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 |
|---|---|---|---|---|---|---|---|---|
26579
|
Peter Cardew designed the Canadian National Pavilion for Expo 86 in Vancouver.
|
{
"evidence": [
"Peter Cardew_sentence_27",
"Peter Cardew_sentence_28"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPeter Cardew designed the Canadian National Pavilion for Expo 86 in Vancouver.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Peter Cardew_sentence_27",
"Peter Cardew_sentence_28"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 884,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.654",
"value": "?search="
},
{
"operation": "search",
"time": "63.415",
"value": "Peter Cardew"
},
{
"operation": "Now on",
"time": "64.465",
"value": "Peter Cardew"
},
{
"operation": "Highlighting",
"time": "84.836",
"value": "Peter Cardew_header_cell_0_7_0"
},
{
"operation": "Highlighting deleted",
"time": "91.488",
"value": "Peter Cardew_header_cell_0_7_0"
},
{
"operation": "Page search",
"time": "95.98",
"value": "pavillion"
},
{
"operation": "Page search",
"time": "117.165",
"value": "pavillon"
},
{
"operation": "Page search",
"time": "119.454",
"value": "pavilion"
},
{
"operation": "Highlighting",
"time": "136.925",
"value": "Peter Cardew_sentence_27"
},
{
"operation": "Highlighting",
"time": "139.45",
"value": "Peter Cardew_sentence_28"
},
{
"operation": "finish",
"time": "359.637",
"value": "finish"
}
]
}
|
60201
|
Sepia mascarensis is a cuttlefish that can be found in the western Indian Ocean at depths of 87 to 325 meters.
|
{
"evidence": [
"Sepia mascarensis_sentence_0",
"Sepia mascarensis_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:\nSepia mascarensis is a cuttlefish that can be found in the western Indian Ocean at depths of 87 to 325 meters.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Sepia mascarensis_sentence_0",
"Sepia mascarensis_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 885,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.887",
"value": "undefined"
},
{
"operation": "search",
"time": "150.119",
"value": "Sepia mascarensis"
},
{
"operation": "Now on",
"time": "151.328",
"value": "Sepia mascarensis"
},
{
"operation": "Highlighting",
"time": "153.885",
"value": "Sepia mascarensis_sentence_0"
},
{
"operation": "Highlighting",
"time": "161.304",
"value": "Sepia mascarensis_sentence_1"
},
{
"operation": "finish",
"time": "178.215",
"value": "finish"
}
]
}
|
exfever_5786
|
Patrick Dempsey is known for his role in a series that focuses on the lives of surgical residents and a physician in training who has completed medical school and has a medical degree but does not yet have a license to practice medicine unsupervised.
|
{
"evidence": [
"Patrick Dempsey",
"Grey's Anatomy",
"Internship (medicine)"
],
"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:\nPatrick Dempsey is known for his role in a series that focuses on the lives of surgical residents and a physician in training who has completed medical school and has a medical degree but does not yet have a license to practice medicine unsupervised.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Patrick Dempsey",
"Grey's Anatomy",
"Internship (medicine)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 886,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6697
|
Lana Wachowski are American film and television directors that wrote the 2006 film V for Vendetta, which was adapted from the graphic novel by an English writer who achieved success by publishing comic strips in magazines such as the weekly British science fiction-oriented comic magazine.
|
{
"evidence": [
"The Wachowskis",
"Alan Moore",
"2000 AD (comics)"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLana Wachowski are American film and television directors that wrote the 2006 film V for Vendetta, which was adapted from the graphic novel by an English writer who achieved success by publishing comic strips in magazines such as the weekly British science fiction-oriented comic magazine.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Wachowskis",
"Alan Moore",
"2000 AD (comics)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 887,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2317
|
Samuel Frederick Smith rose to prominence after being featured on the single "Latch" by an English electronic music duo consisting of brothers Howard and Guy Lawrence who grew up in a town in Surrey.
|
{
"evidence": [
"Sam Smith (singer)",
"Disclosure (band)",
"Reigate"
],
"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:\nSamuel Frederick Smith rose to prominence after being featured on the single \"Latch\" by an English electronic music duo consisting of brothers Howard and Guy Lawrence who grew up in a town in Surrey.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sam Smith (singer)",
"Disclosure (band)",
"Reigate"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 888,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4764
|
Henry V is a 2019 British television film directed by Thea Sharrock and stars Tom Hiddleston as King of England from 1413 until his death in 1422 and is known and celebrated as one of the greatest warrior-kings of medieval England.
|
{
"evidence": [
"Henry V (2012 film)",
"Henry V of England"
],
"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:\nHenry V is a 2019 British television film directed by Thea Sharrock and stars Tom Hiddleston as King of England from 1413 until his death in 1422 and is known and celebrated as one of the greatest warrior-kings of medieval England.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Henry V (2012 film)",
"Henry V of England"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 889,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
51233
|
The engagement was celebrated at the royal residence in Bergen in 1219, and the wedding took place there on May 25, 1225; however, in 1239, the conflict between her father and brother erupted into open warfare.
|
{
"evidence": [
"Margaret Skulesdatter_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe engagement was celebrated at the royal residence in Bergen in 1219, and the wedding took place there on May 25, 1225; however, in 1239, the conflict between her father and brother erupted into open warfare.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Margaret Skulesdatter_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 890,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.108",
"value": "?search="
},
{
"operation": "search",
"time": "46.742",
"value": "wedding took place there on May 25, 1225 in bergen"
},
{
"operation": "Now on",
"time": "48.335",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "54.097",
"value": "Margaret Skulesdatter"
},
{
"operation": "Now on",
"time": "54.809",
"value": "Margaret Skulesdatter"
},
{
"operation": "Page search",
"time": "61.612",
"value": "residence"
},
{
"operation": "Highlighting",
"time": "64.527",
"value": "Margaret Skulesdatter_sentence_4"
},
{
"operation": "Page search",
"time": "75.298",
"value": "1239"
},
{
"operation": "Highlighting",
"time": "79.209",
"value": "Margaret Skulesdatter_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "84.701",
"value": "Margaret Skulesdatter_sentence_4"
},
{
"operation": "finish",
"time": "222.964",
"value": "finish"
}
]
}
|
exfever_3437
|
Grand Canyon is Mary-Louise Parker's fourth film.
|
{
"evidence": [
"Mary-Louise Parker",
"Prelude to a Kiss (play)",
"Duke Ellington"
],
"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:\nGrand Canyon is Mary-Louise Parker's fourth film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mary-Louise Parker",
"Prelude to a Kiss (play)",
"Duke Ellington"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 891,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8013
|
Guday is set in 1971–1988 in the capital of the Indian state of West Bengal.
|
{
"evidence": [
"Gunday",
"Kolkata"
],
"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:\nGuday is set in 1971–1988 in the capital of the Indian state of West Bengal.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Gunday",
"Kolkata"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 892,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
60792
|
Braintree branch line was originally constructed from Maldon to Braintree, the line was proposed by MWBR and the Bishop's Stortford–Braintree branch line was opened in 1869.
|
{
"evidence": [
"Braintree branch line_sentence_7",
"Braintree branch line_sentence_8",
"Braintree branch line_sentence_11"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBraintree branch line was originally constructed from Maldon to Braintree, the line was proposed by MWBR and the Bishop's Stortford–Braintree branch line was opened in 1869.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Braintree branch line_sentence_7",
"Braintree branch line_sentence_8",
"Braintree branch line_sentence_11"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 893,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.252",
"value": "?search="
},
{
"operation": "search",
"time": "6.301",
"value": "Braintree branch line"
},
{
"operation": "Now on",
"time": "7.134",
"value": "Braintree branch line"
},
{
"operation": "Highlighting",
"time": "11.394",
"value": "Braintree branch line_sentence_7"
},
{
"operation": "Highlighting",
"time": "15.338",
"value": "Braintree branch line_sentence_8"
},
{
"operation": "Highlighting",
"time": "22.552",
"value": "Braintree branch line_sentence_11"
},
{
"operation": "finish",
"time": "30.657",
"value": "finish"
}
]
}
|
70947
|
Hamish MacInnes, a Scottish mountaineer, explorer, mountain search and rescuer, and author, pioneered the exploration of the Everest cliffs for winter work with the Glencoe School of Winter Climbing and led the area's mountain rescue team from 1961, and is recognised as having developed modern mountain rescue in Scotland, setting up the Search and Rescue Dog Association with his wife in 1965 and the Scottish Avalanche Information Service in 1998.
|
{
"evidence": [
"Hamish MacInnes_sentence_19"
],
"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:\nHamish MacInnes, a Scottish mountaineer, explorer, mountain search and rescuer, and author, pioneered the exploration of the Everest cliffs for winter work with the Glencoe School of Winter Climbing and led the area's mountain rescue team from 1961, and is recognised as having developed modern mountain rescue in Scotland, setting up the Search and Rescue Dog Association with his wife in 1965 and the Scottish Avalanche Information Service in 1998.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Hamish MacInnes_sentence_19"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 894,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "6.172",
"value": "?search="
},
{
"operation": "search",
"time": "110.219",
"value": "Hamish MacInnes"
},
{
"operation": "Now on",
"time": "112.192",
"value": "Hamish MacInnes"
},
{
"operation": "Highlighting",
"time": "150.748",
"value": "Hamish MacInnes_sentence_0"
},
{
"operation": "Highlighting",
"time": "216.823",
"value": "Hamish MacInnes_sentence_19"
},
{
"operation": "finish",
"time": "243.236",
"value": "finish"
}
]
}
|
24454
|
Diallo was the secretary-general of the District Workers Committee in Conakry from 1981 to 1984, but before this, her work as a labor unionist (workplace volunteers who are appointed by members in democratic elections) began early.
|
{
"evidence": [
"Rabiatou Sérah Diallo_sentence_7",
"Rabiatou Sérah Diallo_sentence_5",
"Trade union_sentence_5",
"Rabiatou Sérah Diallo_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:\nDiallo was the secretary-general of the District Workers Committee in Conakry from 1981 to 1984, but before this, her work as a labor unionist (workplace volunteers who are appointed by members in democratic elections) began early.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Rabiatou Sérah Diallo_sentence_7",
"Rabiatou Sérah Diallo_sentence_5",
"Trade union_sentence_5",
"Rabiatou Sérah Diallo_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 895,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.508",
"value": "?search="
},
{
"operation": "search",
"time": "5.9",
"value": "Diallo"
},
{
"operation": "Now on",
"time": "6.434",
"value": "Diallo"
},
{
"operation": "Page search",
"time": "15.639",
"value": "secretary"
},
{
"operation": "search",
"time": "55.847",
"value": "Contains...Diallo was the secretary-general of the District Workers Committee in Conakry from 1981 to 1984"
},
{
"operation": "Now on",
"time": "57.984",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "59.483",
"value": "Rabiatou Sérah Diallo"
},
{
"operation": "Now on",
"time": "59.918",
"value": "Rabiatou Sérah Diallo"
},
{
"operation": "Highlighting",
"time": "69.923",
"value": "Rabiatou Sérah Diallo_sentence_1"
},
{
"operation": "Page search",
"time": "97.538",
"value": "union"
},
{
"operation": "Highlighting",
"time": "115.513",
"value": "Rabiatou Sérah Diallo_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "117.031",
"value": "Rabiatou Sérah Diallo_sentence_1"
},
{
"operation": "Highlighting",
"time": "134.891",
"value": "Rabiatou Sérah Diallo_sentence_5"
},
{
"operation": "Page search",
"time": "142.339",
"value": "appointed"
},
{
"operation": "hyperlink",
"time": "152.619",
"value": "trade unionist"
},
{
"operation": "Now on",
"time": "153.282",
"value": "Trade union"
},
{
"operation": "Page search",
"time": "159.368",
"value": "volun"
},
{
"operation": "Highlighting",
"time": "161.905",
"value": "Trade union_sentence_5"
},
{
"operation": "back-button-clicked",
"time": "169.479",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "169.552",
"value": "Rabiatou Sérah Diallo"
},
{
"operation": "Highlighting",
"time": "202.957",
"value": "Rabiatou Sérah Diallo_sentence_6"
},
{
"operation": "finish",
"time": "243.111",
"value": "finish"
}
]
}
|
exfever_2832
|
Gorgeous Ladies of Wrestling is a women's professional wrestling promotion that began in 1986 and has continued in various forms after it left a telecommunication medium used for transmitting moving images in black-and-white or in color, and two or three dimensions and sound.
|
{
"evidence": [
"Gorgeous Ladies of Wrestling",
"Television"
],
"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:\nGorgeous Ladies of Wrestling is a women's professional wrestling promotion that began in 1986 and has continued in various forms after it left a telecommunication medium used for transmitting moving images in black-and-white or in color, and two or three dimensions and sound.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Gorgeous Ladies of Wrestling",
"Television"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 896,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
52205
|
The impact factor (the scientometric index calculated by Clarivate that reflects the yearly average number of citations that articles published in the last two years in a given journal received) of the European Journal of Neuroscience is 3.115, ranking it 128 out of 272 journals.
|
{
"evidence": [
"European Journal of Neuroscience_sentence_8",
"Impact factor_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 impact factor (the scientometric index calculated by Clarivate that reflects the yearly average number of citations that articles published in the last two years in a given journal received) of the European Journal of Neuroscience is 3.115, ranking it 128 out of 272 journals.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"European Journal of Neuroscience_sentence_8",
"Impact factor_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 897,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.3",
"value": "?search="
},
{
"operation": "search",
"time": "21.16",
"value": "European Journal of Neuroscience"
},
{
"operation": "Now on",
"time": "21.873",
"value": "European Journal of Neuroscience"
},
{
"operation": "Highlighting",
"time": "38.313",
"value": "European Journal of Neuroscience_header_cell_0_8_0"
},
{
"operation": "Highlighting",
"time": "42.231",
"value": "European Journal of Neuroscience_cell_0_8_1"
},
{
"operation": "Highlighting",
"time": "42.238",
"value": "European Journal of Neuroscience_header_cell_0_8_0"
},
{
"operation": "Highlighting",
"time": "42.239",
"value": "European Journal of Neuroscience_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "50.151",
"value": "European Journal of Neuroscience_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "53.984",
"value": "European Journal of Neuroscience_cell_0_8_1"
},
{
"operation": "Highlighting deleted",
"time": "53.99",
"value": "European Journal of Neuroscience_header_cell_0_8_0"
},
{
"operation": "Highlighting deleted",
"time": "53.99",
"value": "European Journal of Neuroscience_header_cell_0_3_0"
},
{
"operation": "hyperlink",
"time": "64.895",
"value": "Impact factor"
},
{
"operation": "Now on",
"time": "65.828",
"value": "Impact factor"
},
{
"operation": "Highlighting",
"time": "88.567",
"value": "Impact factor_sentence_2"
},
{
"operation": "finish",
"time": "99.903",
"value": "finish"
}
]
}
|
35691
|
The professional wrestling pay-per-view event put on by Juggalo Championship Wrestling took place in 2011 and featured nine different matches with wrestlers such as X-Pac, Corporal Robinson, Bill Martel and Officer Colt Cabana.
|
{
"evidence": [
"Juggalo Championship Wrestling_sentence_11"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe professional wrestling pay-per-view event put on by Juggalo Championship Wrestling took place in 2011 and featured nine different matches with wrestlers such as X-Pac, Corporal Robinson, Bill Martel and Officer Colt Cabana.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Juggalo Championship Wrestling_sentence_11"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 898,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.099",
"value": "?search="
},
{
"operation": "search",
"time": "40.252",
"value": "Juggalo Championship Wrestling"
},
{
"operation": "Now on",
"time": "41.121",
"value": "Juggalo Championship Wrestling"
},
{
"operation": "hyperlink",
"time": "121.534",
"value": "List of Juggalo Championship Wrestling employees"
},
{
"operation": "Now on",
"time": "122.21",
"value": "List of Juggalo Championship Wrestling personnel"
},
{
"operation": "back-button-clicked",
"time": "159.119",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "159.23",
"value": "Juggalo Championship Wrestling"
},
{
"operation": "Highlighting",
"time": "177.654",
"value": "Juggalo Championship Wrestling_sentence_11"
},
{
"operation": "hyperlink",
"time": "179.379",
"value": "internet"
},
{
"operation": "Now on",
"time": "180.465",
"value": "Internet"
},
{
"operation": "back-button-clicked",
"time": "182.42",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "182.534",
"value": "Juggalo Championship Wrestling"
},
{
"operation": "hyperlink",
"time": "183.538",
"value": "pay-per-view"
},
{
"operation": "Now on",
"time": "185.089",
"value": "Pay-per-view"
},
{
"operation": "back-button-clicked",
"time": "204.22",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "204.351",
"value": "Juggalo Championship Wrestling"
},
{
"operation": "search",
"time": "243.804",
"value": "X-Pac"
},
{
"operation": "Now on",
"time": "244.918",
"value": "Sean Waltman"
},
{
"operation": "finish",
"time": "267.703",
"value": "finish"
}
]
}
|
88736
|
Joseph Boatner led various groups using the "Ink Spots" name until his death in Laconia, New Hampshire; home of the Laconia World Championship Sled Dog Derby.
|
{
"evidence": [
"Joseph Boatner_sentence_2",
"Joseph Boatner_sentence_3",
"Laconia, New Hampshire_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:\nJoseph Boatner led various groups using the \"Ink Spots\" name until his death in Laconia, New Hampshire; home of the Laconia World Championship Sled Dog Derby.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Joseph Boatner_sentence_2",
"Joseph Boatner_sentence_3",
"Laconia, New Hampshire_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 899,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.548",
"value": "?search="
},
{
"operation": "search",
"time": "16.555",
"value": "Joseph Boatner"
},
{
"operation": "Now on",
"time": "17.423",
"value": "Joseph Boatner"
},
{
"operation": "Highlighting",
"time": "260.584",
"value": "Joseph Boatner_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "271.699",
"value": "Joseph Boatner_sentence_0"
},
{
"operation": "Highlighting",
"time": "284.726",
"value": "Joseph Boatner_sentence_0"
},
{
"operation": "Highlighting",
"time": "287.089",
"value": "Joseph Boatner_sentence_1"
},
{
"operation": "Highlighting",
"time": "292.05",
"value": "Joseph Boatner_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "293.378",
"value": "Joseph Boatner_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "297.373",
"value": "Joseph Boatner_sentence_0"
},
{
"operation": "Highlighting",
"time": "300.048",
"value": "Joseph Boatner_sentence_3"
},
{
"operation": "hyperlink",
"time": "303.783",
"value": "Laconia, New Hampshire"
},
{
"operation": "Now on",
"time": "304.921",
"value": "Laconia, New Hampshire"
},
{
"operation": "Page search",
"time": "342.464",
"value": "Sled Dog Derby"
},
{
"operation": "Highlighting",
"time": "346.095",
"value": "Laconia, New Hampshire_sentence_4"
},
{
"operation": "back-button-clicked",
"time": "348.318",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "348.376",
"value": "Joseph Boatner"
},
{
"operation": "finish",
"time": "405.707",
"value": "finish"
}
]
}
|
exfever_7923
|
Joni Mitchell is an Indian-American singer-songwriter who began singing in small nightclubs in western Canada, before moving on to the nightclubs of the capital city of the Canadian province of Ontario which is the most populous city in Canada and the fourth most populous city in North America.
|
{
"evidence": [
"Joni Mitchell",
"Toronto"
],
"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:\nJoni Mitchell is an Indian-American singer-songwriter who began singing in small nightclubs in western Canada, before moving on to the nightclubs of the capital city of the Canadian province of Ontario which is the most populous city in Canada and the fourth most populous city in North America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Joni Mitchell",
"Toronto"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 900,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3112
|
Ben Stiller is the youngest son of comedian Jerry Stiller.
|
{
"evidence": [
"Ben Stiller",
"The Secret Life of Walter Mitty (2013 film)"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBen Stiller is the youngest son of comedian Jerry Stiller.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ben Stiller",
"The Secret Life of Walter Mitty (2013 film)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 901,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1992
|
The Hundred-Foot Journey is a 2014 French comedy-drama film directed by Lasse Hallström and starring an English actress who received an Academy Award and a British Academy Film Award for her portrayal of Queen Elizabeth II in The Queen.
|
{
"evidence": [
"The Hundred-Foot Journey (film)",
"Helen Mirren"
],
"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 Hundred-Foot Journey is a 2014 French comedy-drama film directed by Lasse Hallström and starring an English actress who received an Academy Award and a British Academy Film Award for her portrayal of Queen Elizabeth II in The Queen.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Hundred-Foot Journey (film)",
"Helen Mirren"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 902,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
80302
|
The Lions were to play in every one of the next twelve Eurobowls; wherein they won the first three of those, the first one against the Cologne Crocodiles and title number two and three against the Hamburg Blue Devils, the latest in front of a record crowd of over 30,000 in Hamburg.
|
{
"evidence": [
"New Yorker Lions_sentence_12",
"New Yorker Lions_sentence_13",
"New Yorker Lions_sentence_14"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Lions were to play in every one of the next twelve Eurobowls; wherein they won the first three of those, the first one against the Cologne Crocodiles and title number two and three against the Hamburg Blue Devils, the latest in front of a record crowd of over 30,000 in Hamburg.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"New Yorker Lions_sentence_12",
"New Yorker Lions_sentence_13",
"New Yorker Lions_sentence_14"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 903,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.799",
"value": "?search="
},
{
"operation": "search",
"time": "32.494",
"value": "Eurobowl"
},
{
"operation": "Now on",
"time": "33.858",
"value": "Eurobowl"
},
{
"operation": "hyperlink",
"time": "47.586",
"value": "Braunschweig Lions"
},
{
"operation": "Now on",
"time": "49.194",
"value": "New Yorker Lions"
},
{
"operation": "Highlighting",
"time": "84.41",
"value": "New Yorker Lions_sentence_12"
},
{
"operation": "Highlighting",
"time": "95.523",
"value": "New Yorker Lions_sentence_13"
},
{
"operation": "Highlighting",
"time": "111.49",
"value": "New Yorker Lions_sentence_14"
},
{
"operation": "finish",
"time": "122.46",
"value": "finish"
}
]
}
|
exfever_2102
|
Beauty and the Beast is a 2017 American musical romantic fantasy film directed by Bill Condon from a screenplay by Stephen Chbosky and Evan Spiliotopoulos, the film is a live-action adaptation of Disney's 1991 American animated musical romantic fantasy film produced by Walt Disney Feature Animation and released by Walt Disney Pictures.
|
{
"evidence": [
"Beauty and the Beast (2017 film)",
"Beauty and the Beast (1991 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:\nBeauty and the Beast is a 2017 American musical romantic fantasy film directed by Bill Condon from a screenplay by Stephen Chbosky and Evan Spiliotopoulos, the film is a live-action adaptation of Disney's 1991 American animated musical romantic fantasy film produced by Walt Disney Feature Animation and released by Walt Disney Pictures.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Beauty and the Beast (2017 film)",
"Beauty and the Beast (1991 film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 904,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5674
|
Shifty is an American urban crime thriller filmed in Borehamwood, starring a veteran actress who is a six-time BAFTA TV Award nominee and won the 1979 BAFTA TV Award for Best Actress for the ITV series Lillie.
|
{
"evidence": [
"Shifty (film)",
"Francesca Annis"
],
"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:\nShifty is an American urban crime thriller filmed in Borehamwood, starring a veteran actress who is a six-time BAFTA TV Award nominee and won the 1979 BAFTA TV Award for Best Actress for the ITV series Lillie.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Shifty (film)",
"Francesca Annis"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 905,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
19572
|
Leonard Ross became known as the whiz kid who won $100,000 on The Big Surprise, a television quiz show, at age ten, and also as a prize, he became the first guest to ring the opening bell at the New York Stock Exchange.
|
{
"evidence": [
"Leonard Ross_sentence_7",
"Leonard Ross_sentence_9"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLeonard Ross became known as the whiz kid who won $100,000 on The Big Surprise, a television quiz show, at age ten, and also as a prize, he became the first guest to ring the opening bell at the New York Stock Exchange.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Leonard Ross_sentence_7",
"Leonard Ross_sentence_9"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 906,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.487",
"value": "?search="
},
{
"operation": "search",
"time": "10.478",
"value": "Leonard Ross"
},
{
"operation": "Now on",
"time": "11.469",
"value": "Leonard Ross"
},
{
"operation": "Highlighting",
"time": "13.36",
"value": "Leonard Ross_sentence_7"
},
{
"operation": "Highlighting",
"time": "21.616",
"value": "Leonard Ross_sentence_9"
},
{
"operation": "finish",
"time": "56.763",
"value": "finish"
}
]
}
|
exfever_3395
|
Jeffrey Lionel Dahmer was one of the most wanted criminals in America.
|
{
"evidence": [
"Jeffrey Dahmer",
"Serial killer"
],
"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:\nJeffrey Lionel Dahmer was one of the most wanted criminals in America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jeffrey Dahmer",
"Serial killer"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 907,
"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": 908,
"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"
}
]
}
|
exfever_3152
|
The Bostonians is based on an American novel starring an American actor, director, and activist, best known for playing the titular main character in the film Batman and its first three sequels.
|
{
"evidence": [
"The Bostonians (film)",
"Christopher Reeve"
],
"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 Bostonians is based on an American novel starring an American actor, director, and activist, best known for playing the titular main character in the film Batman and its first three sequels.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Bostonians (film)",
"Christopher Reeve"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 909,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_178
|
Chris Hemsworth rose to prominence with the television series Red Dawn and appeared in the American science fiction action film that follows James T. Kirk, played by the American actor best known as Will Colson in Unstoppable.
|
{
"evidence": [
"Chris Hemsworth",
"Star Trek (film)",
"Chris Pine"
],
"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:\nChris Hemsworth rose to prominence with the television series Red Dawn and appeared in the American science fiction action film that follows James T. Kirk, played by the American actor best known as Will Colson in Unstoppable.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Chris Hemsworth",
"Star Trek (film)",
"Chris Pine"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 910,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1105
|
How to Train Your Dragon is a 2011 American computer-animated action film that belong to the fantasy genre with fantastic themes, usually magic, supernatural events,
|
{
"evidence": [
"How to Train Your Dragon (film)",
"Fantasy 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:\nHow to Train Your Dragon is a 2011 American computer-animated action film that belong to the fantasy genre with fantastic themes, usually magic, supernatural events,\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"How to Train Your Dragon (film)",
"Fantasy film"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 911,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6001
|
The Big Chill is an American comedy-drama film starring an American actor who has starred in some of the lowest-grossing films.
|
{
"evidence": [
"The Big Chill (film)",
"Jeff Goldblum"
],
"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 Big Chill is an American comedy-drama film starring an American actor who has starred in some of the lowest-grossing films.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Big Chill (film)",
"Jeff Goldblum"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 912,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4290
|
Miley Cyrus produced and directed the last episode of Smallville.
|
{
"evidence": [
"See You Again (Miley Cyrus song)",
"Hannah Montana"
],
"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:\nMiley Cyrus produced and directed the last episode of Smallville.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"See You Again (Miley Cyrus song)",
"Hannah Montana"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 913,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
90562
|
Ibraheim Campbell grew up in the Philippines, where he started playing tumbang preso.
|
{
"evidence": [
"Ibraheim Campbell_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:\nIbraheim Campbell grew up in the Philippines, where he started playing tumbang preso.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ibraheim Campbell_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 914,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.37",
"value": "?search="
},
{
"operation": "search",
"time": "16.87",
"value": "Ibraheim Campbell"
},
{
"operation": "Now on",
"time": "17.655",
"value": "Ibraheim Campbell"
},
{
"operation": "Highlighting",
"time": "463.387",
"value": "Ibraheim Campbell_sentence_3"
},
{
"operation": "finish",
"time": "472.8",
"value": "finish"
}
]
}
|
75646
|
The winner in Uva Province, Viyaluwa Polling Division during the 2010 Sri Lankan Parliamentary elections was UPFA (a political alliance in Sri Lanka founded by former Sri Lankan president Chandrika Kumaratunga in 2004).
|
{
"evidence": [
"Viyaluwa Polling Division_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe winner in Uva Province, Viyaluwa Polling Division during the 2010 Sri Lankan Parliamentary elections was UPFA (a political alliance in Sri Lanka founded by former Sri Lankan president Chandrika Kumaratunga in 2004).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Viyaluwa Polling Division_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 915,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "8.379",
"value": "undefined"
},
{
"operation": "search",
"time": "43.313",
"value": "2010 Sri Lankan parliamentary election"
},
{
"operation": "Now on",
"time": "45.957",
"value": "2010 Sri Lankan parliamentary election"
},
{
"operation": "hyperlink",
"time": "56.267",
"value": "5.1 District by district 2010 Sri Lankan parliamentary election_section_5"
},
{
"operation": "search",
"time": "197.092",
"value": "Viyaluwa Polling Divisio"
},
{
"operation": "Now on",
"time": "201.714",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "209.063",
"value": "Viyaluwa Polling Division"
},
{
"operation": "Now on",
"time": "212.396",
"value": "Viyaluwa Polling Division"
},
{
"operation": "Highlighting",
"time": "217.161",
"value": "Viyaluwa Polling Division_sentence_0"
},
{
"operation": "hyperlink",
"time": "226.147",
"value": "2.3 2010 Sri Lankan Parliamentary Election Viyaluwa Polling Division_section_13"
},
{
"operation": "hyperlink",
"time": "256.784",
"value": "2.3 2010 Sri Lankan Parliamentary Election Viyaluwa Polling Division_section_13"
},
{
"operation": "back-button-clicked",
"time": "266.634",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "274.242",
"value": "undefined"
},
{
"operation": "finish",
"time": "280.866",
"value": "finish"
}
]
}
|
exfever_4841
|
Klute stars an American actor who plays Lt. Matt Reardon in Get Christie Love!, the opposite of an American actress.
|
{
"evidence": [
"Klute",
"Charles Cioffi",
"Teresa Graves"
],
"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:\nKlute stars an American actor who plays Lt. Matt Reardon in Get Christie Love!, the opposite of an American actress.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Klute",
"Charles Cioffi",
"Teresa Graves"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 916,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
23492
|
Most Somali men (residents of Australia who are of culturally and linguistically Arabized people native to Greater Somalia ancestry) have found jobs in the transport and production industries, and a rising number of women in the community do clerical work; while communal support networks have also expanded.
|
{
"evidence": [
"Somali Australians_sentence_11"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMost Somali men (residents of Australia who are of culturally and linguistically Arabized people native to Greater Somalia ancestry) have found jobs in the transport and production industries, and a rising number of women in the community do clerical work; while communal support networks have also expanded.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Somali Australians_sentence_11"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 917,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.27",
"value": "?search="
},
{
"operation": "search",
"time": "29.858",
"value": "Contains...Somali men"
},
{
"operation": "Now on",
"time": "31.292",
"value": "undefined"
},
{
"operation": "search",
"time": "51.417",
"value": "Contains...have found jobs in the transport and production industries, and a rising number of women in the community do clerical work"
},
{
"operation": "Now on",
"time": "54.523",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "59.415",
"value": "Somali Australians"
},
{
"operation": "Now on",
"time": "60.555",
"value": "Somali Australians"
},
{
"operation": "Page search",
"time": "86.262",
"value": "nativ"
},
{
"operation": "search",
"time": "161.515",
"value": "Contains...residents of Australia who are of culturally and linguistically Arabized people native to Greater Somalia ancestry"
},
{
"operation": "Now on",
"time": "163.81",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "177.898",
"value": "Autosomal ancestry Somalis_section_27"
},
{
"operation": "Now on",
"time": "180.389",
"value": "Somalis#Autosomal ancestry Somalis section 27"
},
{
"operation": "search",
"time": "282.641",
"value": "Somali Australians"
},
{
"operation": "Now on",
"time": "283.513",
"value": "Somali Australians"
},
{
"operation": "Page search",
"time": "296.534",
"value": "transport"
},
{
"operation": "Highlighting",
"time": "305.2",
"value": "Somali Australians_sentence_11"
},
{
"operation": "Page search",
"time": "330.831",
"value": "cultural"
},
{
"operation": "Page search",
"time": "353.364",
"value": "Arabized"
},
{
"operation": "hyperlink",
"time": "360.249",
"value": "Somali diaspora"
},
{
"operation": "Now on",
"time": "361.191",
"value": "Somali diaspora"
},
{
"operation": "Page search",
"time": "366.319",
"value": "Arabized"
},
{
"operation": "Page search",
"time": "368.961",
"value": "Arabized"
},
{
"operation": "search",
"time": "387.18",
"value": "Contains...residents of Australia who are of culturally and linguistically Arabized"
},
{
"operation": "Now on",
"time": "388.912",
"value": "undefined"
},
{
"operation": "search",
"time": "436.791",
"value": "Contains...people native to Greater Somalia ancestry"
},
{
"operation": "Now on",
"time": "438.255",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "445.355",
"value": "Autosomal ancestry Somalis_section_27"
},
{
"operation": "Now on",
"time": "446.667",
"value": "Somalis#Autosomal ancestry Somalis section 27"
},
{
"operation": "Page search",
"time": "465.614",
"value": "people native to Greater Somalia ancestry"
},
{
"operation": "Page search",
"time": "518.035",
"value": "native to Greater Somalia ancestry"
},
{
"operation": "search",
"time": "530.033",
"value": "Contains...culturally and linguistically Arabized"
},
{
"operation": "Now on",
"time": "531.246",
"value": "undefined"
},
{
"operation": "search",
"time": "589.712",
"value": "somali arabs"
},
{
"operation": "Now on",
"time": "590.512",
"value": "undefined"
},
{
"operation": "search",
"time": "655.119",
"value": "Contains... have found jobs in the transport and production industries"
},
{
"operation": "Now on",
"time": "658.068",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "667.468",
"value": "Somali Australians"
},
{
"operation": "Now on",
"time": "668.289",
"value": "Somali Australians"
},
{
"operation": "Page search",
"time": "690.17",
"value": "clerical"
},
{
"operation": "Highlighting deleted",
"time": "695.333",
"value": "Somali Australians_sentence_11"
},
{
"operation": "Highlighting",
"time": "696.755",
"value": "Somali Australians_sentence_11"
},
{
"operation": "finish",
"time": "750.653",
"value": "finish"
}
]
}
|
exfever_563
|
Louis Tomlinson released the second solo single.
|
{
"evidence": [
"Louis Tomlinson",
"Back to You (Louis Tomlinson song)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLouis Tomlinson released the second solo single.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Louis Tomlinson",
"Back to You (Louis Tomlinson song)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 918,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7471
|
Attack the Block is the first comedy horror film of Jodie Whittaker.
|
{
"evidence": [
"Attack the Block",
"Jodie Whittaker"
],
"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:\nAttack the Block is the first comedy horror film of Jodie Whittaker.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Attack the Block",
"Jodie Whittaker"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 919,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
47246
|
NFATC2 a member of the nuclear factor of activated T cells and the product of this gene is a DNA-binding protein (have a specific or general affinity for single- or double-stranded DNA).
|
{
"evidence": [
"NFATC2_sentence_1",
"NFATC2_sentence_2",
"DNA-binding protein_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:\nNFATC2 a member of the nuclear factor of activated T cells and the product of this gene is a DNA-binding protein (have a specific or general affinity for single- or double-stranded DNA).\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"NFATC2_sentence_1",
"NFATC2_sentence_2",
"DNA-binding protein_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 920,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.799",
"value": "?search="
},
{
"operation": "search",
"time": "9.816",
"value": "NFATC2"
},
{
"operation": "Now on",
"time": "10.499",
"value": "NFATC2"
},
{
"operation": "Highlighting",
"time": "29.5",
"value": "NFATC2_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "37.732",
"value": "NFATC2_sentence_0"
},
{
"operation": "Highlighting",
"time": "42.605",
"value": "NFATC2_sentence_1"
},
{
"operation": "Highlighting",
"time": "52.311",
"value": "NFATC2_sentence_2"
},
{
"operation": "hyperlink",
"time": "54.095",
"value": "DNA-binding protein"
},
{
"operation": "Now on",
"time": "54.866",
"value": "DNA-binding protein"
},
{
"operation": "Highlighting",
"time": "62.146",
"value": "DNA-binding protein_sentence_0"
},
{
"operation": "finish",
"time": "86.44",
"value": "finish"
}
]
}
|
exfever_2957
|
Emilio Estévez is the older brother of American actor Carlos Irwin Estévez.
|
{
"evidence": [
"Emilio Estevez",
"Charlie Sheen"
],
"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:\nEmilio Estévez is the older brother of American actor Carlos Irwin Estévez.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Emilio Estevez",
"Charlie Sheen"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 921,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4008
|
Harvey Kurtzman was an American cartoonist and editor who began to work on editing the anthology of war comics published bi-monthly by EC Comics in the early 1960s.
|
{
"evidence": [
"Harvey Kurtzman",
"Two-Fisted Tales"
],
"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:\nHarvey Kurtzman was an American cartoonist and editor who began to work on editing the anthology of war comics published bi-monthly by EC Comics in the early 1960s.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Harvey Kurtzman",
"Two-Fisted Tales"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 922,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_848
|
Bruce Willis is an American retired actor who began his career in the theatre, which is any professional theatre venue in Mexico City.
|
{
"evidence": [
"Bruce Willis",
"Off-Broadway"
],
"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:\nBruce Willis is an American retired actor who began his career in the theatre, which is any professional theatre venue in Mexico City.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Bruce Willis",
"Off-Broadway"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 923,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6740
|
Alani Nicole "La La" Anthony married NBA player Carmelo Anthony in 2010.
|
{
"evidence": [
"La La Anthony",
"Charm School with Ricki Lake",
"Stryker (disc jockey)"
],
"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:\nAlani Nicole \"La La\" Anthony married NBA player Carmelo Anthony in 2010.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"La La Anthony",
"Charm School with Ricki Lake",
"Stryker (disc jockey)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 924,
"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": 925,
"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"
}
]
}
|
9934
|
Unseen character are causal figures included in dramatic works but are seen as more powerful because they are only known by inference.
|
{
"evidence": [
"Unseen character_sentence_1",
"Unseen character_sentence_7",
"Unseen character_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:\nUnseen character are causal figures included in dramatic works but are seen as more powerful because they are only known by inference.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Unseen character_sentence_1",
"Unseen character_sentence_7",
"Unseen character_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 926,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.746",
"value": "?search="
},
{
"operation": "search",
"time": "36.022",
"value": "Unseen character"
},
{
"operation": "Now on",
"time": "37.03",
"value": "Unseen character"
},
{
"operation": "Highlighting",
"time": "48.051",
"value": "Unseen character_sentence_1"
},
{
"operation": "Highlighting",
"time": "92.778",
"value": "Unseen character_sentence_7"
},
{
"operation": "Highlighting",
"time": "93.853",
"value": "Unseen character_sentence_8"
},
{
"operation": "finish",
"time": "178.961",
"value": "finish"
}
]
}
|
exfever_90
|
"Stars Are Blind" is a song recorded by an American socialite, who began working on the album with Rob Cavallo.
|
{
"evidence": [
"Stars Are Blind",
"Paris (Paris Hilton album)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n\"Stars Are Blind\" is a song recorded by an American socialite, who began working on the album with Rob Cavallo.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Stars Are Blind",
"Paris (Paris Hilton album)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 927,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2942
|
Game of Thrones created by David Benioff is a film series that has a genre of television programming featuring elements of the fantastic, often including magic, supernatural forces, or exotic fantasy worlds.
|
{
"evidence": [
"Game of Thrones",
"Fantasy television"
],
"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:\nGame of Thrones created by David Benioff is a film series that has a genre of television programming featuring elements of the fantastic, often including magic, supernatural forces, or exotic fantasy worlds.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Game of Thrones",
"Fantasy television"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 928,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_196
|
Sarrainodu is a 2016 Indian Telugu-language film that stars an Indian actor known for his work predominantly in Telugu cinema.
|
{
"evidence": [
"Sarrainodu",
"Srikanth (Telugu actor)"
],
"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:\nSarrainodu is a 2016 Indian Telugu-language film that stars an Indian actor known for his work predominantly in Telugu cinema.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sarrainodu",
"Srikanth (Telugu actor)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 929,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8235
|
Roger Donaldson directed The Bank Job and The Bounty.
|
{
"evidence": [
"The Bounty (1984 film)",
"Laurence Olivier"
],
"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:\nRoger Donaldson directed The Bank Job and The Bounty.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Bounty (1984 film)",
"Laurence Olivier"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 930,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
94457
|
Lake Phalen is a lake located in Saint Paul, Minnesota, its primary outflow is Phalen Creek and is one of the Twin Cities most popular fishing lakes.
|
{
"evidence": [
"Lake Phalen_sentence_0",
"Lake Phalen_sentence_4",
"Lake Phalen_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:\nLake Phalen is a lake located in Saint Paul, Minnesota, its primary outflow is Phalen Creek and is one of the Twin Cities most popular fishing lakes.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Lake Phalen_sentence_0",
"Lake Phalen_sentence_4",
"Lake Phalen_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 931,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.435",
"value": "?search="
},
{
"operation": "search",
"time": "9.29",
"value": "Lake Phalen"
},
{
"operation": "Now on",
"time": "10.198",
"value": "Lake Phalen"
},
{
"operation": "Highlighting",
"time": "13.419",
"value": "Lake Phalen_sentence_0"
},
{
"operation": "Highlighting",
"time": "30.686",
"value": "Lake Phalen_sentence_4"
},
{
"operation": "Highlighting",
"time": "34.345",
"value": "Lake Phalen_sentence_2"
},
{
"operation": "finish",
"time": "51.484",
"value": "finish"
}
]
}
|
exfever_1293
|
Jurassic World is a 2015 American science fiction action film starring an American singer who also starred in a 2012 romantic comedy film written, directed, and produced by Nicholas Stoller.
|
{
"evidence": [
"Jurassic World",
"Chris Pratt",
"The Five-Year Engagement"
],
"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:\nJurassic World is a 2015 American science fiction action film starring an American singer who also starred in a 2012 romantic comedy film written, directed, and produced by Nicholas Stoller.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jurassic World",
"Chris Pratt",
"The Five-Year Engagement"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 932,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6781
|
Touch My Body is a song by Mariah Carey, which was co-written by an American songwriter whose other co-writing credits include a song by American recording artist Beyonce that features American rapper Ludacris.
|
{
"evidence": [
"Touch My Body",
"The-Dream",
"Ride (Ciara song)"
],
"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:\nTouch My Body is a song by Mariah Carey, which was co-written by an American songwriter whose other co-writing credits include a song by American recording artist Beyonce that features American rapper Ludacris.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Touch My Body",
"The-Dream",
"Ride (Ciara song)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 933,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
22857
|
Ursula Holliger was an important harpist who taught at Hochschule für Musik Freiburg.
|
{
"evidence": [
"Ursula Holliger_sentence_6",
"Ursula Holliger_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:\nUrsula Holliger was an important harpist who taught at Hochschule für Musik Freiburg.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ursula Holliger_sentence_6",
"Ursula Holliger_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 934,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.118",
"value": "?search="
},
{
"operation": "search",
"time": "4.191",
"value": "Ursula Holliger"
},
{
"operation": "Now on",
"time": "5.104",
"value": "Ursula Holliger"
},
{
"operation": "Highlighting",
"time": "26.502",
"value": "Ursula Holliger_sentence_0"
},
{
"operation": "Highlighting",
"time": "34.994",
"value": "Ursula Holliger_sentence_6"
},
{
"operation": "Highlighting",
"time": "44.794",
"value": "Ursula Holliger_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "45.704",
"value": "Ursula Holliger_sentence_0"
},
{
"operation": "finish",
"time": "53.697",
"value": "finish"
}
]
}
|
exfever_7874
|
Denmark belongs to the Boreal Kingdom.
|
{
"evidence": [
"Denmark",
"Europe",
"Eurasia"
],
"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:\nDenmark belongs to the Boreal Kingdom.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Denmark",
"Europe",
"Eurasia"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 935,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5189
|
Dennis Quaid received a nomination for Screen Actors Guild for the film Traffic.
|
{
"evidence": [
"Dennis Quaid",
"The Parent Trap (1998 film)",
"Lindsay Lohan"
],
"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:\nDennis Quaid received a nomination for Screen Actors Guild for the film Traffic.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dennis Quaid",
"The Parent Trap (1998 film)",
"Lindsay Lohan"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 936,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6629
|
The Contender is a 2000 American comedy-drama film starring an American actress who won the 1988 Tony Award for Best Actress in a Play for her theater debut, a theatrical performance located in the Theater District in Midtown Manhattan, New York City.
|
{
"evidence": [
"The Contender (2000 film)",
"Joan Allen",
"Broadway theatre"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Contender is a 2000 American comedy-drama film starring an American actress who won the 1988 Tony Award for Best Actress in a Play for her theater debut, a theatrical performance located in the Theater District in Midtown Manhattan, New York City.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Contender (2000 film)",
"Joan Allen",
"Broadway theatre"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 937,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
77889
|
Hessie Donahue declined to spar with Sullivan Stapleton in a vaudeville act until the curtain came down.
|
{
"evidence": [
"Hessie Donahue_sentence_3",
"Hessie Donahue_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHessie Donahue declined to spar with Sullivan Stapleton in a vaudeville act until the curtain came down.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Hessie Donahue_sentence_3",
"Hessie Donahue_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 938,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.806",
"value": "undefined"
},
{
"operation": "search",
"time": "11.395",
"value": "Hessie Donahue"
},
{
"operation": "Now on",
"time": "12.215",
"value": "Hessie Donahue"
},
{
"operation": "Page search",
"time": "23.16",
"value": "sullivan"
},
{
"operation": "Page search",
"time": "31.551",
"value": "sullivan staple"
},
{
"operation": "Highlighting",
"time": "45.328",
"value": "Hessie Donahue_sentence_3"
},
{
"operation": "Highlighting",
"time": "55.329",
"value": "Hessie Donahue_sentence_0"
},
{
"operation": "finish",
"time": "182.276",
"value": "finish"
}
]
}
|
exfever_3430
|
iZombie is an Asian supernatural procedural crime drama television series that is a loose adaptation of a comic book series published by DC Comics' Vertigo imprint beginning in 2010, created by an American science fiction author and publisher based in Portland, Oregon, best known for alternate history novels.
|
{
"evidence": [
"IZombie (TV series)",
"IZOMBIE",
"Chris Roberson (author)"
],
"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:\niZombie is an Asian supernatural procedural crime drama television series that is a loose adaptation of a comic book series published by DC Comics' Vertigo imprint beginning in 2010, created by an American science fiction author and publisher based in Portland, Oregon, best known for alternate history novels.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"IZombie (TV series)",
"IZOMBIE",
"Chris Roberson (author)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 939,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2890
|
Metallica is an American heavy metal band consisting of Hetfield, Ulrich, Robert Trujillo, and the American musician and lead guitarist who formed the American thrash metal band in Alameda, California.
|
{
"evidence": [
"Metallica",
"Kirk Hammett",
"Exodus (American band)"
],
"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:\nMetallica is an American heavy metal band consisting of Hetfield, Ulrich, Robert Trujillo, and the American musician and lead guitarist who formed the American thrash metal band in Alameda, California.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Metallica",
"Kirk Hammett",
"Exodus (American band)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 940,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
16449
|
Copepteryx is an extant genus of the family Plotopteridae within the kingdom Animalia.
|
{
"evidence": [
"Copepteryx_sentence_0",
"Extant_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:\nCopepteryx is an extant genus of the family Plotopteridae within the kingdom Animalia.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Copepteryx_sentence_0",
"Extant_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 941,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.047",
"value": "?search="
},
{
"operation": "search",
"time": "10.323",
"value": "Copepteryx"
},
{
"operation": "Now on",
"time": "11.222",
"value": "Copepteryx"
},
{
"operation": "Highlighting",
"time": "17.412",
"value": "Copepteryx_sentence_0"
},
{
"operation": "search",
"time": "36.919",
"value": "extant"
},
{
"operation": "Now on",
"time": "37.707",
"value": "Extant"
},
{
"operation": "Highlighting",
"time": "39.748",
"value": "Extant_sentence_0"
},
{
"operation": "finish",
"time": "52.718",
"value": "finish"
}
]
}
|
exfever_4364
|
Wi-Fi is a family of wireless network protocols commonly used for local area networking of devices and a global system of interconnected mobile networks that uses the Internet protocol suite to communicate between networks and devices.
|
{
"evidence": [
"Wi-Fi",
"Internet"
],
"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:\nWi-Fi is a family of wireless network protocols commonly used for local area networking of devices and a global system of interconnected mobile networks that uses the Internet protocol suite to communicate between networks and devices.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Wi-Fi",
"Internet"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 942,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7597
|
Scooby-Doo is an American animated franchise series created by Joe Ruby and Ken Spears, who also created the original series, which was produced by Hanna-Barbera for CBS in 1959.
|
{
"evidence": [
"Scooby-Doo",
"Scooby-Doo, Where Are You!"
],
"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:\nScooby-Doo is an American animated franchise series created by Joe Ruby and Ken Spears, who also created the original series, which was produced by Hanna-Barbera for CBS in 1959.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Scooby-Doo",
"Scooby-Doo, Where Are You!"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 943,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4755
|
Andrew Ayers Stanton's film work includes directing a 2016 computer-animated adventure film released by an American film production studio and subsidiary of Walt Disney Studios, which is owned by The Warner Bros. Company.
|
{
"evidence": [
"Andrew Stanton",
"Finding Dory",
"Walt Disney Pictures"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAndrew Ayers Stanton's film work includes directing a 2016 computer-animated adventure film released by an American film production studio and subsidiary of Walt Disney Studios, which is owned by The Warner Bros. Company.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Andrew Stanton",
"Finding Dory",
"Walt Disney Pictures"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 944,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
22934
|
To make traditional ginger ale, yeast which contains zymase is added with sugar.
|
{
"evidence": [
"Ginger ale_sentence_28",
"Ginger ale_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:\nTo make traditional ginger ale, yeast which contains zymase is added with sugar.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ginger ale_sentence_28",
"Ginger ale_sentence_25"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 945,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.709",
"value": "?search="
},
{
"operation": "search",
"time": "30.086",
"value": "traditional ginger ale"
},
{
"operation": "Now on",
"time": "32.871",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "37.249",
"value": "Ginger ale"
},
{
"operation": "Now on",
"time": "38.618",
"value": "Ginger ale"
},
{
"operation": "Highlighting",
"time": "54.845",
"value": "Ginger ale_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "56.192",
"value": "Ginger ale_sentence_2"
},
{
"operation": "Page search",
"time": "58.629",
"value": "traditional"
},
{
"operation": "Highlighting",
"time": "61.776",
"value": "Ginger ale_sentence_21"
},
{
"operation": "Highlighting",
"time": "61.777",
"value": "Ginger ale_section_0"
},
{
"operation": "Highlighting",
"time": "61.778",
"value": "Ginger ale_section_1"
},
{
"operation": "Page search",
"time": "84.444",
"value": "zymase"
},
{
"operation": "Highlighting",
"time": "89.812",
"value": "Ginger ale_sentence_28"
},
{
"operation": "Highlighting",
"time": "134.678",
"value": "Ginger ale_sentence_25"
},
{
"operation": "Highlighting deleted",
"time": "140.63",
"value": "Ginger ale_sentence_21"
},
{
"operation": "Highlighting deleted",
"time": "140.633",
"value": "Ginger ale_section_0"
},
{
"operation": "Highlighting deleted",
"time": "140.635",
"value": "Ginger ale_section_1"
},
{
"operation": "finish",
"time": "145.589",
"value": "finish"
}
]
}
|
exfever_5036
|
Ileana D'Cruz is an Indian-born Portuguese actress made her screen debut in 2007 with Telugu-language film which is a romantic action film directed and produced by Y. V. S. Chowdary.
|
{
"evidence": [
"Ileana D'Cruz",
"Devadasu (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:\nIleana D'Cruz is an Indian-born Portuguese actress made her screen debut in 2007 with Telugu-language film which is a romantic action film directed and produced by Y. V. S. Chowdary.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ileana D'Cruz",
"Devadasu (2006 film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 946,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4115
|
Denis Leary began taking roles in film and television starting in the early 1990s, including a substantial role in a 1994 American black comedy film directed by an American director.
|
{
"evidence": [
"Denis Leary",
"The Ref",
"Ted Demme"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDenis Leary began taking roles in film and television starting in the early 1990s, including a substantial role in a 1994 American black comedy film directed by an American director.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Denis Leary",
"The Ref",
"Ted Demme"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 947,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
65410
|
Wilkinson is a moth in the family Drepanidae which is found in New Ireland, New Britain, Rook Island and New Hanover.
|
{
"evidence": [
"Teldenia apata_sentence_0",
"Teldenia apata_sentence_2",
"Teldenia apata_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:\nWilkinson is a moth in the family Drepanidae which is found in New Ireland, New Britain, Rook Island and New Hanover.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Teldenia apata_sentence_0",
"Teldenia apata_sentence_2",
"Teldenia apata_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 948,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.125",
"value": "?search="
},
{
"operation": "search",
"time": "9.502",
"value": "Wilkinson"
},
{
"operation": "Now on",
"time": "13.462",
"value": "Wilkinson"
},
{
"operation": "search",
"time": "37.186",
"value": "Contains...Wilkinson is a moth"
},
{
"operation": "Now on",
"time": "40.122",
"value": "undefined"
},
{
"operation": "search",
"time": "64.476",
"value": "found in New Ireland, New Britain, Rook Island and New Hanover"
},
{
"operation": "Now on",
"time": "65.938",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "70.531",
"value": "Teldenia apata"
},
{
"operation": "Now on",
"time": "71.209",
"value": "Teldenia apata"
},
{
"operation": "Highlighting",
"time": "74.898",
"value": "Teldenia apata_sentence_0"
},
{
"operation": "Highlighting",
"time": "77.274",
"value": "Teldenia apata_sentence_2"
},
{
"operation": "Highlighting",
"time": "86.42",
"value": "Teldenia apata_sentence_1"
},
{
"operation": "finish",
"time": "95.181",
"value": "finish"
}
]
}
|
exfever_193
|
Alessia Cara, also known as Alessia Caracciolo, is a French singer-songwriter who signed with an American record label founded by Robert Eleazer and Tony Perez.
|
{
"evidence": [
"Alessia Cara",
"EP Entertainment"
],
"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:\nAlessia Cara, also known as Alessia Caracciolo, is a French singer-songwriter who signed with an American record label founded by Robert Eleazer and Tony Perez.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Alessia Cara",
"EP Entertainment"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 949,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
50519
|
The brickworks, along with the local pottery, was served by a branch of the South Yorkshire Railway (absorbed neither by the Manchester, Sheffield nor the Lincolnshire Railway in 1864) from 1850; wherein the railway junction from the main line was known as Thrybergh Colliery Junction.
|
{
"evidence": [
"South Yorkshire Railway_sentence_6"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe brickworks, along with the local pottery, was served by a branch of the South Yorkshire Railway (absorbed neither by the Manchester, Sheffield nor the Lincolnshire Railway in 1864) from 1850; wherein the railway junction from the main line was known as Thrybergh Colliery Junction.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"South Yorkshire Railway_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 950,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "25.003",
"value": "undefined"
},
{
"operation": "search",
"time": "31.388",
"value": "South Yorkshire Railway"
},
{
"operation": "Now on",
"time": "37.62",
"value": "South Yorkshire Railway"
},
{
"operation": "Highlighting",
"time": "44.681",
"value": "South Yorkshire Railway_sentence_6"
},
{
"operation": "finish",
"time": "82.439",
"value": "finish"
}
]
}
|
85881
|
Moonia belongs to the tribe Coreopsideae under the family Asteraceae which includes over 40,000 currently accepted species, in over 1,900 genera (list) in 13 subfamilies.
|
{
"evidence": [
"Asteraceae_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:\nMoonia belongs to the tribe Coreopsideae under the family Asteraceae which includes over 40,000 currently accepted species, in over 1,900 genera (list) in 13 subfamilies.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Asteraceae_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 951,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.245",
"value": "?search="
},
{
"operation": "search",
"time": "8.498",
"value": "Moonia"
},
{
"operation": "Now on",
"time": "9.401",
"value": "Moonia"
},
{
"operation": "Highlighting",
"time": "13.175",
"value": "Moonia_cell_0_8_0"
},
{
"operation": "Highlighting",
"time": "13.176",
"value": "Moonia_title"
},
{
"operation": "Highlighting",
"time": "13.179",
"value": "Moonia_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "13.18",
"value": "Moonia_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "14.781",
"value": "Moonia_cell_0_8_1"
},
{
"operation": "Highlighting",
"time": "16.271",
"value": "Moonia_cell_0_7_0"
},
{
"operation": "Highlighting",
"time": "17.842",
"value": "Moonia_cell_0_7_1"
},
{
"operation": "hyperlink",
"time": "19.139",
"value": "Asteraceae"
},
{
"operation": "Now on",
"time": "20.106",
"value": "Asteraceae"
},
{
"operation": "Highlighting",
"time": "22.479",
"value": "Asteraceae_sentence_1"
},
{
"operation": "finish",
"time": "36.2",
"value": "finish"
}
]
}
|
85196
|
The San Lorenzo Creek runs through the town of San Lorenzo and empties into San Francisco Bay, and a portion of the San Francisco Bay Trail (a bicycle and pedestrian trail which is a project of the Association of Bay Area Governments and the Metropolitan Transportation Commission) runs along with the mouth of the creek.
|
{
"evidence": [
"San Lorenzo Creek_sentence_3",
"San Lorenzo Creek_sentence_4",
"San Francisco Bay Trail_sentence_3",
"San Francisco Bay Trail_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 San Lorenzo Creek runs through the town of San Lorenzo and empties into San Francisco Bay, and a portion of the San Francisco Bay Trail (a bicycle and pedestrian trail which is a project of the Association of Bay Area Governments and the Metropolitan Transportation Commission) runs along with the mouth of the creek.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"San Lorenzo Creek_sentence_3",
"San Lorenzo Creek_sentence_4",
"San Francisco Bay Trail_sentence_3",
"San Francisco Bay Trail_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 952,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.084",
"value": "?search="
},
{
"operation": "search",
"time": "16.866",
"value": "San Lorenzo Creek"
},
{
"operation": "Now on",
"time": "18.957",
"value": "San Lorenzo Creek"
},
{
"operation": "Highlighting",
"time": "34.52",
"value": "San Lorenzo Creek_sentence_3"
},
{
"operation": "Highlighting",
"time": "39.531",
"value": "San Lorenzo Creek_sentence_4"
},
{
"operation": "hyperlink",
"time": "54.367",
"value": "San Francisco Bay Trail"
},
{
"operation": "Now on",
"time": "55.547",
"value": "San Francisco Bay Trail"
},
{
"operation": "Highlighting",
"time": "62.02",
"value": "San Francisco Bay Trail_sentence_0"
},
{
"operation": "Highlighting",
"time": "71.564",
"value": "San Francisco Bay Trail_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "92.352",
"value": "San Francisco Bay Trail_sentence_0"
},
{
"operation": "Highlighting",
"time": "92.995",
"value": "San Francisco Bay Trail_sentence_0"
},
{
"operation": "Now on",
"time": "102.117",
"value": "San Francisco Bay Trail"
},
{
"operation": "finish",
"time": "114.905",
"value": "finish"
}
]
}
|
33298
|
Dobson was born in the small fishing village of Keiss in Caithness, then in the 1890s, Dobson had moved to the Cardiff Docks area (a port in southern Cardiff) where he had become a coal trimmer.
|
{
"evidence": [
"Cardiff Docks_sentence_0",
"Tom Dobson_sentence_5",
"Tom Dobson_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:\nDobson was born in the small fishing village of Keiss in Caithness, then in the 1890s, Dobson had moved to the Cardiff Docks area (a port in southern Cardiff) where he had become a coal trimmer.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Cardiff Docks_sentence_0",
"Tom Dobson_sentence_5",
"Tom Dobson_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 953,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.011",
"value": "?search="
},
{
"operation": "search",
"time": "17.502",
"value": "Cardiff Docks"
},
{
"operation": "Now on",
"time": "18.218",
"value": "Cardiff Docks"
},
{
"operation": "Highlighting",
"time": "20.024",
"value": "Cardiff Docks_sentence_0"
},
{
"operation": "search",
"time": "35.399",
"value": "\"small fishing village of Keiss in Caithness\""
},
{
"operation": "Now on",
"time": "36.574",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "38.221",
"value": "Tom Dobson"
},
{
"operation": "Now on",
"time": "38.667",
"value": "Tom Dobson"
},
{
"operation": "Highlighting",
"time": "40.985",
"value": "Tom Dobson_sentence_5"
},
{
"operation": "Highlighting",
"time": "46.704",
"value": "Tom Dobson_sentence_6"
},
{
"operation": "finish",
"time": "66.536",
"value": "finish"
}
]
}
|
61494
|
Bimbisāra (558 BC-558 BC) was one of the early kings of the Indian kingdom of Magadha who ascended to throne at the age of 15 in 543 BC and established the Anga dynasty.
|
{
"evidence": [
"Bimbisara_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:\nBimbisāra (558 BC-558 BC) was one of the early kings of the Indian kingdom of Magadha who ascended to throne at the age of 15 in 543 BC and established the Anga dynasty.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Bimbisara_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 954,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.16",
"value": "?search="
},
{
"operation": "search",
"time": "7.533",
"value": "Bimbisara"
},
{
"operation": "Now on",
"time": "8.815",
"value": "Bimbisara"
},
{
"operation": "Highlighting",
"time": "22.476",
"value": "Bimbisara_cell_0_2_1"
},
{
"operation": "Highlighting",
"time": "22.482",
"value": "Bimbisara_header_cell_0_2_0"
},
{
"operation": "Highlighting",
"time": "22.482",
"value": "Bimbisara_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "22.482",
"value": "Bimbisara_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "23.811",
"value": "Bimbisara_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "41.359",
"value": "Bimbisara_cell_0_2_1"
},
{
"operation": "Highlighting deleted",
"time": "41.366",
"value": "Bimbisara_header_cell_0_2_0"
},
{
"operation": "Highlighting deleted",
"time": "41.367",
"value": "Bimbisara_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "41.368",
"value": "Bimbisara_header_cell_0_0_0"
},
{
"operation": "finish",
"time": "48.214",
"value": "finish"
}
]
}
|
69388
|
Monte Carlo Casino officially named Casino de Monte-Carlo includes a casino (built in the early 1980s) and owns hotels.
|
{
"evidence": [
"Monte Carlo Casino_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:\nMonte Carlo Casino officially named Casino de Monte-Carlo includes a casino (built in the early 1980s) and owns hotels.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Monte Carlo Casino_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 955,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.211",
"value": "?search="
},
{
"operation": "search",
"time": "9.63",
"value": "Monte Carlo Casino"
},
{
"operation": "Now on",
"time": "10.5",
"value": "Monte Carlo Casino"
},
{
"operation": "Highlighting",
"time": "14.013",
"value": "Monte Carlo Casino_sentence_2"
},
{
"operation": "Highlighting",
"time": "18.739",
"value": "Monte Carlo Casino_sentence_3"
},
{
"operation": "hyperlink",
"time": "44.41",
"value": "casino"
},
{
"operation": "Now on",
"time": "45.141",
"value": "Casino"
},
{
"operation": "back-button-clicked",
"time": "49.511",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "49.61",
"value": "Monte Carlo Casino"
},
{
"operation": "Highlighting deleted",
"time": "61.052",
"value": "Monte Carlo Casino_sentence_2"
},
{
"operation": "finish",
"time": "68.602",
"value": "finish"
}
]
}
|
76322
|
On returning home and ascending the throne on the death of his father, Izates, son of Queen Helena of Adiabene, discovered the conversion of his mother who had been previously won over to Judaism without his knowledge.
|
{
"evidence": [
"Izates II_sentence_2",
"Izates II_sentence_7",
"Izates II_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:\nOn returning home and ascending the throne on the death of his father, Izates, son of Queen Helena of Adiabene, discovered the conversion of his mother who had been previously won over to Judaism without his knowledge.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Izates II_sentence_2",
"Izates II_sentence_7",
"Izates II_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 956,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.629",
"value": "?search="
},
{
"operation": "search",
"time": "49.716",
"value": "izates"
},
{
"operation": "Now on",
"time": "50.533",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "68.395",
"value": "Ananias of Adiabene"
},
{
"operation": "Now on",
"time": "69.082",
"value": "Ananias of Adiabene"
},
{
"operation": "Page search",
"time": "78.506",
"value": "helena"
},
{
"operation": "back-button-clicked",
"time": "204.757",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "205.45",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "209.433",
"value": "Izates II"
},
{
"operation": "Now on",
"time": "210.344",
"value": "Izates II"
},
{
"operation": "Highlighting",
"time": "220.445",
"value": "Izates II_sentence_2"
},
{
"operation": "Page search",
"time": "232.13",
"value": "convert"
},
{
"operation": "Highlighting",
"time": "272.637",
"value": "Izates II_sentence_8"
},
{
"operation": "page-search-reset",
"time": "314.866",
"value": "page-search-reset"
},
{
"operation": "Highlighting deleted",
"time": "333.931",
"value": "Izates II_sentence_8"
},
{
"operation": "Highlighting",
"time": "335.413",
"value": "Izates II_sentence_8"
},
{
"operation": "Highlighting",
"time": "336.361",
"value": "Izates II_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "351.818",
"value": "Izates II_sentence_8"
},
{
"operation": "Highlighting",
"time": "352.397",
"value": "Izates II_sentence_8"
},
{
"operation": "finish",
"time": "373.016",
"value": "finish"
}
]
}
|
18200
|
Hesse worked in the spirit of Karl August Möbius with biogeography and ecology of vertebrates and in 1945, he published his Tiergeographie auf Ökologischer Grundlage.
|
{
"evidence": [
"Richard Hesse_sentence_6"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHesse worked in the spirit of Karl August Möbius with biogeography and ecology of vertebrates and in 1945, he published his Tiergeographie auf Ökologischer Grundlage.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Richard Hesse_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 957,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.26",
"value": "?search="
},
{
"operation": "search",
"time": "13.829",
"value": "hesse"
},
{
"operation": "Now on",
"time": "17.43",
"value": "Hesse"
},
{
"operation": "Page search",
"time": "36.715",
"value": "spirit"
},
{
"operation": "Page search",
"time": "644.461",
"value": "biology"
},
{
"operation": "Page search",
"time": "658.09",
"value": "biogeography"
},
{
"operation": "hyperlink",
"time": "683.125",
"value": "Hesse (disambiguation)"
},
{
"operation": "Now on",
"time": "685.537",
"value": "Hesse (disambiguation)"
},
{
"operation": "search",
"time": "701.214",
"value": "Tiergeographie auf Ökologischer Grundlage"
},
{
"operation": "Now on",
"time": "703.291",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "707.072",
"value": "Richard Hesse"
},
{
"operation": "Now on",
"time": "708.16",
"value": "Richard Hesse"
},
{
"operation": "Page search",
"time": "758.115",
"value": "karl"
},
{
"operation": "Highlighting",
"time": "761.359",
"value": "Richard Hesse_sentence_4"
},
{
"operation": "Highlighting",
"time": "808.036",
"value": "Richard Hesse_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "809.844",
"value": "Richard Hesse_sentence_4"
},
{
"operation": "finish",
"time": "819.799",
"value": "finish"
}
]
}
|
27619
|
The majority of Maryland residents did not approve of LGBT anti-discrimination laws as indicated in opinion polls.
|
{
"evidence": [
"Maryland_sentence_377"
],
"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 majority of Maryland residents did not approve of LGBT anti-discrimination laws as indicated in opinion polls.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Maryland_sentence_377"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 958,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.26",
"value": "?search="
},
{
"operation": "search",
"time": "8.321",
"value": "Maryland"
},
{
"operation": "Now on",
"time": "14.062",
"value": "Maryland"
},
{
"operation": "Page search",
"time": "36.623",
"value": "lgbt"
},
{
"operation": "Highlighting",
"time": "120.197",
"value": "Maryland_sentence_375"
},
{
"operation": "Highlighting",
"time": "190.841",
"value": "Maryland_sentence_370"
},
{
"operation": "Highlighting deleted",
"time": "200.827",
"value": "Maryland_sentence_375"
},
{
"operation": "Highlighting",
"time": "201.747",
"value": "Maryland_sentence_373"
},
{
"operation": "Highlighting deleted",
"time": "202.816",
"value": "Maryland_sentence_373"
},
{
"operation": "Highlighting",
"time": "205.167",
"value": "Maryland_sentence_372"
},
{
"operation": "Highlighting",
"time": "209.006",
"value": "Maryland_sentence_373"
},
{
"operation": "Highlighting",
"time": "210.496",
"value": "Maryland_sentence_375"
},
{
"operation": "Highlighting deleted",
"time": "230.623",
"value": "Maryland_sentence_375"
},
{
"operation": "Highlighting deleted",
"time": "232.576",
"value": "Maryland_sentence_373"
},
{
"operation": "Highlighting deleted",
"time": "233.427",
"value": "Maryland_sentence_372"
},
{
"operation": "Highlighting deleted",
"time": "234.568",
"value": "Maryland_sentence_370"
},
{
"operation": "Highlighting",
"time": "236.03",
"value": "Maryland_sentence_377"
},
{
"operation": "finish",
"time": "245.993",
"value": "finish"
}
]
}
|
exfever_948
|
Kingdom Hearts III is a 2019 action role-playing game developed and published by Square Enix for a home video game console developed by Nintendo.
|
{
"evidence": [
"Kingdom Hearts III",
"Xbox One"
],
"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:\nKingdom Hearts III is a 2019 action role-playing game developed and published by Square Enix for a home video game console developed by Nintendo.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Kingdom Hearts III",
"Xbox One"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 959,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_201
|
The Challenge is a reality competition show on an American cable channel that launched on August 1, 1981.
|
{
"evidence": [
"The Challenge (TV series)",
"MTV"
],
"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 Challenge is a reality competition show on an American cable channel that launched on August 1, 1981.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Challenge (TV series)",
"MTV"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 960,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
88736
|
Joseph Boatner led various groups using the "Ink Spots" name until his death in Laconia, New Hampshire; home of the Laconia World Championship Sled Dog Derby.
|
{
"evidence": [
"Joseph Boatner_sentence_2",
"Joseph Boatner_sentence_3",
"Laconia, New Hampshire_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:\nJoseph Boatner led various groups using the \"Ink Spots\" name until his death in Laconia, New Hampshire; home of the Laconia World Championship Sled Dog Derby.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Joseph Boatner_sentence_2",
"Joseph Boatner_sentence_3",
"Laconia, New Hampshire_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 961,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.548",
"value": "?search="
},
{
"operation": "search",
"time": "16.555",
"value": "Joseph Boatner"
},
{
"operation": "Now on",
"time": "17.423",
"value": "Joseph Boatner"
},
{
"operation": "Highlighting",
"time": "260.584",
"value": "Joseph Boatner_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "271.699",
"value": "Joseph Boatner_sentence_0"
},
{
"operation": "Highlighting",
"time": "284.726",
"value": "Joseph Boatner_sentence_0"
},
{
"operation": "Highlighting",
"time": "287.089",
"value": "Joseph Boatner_sentence_1"
},
{
"operation": "Highlighting",
"time": "292.05",
"value": "Joseph Boatner_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "293.378",
"value": "Joseph Boatner_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "297.373",
"value": "Joseph Boatner_sentence_0"
},
{
"operation": "Highlighting",
"time": "300.048",
"value": "Joseph Boatner_sentence_3"
},
{
"operation": "hyperlink",
"time": "303.783",
"value": "Laconia, New Hampshire"
},
{
"operation": "Now on",
"time": "304.921",
"value": "Laconia, New Hampshire"
},
{
"operation": "Page search",
"time": "342.464",
"value": "Sled Dog Derby"
},
{
"operation": "Highlighting",
"time": "346.095",
"value": "Laconia, New Hampshire_sentence_4"
},
{
"operation": "back-button-clicked",
"time": "348.318",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "348.376",
"value": "Joseph Boatner"
},
{
"operation": "finish",
"time": "405.707",
"value": "finish"
}
]
}
|
exfever_5782
|
Christopher Frank Carandini Lee was an English actor with a film role as a fictional character in the Star Wars franchise that was created by George Lucas.
|
{
"evidence": [
"Christopher Lee",
"Count Dooku"
],
"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:\nChristopher Frank Carandini Lee was an English actor with a film role as a fictional character in the Star Wars franchise that was created by George Lucas.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Christopher Lee",
"Count Dooku"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 962,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
18307
|
Fantasia 2000 is a 1999 American animated film produced by Roy E. Disney, a longtime senior executive for the Walt Disney Company which was founded by his father.
|
{
"evidence": [
"Fantasia 2000_sentence_0",
"Fantasia 2000_sentence_1",
"Roy E. Disney_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:\nFantasia 2000 is a 1999 American animated film produced by Roy E. Disney, a longtime senior executive for the Walt Disney Company which was founded by his father.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Fantasia 2000_sentence_0",
"Fantasia 2000_sentence_1",
"Roy E. Disney_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 963,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.503",
"value": "undefined"
},
{
"operation": "search",
"time": "8.363",
"value": "Fantasia 2000"
},
{
"operation": "Now on",
"time": "9.414",
"value": "Fantasia 2000"
},
{
"operation": "Highlighting",
"time": "10.527",
"value": "Fantasia 2000_sentence_0"
},
{
"operation": "Highlighting",
"time": "19.234",
"value": "Fantasia 2000_sentence_1"
},
{
"operation": "hyperlink",
"time": "25.025",
"value": "Roy E. Disney"
},
{
"operation": "Now on",
"time": "25.591",
"value": "Roy E. Disney"
},
{
"operation": "Highlighting",
"time": "34.556",
"value": "Roy E. Disney_sentence_2"
},
{
"operation": "finish",
"time": "46.245",
"value": "finish"
}
]
}
|
3163
|
Samantha Buck and Marie Schlingmann worked on nine projects together over eleven years.
|
{
"evidence": [
"Samantha Buck and Marie Schlingmann_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSamantha Buck and Marie Schlingmann worked on nine projects together over eleven years.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Samantha Buck and Marie Schlingmann_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 964,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "11.981",
"value": "Samantha Buck"
},
{
"operation": "Highlighting",
"time": "93.692",
"value": "Samantha Buck and Marie Schlingmann_sentence_0"
},
{
"operation": "finish",
"time": "128.626",
"value": "finish"
}
]
}
|
82524
|
Commander Jaan Maide, chief of the Estonian Defence Forces, took office on September 18, 1944 and left office on October 24, 1944, with a total of 36 days in office in the Army Defence branch.
|
{
"evidence": [
"Jaan Maide_sentence_1",
"Jaan Maide_sentence_15",
"Jaan Maide_sentence_17",
"Jaan Maide_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:\nCommander Jaan Maide, chief of the Estonian Defence Forces, took office on September 18, 1944 and left office on October 24, 1944, with a total of 36 days in office in the Army Defence branch.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Jaan Maide_sentence_1",
"Jaan Maide_sentence_15",
"Jaan Maide_sentence_17",
"Jaan Maide_sentence_19"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 965,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.086",
"value": "?search="
},
{
"operation": "search",
"time": "29.044",
"value": "Jaan Maide"
},
{
"operation": "Now on",
"time": "29.684",
"value": "Jaan Maide"
},
{
"operation": "Highlighting",
"time": "36.611",
"value": "Jaan Maide_sentence_1"
},
{
"operation": "Page search",
"time": "64.866",
"value": "estonian defen"
},
{
"operation": "Highlighting",
"time": "81.48",
"value": "Jaan Maide_sentence_10"
},
{
"operation": "Highlighting",
"time": "113.024",
"value": "Jaan Maide_sentence_15"
},
{
"operation": "Highlighting deleted",
"time": "136.069",
"value": "Jaan Maide_sentence_10"
},
{
"operation": "Highlighting",
"time": "150.695",
"value": "Jaan Maide_sentence_17"
},
{
"operation": "Page search",
"time": "238.146",
"value": "october"
},
{
"operation": "Highlighting",
"time": "259.131",
"value": "Jaan Maide_sentence_19"
},
{
"operation": "Page search",
"time": "285.355",
"value": "days"
},
{
"operation": "finish",
"time": "359.852",
"value": "finish"
}
]
}
|
78849
|
Isyana Sarasvati (born Bandung, May 2, 1993) is an alumna of Royal College of Music, United Kingdom, and once was an opera singer in Singapore.
|
{
"evidence": [
"Isyana Sarasvati_sentence_0",
"Isyana Sarasvati_sentence_1",
"Isyana Sarasvati_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:\nIsyana Sarasvati (born Bandung, May 2, 1993) is an alumna of Royal College of Music, United Kingdom, and once was an opera singer in Singapore.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Isyana Sarasvati_sentence_0",
"Isyana Sarasvati_sentence_1",
"Isyana Sarasvati_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 966,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.103",
"value": "?search="
},
{
"operation": "search",
"time": "11.992",
"value": "Isyana Sarasvati"
},
{
"operation": "Now on",
"time": "13.562",
"value": "Isyana Sarasvati"
},
{
"operation": "Highlighting",
"time": "20.637",
"value": "Isyana Sarasvati_sentence_0"
},
{
"operation": "Highlighting",
"time": "24.984",
"value": "Isyana Sarasvati_sentence_1"
},
{
"operation": "Highlighting",
"time": "30.692",
"value": "Isyana Sarasvati_sentence_2"
},
{
"operation": "finish",
"time": "41.599",
"value": "finish"
}
]
}
|
37555
|
In 1992, Gaviria joined Colgate-Palmolive as the director of multicultural markets, wherein Advertising Age recognized her work on its "The Marketing 100" list.
|
{
"evidence": [
"Ruth Gaviria_sentence_4",
"Ruth Gaviria_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:\nIn 1992, Gaviria joined Colgate-Palmolive as the director of multicultural markets, wherein Advertising Age recognized her work on its \"The Marketing 100\" list.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Ruth Gaviria_sentence_4",
"Ruth Gaviria_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 967,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.667",
"value": "?search="
},
{
"operation": "search",
"time": "16.965",
"value": "Colgate-Palmolive"
},
{
"operation": "Now on",
"time": "17.905",
"value": "Colgate-Palmolive"
},
{
"operation": "search",
"time": "34.653",
"value": "Ruth Gaviria"
},
{
"operation": "Now on",
"time": "35.519",
"value": "Ruth Gaviria"
},
{
"operation": "Page search",
"time": "47.466",
"value": "lgate-Palmolive"
},
{
"operation": "Highlighting",
"time": "50.405",
"value": "Ruth Gaviria_sentence_4"
},
{
"operation": "Highlighting",
"time": "68.276",
"value": "Ruth Gaviria_sentence_5"
},
{
"operation": "finish",
"time": "189.222",
"value": "finish"
}
]
}
|
86137
|
Samuel Denoff, born on July 1, 1928 in Brooklyn, New York, U.S., was a screenwriter, television producer and winner of the 1969 Outstanding Variety or Musical Program Emmy Award for the Bill Cosby Special.
|
{
"evidence": [
"Sam Denoff_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSamuel Denoff, born on July 1, 1928 in Brooklyn, New York, U.S., was a screenwriter, television producer and winner of the 1969 Outstanding Variety or Musical Program Emmy Award for the Bill Cosby Special.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sam Denoff_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 968,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.722",
"value": "?search="
},
{
"operation": "search",
"time": "23.044",
"value": "Contains...Samuel Denoff"
},
{
"operation": "Now on",
"time": "23.811",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "27.363",
"value": "Sam Denoff"
},
{
"operation": "Now on",
"time": "27.865",
"value": "Sam Denoff"
},
{
"operation": "Highlighting",
"time": "30.334",
"value": "Sam Denoff_sentence_0"
},
{
"operation": "search",
"time": "104.769",
"value": "1969 Emmys"
},
{
"operation": "Now on",
"time": "105.628",
"value": "1969 Emmys"
},
{
"operation": "Page search",
"time": "117.967",
"value": "Denoff"
},
{
"operation": "search",
"time": "161.132",
"value": "Contains...1969 Outstanding Variety or Musical Program"
},
{
"operation": "Now on",
"time": "163.445",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "176.851",
"value": "Bill Cosby"
},
{
"operation": "Now on",
"time": "178.082",
"value": "Bill Cosby"
},
{
"operation": "Page search",
"time": "182.656",
"value": "1969 Outstanding Variety or Musical Program"
},
{
"operation": "Page search",
"time": "203.496",
"value": "Denoff"
},
{
"operation": "back-button-clicked",
"time": "215.196",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "215.951",
"value": "undefined"
},
{
"operation": "search",
"time": "342.296",
"value": "Outstanding Variety or Musical Program"
},
{
"operation": "Now on",
"time": "343.338",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "346.982",
"value": "Primetime Emmy Award for Outstanding Variety Special (Pre-Recorded)"
},
{
"operation": "Now on",
"time": "347.484",
"value": "Primetime Emmy Award for Outstanding Variety Special (Pre-Recorded)"
},
{
"operation": "finish",
"time": "369.726",
"value": "finish"
}
]
}
|
82596
|
The Iranian Cyber Army, an Iranian white hacker group which was directly hired by the Iranian government, pledged loyalty to the Supreme Leader of Iran and was assumed to have been managed by Mohamad Hussein Tajik.
|
{
"evidence": [
"Iranian Cyber Army_sentence_0",
"Iranian Cyber Army_sentence_1",
"Iranian Cyber Army_sentence_2",
"Iranian Cyber Army_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:\nThe Iranian Cyber Army, an Iranian white hacker group which was directly hired by the Iranian government, pledged loyalty to the Supreme Leader of Iran and was assumed to have been managed by Mohamad Hussein Tajik.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Iranian Cyber Army_sentence_0",
"Iranian Cyber Army_sentence_1",
"Iranian Cyber Army_sentence_2",
"Iranian Cyber Army_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 969,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.004",
"value": "?search="
},
{
"operation": "search",
"time": "16.239",
"value": "Iranian Cyber Army"
},
{
"operation": "Now on",
"time": "16.949",
"value": "Iranian Cyber Army"
},
{
"operation": "Highlighting",
"time": "82.221",
"value": "Iranian Cyber Army_sentence_0"
},
{
"operation": "Highlighting",
"time": "86.432",
"value": "Iranian Cyber Army_sentence_1"
},
{
"operation": "Highlighting",
"time": "98.028",
"value": "Iranian Cyber Army_sentence_2"
},
{
"operation": "Highlighting",
"time": "117.009",
"value": "Iranian Cyber Army_sentence_4"
},
{
"operation": "hyperlink",
"time": "189.06",
"value": "Iranian"
},
{
"operation": "Now on",
"time": "190.445",
"value": "Iran"
},
{
"operation": "finish",
"time": "200.45",
"value": "finish"
}
]
}
|
exfever_7001
|
Paul Walker was an American actor who starred in the 2002 American horror thriller film directed by John Dahl and written by an American filmmaker and composer best known for his works in the genres of action, drama, and science fiction.
|
{
"evidence": [
"Paul Walker",
"Joy Ride (2001 film)",
"J. J. Abrams"
],
"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:\nPaul Walker was an American actor who starred in the 2002 American horror thriller film directed by John Dahl and written by an American filmmaker and composer best known for his works in the genres of action, drama, and science fiction.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Paul Walker",
"Joy Ride (2001 film)",
"J. J. Abrams"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 970,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8460
|
Shakira is a Colombian singer who rose to prominence in Hispanic countries with her third studio album, released by Sony Music and Warner Bros.
|
{
"evidence": [
"Shakira",
"Pies Descalzos"
],
"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:\nShakira is a Colombian singer who rose to prominence in Hispanic countries with her third studio album, released by Sony Music and Warner Bros.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Shakira",
"Pies Descalzos"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 971,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_84
|
The Beverly Hillbillies show was produced by a television and film production company founded by film executives Martin Ransohoff and Edwin Kasper.
|
{
"evidence": [
"The Beverly Hillbillies",
"Filmways"
],
"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 Beverly Hillbillies show was produced by a television and film production company founded by film executives Martin Ransohoff and Edwin Kasper.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Beverly Hillbillies",
"Filmways"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 972,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6485
|
Clubber Lang is a fictional character created by Sylvester Stallone for an American film genre that uses sport as the theme.
|
{
"evidence": [
"Clubber Lang",
"Rocky III",
"Sports 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:\nClubber Lang is a fictional character created by Sylvester Stallone for an American film genre that uses sport as the theme.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Clubber Lang",
"Rocky III",
"Sports film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 973,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
46315
|
Zvi Lurie, a member of Hashomer Hatzair, serving as its general secretary between 1935 and 1937, helped establish Kol Yisrael, Israel's public domestic and international radio service, which broadcast the declaration.
|
{
"evidence": [
"Zvi Lurie_sentence_5",
"Zvi Lurie_sentence_7",
"Kol Yisrael_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:\nZvi Lurie, a member of Hashomer Hatzair, serving as its general secretary between 1935 and 1937, helped establish Kol Yisrael, Israel's public domestic and international radio service, which broadcast the declaration.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Zvi Lurie_sentence_5",
"Zvi Lurie_sentence_7",
"Kol Yisrael_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 974,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.673",
"value": "?search="
},
{
"operation": "search",
"time": "165.123",
"value": "Zvi Lurie"
},
{
"operation": "Now on",
"time": "166.007",
"value": "Zvi Lurie"
},
{
"operation": "Highlighting",
"time": "183.651",
"value": "Zvi Lurie_sentence_5"
},
{
"operation": "Highlighting",
"time": "212.761",
"value": "Zvi Lurie_sentence_7"
},
{
"operation": "hyperlink",
"time": "221.24",
"value": "Kol Yisrael"
},
{
"operation": "Now on",
"time": "222.06",
"value": "Kol Yisrael"
},
{
"operation": "Highlighting",
"time": "224.8",
"value": "Kol Yisrael_sentence_2"
},
{
"operation": "finish",
"time": "232.997",
"value": "finish"
}
]
}
|
4647
|
In 2013, Curtis Johnson was investigated by ASADA due to an allegation regarding peptides use but was cleared afterwards where in his return game against the Leigh Centurions, he scored 4 tries.
|
{
"evidence": [
"Curtis Johnston_sentence_9"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn 2013, Curtis Johnson was investigated by ASADA due to an allegation regarding peptides use but was cleared afterwards where in his return game against the Leigh Centurions, he scored 4 tries.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Curtis Johnston_sentence_9"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 975,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "14.807",
"value": "Leigh Centurions"
},
{
"operation": "search",
"time": "28.993",
"value": "Curtis Johnson (athlete)"
},
{
"operation": "search",
"time": "39.759",
"value": "Curtis Johnson"
},
{
"operation": "Highlighting",
"time": "50.153",
"value": "Curtis Johnson_sentence_0"
},
{
"operation": "Highlighting",
"time": "50.156",
"value": "Curtis Johnson_title"
},
{
"operation": "Highlighting",
"time": "51.363",
"value": "Curtis Johnson_item_0_0"
},
{
"operation": "Highlighting",
"time": "52.075",
"value": "Curtis Johnson_item_0_1"
},
{
"operation": "Highlighting",
"time": "52.886",
"value": "Curtis Johnson_item_0_2"
},
{
"operation": "Highlighting",
"time": "53.854",
"value": "Curtis Johnson_item_0_3"
},
{
"operation": "Highlighting",
"time": "55.598",
"value": "Curtis Johnson_item_0_4"
},
{
"operation": "Highlighting",
"time": "56.38",
"value": "Curtis Johnson_item_0_5"
},
{
"operation": "Highlighting",
"time": "57.249",
"value": "Curtis Johnson_item_0_6"
},
{
"operation": "Highlighting",
"time": "58.118",
"value": "Curtis Johnson_item_0_7"
},
{
"operation": "search",
"time": "68.717",
"value": "Leigh Centurions"
},
{
"operation": "Highlighting",
"time": "73.205",
"value": "Leigh Centurions_sentence_0"
},
{
"operation": "Highlighting",
"time": "73.236",
"value": "Leigh Centurions_title"
},
{
"operation": "Page search",
"time": "89.897",
"value": "leigh"
},
{
"operation": "Page search",
"time": "96.445",
"value": "johnson"
},
{
"operation": "Highlighting deleted",
"time": "102.974",
"value": "Leigh Centurions_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "103.009",
"value": "Leigh Centurions_title"
},
{
"operation": "Highlighting",
"time": "103.736",
"value": "Leigh Centurions_sentence_0"
},
{
"operation": "Highlighting",
"time": "103.763",
"value": "Leigh Centurions_title"
},
{
"operation": "search",
"time": "133.939",
"value": "ASADA"
},
{
"operation": "Page search",
"time": "141.174",
"value": "johnson"
},
{
"operation": "hyperlink",
"time": "154.84",
"value": "Drugs in sport in Australia"
},
{
"operation": "Page search",
"time": "160.792",
"value": "johnson"
},
{
"operation": "Highlighting",
"time": "194.739",
"value": "Drugs in sport in Australia_sentence_58"
},
{
"operation": "Highlighting",
"time": "194.759",
"value": "Drugs in sport in Australia_title"
},
{
"operation": "Highlighting",
"time": "194.785",
"value": "Drugs in sport in Australia_section_3"
},
{
"operation": "Highlighting",
"time": "195.807",
"value": "Drugs in sport in Australia_sentence_59"
},
{
"operation": "Highlighting",
"time": "196.846",
"value": "Drugs in sport in Australia_sentence_60"
},
{
"operation": "Highlighting",
"time": "198.969",
"value": "Drugs in sport in Australia_sentence_61"
},
{
"operation": "Highlighting",
"time": "202.264",
"value": "Drugs in sport in Australia_sentence_62"
},
{
"operation": "Highlighting",
"time": "204.428",
"value": "Drugs in sport in Australia_sentence_63"
},
{
"operation": "Highlighting",
"time": "205.78",
"value": "Drugs in sport in Australia_sentence_64"
},
{
"operation": "Highlighting",
"time": "209.062",
"value": "Drugs in sport in Australia_sentence_65"
},
{
"operation": "Highlighting",
"time": "209.995",
"value": "Drugs in sport in Australia_sentence_66"
},
{
"operation": "Highlighting",
"time": "213.12",
"value": "Drugs in sport in Australia_sentence_67"
},
{
"operation": "Highlighting",
"time": "213.832",
"value": "Drugs in sport in Australia_sentence_68"
},
{
"operation": "Highlighting",
"time": "216.992",
"value": "Drugs in sport in Australia_sentence_69"
},
{
"operation": "Highlighting",
"time": "220.423",
"value": "Drugs in sport in Australia_sentence_70"
},
{
"operation": "Highlighting",
"time": "221.883",
"value": "Drugs in sport in Australia_sentence_71"
},
{
"operation": "Highlighting",
"time": "222.717",
"value": "Drugs in sport in Australia_sentence_72"
},
{
"operation": "Highlighting",
"time": "226.383",
"value": "Drugs in sport in Australia_sentence_73"
},
{
"operation": "hyperlink",
"time": "236.91",
"value": "peptides"
},
{
"operation": "search",
"time": "257.486",
"value": "Curtis Johnson"
},
{
"operation": "hyperlink",
"time": "271.407",
"value": "Samuel Curtis Johnson (disambiguation)"
},
{
"operation": "search",
"time": "284.496",
"value": "Curtis Johnston"
},
{
"operation": "Highlighting",
"time": "301.534",
"value": "Curtis Johnston_sentence_0"
},
{
"operation": "Highlighting",
"time": "310.363",
"value": "Curtis Johnston_sentence_7"
},
{
"operation": "Highlighting",
"time": "312.591",
"value": "Curtis Johnston_sentence_8"
},
{
"operation": "Highlighting",
"time": "322.571",
"value": "Curtis Johnston_sentence_9"
},
{
"operation": "Highlighting deleted",
"time": "323.762",
"value": "Curtis Johnston_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "325.056",
"value": "Curtis Johnston_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "343.643",
"value": "Curtis Johnston_sentence_0"
},
{
"operation": "finish",
"time": "361.441",
"value": "finish"
}
]
}
|
exfever_4555
|
Legion was a fictional character that was the son of Bruce Lee, who was referred to as the human subspecies, Homo sapiens superior.
|
{
"evidence": [
"Legion (Marvel Comics)",
"Mutant (Marvel Comics)"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLegion was a fictional character that was the son of Bruce Lee, who was referred to as the human subspecies, Homo sapiens superior.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Legion (Marvel Comics)",
"Mutant (Marvel Comics)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 976,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1390
|
Roswell is the last American science fiction television series developed, produced, and co-written by Jason Katims.
|
{
"evidence": [
"Roswell (TV series)",
"UPN"
],
"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:\nRoswell is the last American science fiction television series developed, produced, and co-written by Jason Katims.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Roswell (TV series)",
"UPN"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 977,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
7502
|
The Sports in Minnesota no longer have any active teams.
|
{
"evidence": [
"Sports in Minnesota_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 Sports in Minnesota no longer have any active teams.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sports in Minnesota_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 978,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.545",
"value": "?search="
},
{
"operation": "search",
"time": "18.544",
"value": "Sports in Minnesota"
},
{
"operation": "Now on",
"time": "19.75",
"value": "Sports in Minnesota"
},
{
"operation": "Highlighting",
"time": "77.499",
"value": "Sports in Minnesota_sentence_0"
},
{
"operation": "finish",
"time": "369.077",
"value": "finish"
}
]
}
|
46144
|
The history of Michigan State University includes five Rose Bowl wins from 1954 to 2014.
|
{
"evidence": [
"Michigan State University_sentence_15"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe history of Michigan State University includes five Rose Bowl wins from 1954 to 2014.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Michigan State University_sentence_15"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 979,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.586",
"value": "?search="
},
{
"operation": "search",
"time": "15.947",
"value": "Michigan State University"
},
{
"operation": "Now on",
"time": "17.568",
"value": "Michigan State University"
},
{
"operation": "hyperlink",
"time": "31.682",
"value": "1 History Michigan State University_section_0"
},
{
"operation": "Page search",
"time": "46.675",
"value": "Rose Bowl"
},
{
"operation": "Highlighting",
"time": "59.553",
"value": "Michigan State University_sentence_15"
},
{
"operation": "finish",
"time": "77.208",
"value": "finish"
}
]
}
|
63590
|
Opera soprano singer Marilyn Mims made her debut in 1987 at the New York City Opera as Violetta in Verdi's La traviata and in 1006 won the Metropolitan Opera National Council Auditions/
|
{
"evidence": [
"Marilyn Mims_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:\nOpera soprano singer Marilyn Mims made her debut in 1987 at the New York City Opera as Violetta in Verdi's La traviata and in 1006 won the Metropolitan Opera National Council Auditions/\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Marilyn Mims_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 980,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.763",
"value": "?search="
},
{
"operation": "search",
"time": "183.042",
"value": "Marilyn Mims"
},
{
"operation": "Now on",
"time": "183.988",
"value": "Marilyn Mims"
},
{
"operation": "Highlighting",
"time": "191.656",
"value": "Marilyn Mims_sentence_8"
},
{
"operation": "Highlighting",
"time": "192.6",
"value": "Marilyn Mims_sentence_7"
},
{
"operation": "Highlighting",
"time": "212.411",
"value": "Marilyn Mims_sentence_7"
},
{
"operation": "finish",
"time": "217.313",
"value": "finish"
}
]
}
|
exfever_8878
|
Prince was the youngest musician and one?of the greatest musicians of his generation.
|
{
"evidence": [
"Prince (musician)",
"Minneapolis sound"
],
"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:\nPrince was the youngest musician and one?of the greatest musicians of his generation.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Prince (musician)",
"Minneapolis sound"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 981,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2211
|
42 is a biographical sports film about the first black athlete to play in a professional baseball organization.
|
{
"evidence": [
"42 (film)",
"Major League Baseball"
],
"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:\n42 is a biographical sports film about the first black athlete to play in a professional baseball organization.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"42 (film)",
"Major League Baseball"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 982,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
29336
|
Karen, who currently works full-time as the receptionist at a cemetery, wanted to apply for art college, but missed the deadline.
|
{
"evidence": [
"Planet Karen_sentence_22"
],
"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:\nKaren, who currently works full-time as the receptionist at a cemetery, wanted to apply for art college, but missed the deadline.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Planet Karen_sentence_22"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 983,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.892",
"value": "?search="
},
{
"operation": "search",
"time": "43.041",
"value": "Contains...Karen, who currently works full-time as the receptionist at a cemetery, wanted to apply for art college, but missed the deadline"
},
{
"operation": "Now on",
"time": "46.664",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "51.132",
"value": "Planet Karen"
},
{
"operation": "Now on",
"time": "52.018",
"value": "Planet Karen"
},
{
"operation": "Page search",
"time": "86.006",
"value": "receptionist at a cemetery"
},
{
"operation": "Highlighting",
"time": "95.652",
"value": "Planet Karen_sentence_22"
},
{
"operation": "Highlighting",
"time": "99.554",
"value": "Planet Karen_sentence_25"
},
{
"operation": "Highlighting deleted",
"time": "102.878",
"value": "Planet Karen_sentence_25"
},
{
"operation": "finish",
"time": "129.998",
"value": "finish"
}
]
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.