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 |
|---|---|---|---|---|---|---|---|---|
27034
|
Little My, adopted by the Moomin family, is a small, independent, brash, mischievous girl, who can also be a good friend.
|
{
"evidence": [
"Little My_sentence_5",
"Little My_sentence_2",
"Little My_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:\nLittle My, adopted by the Moomin family, is a small, independent, brash, mischievous girl, who can also be a good friend.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Little My_sentence_5",
"Little My_sentence_2",
"Little My_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7884,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.267",
"value": "?search="
},
{
"operation": "search",
"time": "14.941",
"value": "Contains...Little My, adopted by the Moomin family"
},
{
"operation": "Now on",
"time": "16.285",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "23.111",
"value": "Little My"
},
{
"operation": "Now on",
"time": "24.018",
"value": "Little My"
},
{
"operation": "Highlighting",
"time": "38.371",
"value": "Little My_sentence_5"
},
{
"operation": "Highlighting",
"time": "46.734",
"value": "Little My_sentence_2"
},
{
"operation": "Highlighting",
"time": "49.814",
"value": "Little My_sentence_3"
},
{
"operation": "finish",
"time": "80.855",
"value": "finish"
}
]
}
|
exfever_5856
|
Lonesome Dove is an American epic Western adventure television miniseries directed by an Australian director and producer of film and television.
|
{
"evidence": [
"Lonesome Dove (miniseries)",
"Simon Wincer"
],
"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:\nLonesome Dove is an American epic Western adventure television miniseries directed by an Australian director and producer of film and television.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lonesome Dove (miniseries)",
"Simon Wincer"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7885,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3156
|
Wilton Norman Chamberlain was an American professional player who played as a center?in a team sport in which three teams, most commonly of five players each, competed.
|
{
"evidence": [
"Wilt Chamberlain",
"Basketball"
],
"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:\nWilton Norman Chamberlain was an American professional player who played as a center?in a team sport in which three teams, most commonly of five players each, competed.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Wilt Chamberlain",
"Basketball"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7886,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
83891
|
Iraqis in Germany include migrants from Iraq to Germany, as well as their descendants, (estimated at around +300,000 people) and in 2015, out of 2,727 asylum applications for Iraqi refugees, only 8.3 percent or 122,000 people were accepted.
|
{
"evidence": [
"Iraqis in Germany_sentence_0",
"Iraqis in Germany_sentence_5",
"Iraqis in Germany_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIraqis in Germany include migrants from Iraq to Germany, as well as their descendants, (estimated at around +300,000 people) and in 2015, out of 2,727 asylum applications for Iraqi refugees, only 8.3 percent or 122,000 people were accepted.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Iraqis in Germany_sentence_0",
"Iraqis in Germany_sentence_5",
"Iraqis in Germany_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7887,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.779",
"value": "?search="
},
{
"operation": "search",
"time": "63.437",
"value": "Iraqis in Germany"
},
{
"operation": "Now on",
"time": "64.144",
"value": "Iraqis in Germany"
},
{
"operation": "Highlighting",
"time": "190.44",
"value": "Iraqis in Germany_sentence_0"
},
{
"operation": "Highlighting",
"time": "193.755",
"value": "Iraqis in Germany_sentence_1"
},
{
"operation": "Highlighting",
"time": "205.054",
"value": "Iraqis in Germany_sentence_5"
},
{
"operation": "Highlighting",
"time": "206.267",
"value": "Iraqis in Germany_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "243.93",
"value": "Iraqis in Germany_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "245.066",
"value": "Iraqis in Germany_sentence_2"
},
{
"operation": "Highlighting",
"time": "255.671",
"value": "Iraqis in Germany_sentence_2"
},
{
"operation": "finish",
"time": "262.134",
"value": "finish"
}
]
}
|
exfever_5877
|
Shantel VanSanten is an American actress who played the role of Julie Swagger in Shooter, which aired on the American basic cable channel owned by NBCUniversal Television.
|
{
"evidence": [
"Shantel VanSanten",
"USA Network"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nShantel VanSanten is an American actress who played the role of Julie Swagger in Shooter, which aired on the American basic cable channel owned by NBCUniversal Television.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Shantel VanSanten",
"USA Network"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7888,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3561
|
Adidas AG is a German multinational corporation that was founded and headquartered in a state in the southeast of Germany with an area of 70,550.19 km2.
|
{
"evidence": [
"Adidas",
"Bavaria"
],
"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:\nAdidas AG is a German multinational corporation that was founded and headquartered in a state in the southeast of Germany with an area of 70,550.19 km2.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Adidas",
"Bavaria"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7889,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
38359
|
Congregation Dorshei Emet is a Reconstructionist synagogue which is a modern Jewish movement that views Judiaism as a progressively evolving civilization.
|
{
"evidence": [
"Congregation Dorshei Emet_sentence_0",
"Reconstructionist Judaism_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:\nCongregation Dorshei Emet is a Reconstructionist synagogue which is a modern Jewish movement that views Judiaism as a progressively evolving civilization.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Congregation Dorshei Emet_sentence_0",
"Reconstructionist Judaism_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7890,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.58",
"value": "?search="
},
{
"operation": "Now on",
"time": "519.686",
"value": "?search="
},
{
"operation": "search",
"time": "677.266",
"value": "Congregation Dorshei Emet"
},
{
"operation": "Now on",
"time": "677.975",
"value": "Congregation Dorshei Emet"
},
{
"operation": "Highlighting",
"time": "684.161",
"value": "Congregation Dorshei Emet_sentence_0"
},
{
"operation": "Page search",
"time": "691.215",
"value": "Jew"
},
{
"operation": "hyperlink",
"time": "731.805",
"value": "Reconstructionist"
},
{
"operation": "Now on",
"time": "732.326",
"value": "Reconstructionist Judaism"
},
{
"operation": "Highlighting",
"time": "734.09",
"value": "Reconstructionist Judaism_sentence_0"
},
{
"operation": "finish",
"time": "824.51",
"value": "finish"
}
]
}
|
exfever_7484
|
The Chainsmokers have won twelve music awards from the event that recognizes the most popular artists and music over the past year.
|
{
"evidence": [
"The Chainsmokers",
"IHeartRadio Music Awards"
],
"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 Chainsmokers have won twelve music awards from the event that recognizes the most popular artists and music over the past year.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Chainsmokers",
"IHeartRadio Music Awards"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7891,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4789
|
Felicity Rose Hadley Jones won an award in the Empire Awards.
|
{
"evidence": [
"Felicity Jones",
"The Treasure Seekers (1996 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:\nFelicity Rose Hadley Jones won an award in the Empire Awards.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Felicity Jones",
"The Treasure Seekers (1996 film)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7892,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
70127
|
The Oxford University Handball Club men's team won the British Student Championship title in 2004, 2005, 2006, 2007, 2009, 2013 and 2014, and also became the first university team to compete in the EHF Challenge Cup (2006–07 and 2007–08).
|
{
"evidence": [
"Oxford University Handball Club_sentence_3",
"Oxford University Handball Club_sentence_4"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Oxford University Handball Club men's team won the British Student Championship title in 2004, 2005, 2006, 2007, 2009, 2013 and 2014, and also became the first university team to compete in the EHF Challenge Cup (2006–07 and 2007–08).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Oxford University Handball Club_sentence_3",
"Oxford University Handball Club_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7893,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.146",
"value": "?search="
},
{
"operation": "search",
"time": "10.641",
"value": "Oxford University Handball Club"
},
{
"operation": "Now on",
"time": "11.345",
"value": "Oxford University Handball Club"
},
{
"operation": "Page search",
"time": "20.119",
"value": "british student"
},
{
"operation": "Highlighting",
"time": "22.697",
"value": "Oxford University Handball Club_sentence_3"
},
{
"operation": "Highlighting",
"time": "32",
"value": "Oxford University Handball Club_sentence_4"
},
{
"operation": "finish",
"time": "45.49",
"value": "finish"
}
]
}
|
80023
|
The Congress elected Freire as the new President with 37 votes, but the Congress knew Freire would resign after Pinto's arrival in Chile, so the election was only a sign of appreciation towards him.
|
{
"evidence": [
"1827 Chilean presidential election_sentence_4",
"1827 Chilean presidential election_sentence_3"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Congress elected Freire as the new President with 37 votes, but the Congress knew Freire would resign after Pinto's arrival in Chile, so the election was only a sign of appreciation towards him.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"1827 Chilean presidential election_sentence_4",
"1827 Chilean presidential election_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7894,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.717",
"value": "?search="
},
{
"operation": "search",
"time": "11.037",
"value": "Congress knew Freire would resign after Pinto's arrival in Chile, so the election was only a sign of appreciation towards him."
},
{
"operation": "Now on",
"time": "14.03",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "20.736",
"value": "1827 Chilean presidential election"
},
{
"operation": "Now on",
"time": "21.115",
"value": "1827 Chilean presidential election"
},
{
"operation": "Highlighting",
"time": "25.475",
"value": "1827 Chilean presidential election_sentence_4"
},
{
"operation": "Highlighting",
"time": "35.336",
"value": "1827 Chilean presidential election_sentence_3"
},
{
"operation": "finish",
"time": "45.181",
"value": "finish"
}
]
}
|
16685
|
Gabrovo Municipality, named after the Gabrovo Province (and its capital city, Gabrovo), is located in the Balkan Mountains, which house some of Europe's largest animals.
|
{
"evidence": [
"Gabrovo Municipality_sentence_0",
"Gabrovo Municipality_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:\nGabrovo Municipality, named after the Gabrovo Province (and its capital city, Gabrovo), is located in the Balkan Mountains, which house some of Europe's largest animals.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Gabrovo Municipality_sentence_0",
"Gabrovo Municipality_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7895,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.428",
"value": "?search="
},
{
"operation": "search",
"time": "21.607",
"value": "Gabrovo Municipality"
},
{
"operation": "Now on",
"time": "23.213",
"value": "Gabrovo Municipality"
},
{
"operation": "Highlighting",
"time": "27.588",
"value": "Gabrovo Municipality_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "29.794",
"value": "Gabrovo Municipality_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "35.454",
"value": "Gabrovo Municipality_sentence_0"
},
{
"operation": "Highlighting",
"time": "43.903",
"value": "Gabrovo Municipality_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "58.44",
"value": "Gabrovo Municipality_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "87.551",
"value": "Gabrovo Municipality_sentence_1"
},
{
"operation": "Highlighting",
"time": "88.204",
"value": "Gabrovo Municipality_sentence_0"
},
{
"operation": "Highlighting",
"time": "88.816",
"value": "Gabrovo Municipality_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "159.966",
"value": "Gabrovo Municipality_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "160.817",
"value": "Gabrovo Municipality_sentence_1"
},
{
"operation": "Highlighting",
"time": "179.212",
"value": "Gabrovo Municipality_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "196.695",
"value": "Gabrovo Municipality_sentence_1"
},
{
"operation": "Highlighting",
"time": "198.722",
"value": "Gabrovo Municipality_sentence_0"
},
{
"operation": "hyperlink",
"time": "200.589",
"value": "Stara planina"
},
{
"operation": "Now on",
"time": "202.937",
"value": "Balkan Mountains"
},
{
"operation": "back-button-clicked",
"time": "204.603",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "204.863",
"value": "Gabrovo Municipality"
},
{
"operation": "Highlighting",
"time": "206.25",
"value": "Gabrovo Municipality_sentence_1"
},
{
"operation": "hyperlink",
"time": "221.4",
"value": "Gabrovo"
},
{
"operation": "Now on",
"time": "222.944",
"value": "Gabrovo"
},
{
"operation": "finish",
"time": "238.545",
"value": "finish"
}
]
}
|
56635
|
Reflections, released in 2008, was his first and only album to date where he got a Grammy Award nomination at the 51st Annual Grammy Awards in the Best Contemporary R&B Gospel Album category.
|
{
"evidence": [
"Jason Champion_sentence_4",
"Jason Champion_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:\nReflections, released in 2008, was his first and only album to date where he got a Grammy Award nomination at the 51st Annual Grammy Awards in the Best Contemporary R&B Gospel Album category.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Jason Champion_sentence_4",
"Jason Champion_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7896,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.766",
"value": "?search="
},
{
"operation": "search",
"time": "28.45",
"value": "51st Annual Grammy Awards"
},
{
"operation": "Now on",
"time": "30.372",
"value": "undefined"
},
{
"operation": "search",
"time": "49.791",
"value": "Reflections"
},
{
"operation": "Now on",
"time": "50.452",
"value": "Reflections"
},
{
"operation": "search",
"time": "67.262",
"value": "Reflections (The Supremes song)"
},
{
"operation": "Now on",
"time": "68.142",
"value": "Reflections (The Supremes song)"
},
{
"operation": "search",
"time": "106.222",
"value": "Grammy Award for Best Contemporary R&B Gospel Album"
},
{
"operation": "Now on",
"time": "107.26",
"value": "Grammy Award for Best Contemporary R&B Gospel Album"
},
{
"operation": "search",
"time": "162.246",
"value": "Grammy Award for Best Contemporary R&B Gospel Album"
},
{
"operation": "Now on",
"time": "162.751",
"value": "Grammy Award for Best Contemporary R&B Gospel Album"
},
{
"operation": "Page search",
"time": "167.863",
"value": "reflections"
},
{
"operation": "hyperlink",
"time": "183.251",
"value": "Grammy Award"
},
{
"operation": "Now on",
"time": "186.681",
"value": "Grammy Award"
},
{
"operation": "search",
"time": "195.996",
"value": "51st Annual Grammy Awards"
},
{
"operation": "Now on",
"time": "196.728",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "199.179",
"value": "Lost+ / Viva la Vida (Live At the 51st Annual Grammy Awards)"
},
{
"operation": "Now on",
"time": "200.317",
"value": "Viva la Vida"
},
{
"operation": "hyperlink",
"time": "206.083",
"value": "Lost+ / Viva la Vida (Live At the 51st Annual Grammy Awards)"
},
{
"operation": "Now on",
"time": "206.698",
"value": "undefined"
},
{
"operation": "search",
"time": "244.851",
"value": "Jason Champion"
},
{
"operation": "Now on",
"time": "246.089",
"value": "Jason Champion"
},
{
"operation": "Page search",
"time": "251.627",
"value": "reflections"
},
{
"operation": "Highlighting",
"time": "254.983",
"value": "Jason Champion_sentence_4"
},
{
"operation": "Highlighting",
"time": "259.484",
"value": "Jason Champion_sentence_5"
},
{
"operation": "finish",
"time": "276.397",
"value": "finish"
}
]
}
|
exfever_5560
|
Robert Arum is an American lawyer and boxing promoter who was employed as a person who practices law.
|
{
"evidence": [
"Bob Arum",
"Lawyer"
],
"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:\nRobert Arum is an American lawyer and boxing promoter who was employed as a person who practices law.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Bob Arum",
"Lawyer"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7897,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6035
|
Abraham Lincoln was an American lawyer and statesman who served as the 16th president of the United States, born in a state in the Southeastern region of the United States.
|
{
"evidence": [
"Abraham Lincoln",
"Kentucky"
],
"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:\nAbraham Lincoln was an American lawyer and statesman who served as the 16th president of the United States, born in a state in the Southeastern region of the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Abraham Lincoln",
"Kentucky"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7898,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_670
|
Susan Abigail Sarandon is the recipient of numerous accolades, including a recognition from an annual award show hosted by the British Academy of Film and Television Arts to honor the best British and international contributions to film.
|
{
"evidence": [
"Susan Sarandon",
"British Academy Film Awards"
],
"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:\nSusan Abigail Sarandon is the recipient of numerous accolades, including a recognition from an annual award show hosted by the British Academy of Film and Television Arts to honor the best British and international contributions to film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Susan Sarandon",
"British Academy Film Awards"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7899,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7985
|
Tall Story is a 1950 American romantic comedy film starring an American actor, director, and singer.
|
{
"evidence": [
"Tall Story",
"Anthony Perkins"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nTall Story is a 1950 American romantic comedy film starring an American actor, director, and singer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tall Story",
"Anthony Perkins"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7900,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
86713
|
Anvar, best known as Alex Kamal on The Expanse (an American science fiction television series developed by Mark Fergus and Hawk Ostby), appeared as Dodi Fayed with Naomi Watts in Diana in 2013.
|
{
"evidence": [
"Cas Anvar_sentence_7",
"Cas Anvar_sentence_6",
"The Expanse (TV series)_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:\nAnvar, best known as Alex Kamal on The Expanse (an American science fiction television series developed by Mark Fergus and Hawk Ostby), appeared as Dodi Fayed with Naomi Watts in Diana in 2013.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Cas Anvar_sentence_7",
"Cas Anvar_sentence_6",
"The Expanse (TV series)_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7901,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.596",
"value": "?search="
},
{
"operation": "search",
"time": "35.881",
"value": "Anvar"
},
{
"operation": "Now on",
"time": "37.203",
"value": "Anvar"
},
{
"operation": "Highlighting",
"time": "64.126",
"value": "Anvar_item_1_15"
},
{
"operation": "Highlighting",
"time": "64.128",
"value": "Anvar_title"
},
{
"operation": "Highlighting",
"time": "64.132",
"value": "Anvar_section_1"
},
{
"operation": "Highlighting deleted",
"time": "67.297",
"value": "Anvar_item_1_15"
},
{
"operation": "Highlighting deleted",
"time": "67.3",
"value": "Anvar_title"
},
{
"operation": "Highlighting deleted",
"time": "67.305",
"value": "Anvar_section_1"
},
{
"operation": "hyperlink",
"time": "68.846",
"value": "Cas Anvar"
},
{
"operation": "Now on",
"time": "69.836",
"value": "Cas Anvar"
},
{
"operation": "Highlighting",
"time": "89.97",
"value": "Cas Anvar_sentence_7"
},
{
"operation": "Highlighting",
"time": "99.385",
"value": "Cas Anvar_sentence_6"
},
{
"operation": "hyperlink",
"time": "130.023",
"value": "The Expanse"
},
{
"operation": "Now on",
"time": "131.209",
"value": "The Expanse (TV series)"
},
{
"operation": "Highlighting",
"time": "139.179",
"value": "The Expanse (TV series)_sentence_2"
},
{
"operation": "finish",
"time": "166.423",
"value": "finish"
}
]
}
|
exfever_4205
|
Inside Man is Denzel Washington's ninth film.
|
{
"evidence": [
"Inside Man",
"New York City Police Department",
"New York City"
],
"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:\nInside Man is Denzel Washington's ninth film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Inside Man",
"New York City Police Department",
"New York City"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7902,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4823
|
Bosnia and Herzegovina, known informally as Bosnia, is a country at the crossroads of the south and southeast in a continent that is also recognized as a part of Eurasia that shares the continental landmass of Afro-Eurasia with both America and the world's second-largest and second-most-populous continent, after Asia in both cases.
|
{
"evidence": [
"Bosnia and Herzegovina",
"Europe",
"Africa"
],
"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:\nBosnia and Herzegovina, known informally as Bosnia, is a country at the crossroads of the south and southeast in a continent that is also recognized as a part of Eurasia that shares the continental landmass of Afro-Eurasia with both America and the world's second-largest and second-most-populous continent, after Asia in both cases.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Bosnia and Herzegovina",
"Europe",
"Africa"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7903,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
27597
|
NETVC was the name given to a planned royalty-free video codec (software or hardware that compresses and decompresses digital video), that was intended to provide a royalty-free alternative to industry standards such as H.264/AVC and HEVC.
|
{
"evidence": [
"NETVC_sentence_2",
"Video codec_sentence_1",
"NETVC_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:\nNETVC was the name given to a planned royalty-free video codec (software or hardware that compresses and decompresses digital video), that was intended to provide a royalty-free alternative to industry standards such as H.264/AVC and HEVC.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"NETVC_sentence_2",
"Video codec_sentence_1",
"NETVC_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7904,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.753",
"value": "?search="
},
{
"operation": "search",
"time": "3.326",
"value": "NETVC"
},
{
"operation": "Now on",
"time": "3.924",
"value": "NETVC"
},
{
"operation": "Highlighting",
"time": "8.8",
"value": "NETVC_sentence_2"
},
{
"operation": "hyperlink",
"time": "10.967",
"value": "video codec"
},
{
"operation": "Now on",
"time": "12.172",
"value": "Video codec"
},
{
"operation": "Highlighting",
"time": "13.816",
"value": "Video codec_sentence_1"
},
{
"operation": "back-button-clicked",
"time": "16.126",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "16.21",
"value": "NETVC"
},
{
"operation": "Highlighting",
"time": "20.766",
"value": "NETVC_sentence_3"
},
{
"operation": "finish",
"time": "28.056",
"value": "finish"
}
]
}
|
37181
|
Karabo Mothibi, a race car driver from Botswana who was born on October 15, 1996, came in 2nd position in the 2015 African Junior Championships in the 4 x 100 meter relay, which was held in Eugene in the United States.
|
{
"evidence": [
"Karabo Mothibi_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:\nKarabo Mothibi, a race car driver from Botswana who was born on October 15, 1996, came in 2nd position in the 2015 African Junior Championships in the 4 x 100 meter relay, which was held in Eugene in the United States.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Karabo Mothibi_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7905,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.075",
"value": "?search="
},
{
"operation": "search",
"time": "18.133",
"value": "Karabo Mothibi"
},
{
"operation": "Now on",
"time": "18.846",
"value": "Karabo Mothibi"
},
{
"operation": "Highlighting",
"time": "24.749",
"value": "Karabo Mothibi_sentence_0"
},
{
"operation": "finish",
"time": "36.344",
"value": "finish"
}
]
}
|
28119
|
The blue tiger moth or blue day moth was described in 1790 and can be found in India and Sri Lanka.
|
{
"evidence": [
"Dysphania percota_sentence_3",
"Dysphania percota_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe blue tiger moth or blue day moth was described in 1790 and can be found in India and Sri Lanka.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Dysphania percota_sentence_3",
"Dysphania percota_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7906,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.946",
"value": "?search="
},
{
"operation": "search",
"time": "10.138",
"value": "Blue Tiger Moth"
},
{
"operation": "Now on",
"time": "11.063",
"value": "Blue Tiger Moth"
},
{
"operation": "Highlighting",
"time": "26.876",
"value": "Dysphania percota_sentence_3"
},
{
"operation": "Highlighting",
"time": "35.897",
"value": "Dysphania percota_sentence_2"
},
{
"operation": "finish",
"time": "42.454",
"value": "finish"
}
]
}
|
exfever_7772
|
The Odyssey is one of two major ancient Greek short narrative poem typically about the extraordinary deeds of extraordinary characters, attributed to Homer.
|
{
"evidence": [
"Odyssey",
"Epic poetry"
],
"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 Odyssey is one of two major ancient Greek short narrative poem typically about the extraordinary deeds of extraordinary characters, attributed to Homer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Odyssey",
"Epic poetry"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7907,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
5597
|
Pendle Hill is in Africa, and it has no parent peak.
|
{
"evidence": [
"Pendle Hill_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPendle Hill is in Africa, and it has no parent peak.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Pendle Hill_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7908,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.948",
"value": "?search="
},
{
"operation": "search",
"time": "14.502",
"value": "Pendle Hill"
},
{
"operation": "Now on",
"time": "15.23",
"value": "Pendle Hill"
},
{
"operation": "Highlighting",
"time": "53.713",
"value": "Pendle Hill_sentence_2"
},
{
"operation": "finish",
"time": "82.951",
"value": "finish"
}
]
}
|
exfever_7539
|
Selena Marie Gomez is a Filipino singer and actress who appeared in a 2009 American made-for-television comedy-drama film directed by Lev L. Spiro that also classified as a film genre with fantastic themes, usually magic, supernatural events, mythology, folklore, or exotic fantasy worlds.
|
{
"evidence": [
"Selena Gomez",
"Wizards of Waverly Place: The Movie",
"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:\nSelena Marie Gomez is a Filipino singer and actress who appeared in a 2009 American made-for-television comedy-drama film directed by Lev L. Spiro that also classified as a film genre with fantastic themes, usually magic, supernatural events, mythology, folklore, or exotic fantasy worlds.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Selena Gomez",
"Wizards of Waverly Place: The Movie",
"Fantasy film"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7909,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6830
|
Boomerang was the debut film starring Eddie Murphy.
|
{
"evidence": [
"Boomerang (1992 film)",
"Martin Lawrence"
],
"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:\nBoomerang was the debut film starring Eddie Murphy.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Boomerang (1992 film)",
"Martin Lawrence"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7910,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6285
|
Super Metroid is an action-adventure game developed and published by a Spanish multinational video game company that gained international recognition with the release of an 8-bit third-generation home video game console.
|
{
"evidence": [
"Super Metroid",
"Nintendo",
"Nintendo Entertainment System"
],
"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:\nSuper Metroid is an action-adventure game developed and published by a Spanish multinational video game company that gained international recognition with the release of an 8-bit third-generation home video game console.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Super Metroid",
"Nintendo",
"Nintendo Entertainment System"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7911,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
61414
|
The 828th Bombardment Squadron, from the United States, was only active for one year.
|
{
"evidence": [
"828th Bombardment Squadron_sentence_6",
"828th Bombardment Squadron_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 828th Bombardment Squadron, from the United States, was only active for one year.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"828th Bombardment Squadron_sentence_6",
"828th Bombardment Squadron_sentence_15"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7912,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.177",
"value": "?search="
},
{
"operation": "search",
"time": "8.816",
"value": "828th Bombardment Squadron"
},
{
"operation": "Now on",
"time": "9.744",
"value": "828th Bombardment Squadron"
},
{
"operation": "Highlighting",
"time": "197.119",
"value": "828th Bombardment Squadron_sentence_6"
},
{
"operation": "Highlighting",
"time": "203.449",
"value": "828th Bombardment Squadron_sentence_15"
},
{
"operation": "finish",
"time": "210.019",
"value": "finish"
}
]
}
|
48789
|
Curtis James Jackson, popularly known as 50 cent, has a sole business in the music industry.
|
{
"evidence": [
"50 Cent_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:\nCurtis James Jackson, popularly known as 50 cent, has a sole business in the music industry.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"50 Cent_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7913,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.393",
"value": "?search="
},
{
"operation": "search",
"time": "19.138",
"value": "Curtis James Jackson"
},
{
"operation": "Now on",
"time": "21.256",
"value": "50 Cent"
},
{
"operation": "Highlighting",
"time": "35.25",
"value": "50 Cent_sentence_5"
},
{
"operation": "finish",
"time": "56.856",
"value": "finish"
}
]
}
|
15831
|
Pretty Shield became the second wife of Goes Ahead and gave birth to four girls and three boys.
|
{
"evidence": [
"Pretty Shield_sentence_14",
"Pretty Shield_sentence_15"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPretty Shield became the second wife of Goes Ahead and gave birth to four girls and three boys.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Pretty Shield_sentence_14",
"Pretty Shield_sentence_15"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7914,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.544",
"value": "?search="
},
{
"operation": "search",
"time": "9.201",
"value": "Pretty Shield"
},
{
"operation": "Now on",
"time": "9.86",
"value": "Pretty Shield"
},
{
"operation": "Page search",
"time": "20.215",
"value": "wife"
},
{
"operation": "Highlighting",
"time": "22.544",
"value": "Pretty Shield_sentence_14"
},
{
"operation": "Highlighting",
"time": "28.907",
"value": "Pretty Shield_sentence_15"
},
{
"operation": "finish",
"time": "314.466",
"value": "finish"
}
]
}
|
exfever_8806
|
Ronin is a 1998 American action thriller film written by a British filmmaker who won the Pulitzer Prize.
|
{
"evidence": [
"Ronin (film)",
"David Mamet"
],
"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:\nRonin is a 1998 American action thriller film written by a British filmmaker who won the Pulitzer Prize.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ronin (film)",
"David Mamet"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7915,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2383
|
The 35th president of the United States proposed the establishment of the Environmental Protection Agency, an independent executive agency of the United States federal government, on July 9, 1970.
|
{
"evidence": [
"United States Environmental Protection Agency",
"Richard Nixon"
],
"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 35th president of the United States proposed the establishment of the Environmental Protection Agency, an independent executive agency of the United States federal government, on July 9, 1970.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"United States Environmental Protection Agency",
"Richard Nixon"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7916,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_57
|
Judith Eva Barsi and her mother, Maria, were killed in July 1988 in an act in which an individual kills one or more people before killing themselves.
|
{
"evidence": [
"Judith Barsi",
"Murder–suicide"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJudith Eva Barsi and her mother, Maria, were killed in July 1988 in an act in which an individual kills one or more people before killing themselves.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Judith Barsi",
"Murder–suicide"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7917,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_727
|
Veronika Decides to Die is a 2009 American psychological drama film starring an American actress who received critical acclaim for her performance as Ray Eddy in the 2008 film Frozen River.
|
{
"evidence": [
"Veronika Decides to Die (film)",
"Melissa Leo"
],
"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:\nVeronika Decides to Die is a 2009 American psychological drama film starring an American actress who received critical acclaim for her performance as Ray Eddy in the 2008 film Frozen River.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Veronika Decides to Die (film)",
"Melissa Leo"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7918,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
64692
|
UC-70, badly damaged by a near miss by a 520 pounds (240 kg) bomb and finished off by the destroyer HMS Ouse, was spotted lying submerged on the sea bottom near Runswick Bay on 28 August 1918, by a Kangaroo flown by Lt E. F. Waring.
|
{
"evidence": [
"Blackburn Kangaroo_sentence_9"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nUC-70, badly damaged by a near miss by a 520 pounds (240 kg) bomb and finished off by the destroyer HMS Ouse, was spotted lying submerged on the sea bottom near Runswick Bay on 28 August 1918, by a Kangaroo flown by Lt E. F. Waring.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Blackburn Kangaroo_sentence_9"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7919,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.331",
"value": "?search="
},
{
"operation": "search",
"time": "10.012",
"value": "UC-70"
},
{
"operation": "Now on",
"time": "20.706",
"value": "undefined"
},
{
"operation": "search",
"time": "70.46",
"value": "Lt E. F. Waring"
},
{
"operation": "Now on",
"time": "81.256",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "85.267",
"value": "Blackburn Kangaroo"
},
{
"operation": "Now on",
"time": "86.227",
"value": "Blackburn Kangaroo"
},
{
"operation": "Page search",
"time": "110.763",
"value": "uc-70"
},
{
"operation": "Highlighting",
"time": "122.696",
"value": "Blackburn Kangaroo_sentence_9"
},
{
"operation": "hyperlink",
"time": "143.336",
"value": "UC-70"
},
{
"operation": "hyperlink",
"time": "157.658",
"value": "UC-70"
},
{
"operation": "Now on",
"time": "158.489",
"value": "SM UC-70"
},
{
"operation": "search",
"time": "254.474",
"value": "HMS Ouse"
},
{
"operation": "Now on",
"time": "256.338",
"value": "HMS Ouse"
},
{
"operation": "finish",
"time": "310.159",
"value": "finish"
}
]
}
|
15379
|
Radosław Janukiewicz signed a contract with Skoda Xanthi that was terminated before joining GKP Gorzów Wielkopolski on 28 February 2009 on a half-year contract.
|
{
"evidence": [
"Radosław Janukiewicz_sentence_1",
"Radosław Janukiewicz_sentence_2",
"Radosław Janukiewicz_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:\nRadosław Janukiewicz signed a contract with Skoda Xanthi that was terminated before joining GKP Gorzów Wielkopolski on 28 February 2009 on a half-year contract.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Radosław Janukiewicz_sentence_1",
"Radosław Janukiewicz_sentence_2",
"Radosław Janukiewicz_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7920,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.538",
"value": "?search="
},
{
"operation": "search",
"time": "12.117",
"value": "Radosław Janukiewicz"
},
{
"operation": "Now on",
"time": "12.816",
"value": "Radosław Janukiewicz"
},
{
"operation": "Page search",
"time": "17.441",
"value": "contract"
},
{
"operation": "page-search-reset",
"time": "20.164",
"value": "page-search-reset"
},
{
"operation": "Highlighting",
"time": "22.519",
"value": "Radosław Janukiewicz_sentence_1"
},
{
"operation": "Highlighting",
"time": "26.317",
"value": "Radosław Janukiewicz_sentence_2"
},
{
"operation": "Highlighting",
"time": "32.779",
"value": "Radosław Janukiewicz_sentence_3"
},
{
"operation": "finish",
"time": "182.008",
"value": "finish"
}
]
}
|
exfever_3667
|
Richard Dreyfuss starred in a 1980 American coming-of-age story directed by Rob Reiner.
|
{
"evidence": [
"Richard Dreyfuss",
"Stand by Me (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:\nRichard Dreyfuss starred in a 1980 American coming-of-age story directed by Rob Reiner.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Richard Dreyfuss",
"Stand by Me (film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7921,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6139
|
Mel Gibson's first acting appearance was in the film Gallipoli.
|
{
"evidence": [
"Mel Gibson",
"Lethal Weapon (franchise)",
"Los Angeles Police Department"
],
"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:\nMel Gibson's first acting appearance was in the film Gallipoli.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mel Gibson",
"Lethal Weapon (franchise)",
"Los Angeles Police Department"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7922,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
53803
|
In the field of passive fluency, commonly known as passive bilingualism, François Grosjean argues that people with no equal competence in multiple languages are not "bilinguals."
|
{
"evidence": [
"Passive fluency_sentence_3",
"Multilingualism_sentence_123",
"François Grosjean_sentence_3",
"François Grosjean_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:\nIn the field of passive fluency, commonly known as passive bilingualism, François Grosjean argues that people with no equal competence in multiple languages are not \"bilinguals.\"\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Passive fluency_sentence_3",
"Multilingualism_sentence_123",
"François Grosjean_sentence_3",
"François Grosjean_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7923,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.3",
"value": "?search="
},
{
"operation": "search",
"time": "11.726",
"value": "Passive fluency"
},
{
"operation": "Now on",
"time": "12.439",
"value": "Passive fluency"
},
{
"operation": "Highlighting",
"time": "20.047",
"value": "Passive fluency_sentence_3"
},
{
"operation": "Page search",
"time": "39.223",
"value": "François Grosjean"
},
{
"operation": "search",
"time": "82.569",
"value": "François Grosjean argues that people with no equal competence in multiple languages are not \"bilinguals"
},
{
"operation": "Now on",
"time": "84.355",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "96.58",
"value": "Multilingualism"
},
{
"operation": "Now on",
"time": "97.905",
"value": "Multilingualism"
},
{
"operation": "Page search",
"time": "104.851",
"value": "François Grosjean"
},
{
"operation": "Page search",
"time": "105.096",
"value": "François Grosjean"
},
{
"operation": "Page search",
"time": "142.116",
"value": "people with no equal competence in multiple languages are not bilinguals"
},
{
"operation": "Page search",
"time": "142.892",
"value": "people with no equal competence in multiple languages are not bilinguals"
},
{
"operation": "Page search",
"time": "170.046",
"value": "François Grosjean"
},
{
"operation": "Highlighting",
"time": "183.216",
"value": "Multilingualism_sentence_123"
},
{
"operation": "search",
"time": "190.506",
"value": "François Grosjean"
},
{
"operation": "Now on",
"time": "191.239",
"value": "François Grosjean"
},
{
"operation": "Page search",
"time": "204.895",
"value": "people with no equal competence"
},
{
"operation": "Highlighting",
"time": "226.032",
"value": "François Grosjean_sentence_3"
},
{
"operation": "Highlighting",
"time": "228.138",
"value": "François Grosjean_sentence_4"
},
{
"operation": "search",
"time": "298.401",
"value": "myths about bilingualism"
},
{
"operation": "Now on",
"time": "299.915",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "309.42",
"value": "Multilingualism"
},
{
"operation": "Now on",
"time": "310.156",
"value": "Multilingualism"
},
{
"operation": "finish",
"time": "341.311",
"value": "finish"
}
]
}
|
exfever_5498
|
Con Man is an American comedy web series directed by Alan Tudyk, who was also one of the stars of the 2002 Western drama television series directed by Joss Whedon.
|
{
"evidence": [
"Con Man (web series)",
"Firefly (TV series)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nCon Man is an American comedy web series directed by Alan Tudyk, who was also one of the stars of the 2002 Western drama television series directed by Joss Whedon.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Con Man (web series)",
"Firefly (TV series)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7924,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
9179
|
Teodoro Fernández received several offers throughout his career from teams in Europe, Argentina, and Chile and, despite a rumor that they would pay him whatever he liked, Fernández did not sign with Chile.
|
{
"evidence": [
"Teodoro Fernández_sentence_18",
"Teodoro Fernández_sentence_19",
"Teodoro Fernández_sentence_20"
],
"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:\nTeodoro Fernández received several offers throughout his career from teams in Europe, Argentina, and Chile and, despite a rumor that they would pay him whatever he liked, Fernández did not sign with Chile.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Teodoro Fernández_sentence_18",
"Teodoro Fernández_sentence_19",
"Teodoro Fernández_sentence_20"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7925,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.502",
"value": "?search="
},
{
"operation": "search",
"time": "10.528",
"value": "Teodoro Fernández"
},
{
"operation": "Now on",
"time": "11.242",
"value": "Teodoro Fernández"
},
{
"operation": "Page search",
"time": "19.031",
"value": "offer"
},
{
"operation": "Highlighting",
"time": "25.86",
"value": "Teodoro Fernández_sentence_18"
},
{
"operation": "Highlighting",
"time": "32.125",
"value": "Teodoro Fernández_sentence_19"
},
{
"operation": "Highlighting",
"time": "35.303",
"value": "Teodoro Fernández_sentence_20"
},
{
"operation": "finish",
"time": "64.943",
"value": "finish"
}
]
}
|
exfever_4260
|
Led Zeppelin group consisted of vocalist Robert Plant, guitarist Jimmy Page, drummer John Bonham, and the bassist/keyboardist who collaborated with musicians including an American musician best known as the founder of the rock band Queens of the Industrial Age.
|
{
"evidence": [
"Led Zeppelin",
"John Paul Jones (musician)",
"Josh Homme"
],
"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:\nLed Zeppelin group consisted of vocalist Robert Plant, guitarist Jimmy Page, drummer John Bonham, and the bassist/keyboardist who collaborated with musicians including an American musician best known as the founder of the rock band Queens of the Industrial Age.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Led Zeppelin",
"John Paul Jones (musician)",
"Josh Homme"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7926,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
86922
|
Albert Louis Deullin was a French Capitaine, serving the aviation branch of the military from 1912 through 1923, when he died at the age of 32.
|
{
"evidence": [
"Albert Louis Deullin_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:\nAlbert Louis Deullin was a French Capitaine, serving the aviation branch of the military from 1912 through 1923, when he died at the age of 32.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Albert Louis Deullin_sentence_9"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7927,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.744",
"value": "?search="
},
{
"operation": "search",
"time": "22.825",
"value": "Albert Louis Deullin"
},
{
"operation": "Now on",
"time": "24.941",
"value": "Albert Louis Deullin"
},
{
"operation": "Highlighting",
"time": "57.199",
"value": "Albert Louis Deullin_sentence_0"
},
{
"operation": "Highlighting",
"time": "84.815",
"value": "Albert Louis Deullin_sentence_9"
},
{
"operation": "Highlighting deleted",
"time": "93.562",
"value": "Albert Louis Deullin_sentence_0"
},
{
"operation": "finish",
"time": "113.012",
"value": "finish"
}
]
}
|
exfever_386
|
Fionn Whitehead's second film is Dunkirk.
|
{
"evidence": [
"Dunkirk (2017 film)",
"Mark Rylance"
],
"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:\nFionn Whitehead's second film is Dunkirk.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dunkirk (2017 film)",
"Mark Rylance"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7928,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5450
|
Lion is a large cat of the genus Panthera, native to the continent that was surrounded by seawater inlet of the Indian Ocean on its northeastern part.
|
{
"evidence": [
"Lion",
"Africa",
"Red Sea"
],
"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:\nLion is a large cat of the genus Panthera, native to the continent that was surrounded by seawater inlet of the Indian Ocean on its northeastern part.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lion",
"Africa",
"Red Sea"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7929,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2615
|
Noam Chomsky is an American political activist and a professor at a public land-grant research university governed by the governing body of Arizona's public university system.
|
{
"evidence": [
"Noam Chomsky",
"University of Arizona",
"Arizona Board of Regents"
],
"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:\nNoam Chomsky is an American political activist and a professor at a public land-grant research university governed by the governing body of Arizona's public university system.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Noam Chomsky",
"University of Arizona",
"Arizona Board of Regents"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7930,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6078
|
Azerbaijan, officially the Republic of Azerbaijan, is a transcontinental country located in the Transcaucasia, a geographical region on the border of Eastern Europe and Southeast Asia.
|
{
"evidence": [
"Azerbaijan",
"Transcaucasia"
],
"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:\nAzerbaijan, officially the Republic of Azerbaijan, is a transcontinental country located in the Transcaucasia, a geographical region on the border of Eastern Europe and Southeast Asia.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Azerbaijan",
"Transcaucasia"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7931,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
71187
|
The Victorian Centenary Grand Prix was a 230 mile motor race held on 27 October 1934 in Australia.
|
{
"evidence": [
"1934 Victorian Centenary Grand Prix_sentence_1",
"1934 Victorian Centenary Grand Prix_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 Victorian Centenary Grand Prix was a 230 mile motor race held on 27 October 1934 in Australia.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"1934 Victorian Centenary Grand Prix_sentence_1",
"1934 Victorian Centenary Grand Prix_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7932,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.424",
"value": "?search="
},
{
"operation": "search",
"time": "11.878",
"value": "Victorian Centenary Grand Prix (1934)"
},
{
"operation": "Now on",
"time": "12.985",
"value": "1934 Victorian Centenary Grand Prix"
},
{
"operation": "Highlighting",
"time": "18.388",
"value": "1934 Victorian Centenary Grand Prix_sentence_1"
},
{
"operation": "Highlighting",
"time": "19.563",
"value": "1934 Victorian Centenary Grand Prix_sentence_0"
},
{
"operation": "finish",
"time": "42.701",
"value": "finish"
}
]
}
|
48311
|
The song "Sugar" by Brockhampton ( an American hip-hop musical group) didn't make it onto the top charts in 2019.
|
{
"evidence": [
"Sugar (Brockhampton song)_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 song \"Sugar\" by Brockhampton ( an American hip-hop musical group) didn't make it onto the top charts in 2019.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Sugar (Brockhampton song)_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7933,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.991",
"value": "?search="
},
{
"operation": "search",
"time": "14.096",
"value": "Sugar Brockhampton"
},
{
"operation": "Now on",
"time": "15.25",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "18.991",
"value": "Sugar (Brockhampton song)"
},
{
"operation": "Now on",
"time": "19.651",
"value": "Sugar (Brockhampton song)"
},
{
"operation": "Highlighting",
"time": "21.387",
"value": "Sugar (Brockhampton song)_sentence_0"
},
{
"operation": "hyperlink",
"time": "32.983",
"value": "Brockhampton"
},
{
"operation": "Now on",
"time": "33.648",
"value": "Brockhampton (band)"
},
{
"operation": "back-button-clicked",
"time": "40.914",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "40.965",
"value": "Sugar (Brockhampton song)"
},
{
"operation": "hyperlink",
"time": "164.418",
"value": "Ginger"
},
{
"operation": "Now on",
"time": "165.091",
"value": "Ginger (Brockhampton album)"
},
{
"operation": "Page search",
"time": "224.366",
"value": "sugar"
},
{
"operation": "Highlighting",
"time": "238.012",
"value": "Ginger (Brockhampton album)_sentence_55"
},
{
"operation": "Highlighting",
"time": "238.012",
"value": "Ginger (Brockhampton album)_title"
},
{
"operation": "Highlighting",
"time": "238.014",
"value": "Ginger (Brockhampton album)_section_3"
},
{
"operation": "Highlighting",
"time": "238.589",
"value": "Ginger (Brockhampton album)_sentence_56"
},
{
"operation": "Highlighting deleted",
"time": "258.475",
"value": "Ginger (Brockhampton album)_sentence_56"
},
{
"operation": "Highlighting deleted",
"time": "412.049",
"value": "Ginger (Brockhampton album)_sentence_55"
},
{
"operation": "Highlighting deleted",
"time": "412.051",
"value": "Ginger (Brockhampton album)_title"
},
{
"operation": "Highlighting deleted",
"time": "412.053",
"value": "Ginger (Brockhampton album)_section_3"
},
{
"operation": "back-button-clicked",
"time": "491.14",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "491.198",
"value": "Sugar (Brockhampton song)"
},
{
"operation": "finish",
"time": "521.589",
"value": "finish"
}
]
}
|
exfever_7059
|
30 Seconds to Mars was the highest-grossing album in 2002.
|
{
"evidence": [
"30 Seconds to Mars (album)",
"Thirty Seconds to Mars"
],
"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:\n30 Seconds to Mars was the highest-grossing album in 2002.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"30 Seconds to Mars (album)",
"Thirty Seconds to Mars"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7934,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
79908
|
During the 2017-2018 season, the Liberty Lady Flames basketball team won the Big South Women's Basketball Tournament for Liberty University and proceeded to the NCAA Women's Tournament; this was the final season for the team in the Atlantic Sun Conference.
|
{
"evidence": [
"2017–18 Liberty Lady Flames basketball team_sentence_4"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDuring the 2017-2018 season, the Liberty Lady Flames basketball team won the Big South Women's Basketball Tournament for Liberty University and proceeded to the NCAA Women's Tournament; this was the final season for the team in the Atlantic Sun Conference.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"2017–18 Liberty Lady Flames basketball team_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7935,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.043",
"value": "?search="
},
{
"operation": "search",
"time": "21.87",
"value": "Contains...2017-2018 Liberty Lady Flames basketball team"
},
{
"operation": "Now on",
"time": "24.031",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "28.857",
"value": "2017–18 Liberty Lady Flames basketball team"
},
{
"operation": "Now on",
"time": "29.762",
"value": "2017–18 Liberty Lady Flames basketball team"
},
{
"operation": "Highlighting",
"time": "42.758",
"value": "2017–18 Liberty Lady Flames basketball team_sentence_0"
},
{
"operation": "Highlighting",
"time": "48.337",
"value": "2017–18 Liberty Lady Flames basketball team_sentence_3"
},
{
"operation": "Highlighting",
"time": "61.313",
"value": "2017–18 Liberty Lady Flames basketball team_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "65.697",
"value": "2017–18 Liberty Lady Flames basketball team_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "66.593",
"value": "2017–18 Liberty Lady Flames basketball team_sentence_0"
},
{
"operation": "finish",
"time": "75.249",
"value": "finish"
}
]
}
|
47595
|
Lil Uzi Vert's second studio album was released in 2020 and had no guest singer and was given a B+ by Consequence of Sound and a 7/10 from Exclaim!,
|
{
"evidence": [
"Lil Uzi Vert_sentence_9",
"Eternal Atake_sentence_4"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLil Uzi Vert's second studio album was released in 2020 and had no guest singer and was given a B+ by Consequence of Sound and a 7/10 from Exclaim!,\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Lil Uzi Vert_sentence_9",
"Eternal Atake_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7936,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.699",
"value": "?search="
},
{
"operation": "search",
"time": "13.664",
"value": "Lil Uzi Vert"
},
{
"operation": "Now on",
"time": "14.836",
"value": "Lil Uzi Vert"
},
{
"operation": "Highlighting",
"time": "32.343",
"value": "Lil Uzi Vert_sentence_9"
},
{
"operation": "hyperlink",
"time": "42.189",
"value": "Eternal Atake"
},
{
"operation": "Now on",
"time": "43.052",
"value": "Eternal Atake"
},
{
"operation": "Highlighting",
"time": "49.333",
"value": "Eternal Atake_sentence_4"
},
{
"operation": "finish",
"time": "73.372",
"value": "finish"
}
]
}
|
34501
|
Matchday two happened on July 28, 2009 against Eintracht Braunschweig (a German football and sports club based in Lower Saxony), then Matchday three happened on August 7, 2009 against Werder Bremen II.
|
{
"evidence": [
"2009–10 Borussia Dortmund II season_sentence_5",
"2009–10 Borussia Dortmund II season_sentence_7",
"Eintracht Braunschweig_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:\nMatchday two happened on July 28, 2009 against Eintracht Braunschweig (a German football and sports club based in Lower Saxony), then Matchday three happened on August 7, 2009 against Werder Bremen II.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"2009–10 Borussia Dortmund II season_sentence_5",
"2009–10 Borussia Dortmund II season_sentence_7",
"Eintracht Braunschweig_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7937,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.552",
"value": "?search="
},
{
"operation": "search",
"time": "15.3",
"value": "Matchday"
},
{
"operation": "Now on",
"time": "16.598",
"value": "Match Day"
},
{
"operation": "search",
"time": "24.967",
"value": "Matchday II"
},
{
"operation": "Now on",
"time": "25.766",
"value": "Match Day II"
},
{
"operation": "search",
"time": "42.161",
"value": "Matchday 2"
},
{
"operation": "Now on",
"time": "43.515",
"value": "Match Day II"
},
{
"operation": "Page search",
"time": "60.706",
"value": "eintr"
},
{
"operation": "search",
"time": "100.189",
"value": "Matchday two happened on July 28, 2009 against Eintracht Braunschweig"
},
{
"operation": "Now on",
"time": "102.635",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "105.545",
"value": "July and August 2009–10 Borussia Dortmund II season_section_2"
},
{
"operation": "Now on",
"time": "109.728",
"value": "2009–10 Borussia Dortmund II season#July and August 2009.E2.80.9310 Borussia Dortmund II season section 2"
},
{
"operation": "Highlighting",
"time": "119.934",
"value": "2009–10 Borussia Dortmund II season_sentence_5"
},
{
"operation": "Highlighting",
"time": "127.461",
"value": "2009–10 Borussia Dortmund II season_sentence_7"
},
{
"operation": "hyperlink",
"time": "129.157",
"value": "Eintracht Braunschweig"
},
{
"operation": "Now on",
"time": "133.143",
"value": "Eintracht Braunschweig"
},
{
"operation": "Highlighting",
"time": "136.428",
"value": "Eintracht Braunschweig_sentence_0"
},
{
"operation": "finish",
"time": "146.297",
"value": "finish"
}
]
}
|
exfever_7589
|
South America is a continent entirely in the Western Hemisphere and mostly in the Southern Hemisphere, with a relatively small portion in the half of the third planet from the Sun, that is north of the Equator.
|
{
"evidence": [
"South America",
"Northern Hemisphere",
"Earth"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSouth America is a continent entirely in the Western Hemisphere and mostly in the Southern Hemisphere, with a relatively small portion in the half of the third planet from the Sun, that is north of the Equator.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"South America",
"Northern Hemisphere",
"Earth"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7938,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8403
|
The Semitic languages are a branch of the Afroasiatic language family spoken by people in a region encompassing the northern portion of the African continent.
|
{
"evidence": [
"Semitic languages",
"North Africa"
],
"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 Semitic languages are a branch of the Afroasiatic language family spoken by people in a region encompassing the northern portion of the African continent.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Semitic languages",
"North Africa"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7939,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
33720
|
David Candy was a persona developed to mock the indie rock scene.
|
{
"evidence": [
"David Candy_sentence_3",
"David Candy_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:\nDavid Candy was a persona developed to mock the indie rock scene.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"David Candy_sentence_3",
"David Candy_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7940,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.74",
"value": "?search="
},
{
"operation": "search",
"time": "9.68",
"value": "David Candy"
},
{
"operation": "Now on",
"time": "10.994",
"value": "David Candy"
},
{
"operation": "Highlighting",
"time": "22.18",
"value": "David Candy_sentence_3"
},
{
"operation": "Highlighting",
"time": "25.767",
"value": "David Candy_sentence_4"
},
{
"operation": "finish",
"time": "75.583",
"value": "finish"
}
]
}
|
22305
|
CareSource, headquartered in Columbus, Ohio, acquired a United Healthcare plan to become the largest Medicaid HMO in Ohio and the second largest in the U.S.
|
{
"evidence": [
"CareSource_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nCareSource, headquartered in Columbus, Ohio, acquired a United Healthcare plan to become the largest Medicaid HMO in Ohio and the second largest in the U.S.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"CareSource_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7941,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.005",
"value": "?search="
},
{
"operation": "search",
"time": "12.059",
"value": "CareSource"
},
{
"operation": "Now on",
"time": "12.707",
"value": "CareSource"
},
{
"operation": "Highlighting",
"time": "22.386",
"value": "CareSource_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "26.308",
"value": "CareSource_sentence_0"
},
{
"operation": "Highlighting",
"time": "27.514",
"value": "CareSource_sentence_2"
},
{
"operation": "finish",
"time": "163.775",
"value": "finish"
}
]
}
|
exfever_6270
|
Within the Doctor Who series' narrative, the Doctor is a millennia-old ancient race of extraterrestrial people named for their command of a technology that has a concept of movement between certain points in time.
|
{
"evidence": [
"Twelfth Doctor",
"Time Lord",
"Time travel"
],
"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:\nWithin the Doctor Who series' narrative, the Doctor is a millennia-old ancient race of extraterrestrial people named for their command of a technology that has a concept of movement between certain points in time.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Twelfth Doctor",
"Time Lord",
"Time travel"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7942,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
85952
|
Ben Swift learned how to drive a motorcycle at the age of 16.
|
{
"evidence": [
"Ben Swift_sentence_3"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBen Swift learned how to drive a motorcycle at the age of 16.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ben Swift_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7943,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.778",
"value": "?search="
},
{
"operation": "search",
"time": "10.431",
"value": "Ben Swift"
},
{
"operation": "Now on",
"time": "11.42",
"value": "Ben Swift"
},
{
"operation": "Highlighting",
"time": "24.173",
"value": "Ben Swift_sentence_3"
},
{
"operation": "Highlighting",
"time": "82.438",
"value": "Ben Swift_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "88.186",
"value": "Ben Swift_sentence_0"
},
{
"operation": "Page search",
"time": "106.881",
"value": "motorcycle"
},
{
"operation": "finish",
"time": "117.57",
"value": "finish"
}
]
}
|
43423
|
Eleanor Calbes (20 February 1940- 19 April 2016) was a Filipina soprano who began her opera career with the Bayanihan Philippine National Folk Dance Company, later appearing in Broadway musicals including South Pacific and West Side Story and then was inducted into the Mississauga Music Walk of Fame in 2013
|
{
"evidence": [
"Eleanor Calbes_sentence_0",
"Eleanor Calbes_sentence_6",
"Eleanor Calbes_sentence_8",
"Eleanor Calbes_sentence_19"
],
"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:\nEleanor Calbes (20 February 1940- 19 April 2016) was a Filipina soprano who began her opera career with the Bayanihan Philippine National Folk Dance Company, later appearing in Broadway musicals including South Pacific and West Side Story and then was inducted into the Mississauga Music Walk of Fame in 2013\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Eleanor Calbes_sentence_0",
"Eleanor Calbes_sentence_6",
"Eleanor Calbes_sentence_8",
"Eleanor Calbes_sentence_19"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7944,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.591",
"value": "?search="
},
{
"operation": "search",
"time": "11.604",
"value": "Eleanor Calbes"
},
{
"operation": "Now on",
"time": "12.357",
"value": "Eleanor Calbes"
},
{
"operation": "Highlighting",
"time": "22.793",
"value": "Eleanor Calbes_sentence_0"
},
{
"operation": "Highlighting",
"time": "30.586",
"value": "Eleanor Calbes_sentence_6"
},
{
"operation": "Highlighting",
"time": "43.084",
"value": "Eleanor Calbes_sentence_8"
},
{
"operation": "Highlighting",
"time": "63.752",
"value": "Eleanor Calbes_sentence_19"
},
{
"operation": "finish",
"time": "105.687",
"value": "finish"
}
]
}
|
exfever_2817
|
Odia has also an automated translator introduced by Google.
|
{
"evidence": [
"Odia language",
"Jharkhand",
"East India"
],
"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:\nOdia has also an automated translator introduced by Google.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Odia language",
"Jharkhand",
"East India"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7945,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2642
|
Meghan McCain is an American television personality, conservative columnist, and author who is the daughter of an American politician, statesman, and United States Navy officer who served as a United States Senator for Arizona.
|
{
"evidence": [
"Meghan McCain",
"John McCain"
],
"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:\nMeghan McCain is an American television personality, conservative columnist, and author who is the daughter of an American politician, statesman, and United States Navy officer who served as a United States Senator for Arizona.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Meghan McCain",
"John McCain"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7946,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1353
|
South Park is an American animated sitcom created by Matt Stone that revolves around a group of boys together with a fictional character voiced by Trey Parker.
|
{
"evidence": [
"South Park",
"Stan Marsh"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSouth Park is an American animated sitcom created by Matt Stone that revolves around a group of boys together with a fictional character voiced by Trey Parker.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"South Park",
"Stan Marsh"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7947,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5182
|
Richard Dreyfuss is known for starring in popular films, including a 1987 Canadian buddy cop action comedy film directed by John Badham.
|
{
"evidence": [
"Richard Dreyfuss",
"Stakeout (1987 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:\nRichard Dreyfuss is known for starring in popular films, including a 1987 Canadian buddy cop action comedy film directed by John Badham.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Richard Dreyfuss",
"Stakeout (1987 film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7948,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
64335
|
It is a hamlet and civil parish in England that is 8 miles north of Pickering with an estimate of 120 inhabitants (2015).
|
{
"evidence": [
"Stape_sentence_1",
"Stape_sentence_4",
"Stape_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:\nIt is a hamlet and civil parish in England that is 8 miles north of Pickering with an estimate of 120 inhabitants (2015).\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Stape_sentence_1",
"Stape_sentence_4",
"Stape_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7949,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.928",
"value": "?search="
},
{
"operation": "search",
"time": "82.042",
"value": "\"north of Pickering\""
},
{
"operation": "Now on",
"time": "83.17",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "98.615",
"value": "Stape"
},
{
"operation": "Now on",
"time": "99.259",
"value": "Stape"
},
{
"operation": "Page search",
"time": "107.304",
"value": "8 miles"
},
{
"operation": "Highlighting",
"time": "110.607",
"value": "Stape_sentence_1"
},
{
"operation": "Highlighting",
"time": "116.995",
"value": "Stape_sentence_4"
},
{
"operation": "Highlighting",
"time": "121.731",
"value": "Stape_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "121.74",
"value": "Stape_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "121.74",
"value": "Stape_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "124.593",
"value": "Stape_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "124.601",
"value": "Stape_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "124.603",
"value": "Stape_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "125.687",
"value": "Stape_sentence_3"
},
{
"operation": "finish",
"time": "197.274",
"value": "finish"
}
]
}
|
exfever_2912
|
Sherlock is the first television series created by Mark Gatiss.
|
{
"evidence": [
"Sherlock (TV series)",
"Steven Moffat"
],
"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:\nSherlock is the first television series created by Mark Gatiss.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sherlock (TV series)",
"Steven Moffat"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7950,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
72727
|
Yousef Jabareen only studied until grade school.
|
{
"evidence": [
"Yousef Jabareen_sentence_2",
"Yousef Jabareen_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:\nYousef Jabareen only studied until grade school.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Yousef Jabareen_sentence_2",
"Yousef Jabareen_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7951,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.228",
"value": "?search="
},
{
"operation": "search",
"time": "31.658",
"value": "youdef jabareen"
},
{
"operation": "Now on",
"time": "32.67",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "38.232",
"value": "yousef jabareen"
},
{
"operation": "Now on",
"time": "46.43",
"value": "Yousef Jabareen"
},
{
"operation": "Highlighting",
"time": "52.44",
"value": "Yousef Jabareen_sentence_2"
},
{
"operation": "Highlighting",
"time": "65.749",
"value": "Yousef Jabareen_sentence_3"
},
{
"operation": "finish",
"time": "81.081",
"value": "finish"
}
]
}
|
exfever_456
|
The United States Naval Academy is a federal service academy in Annapolis, located in a state in the West-Atlantic region of the United States.
|
{
"evidence": [
"United States Naval Academy",
"Maryland"
],
"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 United States Naval Academy is a federal service academy in Annapolis, located in a state in the West-Atlantic region of the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"United States Naval Academy",
"Maryland"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7952,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
18389
|
Mint produced 4,700 gold (1.34% gold, 93% silver and 6% copper and weighs 412g), silver and bronze medals.
|
{
"evidence": [
"Mint (facility)_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:\nMint produced 4,700 gold (1.34% gold, 93% silver and 6% copper and weighs 412g), silver and bronze medals.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Mint (facility)_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7953,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.021",
"value": "?search="
},
{
"operation": "search",
"time": "26.991",
"value": "Mint produced 4,700 gold"
},
{
"operation": "Now on",
"time": "28.755",
"value": "undefined"
},
{
"operation": "search",
"time": "39.191",
"value": "mint"
},
{
"operation": "Now on",
"time": "39.958",
"value": "Mint"
},
{
"operation": "hyperlink",
"time": "211.478",
"value": "2 Coins and collectibles Mint_section_1"
},
{
"operation": "hyperlink",
"time": "214.655",
"value": "Mint (facility)"
},
{
"operation": "Now on",
"time": "215.453",
"value": "Mint (facility)"
},
{
"operation": "Highlighting",
"time": "240.662",
"value": "Mint (facility)_sentence_1"
},
{
"operation": "finish",
"time": "246.388",
"value": "finish"
}
]
}
|
39806
|
Walter Folger Brown was a successful political figure in the 1930s from Toledo, Ohio.
|
{
"evidence": [
"Walter Folger Brown_sentence_6"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWalter Folger Brown was a successful political figure in the 1930s from Toledo, Ohio.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Walter Folger Brown_sentence_6"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7954,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.685",
"value": "?search="
},
{
"operation": "search",
"time": "7.853",
"value": "Walter Folger Brown"
},
{
"operation": "Now on",
"time": "8.51",
"value": "Walter Folger Brown"
},
{
"operation": "Highlighting",
"time": "115.454",
"value": "Walter Folger Brown_sentence_3"
},
{
"operation": "Page search",
"time": "140.837",
"value": "succ"
},
{
"operation": "Highlighting",
"time": "151.359",
"value": "Walter Folger Brown_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "152.944",
"value": "Walter Folger Brown_sentence_3"
},
{
"operation": "finish",
"time": "164.887",
"value": "finish"
}
]
}
|
26696
|
The Ninth Dynasty of Egypt was part of the First Intermediate Period of Egypt, a "dark period" of about 125 years (with very little evidence surviving from it) in which there were two competing power bases; Meryibre Khety was a pharoah during this period.
|
{
"evidence": [
"Ninth Dynasty of Egypt_sentence_0",
"First Intermediate Period of Egypt_sentence_0",
"First Intermediate Period of Egypt_sentence_3",
"First Intermediate Period of Egypt_sentence_4",
"Meryibre Khety_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 Ninth Dynasty of Egypt was part of the First Intermediate Period of Egypt, a \"dark period\" of about 125 years (with very little evidence surviving from it) in which there were two competing power bases; Meryibre Khety was a pharoah during this period.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Ninth Dynasty of Egypt_sentence_0",
"First Intermediate Period of Egypt_sentence_0",
"First Intermediate Period of Egypt_sentence_3",
"First Intermediate Period of Egypt_sentence_4",
"Meryibre Khety_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7955,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.548",
"value": "?search="
},
{
"operation": "search",
"time": "46.063",
"value": "Ninth Dynasty of Egypt"
},
{
"operation": "Now on",
"time": "47.075",
"value": "Ninth Dynasty of Egypt"
},
{
"operation": "Highlighting",
"time": "236.495",
"value": "Ninth Dynasty of Egypt_sentence_0"
},
{
"operation": "hyperlink",
"time": "243.973",
"value": "First Intermediate Period"
},
{
"operation": "Now on",
"time": "245.394",
"value": "First Intermediate Period of Egypt"
},
{
"operation": "Highlighting",
"time": "250.159",
"value": "First Intermediate Period of Egypt_sentence_0"
},
{
"operation": "Highlighting",
"time": "266.458",
"value": "First Intermediate Period of Egypt_sentence_3"
},
{
"operation": "back-button-clicked",
"time": "270.244",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "270.294",
"value": "Ninth Dynasty of Egypt"
},
{
"operation": "hyperlink",
"time": "344.097",
"value": "First Intermediate Period"
},
{
"operation": "Now on",
"time": "345.376",
"value": "First Intermediate Period of Egypt"
},
{
"operation": "Highlighting",
"time": "353.128",
"value": "First Intermediate Period of Egypt_sentence_4"
},
{
"operation": "Page search",
"time": "385.09",
"value": "khety"
},
{
"operation": "search",
"time": "430.318",
"value": "merybre khety"
},
{
"operation": "Now on",
"time": "432.261",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "436.659",
"value": "Meryibre Khety"
},
{
"operation": "Now on",
"time": "437.921",
"value": "Meryibre Khety"
},
{
"operation": "Highlighting",
"time": "444.081",
"value": "Meryibre Khety_sentence_0"
},
{
"operation": "finish",
"time": "630.049",
"value": "finish"
}
]
}
|
93602
|
Portugal at the 2011 World Championships in Athletics was from April 27 to September 4 in Daegu, South Korea and had 25 athletes compete.
|
{
"evidence": [
"Portugal at the 2011 World Championships in Athletics_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:\nPortugal at the 2011 World Championships in Athletics was from April 27 to September 4 in Daegu, South Korea and had 25 athletes compete.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Portugal at the 2011 World Championships in Athletics_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7956,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.016",
"value": "?search="
},
{
"operation": "search",
"time": "23.305",
"value": "Portugal at the 2011 World Championships in Athletics"
},
{
"operation": "Now on",
"time": "24.807",
"value": "Portugal at the 2011 World Championships in Athletics"
},
{
"operation": "Highlighting",
"time": "31.364",
"value": "Portugal at the 2011 World Championships in Athletics_sentence_0"
},
{
"operation": "finish",
"time": "46.56",
"value": "finish"
}
]
}
|
85405
|
There are 30 numbered country roads in Haldimand County
|
{
"evidence": [
"Haldimand_sentence_2"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThere are 30 numbered country roads in Haldimand County\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Haldimand_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7957,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.84",
"value": "?search="
},
{
"operation": "search",
"time": "25.283",
"value": "Contains...country roads in Haldimand County"
},
{
"operation": "Now on",
"time": "27.256",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "31.486",
"value": "Haldimand"
},
{
"operation": "Now on",
"time": "32.416",
"value": "Haldimand"
},
{
"operation": "Highlighting",
"time": "80.851",
"value": "Haldimand_sentence_2"
},
{
"operation": "finish",
"time": "106.44",
"value": "finish"
}
]
}
|
90713
|
Nucleoporin 153, a protein which in reptiles is encoded by the NUP153 gene, is an essential component of the basket of nuclear pore complexes (NPCs) in vertebrates, and required for the anchoring of NPCs.
|
{
"evidence": [
"Nucleoporin 153_sentence_0",
"Nucleoporin 153_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:\nNucleoporin 153, a protein which in reptiles is encoded by the NUP153 gene, is an essential component of the basket of nuclear pore complexes (NPCs) in vertebrates, and required for the anchoring of NPCs.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Nucleoporin 153_sentence_0",
"Nucleoporin 153_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7958,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.522",
"value": "?search="
},
{
"operation": "search",
"time": "14.938",
"value": "Nucleoporin 153"
},
{
"operation": "Now on",
"time": "15.953",
"value": "Nucleoporin 153"
},
{
"operation": "Highlighting",
"time": "38.614",
"value": "Nucleoporin 153_sentence_0"
},
{
"operation": "Highlighting",
"time": "54.235",
"value": "Nucleoporin 153_sentence_1"
},
{
"operation": "search",
"time": "74.057",
"value": "reptiles nucleoporin"
},
{
"operation": "Now on",
"time": "75.098",
"value": "undefined"
},
{
"operation": "search",
"time": "87.019",
"value": "reptiles nup153"
},
{
"operation": "Now on",
"time": "87.794",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "93.33",
"value": "Vanuatu"
},
{
"operation": "Now on",
"time": "94.556",
"value": "Vanuatu"
},
{
"operation": "back-button-clicked",
"time": "113.474",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "114.135",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "114.862",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "114.869",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "115.121",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "116.146",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "116.153",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "116.41",
"value": "Nucleoporin 153"
},
{
"operation": "finish",
"time": "150.701",
"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": 7959,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
88798
|
Manduul Khan was married to the daughter of Beg-Arslan, a warlord, and the husband and wife disliked one another.
|
{
"evidence": [
"Manduul Khan_sentence_2",
"Manduul Khan_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:\nManduul Khan was married to the daughter of Beg-Arslan, a warlord, and the husband and wife disliked one another.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Manduul Khan_sentence_2",
"Manduul Khan_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7960,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.944",
"value": "?search="
},
{
"operation": "search",
"time": "15.518",
"value": "Manduul Khan"
},
{
"operation": "Now on",
"time": "16.487",
"value": "Manduul Khan"
},
{
"operation": "Highlighting",
"time": "22.574",
"value": "Manduul Khan_sentence_2"
},
{
"operation": "Highlighting",
"time": "34.013",
"value": "Manduul Khan_sentence_3"
},
{
"operation": "finish",
"time": "43.516",
"value": "finish"
}
]
}
|
70177
|
The Chaldean Talmud relates the following incident, probably based on an actual occurrence and the emperor (a Christian ruler no doubt being meant) said to Tanhuma, "Let us all become one people."
|
{
"evidence": [
"Tanhuma bar Abba_sentence_5",
"Tanhuma bar Abba_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 Chaldean Talmud relates the following incident, probably based on an actual occurrence and the emperor (a Christian ruler no doubt being meant) said to Tanhuma, \"Let us all become one people.\"\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Tanhuma bar Abba_sentence_5",
"Tanhuma bar Abba_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7961,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.449",
"value": "?search="
},
{
"operation": "search",
"time": "51.83",
"value": "Contains...Chaldean Talmud"
},
{
"operation": "Now on",
"time": "54.036",
"value": "undefined"
},
{
"operation": "search",
"time": "84.698",
"value": "\"Let us all become one people.\""
},
{
"operation": "Now on",
"time": "87.376",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "92.621",
"value": "Tanhuma bar Abba"
},
{
"operation": "Now on",
"time": "101.387",
"value": "Tanhuma bar Abba"
},
{
"operation": "Highlighting",
"time": "117.221",
"value": "Tanhuma bar Abba_sentence_5"
},
{
"operation": "Highlighting",
"time": "119.611",
"value": "Tanhuma bar Abba_sentence_6"
},
{
"operation": "finish",
"time": "214.662",
"value": "finish"
}
]
}
|
exfever_6994
|
Tiffany Hwang is the first artist to sign with Transparent Arts.
|
{
"evidence": [
"Tiffany Hwang",
"I Just Wanna Dance",
"S.M. Entertainment"
],
"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:\nTiffany Hwang is the first artist to sign with Transparent Arts.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tiffany Hwang",
"I Just Wanna Dance",
"S.M. Entertainment"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7962,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7061
|
Ontario is Canada's third-largest jurisdiction in total area when the territories of the Nunavut and the federal territory of Canada that has a land area of approximately 1,144,000 km2 are included.
|
{
"evidence": [
"Ontario",
"Northwest Territories"
],
"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:\nOntario is Canada's third-largest jurisdiction in total area when the territories of the Nunavut and the federal territory of Canada that has a land area of approximately 1,144,000 km2 are included.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ontario",
"Northwest Territories"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7963,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2521
|
John Joseph Nicholson is a retired American dancer best known for his work in films, including a 1975 American psychological comedy-drama film featuring a supporting cast of an American actress, best known for her portrayal of Nurse Ratched in One Flew Over the Cuckoo's Nest.
|
{
"evidence": [
"Jack Nicholson",
"One Flew Over the Cuckoo's Nest (film)",
"Louise Fletcher"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJohn Joseph Nicholson is a retired American dancer best known for his work in films, including a 1975 American psychological comedy-drama film featuring a supporting cast of an American actress, best known for her portrayal of Nurse Ratched in One Flew Over the Cuckoo's Nest.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jack Nicholson",
"One Flew Over the Cuckoo's Nest (film)",
"Louise Fletcher"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7964,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
88791
|
The Alfred M. Landon Lecture Series is a series of speeches on current public affairs, named after Kansas politician Alf Landon (who became an independent oil producer in Lawrence, Kansas), former Governor of Kansas and Republican presidential candidate.
|
{
"evidence": [
"Alf Landon_sentence_18"
],
"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 Alfred M. Landon Lecture Series is a series of speeches on current public affairs, named after Kansas politician Alf Landon (who became an independent oil producer in Lawrence, Kansas), former Governor of Kansas and Republican presidential candidate.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Alf Landon_sentence_18"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7965,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "3.913",
"value": "?search="
},
{
"operation": "search",
"time": "13.254",
"value": "Alfred M. Landon Lecture Series"
},
{
"operation": "Now on",
"time": "14.297",
"value": "Landon Lecture Series"
},
{
"operation": "Highlighting",
"time": "18.369",
"value": "Landon Lecture Series_sentence_0"
},
{
"operation": "Highlighting",
"time": "18.371",
"value": "Landon Lecture Series_title"
},
{
"operation": "Highlighting",
"time": "24.8",
"value": "Landon Lecture Series_sentence_1"
},
{
"operation": "hyperlink",
"time": "31.16",
"value": "Alf Landon"
},
{
"operation": "Now on",
"time": "31.992",
"value": "Alf Landon"
},
{
"operation": "Highlighting",
"time": "49.168",
"value": "Alf Landon_sentence_18"
},
{
"operation": "finish",
"time": "67.694",
"value": "finish"
}
]
}
|
76328
|
Pegolettia is a genus of eastern plants in the Inuleae tribe, within the Asteraceae family, under the Asterales order.
|
{
"evidence": [
"Pegolettia_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:\nPegolettia is a genus of eastern plants in the Inuleae tribe, within the Asteraceae family, under the Asterales order.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Pegolettia_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7966,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.7",
"value": "?search="
},
{
"operation": "search",
"time": "13.388",
"value": "Pegolettia"
},
{
"operation": "Now on",
"time": "14.071",
"value": "Pegolettia"
},
{
"operation": "Highlighting",
"time": "34.653",
"value": "Pegolettia_sentence_0"
},
{
"operation": "hyperlink",
"time": "38.158",
"value": "African"
},
{
"operation": "Now on",
"time": "40.277",
"value": "Africa"
},
{
"operation": "back-button-clicked",
"time": "58.941",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "59.124",
"value": "Pegolettia"
},
{
"operation": "Highlighting",
"time": "64.161",
"value": "Pegolettia_cell_0_8_0"
},
{
"operation": "Highlighting",
"time": "64.166",
"value": "Pegolettia_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "64.166",
"value": "Pegolettia_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "65.771",
"value": "Pegolettia_cell_0_8_1"
},
{
"operation": "Highlighting",
"time": "67.236",
"value": "Pegolettia_cell_0_7_0"
},
{
"operation": "Highlighting",
"time": "68.623",
"value": "Pegolettia_cell_0_7_1"
},
{
"operation": "Highlighting",
"time": "70.188",
"value": "Pegolettia_cell_0_6_0"
},
{
"operation": "Highlighting",
"time": "71.392",
"value": "Pegolettia_cell_0_6_1"
},
{
"operation": "Highlighting deleted",
"time": "87.164",
"value": "Pegolettia_cell_0_6_0"
},
{
"operation": "Highlighting deleted",
"time": "88.17",
"value": "Pegolettia_cell_0_7_0"
},
{
"operation": "Highlighting deleted",
"time": "89.465",
"value": "Pegolettia_cell_0_8_0"
},
{
"operation": "Highlighting deleted",
"time": "90.726",
"value": "Pegolettia_cell_0_6_1"
},
{
"operation": "Highlighting deleted",
"time": "92.84",
"value": "Pegolettia_cell_0_8_1"
},
{
"operation": "Highlighting deleted",
"time": "94.071",
"value": "Pegolettia_cell_0_7_1"
},
{
"operation": "Highlighting deleted",
"time": "94.075",
"value": "Pegolettia_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "94.076",
"value": "Pegolettia_header_cell_0_0_0"
},
{
"operation": "finish",
"time": "103.17",
"value": "finish"
}
]
}
|
exfever_2551
|
e Chernobyl disaster was a nuclear accident that occurred on 26 April 1986 at the No. 4 reactor in the Chernobyl Nuclear Power Plant, with initial emergency response, together with the process of removing contaminants on an object or area achieved by a chemical substance or compound used to inactivate or destroy microorganisms on inert surfaces.
|
{
"evidence": [
"Chernobyl disaster",
"Decontamination",
"Disinfectant"
],
"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:\ne Chernobyl disaster was a nuclear accident that occurred on 26 April 1986 at the No. 4 reactor in the Chernobyl Nuclear Power Plant, with initial emergency response, together with the process of removing contaminants on an object or area achieved by a chemical substance or compound used to inactivate or destroy microorganisms on inert surfaces.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Chernobyl disaster",
"Decontamination",
"Disinfectant"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7967,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
25883
|
ICAM3 transmembrane glycoproteins proteins bind to leukocyte adhesion adhesion LFA-1 proteins, but have no involvement in the initiation of immune response.
|
{
"evidence": [
"ICAM3_sentence_2",
"ICAM3_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:\nICAM3 transmembrane glycoproteins proteins bind to leukocyte adhesion adhesion LFA-1 proteins, but have no involvement in the initiation of immune response.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"ICAM3_sentence_2",
"ICAM3_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7968,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.107",
"value": "?search="
},
{
"operation": "search",
"time": "27.611",
"value": "ICAM3"
},
{
"operation": "Now on",
"time": "28.445",
"value": "ICAM3"
},
{
"operation": "Highlighting",
"time": "31.624",
"value": "ICAM3_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "35.247",
"value": "ICAM3_sentence_0"
},
{
"operation": "Highlighting",
"time": "40.263",
"value": "ICAM3_sentence_2"
},
{
"operation": "Highlighting",
"time": "56.247",
"value": "ICAM3_sentence_3"
},
{
"operation": "finish",
"time": "64.495",
"value": "finish"
}
]
}
|
38566
|
The MAP2K2 functions by a protein with dual specificity protein kinase which plays a role in mitogen growth factor signal transduction.
|
{
"evidence": [
"MAP2K2_sentence_3",
"MAP2K2_sentence_4"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe MAP2K2 functions by a protein with dual specificity protein kinase which plays a role in mitogen growth factor signal transduction.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"MAP2K2_sentence_3",
"MAP2K2_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7969,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.959",
"value": "?search="
},
{
"operation": "search",
"time": "8.792",
"value": "MAP2K2"
},
{
"operation": "Now on",
"time": "9.611",
"value": "MAP2K2"
},
{
"operation": "Highlighting",
"time": "24.175",
"value": "MAP2K2_sentence_3"
},
{
"operation": "Highlighting",
"time": "30.195",
"value": "MAP2K2_sentence_4"
},
{
"operation": "finish",
"time": "47.091",
"value": "finish"
}
]
}
|
51994
|
Julio César Tobar is 50-years-old today.
|
{
"evidence": [
"Julio César Tobar_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:\nJulio César Tobar is 50-years-old today.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Julio César Tobar_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7970,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.187",
"value": "?search="
},
{
"operation": "search",
"time": "9.02",
"value": "Julio César Tobar"
},
{
"operation": "Now on",
"time": "10.438",
"value": "Julio César Tobar"
},
{
"operation": "Highlighting",
"time": "17.69",
"value": "Julio César Tobar_sentence_0"
},
{
"operation": "finish",
"time": "47.759",
"value": "finish"
}
]
}
|
22999
|
Cullingworth is a village and civil parish 7 miles west of Bradford and 3 miles south of Keighley, 2 miles to the east of Haworth and is situated near the head of Harden valley in the town of Bradford, West Yorkshire, England.
|
{
"evidence": [
"Cullingworth_sentence_0",
"Cullingworth_sentence_1",
"Haworth_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:\nCullingworth is a village and civil parish 7 miles west of Bradford and 3 miles south of Keighley, 2 miles to the east of Haworth and is situated near the head of Harden valley in the town of Bradford, West Yorkshire, England.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Cullingworth_sentence_0",
"Cullingworth_sentence_1",
"Haworth_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7971,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.094",
"value": "?search="
},
{
"operation": "search",
"time": "8.19",
"value": "Cullingworth"
},
{
"operation": "Now on",
"time": "9.399",
"value": "Cullingworth"
},
{
"operation": "Highlighting",
"time": "13.511",
"value": "Cullingworth_sentence_0"
},
{
"operation": "Highlighting",
"time": "25.122",
"value": "Cullingworth_sentence_1"
},
{
"operation": "Page search",
"time": "57.953",
"value": "Haworth"
},
{
"operation": "hyperlink",
"time": "92.031",
"value": "South Pennines"
},
{
"operation": "Now on",
"time": "92.82",
"value": "South Pennines"
},
{
"operation": "back-button-clicked",
"time": "101.288",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "101.388",
"value": "Cullingworth"
},
{
"operation": "search",
"time": "108.034",
"value": "haworth"
},
{
"operation": "Now on",
"time": "109.137",
"value": "Haworth"
},
{
"operation": "Highlighting",
"time": "156.47",
"value": "Haworth_sentence_1"
},
{
"operation": "finish",
"time": "188.538",
"value": "finish"
}
]
}
|
27167
|
Produced and directed by Marc Gracie, Surprise Surprise Gotcha pulls practical jokes on Australian celebrities and the first episode aired on September 18, 2007.
|
{
"evidence": [
"Surprise Surprise Gotcha_sentence_0",
"Surprise Surprise Gotcha_sentence_3",
"Surprise Surprise Gotcha_sentence_5",
"Surprise Surprise Gotcha_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:\nProduced and directed by Marc Gracie, Surprise Surprise Gotcha pulls practical jokes on Australian celebrities and the first episode aired on September 18, 2007.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Surprise Surprise Gotcha_sentence_0",
"Surprise Surprise Gotcha_sentence_3",
"Surprise Surprise Gotcha_sentence_5",
"Surprise Surprise Gotcha_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7972,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.001",
"value": "?search="
},
{
"operation": "search",
"time": "0.019",
"value": "Surprise Surprise Gotcha"
},
{
"operation": "Now on",
"time": "0.021",
"value": "Surprise Surprise Gotcha"
},
{
"operation": "Highlighting",
"time": "0.046",
"value": "Surprise Surprise Gotcha_sentence_3"
},
{
"operation": "Highlighting",
"time": "0.059",
"value": "Surprise Surprise Gotcha_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "0.067",
"value": "Surprise Surprise Gotcha_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "0.068",
"value": "Surprise Surprise Gotcha_sentence_3"
},
{
"operation": "Highlighting",
"time": "0.069",
"value": "Surprise Surprise Gotcha_sentence_0"
},
{
"operation": "Highlighting",
"time": "0.074",
"value": "Surprise Surprise Gotcha_sentence_3"
},
{
"operation": "Highlighting",
"time": "0.074",
"value": "Surprise Surprise Gotcha_sentence_5"
},
{
"operation": "Page search",
"time": "0.088",
"value": "episode"
},
{
"operation": "Page search",
"time": "0.09",
"value": "episode"
},
{
"operation": "Highlighting",
"time": "0.111",
"value": "Surprise Surprise Gotcha_sentence_1"
},
{
"operation": "finish",
"time": "0.145",
"value": "finish"
},
{
"operation": "Now on",
"time": "1620140077.474",
"value": "Surprise Surprise Gotcha"
},
{
"operation": "finish",
"time": "1620140085.375",
"value": "finish"
}
]
}
|
76312
|
The enterprise modelling and engineering community have also developed architecture description languages catered for at the enterprise level like ArchiMate, DEMO (an Information systems researcher), and ABACUS.
|
{
"evidence": [
"Architecture description language_sentence_6",
"Architecture description language_sentence_7"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe enterprise modelling and engineering community have also developed architecture description languages catered for at the enterprise level like ArchiMate, DEMO (an Information systems researcher), and ABACUS.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Architecture description language_sentence_6",
"Architecture description language_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7973,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.752",
"value": "?search="
},
{
"operation": "search",
"time": "56.775",
"value": "ABACUS"
},
{
"operation": "Now on",
"time": "57.733",
"value": "Abacus"
},
{
"operation": "search",
"time": "110.922",
"value": "ArchiMate"
},
{
"operation": "Now on",
"time": "111.795",
"value": "ArchiMate"
},
{
"operation": "search",
"time": "167.486",
"value": "Architecture description language"
},
{
"operation": "Now on",
"time": "168.384",
"value": "Architecture description language"
},
{
"operation": "Highlighting",
"time": "172.462",
"value": "Architecture description language_sentence_6"
},
{
"operation": "Highlighting",
"time": "174.029",
"value": "Architecture description language_sentence_7"
},
{
"operation": "finish",
"time": "179.239",
"value": "finish"
}
]
}
|
43671
|
Linear grammar indicates that all linear languages are context-free and that regular languages are its proper subset.
|
{
"evidence": [
"Linear grammar_sentence_11",
"Linear grammar_sentence_12"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLinear grammar indicates that all linear languages are context-free and that regular languages are its proper subset.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Linear grammar_sentence_11",
"Linear grammar_sentence_12"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7974,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.007",
"value": "?search="
},
{
"operation": "search",
"time": "9.782",
"value": "Linear grammar"
},
{
"operation": "Now on",
"time": "11.188",
"value": "Linear grammar"
},
{
"operation": "Highlighting",
"time": "20.272",
"value": "Linear grammar_sentence_0"
},
{
"operation": "Highlighting",
"time": "33.04",
"value": "Linear grammar_sentence_1"
},
{
"operation": "Highlighting",
"time": "53.364",
"value": "Linear grammar_sentence_10"
},
{
"operation": "Highlighting",
"time": "57.522",
"value": "Linear grammar_sentence_11"
},
{
"operation": "Highlighting deleted",
"time": "58.316",
"value": "Linear grammar_sentence_10"
},
{
"operation": "Highlighting",
"time": "74.168",
"value": "Linear grammar_sentence_12"
},
{
"operation": "finish",
"time": "90.301",
"value": "finish"
}
]
}
|
77219
|
Freddie Kearns would combine his football career with his work as a security guard; then he moved to Kent League club Margate in 1957.
|
{
"evidence": [
"Freddie Kearns_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:\nFreddie Kearns would combine his football career with his work as a security guard; then he moved to Kent League club Margate in 1957.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Freddie Kearns_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7975,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.977",
"value": "undefined"
},
{
"operation": "search",
"time": "31.557",
"value": "Freddie Kearns"
},
{
"operation": "Now on",
"time": "32.567",
"value": "Freddie Kearns"
},
{
"operation": "Page search",
"time": "51.827",
"value": "guard"
},
{
"operation": "Page search",
"time": "52.964",
"value": "guard"
},
{
"operation": "Page search",
"time": "73.145",
"value": "sec"
},
{
"operation": "Highlighting",
"time": "95.697",
"value": "Freddie Kearns_sentence_7"
},
{
"operation": "finish",
"time": "103.975",
"value": "finish"
}
]
}
|
48318
|
"Be Okay" was made available for purchase on Spotify on September 19, 2013, then made it onto Billboard's Adult Top 40 chart where it reached number 25.
|
{
"evidence": [
"Be Okay_sentence_14",
"Be Okay_sentence_12"
],
"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:\n\"Be Okay\" was made available for purchase on Spotify on September 19, 2013, then made it onto Billboard's Adult Top 40 chart where it reached number 25.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Be Okay_sentence_14",
"Be Okay_sentence_12"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7976,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.249",
"value": "?search="
},
{
"operation": "search",
"time": "18.016",
"value": "Be Okay"
},
{
"operation": "Now on",
"time": "19.317",
"value": "Be Okay"
},
{
"operation": "Page search",
"time": "33.929",
"value": "spoti"
},
{
"operation": "Page search",
"time": "80.047",
"value": "2013"
},
{
"operation": "Highlighting",
"time": "105.46",
"value": "Be Okay_sentence_14"
},
{
"operation": "Highlighting",
"time": "120.527",
"value": "Be Okay_sentence_12"
},
{
"operation": "search",
"time": "147.211",
"value": "\"Be Okay\" \"spotify\""
},
{
"operation": "Now on",
"time": "148.592",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "159.626",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "159.68",
"value": "Be Okay"
},
{
"operation": "search",
"time": "182.627",
"value": "\"Be Okay\" \"collins\" \"spotify\""
},
{
"operation": "Now on",
"time": "183.442",
"value": "undefined"
},
{
"operation": "finish",
"time": "190.337",
"value": "finish"
}
]
}
|
20569
|
In March 2017, three adolescents accused Arunabh Kumar of sexual harassment and he was booked by police for molestation on March 29, 2017.
|
{
"evidence": [
"Arunabh Kumar_sentence_8",
"Arunabh Kumar_sentence_10",
"Arunabh Kumar_sentence_12",
"Arunabh Kumar_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:\nIn March 2017, three adolescents accused Arunabh Kumar of sexual harassment and he was booked by police for molestation on March 29, 2017.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Arunabh Kumar_sentence_8",
"Arunabh Kumar_sentence_10",
"Arunabh Kumar_sentence_12",
"Arunabh Kumar_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7977,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.642",
"value": "?search="
},
{
"operation": "search",
"time": "90.066",
"value": "Arunabh Kumar"
},
{
"operation": "Now on",
"time": "95.144",
"value": "?search="
},
{
"operation": "search",
"time": "98.327",
"value": "Arunabh Kumar"
},
{
"operation": "Now on",
"time": "98.763",
"value": "Arunabh Kumar"
},
{
"operation": "Page search",
"time": "199.986",
"value": "adolescents"
},
{
"operation": "Highlighting",
"time": "279.687",
"value": "Arunabh Kumar_sentence_8"
},
{
"operation": "Highlighting",
"time": "283.841",
"value": "Arunabh Kumar_sentence_9"
},
{
"operation": "Highlighting",
"time": "289.708",
"value": "Arunabh Kumar_sentence_10"
},
{
"operation": "Highlighting",
"time": "300.102",
"value": "Arunabh Kumar_sentence_11"
},
{
"operation": "Highlighting",
"time": "302.337",
"value": "Arunabh Kumar_sentence_12"
},
{
"operation": "Highlighting",
"time": "324.148",
"value": "Arunabh Kumar_sentence_1"
},
{
"operation": "hyperlink",
"time": "349.809",
"value": "The Viral Fever"
},
{
"operation": "Now on",
"time": "350.535",
"value": "The Viral Fever"
},
{
"operation": "Page search",
"time": "355.669",
"value": "adolescents"
},
{
"operation": "back-button-clicked",
"time": "392.801",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "392.848",
"value": "Arunabh Kumar"
},
{
"operation": "Highlighting deleted",
"time": "416.318",
"value": "Arunabh Kumar_sentence_9"
},
{
"operation": "Highlighting deleted",
"time": "425.667",
"value": "Arunabh Kumar_sentence_11"
},
{
"operation": "finish",
"time": "437.145",
"value": "finish"
}
]
}
|
17653
|
Mohd Amirul Izwan Yaacob was awarded his FIFA badge in 2012 and refereed some matches in AFC Champions League and also AFC Cup.
|
{
"evidence": [
"Mohd Amirul Izwan Yaacob_sentence_2",
"Mohd Amirul Izwan Yaacob_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:\nMohd Amirul Izwan Yaacob was awarded his FIFA badge in 2012 and refereed some matches in AFC Champions League and also AFC Cup.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Mohd Amirul Izwan Yaacob_sentence_2",
"Mohd Amirul Izwan Yaacob_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7978,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.614",
"value": "?search="
},
{
"operation": "search",
"time": "7.265",
"value": "Mohd Amirul Izwan Yaacob"
},
{
"operation": "Now on",
"time": "8.031",
"value": "Mohd Amirul Izwan Yaacob"
},
{
"operation": "Highlighting",
"time": "13.439",
"value": "Mohd Amirul Izwan Yaacob_sentence_2"
},
{
"operation": "Highlighting",
"time": "16.526",
"value": "Mohd Amirul Izwan Yaacob_sentence_3"
},
{
"operation": "finish",
"time": "23.926",
"value": "finish"
}
]
}
|
65444
|
Stuart attended Lincoln Memorial University, located in Harrogate, Tennessee; then after graduating, he taught at Warnock High School in Greenup, Kentucky, his home area.
|
{
"evidence": [
"Jesse Stuart_sentence_7",
"Jesse Stuart_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:\nStuart attended Lincoln Memorial University, located in Harrogate, Tennessee; then after graduating, he taught at Warnock High School in Greenup, Kentucky, his home area.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Jesse Stuart_sentence_7",
"Jesse Stuart_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 7979,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.888",
"value": "?search="
},
{
"operation": "search",
"time": "11.899",
"value": "Contains...Stuart attended Lincoln Memorial University"
},
{
"operation": "Now on",
"time": "16.286",
"value": "undefined"
},
{
"operation": "search",
"time": "31.9",
"value": "Contains...taught at Warnock High School in Greenup, Kentucky,"
},
{
"operation": "Now on",
"time": "32.903",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "35.233",
"value": "Jesse Stuart"
},
{
"operation": "Now on",
"time": "39.355",
"value": "Jesse Stuart"
},
{
"operation": "Highlighting",
"time": "47.209",
"value": "Jesse Stuart_sentence_7"
},
{
"operation": "Highlighting",
"time": "51.632",
"value": "Jesse Stuart_sentence_8"
},
{
"operation": "finish",
"time": "61.74",
"value": "finish"
}
]
}
|
62704
|
Raymond Bass was a general in the US army.
|
{
"evidence": [
"Raymond Bass_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:\nRaymond Bass was a general in the US army.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Raymond Bass_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7980,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.807",
"value": "?search="
},
{
"operation": "search",
"time": "52.542",
"value": "Raymond Bass"
},
{
"operation": "Now on",
"time": "53.345",
"value": "Raymond Bass"
},
{
"operation": "Highlighting",
"time": "59.541",
"value": "Raymond Bass_sentence_0"
},
{
"operation": "finish",
"time": "128.175",
"value": "finish"
}
]
}
|
exfever_4601
|
Adrienne Bailon was the first American actress from Disney to meet the Queen of England.
|
{
"evidence": [
"Adrienne Bailon",
"The Cheetah Girls (group)",
"Sabrina Bryan"
],
"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:\nAdrienne Bailon was the first American actress from Disney to meet the Queen of England.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Adrienne Bailon",
"The Cheetah Girls (group)",
"Sabrina Bryan"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 7981,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7510
|
Shia LaBeouf is an American actor who won an American accolade to recognized excellence in American nighttime television programming.
|
{
"evidence": [
"Shia LaBeouf",
"Daytime Emmy Award"
],
"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:\nShia LaBeouf is an American actor who won an American accolade to recognized excellence in American nighttime television programming.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Shia LaBeouf",
"Daytime Emmy Award"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7982,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5725
|
The Northeast megalopolis is the most populous in the United States, stretching primarily from the northern Boston suburbs to the suburbs of the widespread region radiating westward and southward from Santa Marta.
|
{
"evidence": [
"Northeast megalopolis",
"Northern Virginia"
],
"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 Northeast megalopolis is the most populous in the United States, stretching primarily from the northern Boston suburbs to the suburbs of the widespread region radiating westward and southward from Santa Marta.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Northeast megalopolis",
"Northern Virginia"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 7983,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.