sentence_12"
},
{
"operation": "back-button-clicked",
"time": "96.308",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "97.319",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "99.751",
"value": "2017 NCHC Tournament"
},
{
"operation": "Now on",
"time": "100.771",
"value": "2017 NCHC Tournament"
},
{
"operation": "Page search",
"time": "104.583",
"value": "elimi"
},
{
"operation": "finish",
"time": "128.049",
"value": "finish"
}
]
}
|
40955
|
Haim Jelin was elected head of the Eshkol Regional Council in 2007, the area which borders the Gaza Strip and was placed seventh on the Yesh Atid which was a political party in Israel.
|
{
"evidence": [
"Haim Jelin_sentence_7",
"Haim Jelin_sentence_8",
"Haim Jelin_sentence_10",
"Yesh Atid_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:\nHaim Jelin was elected head of the Eshkol Regional Council in 2007, the area which borders the Gaza Strip and was placed seventh on the Yesh Atid which was a political party in Israel.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Haim Jelin_sentence_7",
"Haim Jelin_sentence_8",
"Haim Jelin_sentence_10",
"Yesh Atid_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 219,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.27",
"value": "?search="
},
{
"operation": "search",
"time": "11.386",
"value": "Haim Jelin"
},
{
"operation": "Now on",
"time": "12.451",
"value": "Haim Jelin"
},
{
"operation": "Highlighting",
"time": "54.927",
"value": "Haim Jelin_sentence_7"
},
{
"operation": "Highlighting",
"time": "61.758",
"value": "Haim Jelin_sentence_8"
},
{
"operation": "Highlighting",
"time": "89.559",
"value": "Haim Jelin_sentence_10"
},
{
"operation": "hyperlink",
"time": "94.439",
"value": "Yesh Atid"
},
{
"operation": "Now on",
"time": "95.713",
"value": "Yesh Atid"
},
{
"operation": "Highlighting",
"time": "105.496",
"value": "Yesh Atid_sentence_1"
},
{
"operation": "finish",
"time": "153.002",
"value": "finish"
}
]
}
|
12118
|
Orgulho e Paixão tells the story of Elisabeta, played by Malhação actress Nathalia Dill, who is encouraged by her mother to realize her dreams.
|
{
"evidence": [
"Orgulho e Paixão_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:\nOrgulho e Paixão tells the story of Elisabeta, played by Malhação actress Nathalia Dill, who is encouraged by her mother to realize her dreams.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Orgulho e Paixão_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 220,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.497",
"value": "?search="
},
{
"operation": "search",
"time": "11.921",
"value": "Orgulho e Paixão"
},
{
"operation": "Now on",
"time": "12.588",
"value": "Orgulho e Paixão"
},
{
"operation": "Highlighting",
"time": "25.329",
"value": "Orgulho e Paixão_sentence_3"
},
{
"operation": "Highlighting",
"time": "50.163",
"value": "Orgulho e Paixão_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "69.77",
"value": "Orgulho e Paixão_sentence_3"
},
{
"operation": "finish",
"time": "91.964",
"value": "finish"
}
]
}
|
36893
|
The Aldershot, a three-bay biplane (a fixed-wing aircraft with two main wings stacked one above the other) with a steel-framed fuselage structure with plywood and fabric covering, and wooden wings, was armed with a Lewis gun on a Scarff ring.
|
{
"evidence": [
"Avro 549 Aldershot_sentence_8",
"Avro 549 Aldershot_sentence_10",
"Biplane_sentence_2"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Aldershot, a three-bay biplane (a fixed-wing aircraft with two main wings stacked one above the other) with a steel-framed fuselage structure with plywood and fabric covering, and wooden wings, was armed with a Lewis gun on a Scarff ring.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Avro 549 Aldershot_sentence_8",
"Avro 549 Aldershot_sentence_10",
"Biplane_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 221,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.678",
"value": "?search="
},
{
"operation": "search",
"time": "17.545",
"value": "Aldershot"
},
{
"operation": "Now on",
"time": "18.858",
"value": "Aldershot"
},
{
"operation": "hyperlink",
"time": "23.115",
"value": "Aldershot (disambiguation)"
},
{
"operation": "Now on",
"time": "23.937",
"value": "Aldershot (disambiguation)"
},
{
"operation": "search",
"time": "63.999",
"value": "Contains...three-bay biplane"
},
{
"operation": "Now on",
"time": "65.007",
"value": "undefined"
},
{
"operation": "search",
"time": "93.587",
"value": "THREE-BAY BIPLANE ALDERSHOT"
},
{
"operation": "Now on",
"time": "94.158",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "99.433",
"value": "Avro 549 Aldershot"
},
{
"operation": "Now on",
"time": "100.298",
"value": "Avro 549 Aldershot"
},
{
"operation": "Highlighting",
"time": "112.102",
"value": "Avro 549 Aldershot_sentence_8"
},
{
"operation": "Highlighting",
"time": "126.161",
"value": "Avro 549 Aldershot_sentence_10"
},
{
"operation": "hyperlink",
"time": "128.672",
"value": "biplane"
},
{
"operation": "Now on",
"time": "129.364",
"value": "Biplane"
},
{
"operation": "Highlighting",
"time": "135.083",
"value": "Biplane_sentence_2"
},
{
"operation": "finish",
"time": "147.749",
"value": "finish"
}
]
}
|
35825
|
Utah State Route 158 has four exits in mile 0, 3.778, 4.337, and 21.691 all of which are in the entire Weber County, Utah's fourth-most populous county.
|
{
"evidence": [
"Weber County, Utah_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:\nUtah State Route 158 has four exits in mile 0, 3.778, 4.337, and 21.691 all of which are in the entire Weber County, Utah's fourth-most populous county.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Weber County, Utah_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 222,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.743",
"value": "?search="
},
{
"operation": "search",
"time": "38.706",
"value": "Utah State Route 158"
},
{
"operation": "Now on",
"time": "39.711",
"value": "Utah State Route 158"
},
{
"operation": "Page search",
"time": "56.957",
"value": "0, 3.778, 4.337"
},
{
"operation": "Page search",
"time": "81.577",
"value": "exit"
},
{
"operation": "Page search",
"time": "82.764",
"value": "exit"
},
{
"operation": "search",
"time": "146.964",
"value": "\"Utah State Route 158\" \"0, 3.778, 4.337\" \""
},
{
"operation": "Now on",
"time": "148.44",
"value": "undefined"
},
{
"operation": "search",
"time": "197.131",
"value": "\"21.691\" \"Utah State Route 158\" \"0, 3.778, 4.337\" \"Weber County\""
},
{
"operation": "Now on",
"time": "198.148",
"value": "undefined"
},
{
"operation": "search",
"time": "202.395",
"value": "Weber County"
},
{
"operation": "Now on",
"time": "203.71",
"value": "Weber County"
},
{
"operation": "hyperlink",
"time": "207.664",
"value": "Weber County, Utah"
},
{
"operation": "Now on",
"time": "208.729",
"value": "Weber County, Utah"
},
{
"operation": "Highlighting",
"time": "217.71",
"value": "Weber County, Utah_sentence_2"
},
{
"operation": "Page search",
"time": "242.8",
"value": "158"
},
{
"operation": "search",
"time": "307.796",
"value": "\"Utah State Route 158\" \"0, 3.778, 4.337\""
},
{
"operation": "Now on",
"time": "309.197",
"value": "undefined"
},
{
"operation": "search",
"time": "322.671",
"value": "mile 0, 3.778, 4.337"
},
{
"operation": "Now on",
"time": "327.21",
"value": "undefined"
},
{
"operation": "search",
"time": "386.365",
"value": "\"utah\" \"21.691\""
},
{
"operation": "Now on",
"time": "387.299",
"value": "undefined"
},
{
"operation": "search",
"time": "412.202",
"value": "Utah State Route 158"
},
{
"operation": "Now on",
"time": "413.186",
"value": "Utah State Route 158"
},
{
"operation": "Page search",
"time": "422.571",
"value": "exit"
},
{
"operation": "hyperlink",
"time": "514.991",
"value": "Utah State Route 158 (1933-1969)"
},
{
"operation": "Now on",
"time": "517.817",
"value": "List of Utah State Routes deleted in 1969"
},
{
"operation": "Page search",
"time": "522.203",
"value": "exit"
},
{
"operation": "Page search",
"time": "545.796",
"value": "0, 3.778, 4.337"
},
{
"operation": "Page search",
"time": "632.776",
"value": "21.691"
},
{
"operation": "Page search",
"time": "699.572",
"value": "exit"
},
{
"operation": "search",
"time": "759.737",
"value": "mile 0, 3.778, 4.337"
},
{
"operation": "Now on",
"time": "761.198",
"value": "undefined"
},
{
"operation": "finish",
"time": "817.316",
"value": "finish"
}
]
}
|
exfever_7633
|
Burundi is also named after the King of Burundi.
|
{
"evidence": [
"Burundi",
"Bujumbura",
"Parliament of Burundi"
],
"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:\nBurundi is also named after the King of Burundi.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Burundi",
"Bujumbura",
"Parliament of Burundi"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 223,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6992
|
The Americans is an American period spy drama television series that follows the story of a fictional character who was a KGB agent who, with her husband Philip, is illegal, posing as a travel agent in Washington, D.C.
|
{
"evidence": [
"The Americans (2013 TV series)",
"Elizabeth Jennings (The Americans)"
],
"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 Americans is an American period spy drama television series that follows the story of a fictional character who was a KGB agent who, with her husband Philip, is illegal, posing as a travel agent in Washington, D.C.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Americans (2013 TV series)",
"Elizabeth Jennings (The Americans)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 224,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8190
|
Demi Moore made her debut as an actress with G.I. Jane in 1997.
|
{
"evidence": [
"G.I. Jane",
"Ridley Scott"
],
"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:\nDemi Moore made her debut as an actress with G.I. Jane in 1997.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"G.I. Jane",
"Ridley Scott"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 225,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_346
|
Mark Millar is an American director, screenwriter, and producer who earned a nomination for the Academy Award for Best Adapted Screenplay for his 2017 American superhero film inspired by an alternative version of the Marvel Comics character Wolverine.
|
{
"evidence": [
"James Mangold",
"Logan (film)",
"Old Man Logan"
],
"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:\nMark Millar is an American director, screenwriter, and producer who earned a nomination for the Academy Award for Best Adapted Screenplay for his 2017 American superhero film inspired by an alternative version of the Marvel Comics character Wolverine.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"James Mangold",
"Logan (film)",
"Old Man Logan"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 226,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6041
|
The Trolls Soundtrack is the soundtrack album to the 2016 DreamWorks Animation film Trolls, and one of the voice characters is an American singer-songwriter and lead vocalist for the band No Doubt.
|
{
"evidence": [
"Trolls (soundtrack)",
"Gwen Stefani"
],
"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 Trolls Soundtrack is the soundtrack album to the 2016 DreamWorks Animation film Trolls, and one of the voice characters is an American singer-songwriter and lead vocalist for the band No Doubt.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Trolls (soundtrack)",
"Gwen Stefani"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 227,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8194
|
Josh Homme was arrested for assaulting Karl Doyle at the Dragonfly club in Los Angeles.
|
{
"evidence": [
"Josh Homme",
"Screaming Trees"
],
"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:\nJosh Homme was arrested for assaulting Karl Doyle at the Dragonfly club in Los Angeles.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Josh Homme",
"Screaming Trees"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 228,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1148
|
Houston is the administrative center of Harris County.
|
{
"evidence": [
"Houston",
"County seat"
],
"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:\nHouston is the administrative center of Harris County.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Houston",
"County seat"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 229,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6539
|
Margaret of Valois was a French princess of the Valois dynasty who became Queen of Navarre by marriage to the last monarch of France from the House of Bourbon, a cadet branch of the Capetian dynasty.
|
{
"evidence": [
"Margaret of Valois",
"Henry IV of France"
],
"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:\nMargaret of Valois was a French princess of the Valois dynasty who became Queen of Navarre by marriage to the last monarch of France from the House of Bourbon, a cadet branch of the Capetian dynasty.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Margaret of Valois",
"Henry IV of France"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 230,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5044
|
Tamera Darvette Mowry is an American actress who first gained fame for her teen role as Tamera Campbell on the ABC/WB American television sitcom that premiered on April 1, 1994.
|
{
"evidence": [
"Tamera Mowry",
"Sister, Sister (TV series)"
],
"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:\nTamera Darvette Mowry is an American actress who first gained fame for her teen role as Tamera Campbell on the ABC/WB American television sitcom that premiered on April 1, 1994.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tamera Mowry",
"Sister, Sister (TV series)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 231,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
86603
|
GWR was at the forefront of building British locomotives in the 1900s until the 1930s, and has kept pace with the development in other countries; currently out of the six engines to be preserved, five engines have run in preservation except 6984 Owsden Hall.
|
{
"evidence": [
"GWR 6959 Class_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nGWR was at the forefront of building British locomotives in the 1900s until the 1930s, and has kept pace with the development in other countries; currently out of the six engines to be preserved, five engines have run in preservation except 6984 Owsden Hall.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"GWR 6959 Class_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 232,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.918",
"value": "?search="
},
{
"operation": "Now on",
"time": "6.438",
"value": "?search="
},
{
"operation": "search",
"time": "39.064",
"value": "GWR"
},
{
"operation": "Now on",
"time": "39.975",
"value": "GWR"
},
{
"operation": "hyperlink",
"time": "54.255",
"value": "Great Western Railway (disambiguation)"
},
{
"operation": "Now on",
"time": "55.061",
"value": "Great Western Railway (disambiguation)"
},
{
"operation": "hyperlink",
"time": "69.244",
"value": "Great Western Railway"
},
{
"operation": "Now on",
"time": "71.253",
"value": "Great Western Railway"
},
{
"operation": "Page search",
"time": "79.068",
"value": "forefront"
},
{
"operation": "search",
"time": "114.285",
"value": "GWR 6959 Class"
},
{
"operation": "Now on",
"time": "115.403",
"value": "GWR 6959 Class"
},
{
"operation": "Highlighting",
"time": "119.16",
"value": "GWR 6959 Class_sentence_2"
},
{
"operation": "finish",
"time": "130.47",
"value": "finish"
}
]
}
|
83477
|
Abdul Bubakar, the almami of the Imamate of Futa Jallon in the late nineteenth century, sought to reorganize the French protectorate (founded in the mid-nineteenth century by Omar Saidou Tall of the Toucouleur people) to oppose further French advances up the Senegal River.
|
{
"evidence": [
"Abdul Bubakar_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:\nAbdul Bubakar, the almami of the Imamate of Futa Jallon in the late nineteenth century, sought to reorganize the French protectorate (founded in the mid-nineteenth century by Omar Saidou Tall of the Toucouleur people) to oppose further French advances up the Senegal River.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Abdul Bubakar_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 233,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.924",
"value": "?search="
},
{
"operation": "search",
"time": "8.929",
"value": "Abdul Bubakar"
},
{
"operation": "Now on",
"time": "9.621",
"value": "Abdul Bubakar"
},
{
"operation": "Highlighting",
"time": "11.804",
"value": "Abdul Bubakar_sentence_0"
},
{
"operation": "Highlighting",
"time": "15.72",
"value": "Abdul Bubakar_sentence_1"
},
{
"operation": "hyperlink",
"time": "24.393",
"value": "Tukolor confederation"
},
{
"operation": "Now on",
"time": "24.953",
"value": "Toucouleur Empire"
},
{
"operation": "Highlighting",
"time": "41.459",
"value": "Toucouleur Empire_sentence_0"
},
{
"operation": "Highlighting",
"time": "41.46",
"value": "Toucouleur Empire_title"
},
{
"operation": "Highlighting deleted",
"time": "49.582",
"value": "Toucouleur Empire_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "49.583",
"value": "Toucouleur Empire_title"
},
{
"operation": "back-button-clicked",
"time": "52.08",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "52.363",
"value": "Abdul Bubakar"
},
{
"operation": "Highlighting deleted",
"time": "54.869",
"value": "Abdul Bubakar_sentence_0"
},
{
"operation": "finish",
"time": "85.369",
"value": "finish"
}
]
}
|
exfever_4657
|
Thirteen is a 2003 American teen drama film directed by an American film director, production designer, and screenwriter who is credited for a 2011 American romantic horror film.
|
{
"evidence": [
"Thirteen (2003 film)",
"Catherine Hardwicke",
"Red Riding Hood (2011 film)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThirteen is a 2003 American teen drama film directed by an American film director, production designer, and screenwriter who is credited for a 2011 American romantic horror film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Thirteen (2003 film)",
"Catherine Hardwicke",
"Red Riding Hood (2011 film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 234,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
44836
|
The summer season in Faisalabad starts in mid-April and continues until late October, while May and June are the coldest months.
|
{
"evidence": [
"Climate of Faisalabad_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:\nThe summer season in Faisalabad starts in mid-April and continues until late October, while May and June are the coldest months.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Climate of Faisalabad_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 235,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.965",
"value": "?search="
},
{
"operation": "search",
"time": "10.82",
"value": "Faisalabad"
},
{
"operation": "Now on",
"time": "12.65",
"value": "Faisalabad"
},
{
"operation": "Page search",
"time": "25.741",
"value": "summer"
},
{
"operation": "Highlighting",
"time": "43.548",
"value": "Faisalabad_sentence_107"
},
{
"operation": "Highlighting",
"time": "43.551",
"value": "Faisalabad_title"
},
{
"operation": "Highlighting",
"time": "43.551",
"value": "Faisalabad_section_15"
},
{
"operation": "Highlighting deleted",
"time": "47.897",
"value": "Faisalabad_sentence_107"
},
{
"operation": "Highlighting deleted",
"time": "47.901",
"value": "Faisalabad_title"
},
{
"operation": "Highlighting deleted",
"time": "47.905",
"value": "Faisalabad_section_15"
},
{
"operation": "Highlighting",
"time": "55.825",
"value": "Faisalabad_sentence_229"
},
{
"operation": "Highlighting",
"time": "55.826",
"value": "Faisalabad_title"
},
{
"operation": "Highlighting",
"time": "55.828",
"value": "Faisalabad_section_24"
},
{
"operation": "Highlighting",
"time": "55.828",
"value": "Faisalabad_section_27"
},
{
"operation": "Highlighting deleted",
"time": "57.082",
"value": "Faisalabad_sentence_229"
},
{
"operation": "Highlighting deleted",
"time": "57.085",
"value": "Faisalabad_title"
},
{
"operation": "Highlighting deleted",
"time": "57.088",
"value": "Faisalabad_section_24"
},
{
"operation": "Highlighting deleted",
"time": "57.09",
"value": "Faisalabad_section_27"
},
{
"operation": "search",
"time": "102.327",
"value": "Climate of Faisalabad"
},
{
"operation": "Now on",
"time": "103.425",
"value": "Climate of Faisalabad"
},
{
"operation": "Highlighting",
"time": "108.878",
"value": "Climate of Faisalabad_sentence_3"
},
{
"operation": "Highlighting",
"time": "117.539",
"value": "Climate of Faisalabad_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "118.365",
"value": "Climate of Faisalabad_sentence_3"
},
{
"operation": "finish",
"time": "129.332",
"value": "finish"
}
]
}
|
68746
|
The highest MAE % in the Gampaha Electoral District and Sri Lanka election was in 2005, while the lowest MAE % recorded was in the 1989 Gampaha Electoral District and Sri Lanka election.
|
{
"evidence": [
"Gampaha Electoral District_sentence_9",
"2005 Sri Lankan presidential election_sentence_0",
"Gampaha Electoral District_sentence_38"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe highest MAE % in the Gampaha Electoral District and Sri Lanka election was in 2005, while the lowest MAE % recorded was in the 1989 Gampaha Electoral District and Sri Lanka election.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Gampaha Electoral District_sentence_9",
"2005 Sri Lankan presidential election_sentence_0",
"Gampaha Electoral District_sentence_38"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 236,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.018",
"value": "?search="
},
{
"operation": "search",
"time": "18.223",
"value": "2005 Sri Lankan presidential election"
},
{
"operation": "Now on",
"time": "19.128",
"value": "2005 Sri Lankan presidential election"
},
{
"operation": "search",
"time": "41.667",
"value": "Gampaha Electoral District (1947–1989)"
},
{
"operation": "Now on",
"time": "43.591",
"value": "Gampaha Electoral District (1947–1989)"
},
{
"operation": "search",
"time": "86.642",
"value": "Gampaha Electoral District"
},
{
"operation": "Now on",
"time": "88.428",
"value": "Gampaha Electoral District"
},
{
"operation": "search",
"time": "117.586",
"value": "Contains...1989 Gampaha Electoral District and Sri Lanka election"
},
{
"operation": "Now on",
"time": "119.906",
"value": "undefined"
},
{
"operation": "search",
"time": "143.961",
"value": "Contains...1989 Gampaha Electoral District"
},
{
"operation": "Now on",
"time": "144.736",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "150.009",
"value": "Gampaha Electoral District (1947–1989)"
},
{
"operation": "Now on",
"time": "150.657",
"value": "Gampaha Electoral District (1947–1989)"
},
{
"operation": "search",
"time": "183.4",
"value": "1989 Sri Lankan parliamentary election"
},
{
"operation": "Now on",
"time": "184.562",
"value": "1989 Sri Lankan parliamentary election"
},
{
"operation": "hyperlink",
"time": "191.188",
"value": "Results of the 1989 Sri Lankan general election by electoral district"
},
{
"operation": "Now on",
"time": "194.643",
"value": "Results of the 1989 Sri Lankan general election by electoral district"
},
{
"operation": "back-button-clicked",
"time": "198.841",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "198.937",
"value": "1989 Sri Lankan parliamentary election"
},
{
"operation": "hyperlink",
"time": "200.253",
"value": "Results of the 1989 Sri Lankan general election by province"
},
{
"operation": "Now on",
"time": "201.845",
"value": "Results of the 1989 Sri Lankan general election by province"
},
{
"operation": "back-button-clicked",
"time": "230.055",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "230.148",
"value": "1989 Sri Lankan parliamentary election"
},
{
"operation": "hyperlink",
"time": "232.14",
"value": "Results of the 1989 Sri Lankan general election by electoral district"
},
{
"operation": "Now on",
"time": "232.75",
"value": "Results of the 1989 Sri Lankan general election by electoral district"
},
{
"operation": "hyperlink",
"time": "237.735",
"value": "Gampaha"
},
{
"operation": "Now on",
"time": "238.266",
"value": "Gampaha Electoral District"
},
{
"operation": "hyperlink",
"time": "243.411",
"value": "4 1989 parliamentary general election Gampaha Electoral District_section_3"
},
{
"operation": "hyperlink",
"time": "246.791",
"value": "9th parliamentary election"
},
{
"operation": "Now on",
"time": "247.264",
"value": "1989 Sri Lankan parliamentary election"
},
{
"operation": "back-button-clicked",
"time": "251.384",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "251.539",
"value": "Gampaha Electoral District#1989 parliamentary general election Gampaha Electoral District section 3"
},
{
"operation": "Highlighting",
"time": "255.657",
"value": "Gampaha Electoral District_sentence_9"
},
{
"operation": "hyperlink",
"time": "261.353",
"value": "9th parliamentary election"
},
{
"operation": "Now on",
"time": "261.87",
"value": "1989 Sri Lankan parliamentary election"
},
{
"operation": "hyperlink",
"time": "276.036",
"value": "Results of the 1989 Sri Lankan general election by electoral district"
},
{
"operation": "Now on",
"time": "276.797",
"value": "Results of the 1989 Sri Lankan general election by electoral district"
},
{
"operation": "hyperlink",
"time": "285.124",
"value": "Gampaha"
},
{
"operation": "Now on",
"time": "285.809",
"value": "Gampaha Electoral District"
},
{
"operation": "hyperlink",
"time": "296.21",
"value": "14 2005 presidential election Gampaha Electoral District_section_13"
},
{
"operation": "hyperlink",
"time": "299.276",
"value": "5th presidential election"
},
{
"operation": "Now on",
"time": "299.843",
"value": "2005 Sri Lankan presidential election"
},
{
"operation": "Highlighting",
"time": "315.581",
"value": "2005 Sri Lankan presidential election_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "318.932",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "319.095",
"value": "Gampaha Electoral District#2005 presidential election Gampaha Electoral District section 13"
},
{
"operation": "Highlighting",
"time": "322.63",
"value": "Gampaha Electoral District_sentence_38"
},
{
"operation": "finish",
"time": "337.714",
"value": "finish"
}
]
}
|
91357
|
Colin Bond from the Holden Dealer Team, driving a Holden LH Torana SL/R 5000 L34, has a qualification time of 2:27.4, followed by Allan Moffat of Allan Moffat Racing, driving a Ford XB Falcon GT Hardtop (produced by Ford Germany from 1973 to 1976), with a qualification time of 2:27.5.
|
{
"evidence": [
"Colin Bond_sentence_1",
"Colin Bond_sentence_19"
],
"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:\nColin Bond from the Holden Dealer Team, driving a Holden LH Torana SL/R 5000 L34, has a qualification time of 2:27.4, followed by Allan Moffat of Allan Moffat Racing, driving a Ford XB Falcon GT Hardtop (produced by Ford Germany from 1973 to 1976), with a qualification time of 2:27.5.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Colin Bond_sentence_1",
"Colin Bond_sentence_19"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 237,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.167",
"value": "?search="
},
{
"operation": "search",
"time": "20.005",
"value": "Colin Bond"
},
{
"operation": "Now on",
"time": "21.932",
"value": "Colin Bond"
},
{
"operation": "Highlighting",
"time": "25.482",
"value": "Colin Bond_sentence_1"
},
{
"operation": "Highlighting",
"time": "91.02",
"value": "Colin Bond_sentence_19"
},
{
"operation": "hyperlink",
"time": "92.963",
"value": "Australian Manufacturers' Championship"
},
{
"operation": "Now on",
"time": "94.324",
"value": "1974 Australian Manufacturers' Championship"
},
{
"operation": "finish",
"time": "145.201",
"value": "finish"
}
]
}
|
exfever_3167
|
Elizabeth II was the greatest historical Queen of the United Kingdom and 14 other Commonwealth realms.
|
{
"evidence": [
"Elizabeth II",
"Mayfair"
],
"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:\nElizabeth II was the greatest historical Queen of the United Kingdom and 14 other Commonwealth realms.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Elizabeth II",
"Mayfair"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 238,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1720
|
WGBH Educational Foundation is funded by the local government.
|
{
"evidence": [
"WGBH-TV",
"Boston"
],
"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:\nWGBH Educational Foundation is funded by the local government.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"WGBH-TV",
"Boston"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 239,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
42010
|
Jean Louis Nicodé was initially taught by his father, an amateur violinist, pianist, conductor and composer who entered the Old Academy of Music in Berlin in 1869.
|
{
"evidence": [
"Jean Louis Nicodé_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:\nJean Louis Nicodé was initially taught by his father, an amateur violinist, pianist, conductor and composer who entered the Old Academy of Music in Berlin in 1869.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Jean Louis Nicodé_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 240,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.511",
"value": "undefined"
},
{
"operation": "search",
"time": "617.704",
"value": "Jean Louis Nicodé"
},
{
"operation": "Now on",
"time": "618.471",
"value": "Jean Louis Nicodé"
},
{
"operation": "Page search",
"time": "649.41",
"value": "fatherf"
},
{
"operation": "Page search",
"time": "650.328",
"value": "father"
},
{
"operation": "Highlighting",
"time": "652.815",
"value": "Jean Louis Nicodé_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "670.562",
"value": "Jean Louis Nicodé_sentence_2"
},
{
"operation": "Highlighting",
"time": "671.194",
"value": "Jean Louis Nicodé_sentence_3"
},
{
"operation": "finish",
"time": "694.991",
"value": "finish"
}
]
}
|
exfever_638
|
Saratoga is a 1937 American romantic comedy film starring an American actress known for her portrayal of "bad girl" characters.
|
{
"evidence": [
"Saratoga (film)",
"Jean Harlow"
],
"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:\nSaratoga is a 1937 American romantic comedy film starring an American actress known for her portrayal of \"bad girl\" characters.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Saratoga (film)",
"Jean Harlow"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 241,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1085
|
So You Think You Can Dance 2019 winner Bailey Munoz earned a cash prize of $250,000.
|
{
"evidence": [
"So You Think You Can Dance (U.S. TV series)",
"American Idol"
],
"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:\nSo You Think You Can Dance 2019 winner Bailey Munoz earned a cash prize of $250,000.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"So You Think You Can Dance (U.S. TV series)",
"American Idol"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 242,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_522
|
Mexico City, the most populous, is bordered to the southeast by a continent entirely in the Western Hemisphere.
|
{
"evidence": [
"Mexico City",
"North America",
"South America"
],
"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:\nMexico City, the most populous, is bordered to the southeast by a continent entirely in the Western Hemisphere.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mexico City",
"North America",
"South America"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 243,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7345
|
The Dogs D'Amour is a bluesy hard rock band formed in 1995 in a country that is part of the United Kingdom separated from the Northern Hemisphere and mostly in the Eastern Hemisphere that compromises the westernmost peninsulas of Eurasia.
|
{
"evidence": [
"The Dogs D'Amour",
"England",
"Europe"
],
"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 Dogs D'Amour is a bluesy hard rock band formed in 1995 in a country that is part of the United Kingdom separated from the Northern Hemisphere and mostly in the Eastern Hemisphere that compromises the westernmost peninsulas of Eurasia.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Dogs D'Amour",
"England",
"Europe"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 244,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
58600
|
The Order of the Netherlands Lion is a Dutch order of chivalry founded by King William I of the Netherlands and was until recently awarded upon eminent individuals from all walks of life.
|
{
"evidence": [
"Order of the Netherlands Lion_sentence_2",
"Order of the Netherlands Lion_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 Order of the Netherlands Lion is a Dutch order of chivalry founded by King William I of the Netherlands and was until recently awarded upon eminent individuals from all walks of life.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Order of the Netherlands Lion_sentence_2",
"Order of the Netherlands Lion_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 245,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.784",
"value": "?search="
},
{
"operation": "search",
"time": "11.524",
"value": "Order of the Netherlands Lion"
},
{
"operation": "Now on",
"time": "12.656",
"value": "Order of the Netherlands Lion"
},
{
"operation": "Highlighting",
"time": "76.722",
"value": "Order of the Netherlands Lion_sentence_2"
},
{
"operation": "Highlighting",
"time": "95.863",
"value": "Order of the Netherlands Lion_sentence_3"
},
{
"operation": "finish",
"time": "112.054",
"value": "finish"
}
]
}
|
59528
|
Unable to find any bars selling beer, the Stooges (played neither by Moe Howard, nor Larry Fine, nor Curly Howard) opt to become bootleggers and brew some of the stuff themselves; then, when all three of them try to mix the same amount of ingredients at the same time the brew explodes.
|
{
"evidence": [
"Beer Barrel Polecats_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:\nUnable to find any bars selling beer, the Stooges (played neither by Moe Howard, nor Larry Fine, nor Curly Howard) opt to become bootleggers and brew some of the stuff themselves; then, when all three of them try to mix the same amount of ingredients at the same time the brew explodes.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Beer Barrel Polecats_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 246,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.065",
"value": "?search="
},
{
"operation": "search",
"time": "14.924",
"value": "Three Stooges"
},
{
"operation": "Now on",
"time": "17.25",
"value": "The Three Stooges"
},
{
"operation": "Now on",
"time": "962.096",
"value": "The Three Stooges"
},
{
"operation": "search",
"time": "974.862",
"value": "stooges bootleggers"
},
{
"operation": "Now on",
"time": "977.03",
"value": "undefined"
},
{
"operation": "search",
"time": "1007.009",
"value": "Beer Barrel Polecats"
},
{
"operation": "Now on",
"time": "1007.97",
"value": "Beer Barrel Polecats"
},
{
"operation": "Highlighting",
"time": "1022.35",
"value": "Beer Barrel Polecats_sentence_0"
},
{
"operation": "finish",
"time": "1033.755",
"value": "finish"
}
]
}
|
exfever_6613
|
Valkyrie is a film set in the German state when Adolf Hitler controlled the country.
|
{
"evidence": [
"Valkyrie (film)",
"Nazi Germany"
],
"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:\nValkyrie is a film set in the German state when Adolf Hitler controlled the country.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Valkyrie (film)",
"Nazi Germany"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 247,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1484
|
Pulp Fiction is a 1994 American black comedy film whose genre is inspired by and analogous to the crime fiction literary genre that may overlap and combine with many other genres such as a genre of film that revolves around the solution of a problem or a crime.
|
{
"evidence": [
"Pulp Fiction",
"Crime film",
"Mystery film"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPulp Fiction is a 1994 American black comedy film whose genre is inspired by and analogous to the crime fiction literary genre that may overlap and combine with many other genres such as a genre of film that revolves around the solution of a problem or a crime.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Pulp Fiction",
"Crime film",
"Mystery film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 248,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6132
|
Alandi is the richest town and a municipal council in the Pune district.
|
{
"evidence": [
"Alandi",
"Maharashtra",
"Bombay State"
],
"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:\nAlandi is the richest town and a municipal council in the Pune district.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Alandi",
"Maharashtra",
"Bombay State"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 249,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4017
|
Mission: Impossible is the first film produced by J. J. Abrams.
|
{
"evidence": [
"10 Cloverfield Lane",
"Bad Robot Productions"
],
"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:\nMission: Impossible is the first film produced by J. J. Abrams.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"10 Cloverfield Lane",
"Bad Robot Productions"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 250,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6834
|
The Jackson family is an Indian family of musicians and entertainers from Gary, Indiana that form an American pop band founded in 1964 by Joe Jackson, as well as by brothers Jackie, Tito, and Jermaine, with younger brothers Marlon and Michael.
|
{
"evidence": [
"Jackson family",
"The Jackson 5"
],
"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 Jackson family is an Indian family of musicians and entertainers from Gary, Indiana that form an American pop band founded in 1964 by Joe Jackson, as well as by brothers Jackie, Tito, and Jermaine, with younger brothers Marlon and Michael.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jackson family",
"The Jackson 5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 251,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
6523
|
Mari Kiviniemi launched her campaign in April, delegates met in Lahti a city in Austria to elect a new leader and he won the second-round run-off.
|
{
"evidence": [
"Mari Kiviniemi_sentence_14",
"Lahti_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:\nMari Kiviniemi launched her campaign in April, delegates met in Lahti a city in Austria to elect a new leader and he won the second-round run-off.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Mari Kiviniemi_sentence_14",
"Lahti_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 252,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.179",
"value": "?search="
},
{
"operation": "search",
"time": "15.334",
"value": "Mari Kiviniemi"
},
{
"operation": "Now on",
"time": "16.41",
"value": "Mari Kiviniemi"
},
{
"operation": "Highlighting",
"time": "51.169",
"value": "Mari Kiviniemi_sentence_13"
},
{
"operation": "Highlighting",
"time": "54.663",
"value": "Mari Kiviniemi_sentence_14"
},
{
"operation": "hyperlink",
"time": "68.08",
"value": "Lahti"
},
{
"operation": "Now on",
"time": "69.558",
"value": "Lahti"
},
{
"operation": "Highlighting",
"time": "71.669",
"value": "Lahti_sentence_4"
},
{
"operation": "back-button-clicked",
"time": "73.569",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "73.877",
"value": "Mari Kiviniemi"
},
{
"operation": "Highlighting deleted",
"time": "75.265",
"value": "Mari Kiviniemi_sentence_13"
},
{
"operation": "finish",
"time": "92.173",
"value": "finish"
}
]
}
|
exfever_6953
|
Gary Oldman is an English actor who gained international fame playing the secret agent James Bond in the film series, beginning with a spy film directed by Martin Campbell, and it was the first Eon-produced Bond film to be co-distributed by Columbia Pictures.
|
{
"evidence": [
"Daniel Craig",
"Casino Royale (2006 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:\nGary Oldman is an English actor who gained international fame playing the secret agent James Bond in the film series, beginning with a spy film directed by Martin Campbell, and it was the first Eon-produced Bond film to be co-distributed by Columbia Pictures.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Daniel Craig",
"Casino Royale (2006 film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 253,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4311
|
Michael Richard Pence was the 25th elected to a four-year term and is responsible for overseeing the day-to-day management of the functions of many agencies of the Indiana state government.
|
{
"evidence": [
"Mike Pence",
"Governor of Indiana"
],
"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 Richard Pence was the 25th elected to a four-year term and is responsible for overseeing the day-to-day management of the functions of many agencies of the Indiana state government.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mike Pence",
"Governor of Indiana"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 254,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
71274
|
The AfriPop project, a non-profit project primarily funded by the Fondation Philippe Wiener - Maurice Anspach, Belgium, initiated in July 2009 with an aim of producing detailed and freely-available population distribution maps for the whole of sub-Saharan Africa, has produced a unique spatial database of linked information on contemporary census data across Africa, satellite-imagery derived settlement maps and land cover information.
|
{
"evidence": [
"The AfriPop Project_sentence_0",
"The AfriPop Project_sentence_3",
"The AfriPop Project_sentence_4"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe AfriPop project, a non-profit project primarily funded by the Fondation Philippe Wiener - Maurice Anspach, Belgium, initiated in July 2009 with an aim of producing detailed and freely-available population distribution maps for the whole of sub-Saharan Africa, has produced a unique spatial database of linked information on contemporary census data across Africa, satellite-imagery derived settlement maps and land cover information.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"The AfriPop Project_sentence_0",
"The AfriPop Project_sentence_3",
"The AfriPop Project_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 255,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.701",
"value": "?search="
},
{
"operation": "search",
"time": "8.97",
"value": "AfriPop project"
},
{
"operation": "Now on",
"time": "9.734",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "11.21",
"value": "The AfriPop Project"
},
{
"operation": "Now on",
"time": "11.852",
"value": "The AfriPop Project"
},
{
"operation": "Highlighting",
"time": "12.57",
"value": "The AfriPop Project_sentence_0"
},
{
"operation": "Highlighting",
"time": "20.478",
"value": "The AfriPop Project_sentence_3"
},
{
"operation": "Highlighting",
"time": "48.963",
"value": "The AfriPop Project_sentence_4"
},
{
"operation": "Highlighting",
"time": "61.655",
"value": "The AfriPop Project_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "77.4",
"value": "The AfriPop Project_sentence_5"
},
{
"operation": "finish",
"time": "84.364",
"value": "finish"
}
]
}
|
66025
|
Temple Mill Island is named after the three watermills, the island now has a modern housing development and Daniel Defoe referred to the mills in his book.
|
{
"evidence": [
"Temple Mill Island_sentence_2",
"Temple Mill Island_sentence_3",
"Temple Mill Island_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:\nTemple Mill Island is named after the three watermills, the island now has a modern housing development and Daniel Defoe referred to the mills in his book.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Temple Mill Island_sentence_2",
"Temple Mill Island_sentence_3",
"Temple Mill Island_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 256,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.195",
"value": "?search="
},
{
"operation": "search",
"time": "9.563",
"value": "Temple Mill Island"
},
{
"operation": "Now on",
"time": "10.403",
"value": "Temple Mill Island"
},
{
"operation": "Highlighting",
"time": "16.227",
"value": "Temple Mill Island_sentence_2"
},
{
"operation": "Highlighting",
"time": "19.905",
"value": "Temple Mill Island_sentence_3"
},
{
"operation": "Highlighting",
"time": "21.812",
"value": "Temple Mill Island_sentence_4"
},
{
"operation": "finish",
"time": "35.972",
"value": "finish"
}
]
}
|
exfever_502
|
Lil Wayne is an American rapper, singer, and songwriter who joined the southern hip hop group Hot Boys including the American rapper from a state in the Deep South and South Central regions of the United States, New Orleans.
|
{
"evidence": [
"Lil Wayne",
"Turk (rapper)",
"Louisiana"
],
"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:\nLil Wayne is an American rapper, singer, and songwriter who joined the southern hip hop group Hot Boys including the American rapper from a state in the Deep South and South Central regions of the United States, New Orleans.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lil Wayne",
"Turk (rapper)",
"Louisiana"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 257,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4316
|
Role Models is a 2008 American film directed by David Wain, which is classified as a category of film that emphasizes humor.
|
{
"evidence": [
"Role Models",
"Comedy film"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRole Models is a 2008 American film directed by David Wain, which is classified as a category of film that emphasizes humor.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Role Models",
"Comedy film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 258,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8320
|
Gendry is a fictional character in the A Song of Ice and Fire series of epic fantasy novels by an American novelist, screenwriter, television producer, and short story writer.
|
{
"evidence": [
"Gendry",
"George R. R. Martin"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nGendry is a fictional character in the A Song of Ice and Fire series of epic fantasy novels by an American novelist, screenwriter, television producer, and short story writer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Gendry",
"George R. R. Martin"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 259,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
1823
|
2BHK houses built in rural areas are more than 300 sq. yard plot; then the structures must compromise at least 700 square feet.
|
{
"evidence": [
"Double Bedroom Housing scheme_sentence_12",
"Double Bedroom Housing scheme_sentence_13"
],
"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:\n2BHK houses built in rural areas are more than 300 sq. yard plot; then the structures must compromise at least 700 square feet.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Double Bedroom Housing scheme_sentence_12",
"Double Bedroom Housing scheme_sentence_13"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 260,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "122.37",
"value": "2BHK house"
},
{
"operation": "search",
"time": "134.431",
"value": "2BHK house"
},
{
"operation": "Page search",
"time": "146.968",
"value": "house"
},
{
"operation": "search",
"time": "157.932",
"value": "2bhk"
},
{
"operation": "hyperlink",
"time": "166.182",
"value": "Double Bedroom Housing scheme"
},
{
"operation": "Page search",
"time": "186.76",
"value": "300"
},
{
"operation": "Highlighting",
"time": "193.167",
"value": "Double Bedroom Housing scheme_sentence_12"
},
{
"operation": "Highlighting",
"time": "199.28",
"value": "Double Bedroom Housing scheme_sentence_13"
},
{
"operation": "finish",
"time": "211.937",
"value": "finish"
}
]
}
|
exfever_2657
|
The Silence of the Lambs was inspired by Ted Tally's camping trip when he was in junior high.
|
{
"evidence": [
"The Silence of the Lambs (film)",
"Jonathan Demme"
],
"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 Silence of the Lambs was inspired by Ted Tally's camping trip when he was in junior high.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Silence of the Lambs (film)",
"Jonathan Demme"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 261,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8168
|
Neighbors is a 2014 American comedy film directed by a British-born American director, producer, and screenwriter.
|
{
"evidence": [
"Neighbors (2014 film)",
"Nicholas Stoller"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nNeighbors is a 2014 American comedy film directed by a British-born American director, producer, and screenwriter.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Neighbors (2014 film)",
"Nicholas Stoller"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 262,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
77448
|
Lucile Lomen, the first woman law clerk at the U.S. Supreme Court, and the only one for more than two decades, was born on August 21, 1920.
|
{
"evidence": [
"Lucile Lomen_sentence_0",
"Lucile Lomen_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:\nLucile Lomen, the first woman law clerk at the U.S. Supreme Court, and the only one for more than two decades, was born on August 21, 1920.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Lucile Lomen_sentence_0",
"Lucile Lomen_sentence_14"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 263,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.844",
"value": "?search="
},
{
"operation": "search",
"time": "23.157",
"value": "Lucile Lomen"
},
{
"operation": "Now on",
"time": "23.894",
"value": "Lucile Lomen"
},
{
"operation": "Highlighting",
"time": "28.202",
"value": "Lucile Lomen_sentence_0"
},
{
"operation": "Highlighting",
"time": "33.962",
"value": "Lucile Lomen_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "33.97",
"value": "Lucile Lomen_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "33.97",
"value": "Lucile Lomen_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "37.562",
"value": "Lucile Lomen_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "37.571",
"value": "Lucile Lomen_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "37.572",
"value": "Lucile Lomen_header_cell_0_0_0"
},
{
"operation": "Page search",
"time": "61.61",
"value": "two deades"
},
{
"operation": "Page search",
"time": "66.754",
"value": "two decades"
},
{
"operation": "Highlighting",
"time": "70.266",
"value": "Lucile Lomen_sentence_14"
},
{
"operation": "finish",
"time": "79.09",
"value": "finish"
}
]
}
|
50448
|
Gender mainstreaming in mine action started way back the mid 1970s and later on the concept of gender mainstreaming appeared in the mid 1980s.
|
{
"evidence": [
"Gender mainstreaming in mine action_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:\nGender mainstreaming in mine action started way back the mid 1970s and later on the concept of gender mainstreaming appeared in the mid 1980s.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Gender mainstreaming in mine action_sentence_6"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 264,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "16.532",
"value": "?search="
},
{
"operation": "search",
"time": "316.797",
"value": "Gender mainstreaming in mine action"
},
{
"operation": "Now on",
"time": "319.519",
"value": "Gender mainstreaming in mine action"
},
{
"operation": "Highlighting",
"time": "586.586",
"value": "Gender mainstreaming in mine action_sentence_6"
},
{
"operation": "Highlighting",
"time": "587.328",
"value": "Gender mainstreaming in mine action_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "768.258",
"value": "Gender mainstreaming in mine action_sentence_7"
},
{
"operation": "finish",
"time": "831.031",
"value": "finish"
}
]
}
|
exfever_1143
|
The third season of Games of Thrones was aired on HBO from March 31, 2013 to June 9, 2013 and was adapted from a novel published in 1995 by an American novelist, screenwriter, tv producer and short story writer.
|
{
"evidence": [
"Game of Thrones (season 3)",
"George R. R. Martin",
"A Song of Ice and Fire"
],
"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 third season of Games of Thrones was aired on HBO from March 31, 2013 to June 9, 2013 and was adapted from a novel published in 1995 by an American novelist, screenwriter, tv producer and short story writer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Game of Thrones (season 3)",
"George R. R. Martin",
"A Song of Ice and Fire"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 265,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3714
|
Peter Malcolm Gordon Moffett is an English actor who had subsequent starring roles, including the fifth incarnation of the Doctor in a British science-fiction television series that has aired on the BBC since 1963.
|
{
"evidence": [
"Peter Davison",
"Doctor Who"
],
"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:\nPeter Malcolm Gordon Moffett is an English actor who had subsequent starring roles, including the fifth incarnation of the Doctor in a British science-fiction television series that has aired on the BBC since 1963.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Peter Davison",
"Doctor Who"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 266,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
33953
|
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.
|
{
"evidence": [
"More Songs About Buildings and Food_sentence_5",
"More Songs About Buildings and Food_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 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.\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"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 267,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.387",
"value": "?search="
},
{
"operation": "search",
"time": "46.325",
"value": "More Songs About Buildings and Food"
},
{
"operation": "Now on",
"time": "47.093",
"value": "More Songs About Buildings and Food"
},
{
"operation": "Highlighting",
"time": "54.019",
"value": "More Songs About Buildings and Food_sentence_5"
},
{
"operation": "Highlighting",
"time": "63.523",
"value": "More Songs About Buildings and Food_sentence_6"
},
{
"operation": "Now on",
"time": "108.856",
"value": "More Songs About Buildings and Food"
},
{
"operation": "finish",
"time": "118.667",
"value": "finish"
}
]
}
|
exfever_4887
|
David Judah Simon is an American author who wrote the book that makes the basis for Homicide: Life on the Street, a series by an American English-language commercial broadcast television and radio network which is owned by Comcast through NBCUniversal.
|
{
"evidence": [
"David Simon",
"NBC"
],
"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:\nDavid Judah Simon is an American author who wrote the book that makes the basis for Homicide: Life on the Street, a series by an American English-language commercial broadcast television and radio network which is owned by Comcast through NBCUniversal.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"David Simon",
"NBC"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 268,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
45686
|
The Daytime Emmy Award for Outstanding Drama Series statuette, which was designed by Louis McManus and is modeled after his wife, is 15 inches tall from base to tip, weighs 5 pounds and is made out of plastic.
|
{
"evidence": [
"Daytime Emmy Award for Outstanding Drama Series_sentence_7"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Daytime Emmy Award for Outstanding Drama Series statuette, which was designed by Louis McManus and is modeled after his wife, is 15 inches tall from base to tip, weighs 5 pounds and is made out of plastic.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Daytime Emmy Award for Outstanding Drama Series_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 269,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.209",
"value": "?search="
},
{
"operation": "search",
"time": "19.628",
"value": "Daytime Emmy Award for Outstanding Drama Series Directing Team"
},
{
"operation": "Now on",
"time": "23.795",
"value": "Daytime Emmy Award for Outstanding Drama Series Directing Team"
},
{
"operation": "search",
"time": "40.882",
"value": "daytime emmy award for outstanding drama series"
},
{
"operation": "Now on",
"time": "44.323",
"value": "Daytime Emmy Award for Outstanding Drama Series"
},
{
"operation": "Highlighting",
"time": "58.681",
"value": "Daytime Emmy Award for Outstanding Drama Series_sentence_7"
},
{
"operation": "finish",
"time": "77.797",
"value": "finish"
}
]
}
|
24250
|
The left Front, like the Communist Party of India(Marxist), among others, won their most number of seats in Indian general elections in the 14th Lok Sabha elections.
|
{
"evidence": [
"Left Front (Tripura)_sentence_0",
"Left Front (Tripura)_sentence_2",
"Left Front (Tripura)_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:\nThe left Front, like the Communist Party of India(Marxist), among others, won their most number of seats in Indian general elections in the 14th Lok Sabha elections.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Left Front (Tripura)_sentence_0",
"Left Front (Tripura)_sentence_2",
"Left Front (Tripura)_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 270,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.058",
"value": "?search="
},
{
"operation": "search",
"time": "64.707",
"value": "left Front"
},
{
"operation": "Now on",
"time": "66.005",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "69.455",
"value": "Left Front"
},
{
"operation": "Now on",
"time": "70.371",
"value": "Left Front"
},
{
"operation": "hyperlink",
"time": "98.592",
"value": "Left Front (Tripura)"
},
{
"operation": "Now on",
"time": "99.546",
"value": "Left Front (Tripura)"
},
{
"operation": "Page search",
"time": "125.011",
"value": "Lok Sabha"
},
{
"operation": "Highlighting",
"time": "159.064",
"value": "Left Front (Tripura)_sentence_0"
},
{
"operation": "Highlighting",
"time": "162.919",
"value": "Left Front (Tripura)_sentence_2"
},
{
"operation": "Highlighting",
"time": "164.118",
"value": "Left Front (Tripura)_sentence_3"
},
{
"operation": "finish",
"time": "176.989",
"value": "finish"
}
]
}
|
exfever_2095
|
Chris McKenna is an American television writer, film producer, screenwriter, and television producer that has written for an American television sitcom created by Dan Harmon.
|
{
"evidence": [
"Chris McKenna (writer)",
"Community (TV series)"
],
"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:\nChris McKenna is an American television writer, film producer, screenwriter, and television producer that has written for an American television sitcom created by Dan Harmon.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Chris McKenna (writer)",
"Community (TV series)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 271,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
71647
|
California magnate granddaughter and American author Harriet Doerr studied at the private research institution Harvard University.
|
{
"evidence": [
"Harriet Doerr_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nCalifornia magnate granddaughter and American author Harriet Doerr studied at the private research institution Harvard University.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Harriet Doerr_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 272,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.749",
"value": "?search="
},
{
"operation": "search",
"time": "1073.377",
"value": "Harriet Doerr"
},
{
"operation": "Now on",
"time": "1075.983",
"value": "Harriet Doerr"
},
{
"operation": "Highlighting",
"time": "1159.024",
"value": "Harriet Doerr_sentence_1"
},
{
"operation": "Page search",
"time": "1179.563",
"value": "Harvard"
},
{
"operation": "finish",
"time": "1218.942",
"value": "finish"
}
]
}
|
28613
|
The Markee Castle was inherited by Bryan Cooper from his father in 1902 and their family lived there until he died in 1930 (except during World War I and when carrying out political duties), then after the Second World War, nobody took after and maintained it, but since 2015, it has been managed by the Corscadden family.
|
{
"evidence": [
"Markree Castle_sentence_20",
"Markree Castle_sentence_21",
"Markree Castle_sentence_25"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Markee Castle was inherited by Bryan Cooper from his father in 1902 and their family lived there until he died in 1930 (except during World War I and when carrying out political duties), then after the Second World War, nobody took after and maintained it, but since 2015, it has been managed by the Corscadden family.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Markree Castle_sentence_20",
"Markree Castle_sentence_21",
"Markree Castle_sentence_25"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 273,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.077",
"value": "?search="
},
{
"operation": "search",
"time": "10.3",
"value": "Markee Castle"
},
{
"operation": "Now on",
"time": "11.2",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "20.227",
"value": "Edward Joshua Cooper"
},
{
"operation": "Now on",
"time": "20.856",
"value": "Edward Joshua Cooper"
},
{
"operation": "hyperlink",
"time": "23.052",
"value": "Markree Castle"
},
{
"operation": "Now on",
"time": "23.356",
"value": "Markree Castle"
},
{
"operation": "Highlighting",
"time": "65.53",
"value": "Markree Castle_sentence_20"
},
{
"operation": "Highlighting",
"time": "85.626",
"value": "Markree Castle_sentence_21"
},
{
"operation": "Highlighting",
"time": "119.326",
"value": "Markree Castle_sentence_25"
},
{
"operation": "finish",
"time": "187.453",
"value": "finish"
}
]
}
|
45525
|
Master Madanwas born in India in 1927, he was a singer from 1937 to 1941 and died in Khan Khana whose capital is New Delhi.
|
{
"evidence": [
"Khan Khana_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:\nMaster Madanwas born in India in 1927, he was a singer from 1937 to 1941 and died in Khan Khana whose capital is New Delhi.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Khan Khana_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 274,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.963",
"value": "?search="
},
{
"operation": "search",
"time": "24.109",
"value": "Master Madan"
},
{
"operation": "Now on",
"time": "25.531",
"value": "Master Madan"
},
{
"operation": "hyperlink",
"time": "44.662",
"value": "Khan Khana"
},
{
"operation": "Now on",
"time": "46.365",
"value": "Khan Khana"
},
{
"operation": "Highlighting",
"time": "60.067",
"value": "Khan Khana_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "73.172",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "73.356",
"value": "Master Madan"
},
{
"operation": "finish",
"time": "104.729",
"value": "finish"
}
]
}
|
25770
|
Jurjen Ferdinand Koksma, a French mathematician, was the first math professor at the Vrije Universiteit Amsterdam.
|
{
"evidence": [
"Jurjen Ferdinand Koksma_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:\nJurjen Ferdinand Koksma, a French mathematician, was the first math professor at the Vrije Universiteit Amsterdam.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Jurjen Ferdinand Koksma_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 275,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "3.775",
"value": "?search="
},
{
"operation": "search",
"time": "10.899",
"value": "Jurjen Ferdinand Koksma"
},
{
"operation": "Now on",
"time": "11.6",
"value": "Jurjen Ferdinand Koksma"
},
{
"operation": "Highlighting",
"time": "18.016",
"value": "Jurjen Ferdinand Koksma_sentence_0"
},
{
"operation": "finish",
"time": "44.391",
"value": "finish"
}
]
}
|
44444
|
The Moroccan government attitudes towards homosexuality tend to be in the interests of the protection of the tradition of the country, in keeping with the culture's traditional gender roles and religious mores; thus it has banned books on heterosexuality and required schools to teach a conservative curriculum.
|
{
"evidence": [
"LGBT rights in Morocco_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:\nThe Moroccan government attitudes towards homosexuality tend to be in the interests of the protection of the tradition of the country, in keeping with the culture's traditional gender roles and religious mores; thus it has banned books on heterosexuality and required schools to teach a conservative curriculum.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"LGBT rights in Morocco_sentence_16"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 276,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.287",
"value": "?search="
},
{
"operation": "search",
"time": "22.41",
"value": "\"Moroccan government attitudes towards homosexuality\""
},
{
"operation": "Now on",
"time": "23.509",
"value": "undefined"
},
{
"operation": "search",
"time": "34.479",
"value": "Contains...moroccan government"
},
{
"operation": "Now on",
"time": "35.973",
"value": "undefined"
},
{
"operation": "search",
"time": "52.699",
"value": "\"in the interests of the protection of the tradition of the country\""
},
{
"operation": "Now on",
"time": "54.302",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "55.881",
"value": "LGBT rights in Morocco"
},
{
"operation": "Now on",
"time": "57.608",
"value": "LGBT rights in Morocco"
},
{
"operation": "Page search",
"time": "60.058",
"value": "in the interests of the protection of the tradition of the country"
},
{
"operation": "Highlighting",
"time": "62.96",
"value": "LGBT rights in Morocco_sentence_15"
},
{
"operation": "Highlighting",
"time": "80.117",
"value": "LGBT rights in Morocco_sentence_16"
},
{
"operation": "Highlighting deleted",
"time": "82.047",
"value": "LGBT rights in Morocco_sentence_15"
},
{
"operation": "finish",
"time": "100.356",
"value": "finish"
}
]
}
|
exfever_7449
|
Cougar Town is an American television sitcom that ran for 102 episodes over six seasons, from September 23, 2009, until March 31, 2015, which aired the first three seasons on an Asian multinational commercial broadcast television network that serves as the flagship property of Walt Disney Television.
|
{
"evidence": [
"Cougar Town",
"American Broadcasting Company"
],
"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:\nCougar Town is an American television sitcom that ran for 102 episodes over six seasons, from September 23, 2009, until March 31, 2015, which aired the first three seasons on an Asian multinational commercial broadcast television network that serves as the flagship property of Walt Disney Television.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Cougar Town",
"American Broadcasting Company"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 277,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1706
|
Ed Sheeran independently released the extended play, the fifth out of five EPs that Ed Sheeran recorded that featured a collection of artists whose genre of electronic music emerged in London in the early 1990s.
|
{
"evidence": [
"Ed Sheeran",
"No. 5 Collaborations Project",
"Grime (music genre)"
],
"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:\nEd Sheeran independently released the extended play, the fifth out of five EPs that Ed Sheeran recorded that featured a collection of artists whose genre of electronic music emerged in London in the early 1990s.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ed Sheeran",
"No. 5 Collaborations Project",
"Grime (music genre)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 278,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4942
|
The Great British Sewing Bee was the first British sewing competition.?
|
{
"evidence": [
"The Great British Sewing Bee",
"The Great British Bake Off"
],
"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 Great British Sewing Bee was the first British sewing competition.?\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Great British Sewing Bee",
"The Great British Bake Off"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 279,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
9011
|
Brinkhill had a parish church dedicated to Saint Philip, one of the Twelve Apostles of Jesus according to New Testament, and a churchyard where an ancient scheduled and Grade II listed churchyard cross stood.
|
{
"evidence": [
"Brinkhill_sentence_4",
"Philip the Apostle_sentence_0",
"Brinkhill_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:\nBrinkhill had a parish church dedicated to Saint Philip, one of the Twelve Apostles of Jesus according to New Testament, and a churchyard where an ancient scheduled and Grade II listed churchyard cross stood.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Brinkhill_sentence_4",
"Philip the Apostle_sentence_0",
"Brinkhill_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 280,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.459",
"value": "?search="
},
{
"operation": "search",
"time": "28.797",
"value": "Brinkhill"
},
{
"operation": "Now on",
"time": "29.271",
"value": "Brinkhill"
},
{
"operation": "Page search",
"time": "76.796",
"value": "Saint Philip"
},
{
"operation": "Highlighting",
"time": "91.802",
"value": "Brinkhill_sentence_4"
},
{
"operation": "hyperlink",
"time": "101.295",
"value": "Saint Philip"
},
{
"operation": "Now on",
"time": "102.023",
"value": "Philip the Apostle"
},
{
"operation": "Highlighting",
"time": "104.764",
"value": "Philip the Apostle_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "110.577",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "110.627",
"value": "Brinkhill"
},
{
"operation": "Highlighting",
"time": "114.574",
"value": "Brinkhill_sentence_5"
},
{
"operation": "hyperlink",
"time": "188.081",
"value": "Saint Philip"
},
{
"operation": "Now on",
"time": "188.949",
"value": "Philip the Apostle"
},
{
"operation": "Page search",
"time": "199.127",
"value": "saint"
},
{
"operation": "finish",
"time": "232.95",
"value": "finish"
}
]
}
|
exfever_8352
|
Carrie Fisher was an actress who appeared in a 1991 American comedy film directed by Michael Hoffman, from a screenplay by an American writer, producer, and film director.
|
{
"evidence": [
"Carrie Fisher",
"Soapdish"
],
"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:\nCarrie Fisher was an actress who appeared in a 1991 American comedy film directed by Michael Hoffman, from a screenplay by an American writer, producer, and film director.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Carrie Fisher",
"Soapdish"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 281,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
17183
|
In 1945 a narrow gauge tramway was built between Meadowood, about 5 miles west of Wanlockhead and other mines and a washery at Leadhills were also connected.
|
{
"evidence": [
"Leadhills and Wanlockhead Branch_sentence_7"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn 1945 a narrow gauge tramway was built between Meadowood, about 5 miles west of Wanlockhead and other mines and a washery at Leadhills were also connected.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Leadhills and Wanlockhead Branch_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 282,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.736",
"value": "undefined"
},
{
"operation": "search",
"time": "38.617",
"value": "washery at leadhills"
},
{
"operation": "Now on",
"time": "40.268",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "44.663",
"value": "Leadhills and Wanlockhead Branch"
},
{
"operation": "Now on",
"time": "45.449",
"value": "Leadhills and Wanlockhead Branch"
},
{
"operation": "Highlighting",
"time": "52.544",
"value": "Leadhills and Wanlockhead Branch_sentence_7"
},
{
"operation": "finish",
"time": "85.254",
"value": "finish"
}
]
}
|
48842
|
The Occult Roots of Nazism is falsely believed to have been written by Nicholas Goodrick-Clarke(15 January 1953 – 29 August 2012) in English in 1985.
|
{
"evidence": [
"The Occult Roots of Nazism_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 Occult Roots of Nazism is falsely believed to have been written by Nicholas Goodrick-Clarke(15 January 1953 – 29 August 2012) in English in 1985.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"The Occult Roots of Nazism_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 283,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.314",
"value": "?search="
},
{
"operation": "search",
"time": "23.948",
"value": "The Occult Roots of Nazism"
},
{
"operation": "Now on",
"time": "24.651",
"value": "The Occult Roots of Nazism"
},
{
"operation": "Page search",
"time": "33.34",
"value": "fals"
},
{
"operation": "Highlighting",
"time": "58.04",
"value": "The Occult Roots of Nazism_sentence_0"
},
{
"operation": "finish",
"time": "66.819",
"value": "finish"
}
]
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.