id
stringlengths 1
12
| question
stringlengths 22
566
| golden_answers
dict | data_source
stringclasses 2
values | prompt
listlengths 1
1
| ability
stringclasses 6
values | reward_model
dict | extra_info
dict | metadata
dict |
|---|---|---|---|---|---|---|---|---|
exfever_1869
|
Michael Giacchino is known for his collaborations with an American screenwriter who wrote the screenplay for a 2013 American action horror film directed by Marc Forster.
|
{
"evidence": [
"Michael Giacchino",
"Drew Goddard",
"World War Z (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:\nMichael Giacchino is known for his collaborations with an American screenwriter who wrote the screenplay for a 2013 American action horror film directed by Marc Forster.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Michael Giacchino",
"Drew Goddard",
"World War Z (film)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1284,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2332
|
Stephen Rider is an Asian film and television actor appearing in the ABC drama television series that ran from September 24, to October 1, 2013.
|
{
"evidence": [
"Stephen Rider",
"Lucky 7 (TV series)"
],
"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:\nStephen Rider is an Asian film and television actor appearing in the ABC drama television series that ran from September 24, to October 1, 2013.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Stephen Rider",
"Lucky 7 (TV series)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1285,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4955
|
Jack Black is an American actor, comedian, musician, and songwriter known for his role in an American comedy film starring Gwyneth Paltrow.
|
{
"evidence": [
"Jack Black",
"Shallow Hal"
],
"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:\nJack Black is an American actor, comedian, musician, and songwriter known for his role in an American comedy film starring Gwyneth Paltrow.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jack Black",
"Shallow Hal"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1286,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2163
|
Kafka directed by Steven Soderbergh is a mystery film whose genre is defined by the moods they elicit, giving viewers heightened feelings of suspense.
|
{
"evidence": [
"Kafka (film)",
"Thriller (genre)"
],
"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:\nKafka directed by Steven Soderbergh is a mystery film whose genre is defined by the moods they elicit, giving viewers heightened feelings of suspense.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Kafka (film)",
"Thriller (genre)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1287,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
55568
|
Carlos Bueno's Uruguay National Team debut was on July 16, 2003, which ended in a two-two draw and he played again the following year, scoring three goals in four matches.
|
{
"evidence": [
"Carlos Bueno_sentence_21",
"Carlos Bueno_sentence_22"
],
"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:\nCarlos Bueno's Uruguay National Team debut was on July 16, 2003, which ended in a two-two draw and he played again the following year, scoring three goals in four matches.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Carlos Bueno_sentence_21",
"Carlos Bueno_sentence_22"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1288,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.282",
"value": "?search="
},
{
"operation": "search",
"time": "8.002",
"value": "Carlos Bueno"
},
{
"operation": "Now on",
"time": "9.469",
"value": "Carlos Bueno"
},
{
"operation": "Highlighting",
"time": "19.381",
"value": "Carlos Bueno_sentence_21"
},
{
"operation": "Highlighting",
"time": "27.166",
"value": "Carlos Bueno_sentence_22"
},
{
"operation": "finish",
"time": "33.202",
"value": "finish"
}
]
}
|
76768
|
Spencerport Methodist Church is a Romanesque style building built between 1870 and 1871 and was added to the NHRP on 2009.
|
{
"evidence": [
"Spencerport Methodist Church_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:\nSpencerport Methodist Church is a Romanesque style building built between 1870 and 1871 and was added to the NHRP on 2009.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Spencerport Methodist Church_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1289,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.241",
"value": "?search="
},
{
"operation": "search",
"time": "10.384",
"value": "Spencerport Methodist Church"
},
{
"operation": "Now on",
"time": "11.127",
"value": "Spencerport Methodist Church"
},
{
"operation": "Highlighting",
"time": "17.903",
"value": "Spencerport Methodist Church_sentence_1"
},
{
"operation": "Highlighting",
"time": "21.951",
"value": "Spencerport Methodist Church_sentence_2"
},
{
"operation": "Highlighting",
"time": "24.879",
"value": "Spencerport Methodist Church_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "26.463",
"value": "Spencerport Methodist Church_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "27.223",
"value": "Spencerport Methodist Church_sentence_1"
},
{
"operation": "finish",
"time": "34.91",
"value": "finish"
}
]
}
|
34472
|
The Kinase insert domain receptor, a VEGF receptor, cannot be encoded by any human genes.
|
{
"evidence": [
"Kinase insert domain receptor_sentence_0",
"Kinase insert domain receptor_sentence_1"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Kinase insert domain receptor, a VEGF receptor, cannot be encoded by any human genes.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Kinase insert domain receptor_sentence_0",
"Kinase insert domain receptor_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1290,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.43",
"value": "?search="
},
{
"operation": "search",
"time": "20.931",
"value": "Kinase insert domain receptor"
},
{
"operation": "Now on",
"time": "22.158",
"value": "Kinase insert domain receptor"
},
{
"operation": "Highlighting",
"time": "35.866",
"value": "Kinase insert domain receptor_sentence_0"
},
{
"operation": "Highlighting",
"time": "44.399",
"value": "Kinase insert domain receptor_sentence_1"
},
{
"operation": "finish",
"time": "57.089",
"value": "finish"
}
]
}
|
exfever_4889
|
Paul Bettany is an English actor known for portraying Vision and voicing J.A.R.V.I.S. in films of an American media franchise and shared universe centered on a series of superhero films produced by Marvel Studios.
|
{
"evidence": [
"Paul Bettany",
"Marvel Cinematic Universe"
],
"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:\nPaul Bettany is an English actor known for portraying Vision and voicing J.A.R.V.I.S. in films of an American media franchise and shared universe centered on a series of superhero films produced by Marvel Studios.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Paul Bettany",
"Marvel Cinematic Universe"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1291,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
52487
|
Ido Nehoshtan served as the 153 Squadron deputy commander as well as the 140 Squadron.
|
{
"evidence": [
"Ido Nehoshtan_sentence_6",
"Ido Nehoshtan_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:\nIdo Nehoshtan served as the 153 Squadron deputy commander as well as the 140 Squadron.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ido Nehoshtan_sentence_6",
"Ido Nehoshtan_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1292,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.832",
"value": "?search="
},
{
"operation": "search",
"time": "11.115",
"value": "Ido Nehoshtan"
},
{
"operation": "Now on",
"time": "11.657",
"value": "Ido Nehoshtan"
},
{
"operation": "Page search",
"time": "15.857",
"value": "153"
},
{
"operation": "search",
"time": "32.482",
"value": "153 Sqn"
},
{
"operation": "Now on",
"time": "33.718",
"value": "No. 153 Squadron RAF"
},
{
"operation": "Page search",
"time": "49.313",
"value": "command"
},
{
"operation": "back-button-clicked",
"time": "90.273",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "90.338",
"value": "Ido Nehoshtan"
},
{
"operation": "Highlighting",
"time": "104.059",
"value": "Ido Nehoshtan_sentence_0"
},
{
"operation": "Highlighting",
"time": "113.914",
"value": "Ido Nehoshtan_sentence_1"
},
{
"operation": "hyperlink",
"time": "124.809",
"value": "140 Squadron"
},
{
"operation": "Now on",
"time": "125.476",
"value": "140 Squadron (Israel)"
},
{
"operation": "back-button-clicked",
"time": "173.768",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "173.837",
"value": "Ido Nehoshtan"
},
{
"operation": "Highlighting",
"time": "176.163",
"value": "Ido Nehoshtan_cell_0_14_1"
},
{
"operation": "Highlighting",
"time": "176.174",
"value": "Ido Nehoshtan_header_cell_0_14_0"
},
{
"operation": "Highlighting",
"time": "176.174",
"value": "Ido Nehoshtan_header_cell_0_9_0"
},
{
"operation": "Highlighting",
"time": "204.694",
"value": "Ido Nehoshtan_sentence_6"
},
{
"operation": "Highlighting",
"time": "205.802",
"value": "Ido Nehoshtan_sentence_7"
},
{
"operation": "hyperlink",
"time": "211.868",
"value": "140 Squadron"
},
{
"operation": "Now on",
"time": "212.207",
"value": "140 Squadron (Israel)"
},
{
"operation": "Highlighting",
"time": "218.458",
"value": "140 Squadron (Israel)_sentence_0"
},
{
"operation": "Highlighting",
"time": "218.459",
"value": "140 Squadron (Israel)_title"
},
{
"operation": "search",
"time": "237.32",
"value": "140 Squadron"
},
{
"operation": "Now on",
"time": "238.303",
"value": "140 Squadron"
},
{
"operation": "search",
"time": "247.12",
"value": "153 Sqn"
},
{
"operation": "Now on",
"time": "247.846",
"value": "153 Sqn"
},
{
"operation": "Highlighting",
"time": "251.665",
"value": "No. 153 Squadron RAF_sentence_1"
},
{
"operation": "Highlighting",
"time": "251.666",
"value": "No. 153 Squadron RAF_title"
},
{
"operation": "Highlighting deleted",
"time": "256.017",
"value": "No. 153 Squadron RAF_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "256.019",
"value": "No. 153 Squadron RAF_title"
},
{
"operation": "Highlighting",
"time": "261.166",
"value": "No. 153 Squadron RAF_cell_0_2_1"
},
{
"operation": "Highlighting",
"time": "261.167",
"value": "No. 153 Squadron RAF_title"
},
{
"operation": "Highlighting",
"time": "261.171",
"value": "No. 153 Squadron RAF_header_cell_0_2_0"
},
{
"operation": "Highlighting",
"time": "261.171",
"value": "No. 153 Squadron RAF_header_cell_0_0_0"
},
{
"operation": "back-button-clicked",
"time": "264.82",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "264.882",
"value": "140 Squadron"
},
{
"operation": "back-button-clicked",
"time": "269.151",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "269.156",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "269.212",
"value": "140 Squadron (Israel)"
},
{
"operation": "back-button-clicked",
"time": "271.937",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "271.941",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "272.006",
"value": "Ido Nehoshtan"
},
{
"operation": "Highlighting",
"time": "285.878",
"value": "Ido Nehoshtan_sentence_6"
},
{
"operation": "Highlighting",
"time": "286.689",
"value": "Ido Nehoshtan_sentence_7"
},
{
"operation": "finish",
"time": "298.284",
"value": "finish"
}
]
}
|
exfever_3782
|
The Walking Dead is an American post-apocalyptic comic book series by an Australian screenwriter.
|
{
"evidence": [
"The Walking Dead (comic book)",
"Robert Kirkman"
],
"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 Walking Dead is an American post-apocalyptic comic book series by an Australian screenwriter.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Walking Dead (comic book)",
"Robert Kirkman"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1293,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_706
|
Alan Rickman was an English actor who became a member of a major British theatre company, based in Stratford-upon-Avon, Warwickshire, Egypt.
|
{
"evidence": [
"Alan Rickman",
"Royal Shakespeare Company"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAlan Rickman was an English actor who became a member of a major British theatre company, based in Stratford-upon-Avon, Warwickshire, Egypt.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Alan Rickman",
"Royal Shakespeare Company"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1294,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4266
|
Natalie Wood was an American actress and model who starred in John Ford's 1956 American Technicolor VistaVision epic film, whose genre of fiction and film is set primarily in the latter half of the early 18th century in the Western United States.
|
{
"evidence": [
"Natalie Wood",
"The Searchers",
"Western (genre)"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nNatalie Wood was an American actress and model who starred in John Ford's 1956 American Technicolor VistaVision epic film, whose genre of fiction and film is set primarily in the latter half of the early 18th century in the Western United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Natalie Wood",
"The Searchers",
"Western (genre)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1295,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
67002
|
Giuseppe Favalli was signed to Cremonese in 1992 then became team captain in 2002 after the departure of the previous captain.
|
{
"evidence": [
"Giuseppe Favalli_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:\nGiuseppe Favalli was signed to Cremonese in 1992 then became team captain in 2002 after the departure of the previous captain.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Giuseppe Favalli_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1296,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.328",
"value": "?search="
},
{
"operation": "search",
"time": "23.727",
"value": "Giuseppe Favalli"
},
{
"operation": "Now on",
"time": "27.509",
"value": "Giuseppe Favalli"
},
{
"operation": "Highlighting",
"time": "35.549",
"value": "Giuseppe Favalli_sentence_2"
},
{
"operation": "Page search",
"time": "54.84",
"value": "1992"
},
{
"operation": "finish",
"time": "103.738",
"value": "finish"
}
]
}
|
exfever_2215
|
Hapalochlaena is the taxonomic rank used in the biological classification of living and fossil organisms for blue-ringed octopuses.
|
{
"evidence": [
"Blue-ringed octopus",
"Genus"
],
"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:\nHapalochlaena is the taxonomic rank used in the biological classification of living and fossil organisms for blue-ringed octopuses.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Blue-ringed octopus",
"Genus"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1297,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6895
|
Mithun Chakraborty is an Indian actor who is the recipient of two awards that honor artistic excellence in the Hindi-language film industry of India.
|
{
"evidence": [
"Mithun Chakraborty",
"Filmfare 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:\nMithun Chakraborty is an Indian actor who is the recipient of two awards that honor artistic excellence in the Hindi-language film industry of India.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mithun Chakraborty",
"Filmfare Awards"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1298,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5712
|
Cirrhosis is the impaired liver function caused by fibrotic or pathological wound healing in which that replaces normal skin after birth.
|
{
"evidence": [
"Cirrhosis",
"Fibrosis",
"Scar"
],
"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:\nCirrhosis is the impaired liver function caused by fibrotic or pathological wound healing in which that replaces normal skin after birth.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Cirrhosis",
"Fibrosis",
"Scar"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1299,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
30455
|
Morandi, appointed professor of bridge design both at the University of Florence and the University of Rome, gained experience in Calabria working with reinforced concrete in earthquake-damaged areas.
|
{
"evidence": [
"Riccardo Morandi_sentence_6",
"Riccardo Morandi_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:\nMorandi, appointed professor of bridge design both at the University of Florence and the University of Rome, gained experience in Calabria working with reinforced concrete in earthquake-damaged areas.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Riccardo Morandi_sentence_6",
"Riccardo Morandi_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1300,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.651",
"value": "?search="
},
{
"operation": "search",
"time": "5.393",
"value": "Morandi"
},
{
"operation": "Now on",
"time": "6.314",
"value": "Morandi"
},
{
"operation": "search",
"time": "19.751",
"value": "Riccardo Morandi"
},
{
"operation": "Now on",
"time": "20.313",
"value": "Riccardo Morandi"
},
{
"operation": "Highlighting",
"time": "28.472",
"value": "Riccardo Morandi_sentence_6"
},
{
"operation": "Highlighting",
"time": "37.056",
"value": "Riccardo Morandi_sentence_3"
},
{
"operation": "finish",
"time": "45.109",
"value": "finish"
}
]
}
|
exfever_6357
|
Leroy Jethro Gibbs is a fictional character in an American police procedural television series released by CBS TV.
|
{
"evidence": [
"Leroy Jethro Gibbs",
"NCIS (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:\nLeroy Jethro Gibbs is a fictional character in an American police procedural television series released by CBS TV.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Leroy Jethro Gibbs",
"NCIS (TV series)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1301,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6904
|
Leslie Howard was an actor who played Ashley Wilkes in a 1939 American epic historical romance film.
|
{
"evidence": [
"Leslie Howard (actor)",
"Gone with the Wind (film)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLeslie Howard was an actor who played Ashley Wilkes in a 1939 American epic historical romance film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Leslie Howard (actor)",
"Gone with the Wind (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1302,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5212
|
The Flash is an American television series developed by the writing team of an American writer, director, and producer of movies, television, video games, and comic books.
|
{
"evidence": [
"The Flash (1990 TV series)",
"Danny Bilson"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Flash is an American television series developed by the writing team of an American writer, director, and producer of movies, television, video games, and comic books.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Flash (1990 TV series)",
"Danny Bilson"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1303,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2818
|
Ontario is one of the thirteen provinces and territories of Europe which is also the Canada's fourth-largest jurisdiction in total area when the territories of the federal territory of Canada that has a land area of approximately 1,144,000 km2.
|
{
"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 one of the thirteen provinces and territories of Europe which is also the Canada's fourth-largest jurisdiction in total area when the territories of the federal territory of Canada that has a land area of approximately 1,144,000 km2.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ontario",
"Northwest Territories"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1304,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
57778
|
Renato Ziggiotti failed to join the army during World War I, so he just studied theology in 1917.
|
{
"evidence": [
"Renato Ziggiotti_sentence_7",
"Renato Ziggiotti_sentence_8"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRenato Ziggiotti failed to join the army during World War I, so he just studied theology in 1917.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Renato Ziggiotti_sentence_7",
"Renato Ziggiotti_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1305,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.021",
"value": "?search="
},
{
"operation": "search",
"time": "16.176",
"value": "Renato Ziggiotti"
},
{
"operation": "Now on",
"time": "16.886",
"value": "Renato Ziggiotti"
},
{
"operation": "Highlighting",
"time": "41.784",
"value": "Renato Ziggiotti_sentence_7"
},
{
"operation": "Highlighting",
"time": "63.879",
"value": "Renato Ziggiotti_sentence_8"
},
{
"operation": "finish",
"time": "93.159",
"value": "finish"
}
]
}
|
exfever_6439
|
Honest is the third studio album released on April 22, 2014, through A1 Recordings and an American record label founded by American rapper Future.
|
{
"evidence": [
"Honest (Future album)",
"Freebandz"
],
"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:\nHonest is the third studio album released on April 22, 2014, through A1 Recordings and an American record label founded by American rapper Future.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Honest (Future album)",
"Freebandz"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1306,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_150
|
Blackhat is a 2015 American action thriller film produced and directed by an American director, screenwriter, and producer of film and television.
|
{
"evidence": [
"Blackhat (film)",
"Michael Mann"
],
"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:\nBlackhat is a 2015 American action thriller film produced and directed by an American director, screenwriter, and producer of film and television.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Blackhat (film)",
"Michael Mann"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1307,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1214
|
Sharon Marie Tate Polanski was an American model who appeared in the American mystery horror film directed by J. Lee Thompson.
|
{
"evidence": [
"Sharon Tate",
"Eye of the Devil"
],
"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:\nSharon Marie Tate Polanski was an American model who appeared in the American mystery horror film directed by J. Lee Thompson.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sharon Tate",
"Eye of the Devil"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1308,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
3234
|
Fingers' last major league appearance was on September 17, 1985, against the Milwaukee Brewers at Memorial Stadium wherein he pitched in relief of Teddy Higuera (a professional baseball pitcher for the Milwaukee Brewers of Major League Baseball) in the bottom of the eighth inning, facing two batters.
|
{
"evidence": [
"Rollie Fingers_sentence_49"
],
"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:\nFingers' last major league appearance was on September 17, 1985, against the Milwaukee Brewers at Memorial Stadium wherein he pitched in relief of Teddy Higuera (a professional baseball pitcher for the Milwaukee Brewers of Major League Baseball) in the bottom of the eighth inning, facing two batters.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Rollie Fingers_sentence_49"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1309,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "48.678",
"value": "Teddy Higuera"
},
{
"operation": "Page search",
"time": "54.749",
"value": "fingers"
},
{
"operation": "Page search",
"time": "93.757",
"value": "1985"
},
{
"operation": "Highlighting",
"time": "110.741",
"value": "Teddy Higuera_sentence_0"
},
{
"operation": "Highlighting",
"time": "110.743",
"value": "Teddy Higuera_title"
},
{
"operation": "Highlighting deleted",
"time": "111.651",
"value": "Teddy Higuera_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "111.653",
"value": "Teddy Higuera_title"
},
{
"operation": "search",
"time": "120.358",
"value": "milwaukee brewers"
},
{
"operation": "Page search",
"time": "128.854",
"value": "1985"
},
{
"operation": "search",
"time": "189.53",
"value": "1985 Milwaukee Brewers season"
},
{
"operation": "Page search",
"time": "194.786",
"value": "fingers"
},
{
"operation": "search",
"time": "220.305",
"value": "major league pitchers"
},
{
"operation": "search",
"time": "234.044",
"value": "pitcher fingers"
},
{
"operation": "hyperlink",
"time": "254.168",
"value": "History of the Oakland Athletics"
},
{
"operation": "Page search",
"time": "259.798",
"value": "fingers"
},
{
"operation": "hyperlink",
"time": "279.186",
"value": "Rollie Fingers"
},
{
"operation": "Highlighting",
"time": "313.571",
"value": "Rollie Fingers_sentence_49"
},
{
"operation": "finish",
"time": "329.323",
"value": "finish"
}
]
}
|
91447
|
Gur has multiple meanings depending on the context but it is commonly used for wisdom and internal conscious mind.
|
{
"evidence": [
"Gurbani_sentence_10",
"Gurbani_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:\nGur has multiple meanings depending on the context but it is commonly used for wisdom and internal conscious mind.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Gurbani_sentence_10",
"Gurbani_sentence_12"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1310,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.449",
"value": "?search="
},
{
"operation": "search",
"time": "15.159",
"value": "Gur"
},
{
"operation": "Now on",
"time": "16.144",
"value": "Gur"
},
{
"operation": "hyperlink",
"time": "57.51",
"value": "Gur languages"
},
{
"operation": "Now on",
"time": "59.786",
"value": "Gur languages"
},
{
"operation": "search",
"time": "113.857",
"value": "Contains... wisdom and internal conscious mind"
},
{
"operation": "Now on",
"time": "116.304",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "122.104",
"value": "Gurbani"
},
{
"operation": "Now on",
"time": "122.65",
"value": "Gurbani"
},
{
"operation": "Highlighting",
"time": "130.379",
"value": "Gurbani_sentence_10"
},
{
"operation": "Highlighting",
"time": "143.818",
"value": "Gurbani_sentence_12"
},
{
"operation": "finish",
"time": "154.459",
"value": "finish"
}
]
}
|
exfever_3848
|
Gone Girl is a 2014 American psychological thriller film directed by an American film director whose films are mostly psychological thrillers and biographical dramas.
|
{
"evidence": [
"Gone Girl (film)",
"David Fincher"
],
"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:\nGone Girl is a 2014 American psychological thriller film directed by an American film director whose films are mostly psychological thrillers and biographical dramas.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Gone Girl (film)",
"David Fincher"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1311,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1760
|
Carole King is an American singer who experienced commercial disappointment with her debut studio album, which included a song co-written by Gerry Goffin and recorded by The Drifters in 1962.
|
{
"evidence": [
"Carole King",
"Writer (album)",
"Up on the Roof (song)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nCarole King is an American singer who experienced commercial disappointment with her debut studio album, which included a song co-written by Gerry Goffin and recorded by The Drifters in 1962.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Carole King",
"Writer (album)",
"Up on the Roof (song)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1312,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3024
|
Grease is a 1978 Canadian film where songs by the characters are interwoven into the narrative.
|
{
"evidence": [
"Grease (film)",
"Musical 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:\nGrease is a 1978 Canadian film where songs by the characters are interwoven into the narrative.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Grease (film)",
"Musical film"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1313,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
19945
|
Doping at the World Athletics Championships is looked for with sampling of athletes blood and urine, if found the athletes performance will be annulled.
|
{
"evidence": [
"Doping at the World Athletics Championships_sentence_2",
"Doping at the World Athletics Championships_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:\nDoping at the World Athletics Championships is looked for with sampling of athletes blood and urine, if found the athletes performance will be annulled.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Doping at the World Athletics Championships_sentence_2",
"Doping at the World Athletics Championships_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1314,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.368",
"value": "?search="
},
{
"operation": "search",
"time": "15.91",
"value": "Doping at the World Athletics Championships"
},
{
"operation": "Now on",
"time": "20.426",
"value": "Doping at the World Athletics Championships"
},
{
"operation": "Page search",
"time": "67.757",
"value": "sampling of athletes blood"
},
{
"operation": "Highlighting",
"time": "69.893",
"value": "Doping at the World Athletics Championships_sentence_2"
},
{
"operation": "Highlighting",
"time": "81.476",
"value": "Doping at the World Athletics Championships_sentence_5"
},
{
"operation": "finish",
"time": "96.267",
"value": "finish"
}
]
}
|
72500
|
Papu Gómez was signed by Italian club Catania and in 2019–20, he set the Serie A record for most assists in a single season with 16 assists.
|
{
"evidence": [
"Papu Gómez_sentence_2",
"Papu Gómez_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:\nPapu Gómez was signed by Italian club Catania and in 2019–20, he set the Serie A record for most assists in a single season with 16 assists.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Papu Gómez_sentence_2",
"Papu Gómez_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1315,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.736",
"value": "?search="
},
{
"operation": "search",
"time": "16.867",
"value": "Papu Gómez"
},
{
"operation": "Now on",
"time": "18.139",
"value": "Papu Gómez"
},
{
"operation": "Page search",
"time": "44.985",
"value": "cantania"
},
{
"operation": "Page search",
"time": "63.254",
"value": "catania"
},
{
"operation": "Page search",
"time": "70.844",
"value": "catania"
},
{
"operation": "Highlighting",
"time": "75.938",
"value": "Papu Gómez_sentence_2"
},
{
"operation": "Highlighting",
"time": "90.691",
"value": "Papu Gómez_sentence_4"
},
{
"operation": "finish",
"time": "354.194",
"value": "finish"
}
]
}
|
exfever_4827
|
Elizabeth II, the Queen of the United Kingdom, was the first child of King George VI and the Dominions of the British Commonwealth from 11 December 1936 to 6 February 1953.
|
{
"evidence": [
"Elizabeth II",
"Queen Elizabeth The Queen Mother"
],
"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:\nElizabeth II, the Queen of the United Kingdom, was the first child of King George VI and the Dominions of the British Commonwealth from 11 December 1936 to 6 February 1953.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Elizabeth II",
"Queen Elizabeth The Queen Mother"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1316,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
32975
|
Franciszek Sulik tied for 2nd-3rd with Henryk Friedman in the 1934 Lviv championship, then tied for 8-9th in Warsaw 20 years after.
|
{
"evidence": [
"Franciszek Sulik_sentence_2",
"Franciszek Sulik_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:\nFranciszek Sulik tied for 2nd-3rd with Henryk Friedman in the 1934 Lviv championship, then tied for 8-9th in Warsaw 20 years after.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Franciszek Sulik_sentence_2",
"Franciszek Sulik_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1317,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.034",
"value": "?search="
},
{
"operation": "search",
"time": "11.11",
"value": "Franciszek Sulik"
},
{
"operation": "Now on",
"time": "11.75",
"value": "Franciszek Sulik"
},
{
"operation": "Highlighting",
"time": "28.623",
"value": "Franciszek Sulik_sentence_2"
},
{
"operation": "Highlighting",
"time": "45.217",
"value": "Franciszek Sulik_sentence_3"
},
{
"operation": "finish",
"time": "81.975",
"value": "finish"
}
]
}
|
exfever_2391
|
Diane Ruggiero-Wright is an American television writer and producer whose credits include an American supernatural procedural crime drama television series that is a loose adaptation of the comic book series of the same name created by an American science fiction author and publisher based in Portland, Oregon.
|
{
"evidence": [
"Diane Ruggiero",
"IZombie (TV series)",
"Chris Roberson (author)"
],
"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:\nDiane Ruggiero-Wright is an American television writer and producer whose credits include an American supernatural procedural crime drama television series that is a loose adaptation of the comic book series of the same name created by an American science fiction author and publisher based in Portland, Oregon.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Diane Ruggiero",
"IZombie (TV series)",
"Chris Roberson (author)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1318,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
65646
|
In 1977, the National Nuclear Data Center was given the additional responsibility for biological structure and zoologic data by the Energy Research and Development Administration (ERDA), then Dr. Charles Xavier served as Center Head from 1992 to 2002.
|
{
"evidence": [
"National Nuclear Data Center_sentence_9"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn 1977, the National Nuclear Data Center was given the additional responsibility for biological structure and zoologic data by the Energy Research and Development Administration (ERDA), then Dr. Charles Xavier served as Center Head from 1992 to 2002.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"National Nuclear Data Center_sentence_9"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1319,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.997",
"value": "?search="
},
{
"operation": "search",
"time": "11.713",
"value": "National Nuclear Data Center"
},
{
"operation": "Now on",
"time": "21.212",
"value": "National Nuclear Data Center"
},
{
"operation": "Page search",
"time": "85.141",
"value": "1977"
},
{
"operation": "Highlighting",
"time": "90.31",
"value": "National Nuclear Data Center_sentence_9"
},
{
"operation": "finish",
"time": "179.846",
"value": "finish"
}
]
}
|
exfever_7842
|
William Alan Shatner is best known for his portrayal of a fictional character in an Australian science fiction media franchise.
|
{
"evidence": [
"William Shatner",
"James T. Kirk",
"Star Trek"
],
"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:\nWilliam Alan Shatner is best known for his portrayal of a fictional character in an Australian science fiction media franchise.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"William Shatner",
"James T. Kirk",
"Star Trek"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1320,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
53023
|
How, born in Busan Metropolitan City in 1956, studied at the National Taiwan University, where he later completed his Bachelor of Business Administration (B.B.A.) degree.
|
{
"evidence": [
"Rex How_sentence_2",
"Rex How_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:\nHow, born in Busan Metropolitan City in 1956, studied at the National Taiwan University, where he later completed his Bachelor of Business Administration (B.B.A.) degree.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Rex How_sentence_2",
"Rex How_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1321,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.907",
"value": "?search="
},
{
"operation": "search",
"time": "18.616",
"value": "How, born in Busan Metropolitan City in 1956, studied at the National Taiwan University"
},
{
"operation": "Now on",
"time": "20.612",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "23.146",
"value": "Rex How"
},
{
"operation": "Now on",
"time": "23.865",
"value": "Rex How"
},
{
"operation": "Highlighting",
"time": "59.797",
"value": "Rex How_sentence_2"
},
{
"operation": "Highlighting",
"time": "63.713",
"value": "Rex How_sentence_4"
},
{
"operation": "finish",
"time": "84.199",
"value": "finish"
}
]
}
|
13091
|
Anton Garcia Caro, a Spanish jurist and politician who was married to Maria Gomez de Sanabria, arrived at Rio de la Plata from Cadiz and was granted lands near Lujan river.
|
{
"evidence": [
"Antón García Caro_sentence_0",
"Antón García Caro_sentence_2",
"Antón García Caro_sentence_3",
"Antón García Caro_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:\nAnton Garcia Caro, a Spanish jurist and politician who was married to Maria Gomez de Sanabria, arrived at Rio de la Plata from Cadiz and was granted lands near Lujan river.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Antón García Caro_sentence_0",
"Antón García Caro_sentence_2",
"Antón García Caro_sentence_3",
"Antón García Caro_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1322,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.947",
"value": "?search="
},
{
"operation": "search",
"time": "8.425",
"value": "Anton Garcia Caro,"
},
{
"operation": "Now on",
"time": "9.769",
"value": "undefined"
},
{
"operation": "search",
"time": "53.265",
"value": "Contains...Maria Gomez de Sanabria"
},
{
"operation": "Now on",
"time": "54.606",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "77.991",
"value": "Antón García Caro"
},
{
"operation": "Now on",
"time": "78.564",
"value": "Antón García Caro"
},
{
"operation": "Highlighting",
"time": "107.429",
"value": "Antón García Caro_sentence_0"
},
{
"operation": "Highlighting",
"time": "111.082",
"value": "Antón García Caro_sentence_2"
},
{
"operation": "Highlighting",
"time": "118.945",
"value": "Antón García Caro_sentence_3"
},
{
"operation": "Highlighting",
"time": "122.653",
"value": "Antón García Caro_sentence_5"
},
{
"operation": "finish",
"time": "256.106",
"value": "finish"
}
]
}
|
86107
|
Second wife of Tsar Paul I of Russia and the founder of the Office of the Institutions of Empress Maria was the mother of Anna Pavlovna of Russia, the queen consort of the Netherlands following the evolution of the Dutch Empire in 1795.
|
{
"evidence": [
"Maria Feodorovna (Sophie Dorothea of Württemberg)_sentence_2",
"Maria Feodorovna (Sophie Dorothea of Württemberg)_sentence_3",
"Maria Feodorovna (Sophie Dorothea of Württemberg)_sentence_128",
"Anna Pavlovna of Russia_sentence_2",
"Netherlands_sentence_219"
],
"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:\nSecond wife of Tsar Paul I of Russia and the founder of the Office of the Institutions of Empress Maria was the mother of Anna Pavlovna of Russia, the queen consort of the Netherlands following the evolution of the Dutch Empire in 1795.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Maria Feodorovna (Sophie Dorothea of Württemberg)_sentence_2",
"Maria Feodorovna (Sophie Dorothea of Württemberg)_sentence_3",
"Maria Feodorovna (Sophie Dorothea of Württemberg)_sentence_128",
"Anna Pavlovna of Russia_sentence_2",
"Netherlands_sentence_219"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1323,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.864",
"value": "?search="
},
{
"operation": "search",
"time": "30.301",
"value": "Second wife of Tsar Paul I of Russia"
},
{
"operation": "Now on",
"time": "33.915",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "41.587",
"value": "Maria Feodorovna (Sophie Dorothea of Württemberg)"
},
{
"operation": "Now on",
"time": "42.446",
"value": "Maria Feodorovna (Sophie Dorothea of Württemberg)"
},
{
"operation": "Highlighting",
"time": "62.04",
"value": "Maria Feodorovna (Sophie Dorothea of Württemberg)_sentence_2"
},
{
"operation": "Highlighting",
"time": "68.402",
"value": "Maria Feodorovna (Sophie Dorothea of Württemberg)_sentence_3"
},
{
"operation": "Page search",
"time": "174.504",
"value": "anna pavlovna"
},
{
"operation": "Highlighting",
"time": "195.2",
"value": "Maria Feodorovna (Sophie Dorothea of Württemberg)_sentence_128"
},
{
"operation": "hyperlink",
"time": "208.64",
"value": "Anna Pavlovna"
},
{
"operation": "Now on",
"time": "209.497",
"value": "Anna Pavlovna of Russia"
},
{
"operation": "Highlighting",
"time": "215.325",
"value": "Anna Pavlovna of Russia_sentence_2"
},
{
"operation": "search",
"time": "310.984",
"value": "Dutch Empire in 1795."
},
{
"operation": "Now on",
"time": "313.849",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "319.818",
"value": "Netherlands"
},
{
"operation": "Now on",
"time": "321.621",
"value": "Netherlands"
},
{
"operation": "hyperlink",
"time": "357.401",
"value": "Evolution of the Dutch Empire"
},
{
"operation": "Now on",
"time": "358.195",
"value": "Evolution of the Dutch Empire"
},
{
"operation": "Page search",
"time": "398.454",
"value": "1795"
},
{
"operation": "back-button-clicked",
"time": "414.497",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "414.96",
"value": "Netherlands"
},
{
"operation": "Page search",
"time": "446.271",
"value": "1795"
},
{
"operation": "Highlighting",
"time": "468.96",
"value": "Netherlands_sentence_219"
},
{
"operation": "finish",
"time": "497.205",
"value": "finish"
}
]
}
|
35170
|
Hosokawa Yoriyuki was the Shogun's Deputy in 1367, and he served as Yoshimitsu's historian for six years
|
{
"evidence": [
"Hosokawa Yoriyuki_sentence_1",
"Hosokawa Yoriyuki_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:\nHosokawa Yoriyuki was the Shogun's Deputy in 1367, and he served as Yoshimitsu's historian for six years\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Hosokawa Yoriyuki_sentence_1",
"Hosokawa Yoriyuki_sentence_9"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1324,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.779",
"value": "?search="
},
{
"operation": "search",
"time": "14.245",
"value": "Hosokawa Yoriyuki"
},
{
"operation": "Now on",
"time": "15.341",
"value": "Hosokawa Yoriyuki"
},
{
"operation": "Highlighting",
"time": "24.383",
"value": "Hosokawa Yoriyuki_sentence_1"
},
{
"operation": "Page search",
"time": "38.832",
"value": "historian"
},
{
"operation": "search",
"time": "130.127",
"value": "Yoshimitsu"
},
{
"operation": "Now on",
"time": "131.561",
"value": "Yoshimitsu"
},
{
"operation": "back-button-clicked",
"time": "163.806",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "163.895",
"value": "Hosokawa Yoriyuki"
},
{
"operation": "Highlighting",
"time": "173.698",
"value": "Hosokawa Yoriyuki_sentence_9"
},
{
"operation": "hyperlink",
"time": "175.568",
"value": "Ashikaga Yoshimitsu"
},
{
"operation": "Now on",
"time": "176.539",
"value": "Ashikaga Yoshimitsu"
},
{
"operation": "Page search",
"time": "185.194",
"value": "historian"
},
{
"operation": "finish",
"time": "210.17",
"value": "finish"
}
]
}
|
exfever_8797
|
Carmelo Anthony, a player for the Boston Celtics, also played for one of the most prestigious intercollegiate men's basketball teams, representing a university in New York.
|
{
"evidence": [
"Carmelo Anthony",
"Syracuse Orange men's 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:\nCarmelo Anthony, a player for the Boston Celtics, also played for one of the most prestigious intercollegiate men's basketball teams, representing a university in New York.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Carmelo Anthony",
"Syracuse Orange men's basketball"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1325,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2249
|
Glenn Frey was an English musician who sang lead vocals on hits such as the second single from their album One of These Nights, reaching No. 2 on the Billboard Hot 100 chart and No. 8 on the Billboard Country chart.
|
{
"evidence": [
"Glenn Frey",
"Lyin' Eyes"
],
"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:\nGlenn Frey was an English musician who sang lead vocals on hits such as the second single from their album One of These Nights, reaching No. 2 on the Billboard Hot 100 chart and No. 8 on the Billboard Country chart.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Glenn Frey",
"Lyin' Eyes"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1326,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3423
|
Taylor Alison Swift is the most successful country pop singer in America.
|
{
"evidence": [
"Taylor Swift",
"Speak Now"
],
"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:\nTaylor Alison Swift is the most successful country pop singer in America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Taylor Swift",
"Speak Now"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1327,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
10948
|
Süleyman Çelebi fought in the Battle of Nigbolu and the Battle of Ankara, the former of which is also known as the Crusade of Nicopolis.
|
{
"evidence": [
"Battle of Nicopolis_sentence_2",
"Süleyman Çelebi_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:\nSüleyman Çelebi fought in the Battle of Nigbolu and the Battle of Ankara, the former of which is also known as the Crusade of Nicopolis.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Battle of Nicopolis_sentence_2",
"Süleyman Çelebi_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1328,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.99",
"value": "?search="
},
{
"operation": "search",
"time": "12.872",
"value": "Süleyman Çelebi"
},
{
"operation": "Now on",
"time": "13.779",
"value": "Süleyman Çelebi"
},
{
"operation": "hyperlink",
"time": "36.049",
"value": "Battle of Ankara"
},
{
"operation": "Now on",
"time": "36.595",
"value": "Battle of Ankara"
},
{
"operation": "Page search",
"time": "44.402",
"value": "nicopolis"
},
{
"operation": "back-button-clicked",
"time": "53.722",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "53.755",
"value": "Süleyman Çelebi"
},
{
"operation": "hyperlink",
"time": "55.98",
"value": "Battle of Niğbolu"
},
{
"operation": "Now on",
"time": "56.945",
"value": "Battle of Nicopolis"
},
{
"operation": "Page search",
"time": "67.626",
"value": "crusade"
},
{
"operation": "Highlighting",
"time": "88.541",
"value": "Battle of Nicopolis_sentence_2"
},
{
"operation": "back-button-clicked",
"time": "90.826",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "90.86",
"value": "Süleyman Çelebi"
},
{
"operation": "Highlighting",
"time": "94.706",
"value": "Süleyman Çelebi_sentence_4"
},
{
"operation": "hyperlink",
"time": "132.235",
"value": "Battle of Niğbolu"
},
{
"operation": "Now on",
"time": "132.635",
"value": "Battle of Niğbolu"
},
{
"operation": "Page search",
"time": "159.698",
"value": "Niğbolu"
},
{
"operation": "finish",
"time": "180.589",
"value": "finish"
}
]
}
|
exfever_6485
|
Clubber Lang is a fictional character created by Sylvester Stallone for an American film genre that uses sport as the theme.
|
{
"evidence": [
"Clubber Lang",
"Rocky III",
"Sports film"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nClubber Lang is a fictional character created by Sylvester Stallone for an American film genre that uses sport as the theme.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Clubber Lang",
"Rocky III",
"Sports film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1329,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7616
|
Taal is a 1999 Indian Hindi musical romantic drama film directed by an Indian film director, known for his works predominantly in Thai cinema.
|
{
"evidence": [
"Taal (film)",
"Subhash Ghai"
],
"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:\nTaal is a 1999 Indian Hindi musical romantic drama film directed by an Indian film director, known for his works predominantly in Thai cinema.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Taal (film)",
"Subhash Ghai"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1330,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
30787
|
Chromothripsis, a process by which up to thousands of clustered chromosomal rearrangements occur in a single event, has high copy states and hence have paired end joining predominating repair mechanisms but requires further study in order to understand the total relative importance of paired end joining and replicative repair.
|
{
"evidence": [
"Chromothripsis_sentence_23"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nChromothripsis, a process by which up to thousands of clustered chromosomal rearrangements occur in a single event, has high copy states and hence have paired end joining predominating repair mechanisms but requires further study in order to understand the total relative importance of paired end joining and replicative repair.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Chromothripsis_sentence_23"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1331,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.655",
"value": "?search="
},
{
"operation": "search",
"time": "20.656",
"value": "Chromothripsis"
},
{
"operation": "Now on",
"time": "21.566",
"value": "Chromothripsis"
},
{
"operation": "Highlighting",
"time": "38.993",
"value": "Chromothripsis_sentence_0"
},
{
"operation": "Page search",
"time": "111.728",
"value": "copy"
},
{
"operation": "Highlighting",
"time": "138.304",
"value": "Chromothripsis_sentence_23"
},
{
"operation": "Highlighting deleted",
"time": "142.791",
"value": "Chromothripsis_sentence_0"
},
{
"operation": "finish",
"time": "158.648",
"value": "finish"
}
]
}
|
exfever_8782
|
Batman Begins also heralded a trend of darker genre films,
|
{
"evidence": [
"Batman Begins",
"Katie Holmes",
"Ang Lee"
],
"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:\nBatman Begins also heralded a trend of darker genre films,\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Batman Begins",
"Katie Holmes",
"Ang Lee"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1332,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
19497
|
Bernard Pagel obtained his Ph.D. in Cambridge in 1955 and became a Research Fellow at Sidney Sussex college in 1953-1956, then later became a Deputy Scientific Officer at Royal Greenwich Observatory, an observatory where the prime meridian passes through it.
|
{
"evidence": [
"Bernard Pagel_sentence_3",
"Bernard Pagel_sentence_4",
"Bernard Pagel_sentence_5"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBernard Pagel obtained his Ph.D. in Cambridge in 1955 and became a Research Fellow at Sidney Sussex college in 1953-1956, then later became a Deputy Scientific Officer at Royal Greenwich Observatory, an observatory where the prime meridian passes through it.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Bernard Pagel_sentence_3",
"Bernard Pagel_sentence_4",
"Bernard Pagel_sentence_5"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1333,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.303",
"value": "?search="
},
{
"operation": "search",
"time": "10.882",
"value": "Bernard Pagel"
},
{
"operation": "Now on",
"time": "12.413",
"value": "Bernard Pagel"
},
{
"operation": "Highlighting",
"time": "28.407",
"value": "Bernard Pagel_sentence_3"
},
{
"operation": "Highlighting",
"time": "32.939",
"value": "Bernard Pagel_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "49.14",
"value": "Bernard Pagel_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "50.477",
"value": "Bernard Pagel_sentence_3"
},
{
"operation": "Highlighting",
"time": "51.731",
"value": "Bernard Pagel_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "59.979",
"value": "Bernard Pagel_sentence_5"
},
{
"operation": "Highlighting",
"time": "61.028",
"value": "Bernard Pagel_sentence_3"
},
{
"operation": "Highlighting",
"time": "61.936",
"value": "Bernard Pagel_sentence_4"
},
{
"operation": "Highlighting",
"time": "63.804",
"value": "Bernard Pagel_sentence_5"
},
{
"operation": "finish",
"time": "90.715",
"value": "finish"
}
]
}
|
31287
|
The Nantou railway station is an elevated station in Japan and the first station in Zhongshan for the rail going to Zhuhai.
|
{
"evidence": [
"Nantou railway station_sentence_1"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Nantou railway station is an elevated station in Japan and the first station in Zhongshan for the rail going to Zhuhai.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Nantou railway station_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1334,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.605",
"value": "?search="
},
{
"operation": "search",
"time": "10.4",
"value": "Nantou railway station"
},
{
"operation": "Now on",
"time": "11.094",
"value": "Nantou railway station"
},
{
"operation": "Highlighting",
"time": "26.663",
"value": "Nantou railway station_sentence_1"
},
{
"operation": "finish",
"time": "33.681",
"value": "finish"
}
]
}
|
23593
|
Desorption electrospray ionization is an ambient ionization technique which is a form of ionization where ions are formed in an ion source outside the mass spectrometer without sample preparation or separation and ions can be formed by extraction into charged electrospray droplets.
|
{
"evidence": [
"Desorption electrospray ionization_sentence_0",
"Ambient ionization_sentence_0",
"Ambient ionization_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:\nDesorption electrospray ionization is an ambient ionization technique which is a form of ionization where ions are formed in an ion source outside the mass spectrometer without sample preparation or separation and ions can be formed by extraction into charged electrospray droplets.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Desorption electrospray ionization_sentence_0",
"Ambient ionization_sentence_0",
"Ambient ionization_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1335,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.876",
"value": "?search="
},
{
"operation": "search",
"time": "10.037",
"value": "Desorption electrospray ionization"
},
{
"operation": "Now on",
"time": "10.727",
"value": "Desorption electrospray ionization"
},
{
"operation": "Highlighting",
"time": "15.585",
"value": "Desorption electrospray ionization_sentence_0"
},
{
"operation": "Page search",
"time": "24.151",
"value": "form"
},
{
"operation": "Page search",
"time": "35.357",
"value": "ion source"
},
{
"operation": "hyperlink",
"time": "119.183",
"value": "ambient ionization"
},
{
"operation": "Now on",
"time": "119.646",
"value": "Ambient ionization"
},
{
"operation": "Highlighting",
"time": "122.641",
"value": "Ambient ionization_sentence_0"
},
{
"operation": "Highlighting",
"time": "135.586",
"value": "Ambient ionization_sentence_1"
},
{
"operation": "finish",
"time": "146.244",
"value": "finish"
}
]
}
|
17094
|
If Only You Were Lonely is a studio album by the Hawthorne Heights and was produced by David Bendeth.
|
{
"evidence": [
"If Only You Were Lonely_sentence_1",
"If Only You Were Lonely_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:\nIf Only You Were Lonely is a studio album by the Hawthorne Heights and was produced by David Bendeth.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"If Only You Were Lonely_sentence_1",
"If Only You Were Lonely_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1336,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.291",
"value": "?search="
},
{
"operation": "search",
"time": "18.021",
"value": "If Only You Were Lonely"
},
{
"operation": "Now on",
"time": "18.906",
"value": "If Only You Were Lonely"
},
{
"operation": "Highlighting",
"time": "25.936",
"value": "If Only You Were Lonely_sentence_1"
},
{
"operation": "Highlighting",
"time": "27.05",
"value": "If Only You Were Lonely_sentence_0"
},
{
"operation": "finish",
"time": "66.835",
"value": "finish"
}
]
}
|
30802
|
Based on the same platform as the Chevrolet Sail, the Baojun 310 was launched on 2017 Beijing Auto Show where the price ranges from 36.800 yuan to 60.800 yuan.
|
{
"evidence": [
"Baojun 310_sentence_1"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBased on the same platform as the Chevrolet Sail, the Baojun 310 was launched on 2017 Beijing Auto Show where the price ranges from 36.800 yuan to 60.800 yuan.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Baojun 310_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1337,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.674",
"value": "?search="
},
{
"operation": "search",
"time": "9.268",
"value": "Baojun 310"
},
{
"operation": "Now on",
"time": "10.996",
"value": "Baojun 310"
},
{
"operation": "Highlighting",
"time": "18.766",
"value": "Baojun 310_sentence_2"
},
{
"operation": "Highlighting",
"time": "26.179",
"value": "Baojun 310_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "29.096",
"value": "Baojun 310_sentence_2"
},
{
"operation": "finish",
"time": "51.942",
"value": "finish"
}
]
}
|
82237
|
David Bowie Is had a 75,000 private collection that Geoffrey Marsh was allowed to access for free, who was born in 1957 beginning his career at the Museum of London.
|
{
"evidence": [
"Geoffrey Marsh (museum director)_sentence_0",
"Geoffrey Marsh (museum director)_sentence_1",
"David Bowie Is_sentence_7",
"David Bowie Is_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:\nDavid Bowie Is had a 75,000 private collection that Geoffrey Marsh was allowed to access for free, who was born in 1957 beginning his career at the Museum of London.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Geoffrey Marsh (museum director)_sentence_0",
"Geoffrey Marsh (museum director)_sentence_1",
"David Bowie Is_sentence_7",
"David Bowie Is_sentence_8"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1338,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.222",
"value": "?search="
},
{
"operation": "search",
"time": "48.227",
"value": "Geoffrey Marsh"
},
{
"operation": "Now on",
"time": "48.922",
"value": "Geoffrey Marsh"
},
{
"operation": "hyperlink",
"time": "59.245",
"value": "Geoffrey Marsh (museum director)"
},
{
"operation": "Now on",
"time": "59.756",
"value": "Geoffrey Marsh (museum director)"
},
{
"operation": "Highlighting",
"time": "66.39",
"value": "Geoffrey Marsh (museum director)_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "69.488",
"value": "Geoffrey Marsh (museum director)_sentence_1"
},
{
"operation": "Highlighting",
"time": "71.692",
"value": "Geoffrey Marsh (museum director)_sentence_0"
},
{
"operation": "Highlighting",
"time": "72.65",
"value": "Geoffrey Marsh (museum director)_sentence_1"
},
{
"operation": "hyperlink",
"time": "105.84",
"value": "David Bowie Is"
},
{
"operation": "Now on",
"time": "106.69",
"value": "David Bowie Is"
},
{
"operation": "Highlighting",
"time": "123.982",
"value": "David Bowie Is_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "134.218",
"value": "David Bowie Is_sentence_8"
},
{
"operation": "Highlighting",
"time": "136.223",
"value": "David Bowie Is_sentence_7"
},
{
"operation": "Highlighting",
"time": "137.837",
"value": "David Bowie Is_sentence_8"
},
{
"operation": "finish",
"time": "162.235",
"value": "finish"
}
]
}
|
exfever_3703
|
Champion is a 1949 American film noir sports film directed by Mark Robson whose genre is in a category of narrative fiction intended to be more serious than humorous in tone.
|
{
"evidence": [
"Champion (1949 film)",
"Drama (film and television)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nChampion is a 1949 American film noir sports film directed by Mark Robson whose genre is in a category of narrative fiction intended to be more serious than humorous in tone.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Champion (1949 film)",
"Drama (film and television)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1339,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
28600
|
Hila Bronstein born July 26, 1983 was a German songwriter who is associated with Bro'Sis.
|
{
"evidence": [
"Hila Bronstein_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:\nHila Bronstein born July 26, 1983 was a German songwriter who is associated with Bro'Sis.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Hila Bronstein_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1340,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.033",
"value": "?search="
},
{
"operation": "search",
"time": "7.188",
"value": "Hila Bronstein"
},
{
"operation": "Now on",
"time": "8.066",
"value": "Hila Bronstein"
},
{
"operation": "Highlighting",
"time": "9.592",
"value": "Hila Bronstein_sentence_0"
},
{
"operation": "Page search",
"time": "19.236",
"value": "Bro'Sis"
},
{
"operation": "page-search-reset",
"time": "21.714",
"value": "page-search-reset"
},
{
"operation": "finish",
"time": "28.701",
"value": "finish"
}
]
}
|
78513
|
Adon Smith House is a historic home located at Hamilton in Madison County, that was built 1850.
|
{
"evidence": [
"Adon Smith House_sentence_0",
"Adon Smith House_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:\nAdon Smith House is a historic home located at Hamilton in Madison County, that was built 1850.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Adon Smith House_sentence_0",
"Adon Smith House_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1341,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.704",
"value": "?search="
},
{
"operation": "search",
"time": "66.51",
"value": "Adon Smith House"
},
{
"operation": "Now on",
"time": "67.179",
"value": "Adon Smith House"
},
{
"operation": "Highlighting",
"time": "69.039",
"value": "Adon Smith House_sentence_0"
},
{
"operation": "Highlighting",
"time": "72.539",
"value": "Adon Smith House_sentence_1"
},
{
"operation": "finish",
"time": "82.144",
"value": "finish"
}
]
}
|
exfever_1327
|
Shonen Jump was a shōnen manga anthology published in North America by the American manga publisher, anime distributor, and entertainment company headquartered in San Francisco, in a state in the Southern United States.
|
{
"evidence": [
"Shonen Jump (magazine)",
"Viz Media",
"California"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nShonen Jump was a shōnen manga anthology published in North America by the American manga publisher, anime distributor, and entertainment company headquartered in San Francisco, in a state in the Southern United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Shonen Jump (magazine)",
"Viz Media",
"California"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1342,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3682
|
Charles I was the wealthiest king of England.
|
{
"evidence": [
"Charles I of England",
"Kingdom of Scotland"
],
"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:\nCharles I was the wealthiest king of England.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Charles I of England",
"Kingdom of Scotland"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1343,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7375
|
The film Just My Luck stars Lindsay Lohan as Ashley Albright, the luckiest person in the most densely populated and geographically smallest of the five boroughs of New York City.
|
{
"evidence": [
"Just My Luck (2006 film)",
"Manhattan"
],
"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 film Just My Luck stars Lindsay Lohan as Ashley Albright, the luckiest person in the most densely populated and geographically smallest of the five boroughs of New York City.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Just My Luck (2006 film)",
"Manhattan"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1344,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1292
|
Ink Master is an American reality competition television series produced by Original Media and the producer of an American reality show that aired on TLC.
|
{
"evidence": [
"Ink Master",
"Miami Ink"
],
"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:\nInk Master is an American reality competition television series produced by Original Media and the producer of an American reality show that aired on TLC.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ink Master",
"Miami Ink"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1345,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
19384
|
Are Næss (born 14 November 1942 in Bergen), was elected to the Norwegian Parliament and studied medicine which encompasses a variety of health care practices.
|
{
"evidence": [
"Are Næss_sentence_0",
"Are Næss_sentence_1",
"Are Næss_sentence_2",
"Medicine_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:\nAre Næss (born 14 November 1942 in Bergen), was elected to the Norwegian Parliament and studied medicine which encompasses a variety of health care practices.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Are Næss_sentence_0",
"Are Næss_sentence_1",
"Are Næss_sentence_2",
"Medicine_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1346,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.49",
"value": "?search="
},
{
"operation": "search",
"time": "9.024",
"value": "Are Næss"
},
{
"operation": "Now on",
"time": "10.012",
"value": "Are Næss"
},
{
"operation": "Highlighting",
"time": "11.205",
"value": "Are Næss_sentence_0"
},
{
"operation": "Highlighting",
"time": "16.774",
"value": "Are Næss_sentence_1"
},
{
"operation": "Highlighting",
"time": "18.908",
"value": "Are Næss_sentence_2"
},
{
"operation": "hyperlink",
"time": "30.673",
"value": "medicine"
},
{
"operation": "Now on",
"time": "31.688",
"value": "Medicine"
},
{
"operation": "Highlighting",
"time": "34.795",
"value": "Medicine_sentence_4"
},
{
"operation": "back-button-clicked",
"time": "47.943",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "47.984",
"value": "Are Næss"
},
{
"operation": "finish",
"time": "66.615",
"value": "finish"
}
]
}
|
exfever_7052
|
Spider-Man is an American comic book villain created by Stan Lee and an American comics artist and writer, best known as the artist and co-creator of the Marvel superhero Doctor Strange.
|
{
"evidence": [
"Spider-Man",
"Steve Ditko"
],
"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:\nSpider-Man is an American comic book villain created by Stan Lee and an American comics artist and writer, best known as the artist and co-creator of the Marvel superhero Doctor Strange.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Spider-Man",
"Steve Ditko"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1347,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4459
|
Leopold II was the second King of the Belgians born in a region of Belgium comprising 20 municipalities, including the City of Brussels, which is the capital of Belgium.
|
{
"evidence": [
"Leopold II of Belgium",
"Brussels"
],
"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:\nLeopold II was the second King of the Belgians born in a region of Belgium comprising 20 municipalities, including the City of Brussels, which is the capital of Belgium.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Leopold II of Belgium",
"Brussels"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1348,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4917
|
Get Out is a 2017 American horror film starring an American actor and musician, best known for his role in X-Men: First Class as a fictional superhero published by Marvel Comics.
|
{
"evidence": [
"Get Out (film)",
"Caleb Landry Jones",
"Banshee (comics)"
],
"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:\nGet Out is a 2017 American horror film starring an American actor and musician, best known for his role in X-Men: First Class as a fictional superhero published by Marvel Comics.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Get Out (film)",
"Caleb Landry Jones",
"Banshee (comics)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1349,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
1923
|
The Elburton Cross railway station was opened by the Great Western Railway on 17 January 1898 and closed on 17 January 1898
|
{
"evidence": [
"Elburton Cross railway station_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 Elburton Cross railway station was opened by the Great Western Railway on 17 January 1898 and closed on 17 January 1898\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Elburton Cross railway station_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1350,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "19.722",
"value": "Elburton Cross railway station"
},
{
"operation": "Highlighting",
"time": "32.997",
"value": "Elburton Cross railway station_sentence_1"
},
{
"operation": "Highlighting",
"time": "67.559",
"value": "Elburton Cross railway station_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "68.84",
"value": "Elburton Cross railway station_sentence_1"
},
{
"operation": "finish",
"time": "88.129",
"value": "finish"
}
]
}
|
35671
|
Todd Helton, an American former professional baseball first baseman, won the Associated Press Major League Baseball Player of the Year for four consecutive years from year 2000 to 2003.
|
{
"evidence": [
"Todd Helton_sentence_0",
"Todd Helton_sentence_60"
],
"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:\nTodd Helton, an American former professional baseball first baseman, won the Associated Press Major League Baseball Player of the Year for four consecutive years from year 2000 to 2003.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Todd Helton_sentence_0",
"Todd Helton_sentence_60"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1351,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.67",
"value": "?search="
},
{
"operation": "search",
"time": "46.764",
"value": "Todd Helton"
},
{
"operation": "Now on",
"time": "47.566",
"value": "Todd Helton"
},
{
"operation": "Highlighting",
"time": "52.503",
"value": "Todd Helton_sentence_0"
},
{
"operation": "Highlighting",
"time": "153.265",
"value": "Todd Helton_sentence_60"
},
{
"operation": "search",
"time": "191.467",
"value": "AP Major League Baseball Player of the Year"
},
{
"operation": "Now on",
"time": "193.937",
"value": "undefined"
},
{
"operation": "search",
"time": "214.192",
"value": "Associated Press Major League Baseball player of the year"
},
{
"operation": "Now on",
"time": "215.521",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "230.231",
"value": "Todd Helton"
},
{
"operation": "Now on",
"time": "230.839",
"value": "Todd Helton"
},
{
"operation": "search",
"time": "282.86",
"value": "AP Baseball player of the year"
},
{
"operation": "Now on",
"time": "283.805",
"value": "undefined"
},
{
"operation": "search",
"time": "300.373",
"value": "Associated Press baseball play of the year"
},
{
"operation": "Now on",
"time": "301.307",
"value": "undefined"
},
{
"operation": "finish",
"time": "332.887",
"value": "finish"
}
]
}
|
exfever_3696
|
Sister Act is a 1992 American comedy film directed by Emile Ardolino that features the American actress who appeared in the State of Play and the 1997 sports comedy film directed by Charles Martin Smith.
|
{
"evidence": [
"Sister Act",
"Wendy Makkena",
"Air Bud"
],
"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:\nSister Act is a 1992 American comedy film directed by Emile Ardolino that features the American actress who appeared in the State of Play and the 1997 sports comedy film directed by Charles Martin Smith.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sister Act",
"Wendy Makkena",
"Air Bud"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1352,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
24996
|
Lenhart Schubert earned his Ph.D. from the University of Toronto in 1970 and became a faculty of the University of Alberta for 15 years.
|
{
"evidence": [
"Lenhart Schubert_sentence_2",
"Lenhart Schubert_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:\nLenhart Schubert earned his Ph.D. from the University of Toronto in 1970 and became a faculty of the University of Alberta for 15 years.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Lenhart Schubert_sentence_2",
"Lenhart Schubert_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1353,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.24",
"value": "?search="
},
{
"operation": "search",
"time": "9.949",
"value": "Lenhart Schubert"
},
{
"operation": "Now on",
"time": "10.815",
"value": "Lenhart Schubert"
},
{
"operation": "Highlighting",
"time": "15.858",
"value": "Lenhart Schubert_sentence_2"
},
{
"operation": "Highlighting",
"time": "20.614",
"value": "Lenhart Schubert_sentence_3"
},
{
"operation": "finish",
"time": "39.105",
"value": "finish"
}
]
}
|
79314
|
Seniors with a life insurance policy are eligible for Medicaid and Orestis created a model for middle-class people to pay for long-term care when they cannot pay out of pocket or through Medicaid.
|
{
"evidence": [
"Medicaid_sentence_210"
],
"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:\nSeniors with a life insurance policy are eligible for Medicaid and Orestis created a model for middle-class people to pay for long-term care when they cannot pay out of pocket or through Medicaid.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Medicaid_sentence_210"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1354,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.97",
"value": "?search="
},
{
"operation": "search",
"time": "14.982",
"value": "Medicaid and Orestis"
},
{
"operation": "Now on",
"time": "16.116",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "22.256",
"value": "medicare and orestes"
},
{
"operation": "Now on",
"time": "22.963",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "26.298",
"value": "medicine and orestes"
},
{
"operation": "Now on",
"time": "27.746",
"value": "undefined"
},
{
"operation": "search",
"time": "32.977",
"value": "medicaid"
},
{
"operation": "Now on",
"time": "36.222",
"value": "Medicaid"
},
{
"operation": "Page search",
"time": "156.785",
"value": "life insurance"
},
{
"operation": "Page search",
"time": "158.479",
"value": "life insurance"
},
{
"operation": "search",
"time": "232.686",
"value": "orestis"
},
{
"operation": "Now on",
"time": "235.02",
"value": "Orestis"
},
{
"operation": "back-button-clicked",
"time": "269.665",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "269.81",
"value": "Medicaid"
},
{
"operation": "hyperlink",
"time": "273.278",
"value": "Medicare (United States)"
},
{
"operation": "Now on",
"time": "276.012",
"value": "Medicare (United States)"
},
{
"operation": "Page search",
"time": "295.448",
"value": "middle-class people"
},
{
"operation": "Page search",
"time": "317.747",
"value": "life insurance"
},
{
"operation": "search",
"time": "358.17",
"value": "life insurance policy"
},
{
"operation": "Now on",
"time": "363.026",
"value": "Life insurance"
},
{
"operation": "Page search",
"time": "373.379",
"value": "medicaid"
},
{
"operation": "search",
"time": "426.176",
"value": "Medicaid"
},
{
"operation": "Now on",
"time": "427.858",
"value": "Medicaid"
},
{
"operation": "Page search",
"time": "610.296",
"value": "long-term care"
},
{
"operation": "Highlighting",
"time": "676.836",
"value": "Medicaid_sentence_210"
},
{
"operation": "finish",
"time": "689.53",
"value": "finish"
}
]
}
|
4077
|
Steinbeck makes the case that with their dignity restored and their sense of "having something of their own, these men are better workers" and observing workers and their families living in these camps, Steinbeck reports that many individuals displayed "a steadiness of gaze and a self-confidence that can only come of restored dignity."
|
{
"evidence": [
"The Harvest Gypsies_sentence_108",
"The Harvest Gypsies_sentence_109"
],
"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:\nSteinbeck makes the case that with their dignity restored and their sense of \"having something of their own, these men are better workers\" and observing workers and their families living in these camps, Steinbeck reports that many individuals displayed \"a steadiness of gaze and a self-confidence that can only come of restored dignity.\"\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"The Harvest Gypsies_sentence_108",
"The Harvest Gypsies_sentence_109"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1355,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "127.319",
"value": "steinbeck"
},
{
"operation": "Page search",
"time": "139.856",
"value": "dignity rest"
},
{
"operation": "search",
"time": "153.886",
"value": "the grapes of wrath"
},
{
"operation": "Page search",
"time": "161.326",
"value": "dignity"
},
{
"operation": "search",
"time": "190.91",
"value": "The Harvest Gypsies"
},
{
"operation": "Page search",
"time": "198.602",
"value": "dignity"
},
{
"operation": "Highlighting",
"time": "204.47",
"value": "The Harvest Gypsies_sentence_108"
},
{
"operation": "Highlighting",
"time": "206.98",
"value": "The Harvest Gypsies_sentence_109"
},
{
"operation": "finish",
"time": "216.037",
"value": "finish"
}
]
}
|
58820
|
Following a renovation, the passenger building of Sondrio railway station, currently managed by Rete Ferroviaria Italiana (the owner of Italy's railway network), consists of the following public spaces: ticketing, bar, kiosk and waiting room, while the station's building containing the public conveniences is now Milanese in style.
|
{
"evidence": [
"Sondrio railway station_sentence_12",
"Sondrio railway station_sentence_2",
"Rete Ferroviaria Italiana_sentence_1",
"Sondrio railway station_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:\nFollowing a renovation, the passenger building of Sondrio railway station, currently managed by Rete Ferroviaria Italiana (the owner of Italy's railway network), consists of the following public spaces: ticketing, bar, kiosk and waiting room, while the station's building containing the public conveniences is now Milanese in style.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Sondrio railway station_sentence_12",
"Sondrio railway station_sentence_2",
"Rete Ferroviaria Italiana_sentence_1",
"Sondrio railway station_sentence_15"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1356,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.388",
"value": "?search="
},
{
"operation": "search",
"time": "32.265",
"value": "Sondrio railway station"
},
{
"operation": "Now on",
"time": "34.114",
"value": "Sondrio railway station"
},
{
"operation": "Highlighting",
"time": "157.964",
"value": "Sondrio railway station_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "175.764",
"value": "Sondrio railway station_sentence_3"
},
{
"operation": "Highlighting",
"time": "178.259",
"value": "Sondrio railway station_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "182.291",
"value": "Sondrio railway station_sentence_4"
},
{
"operation": "Highlighting",
"time": "193.019",
"value": "Sondrio railway station_sentence_12"
},
{
"operation": "Highlighting",
"time": "266.407",
"value": "Sondrio railway station_sentence_2"
},
{
"operation": "hyperlink",
"time": "270.063",
"value": "Rete Ferroviaria Italiana"
},
{
"operation": "Now on",
"time": "271.319",
"value": "Rete Ferroviaria Italiana"
},
{
"operation": "Highlighting",
"time": "273.686",
"value": "Rete Ferroviaria Italiana_sentence_1"
},
{
"operation": "back-button-clicked",
"time": "278.735",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "278.837",
"value": "Sondrio railway station"
},
{
"operation": "Highlighting",
"time": "292.798",
"value": "Sondrio railway station_sentence_15"
},
{
"operation": "finish",
"time": "302.646",
"value": "finish"
}
]
}
|
89508
|
Ugamia is a distant cousin to the wild boar.
|
{
"evidence": [
"Ugamia_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:\nUgamia is a distant cousin to the wild boar.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ugamia_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1357,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.757",
"value": "?search="
},
{
"operation": "search",
"time": "10.605",
"value": "Ugamia"
},
{
"operation": "Now on",
"time": "12.103",
"value": "Ugamia"
},
{
"operation": "Highlighting",
"time": "16.502",
"value": "Ugamia_sentence_0"
},
{
"operation": "finish",
"time": "34.318",
"value": "finish"
}
]
}
|
exfever_3365
|
Tupac Shakur is an African rapper who released his debut album on November 12, 1991, on an American record label owned by Universal Music Group through its Interscope Geffen A&M imprint.
|
{
"evidence": [
"Tupac Shakur",
"2Pacalypse Now",
"Interscope Records"
],
"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:\nTupac Shakur is an African rapper who released his debut album on November 12, 1991, on an American record label owned by Universal Music Group through its Interscope Geffen A&M imprint.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tupac Shakur",
"2Pacalypse Now",
"Interscope Records"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1358,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
57944
|
In 1997, the infield playing surface of the Earl Wilson Stadium was replaced with a new fence that stands 12' high, and in 2007 the stadium received a new playing surface, however, up to this day the locker rooms are still the same.
|
{
"evidence": [
"Earl Wilson Stadium_sentence_5",
"Earl Wilson Stadium_sentence_8"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn 1997, the infield playing surface of the Earl Wilson Stadium was replaced with a new fence that stands 12' high, and in 2007 the stadium received a new playing surface, however, up to this day the locker rooms are still the same.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Earl Wilson Stadium_sentence_5",
"Earl Wilson Stadium_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1359,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.211",
"value": "?search="
},
{
"operation": "search",
"time": "13.659",
"value": "Earl Wilson Stadium"
},
{
"operation": "Now on",
"time": "14.342",
"value": "Earl Wilson Stadium"
},
{
"operation": "Highlighting",
"time": "19.878",
"value": "Earl Wilson Stadium_sentence_5"
},
{
"operation": "Highlighting",
"time": "55.792",
"value": "Earl Wilson Stadium_sentence_8"
},
{
"operation": "finish",
"time": "62.828",
"value": "finish"
}
]
}
|
40871
|
Jacob van der Kerckhoven (Spanish painter) is believed to have started his career as a architect of landscapes and genre scenes and he had a particular interest in scenes and still lifes of fowl.
|
{
"evidence": [
"Jacob van der Kerckhoven_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:\nJacob van der Kerckhoven (Spanish painter) is believed to have started his career as a architect of landscapes and genre scenes and he had a particular interest in scenes and still lifes of fowl.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Jacob van der Kerckhoven_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1360,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.578",
"value": "?search="
},
{
"operation": "search",
"time": "26.008",
"value": "Jacob van der Kerckhoven"
},
{
"operation": "Now on",
"time": "26.548",
"value": "Jacob van der Kerckhoven"
},
{
"operation": "Highlighting",
"time": "32.464",
"value": "Jacob van der Kerckhoven_sentence_0"
},
{
"operation": "finish",
"time": "40.687",
"value": "finish"
}
]
}
|
62577
|
Castle Lake loop is a hiking route in Bailieborough, Limerick, Ireland.
|
{
"evidence": [
"Castle Lake loop_sentence_1",
"Bailieborough_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:\nCastle Lake loop is a hiking route in Bailieborough, Limerick, Ireland.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Castle Lake loop_sentence_1",
"Bailieborough_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1361,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.016",
"value": "?search="
},
{
"operation": "search",
"time": "9.984",
"value": "Castle Lake loop"
},
{
"operation": "Now on",
"time": "10.983",
"value": "Castle Lake loop"
},
{
"operation": "Highlighting",
"time": "12.953",
"value": "Castle Lake loop_sentence_0"
},
{
"operation": "Highlighting",
"time": "13.691",
"value": "Castle Lake loop_sentence_1"
},
{
"operation": "hyperlink",
"time": "26.48",
"value": "Bailieborough"
},
{
"operation": "Now on",
"time": "27.067",
"value": "Bailieborough"
},
{
"operation": "Highlighting",
"time": "32.437",
"value": "Bailieborough_sentence_0"
},
{
"operation": "Page search",
"time": "41.861",
"value": "limerick"
},
{
"operation": "Page search",
"time": "42.254",
"value": "limerick"
},
{
"operation": "Page search",
"time": "42.396",
"value": "limerick"
},
{
"operation": "back-button-clicked",
"time": "45.194",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "45.251",
"value": "Castle Lake loop"
},
{
"operation": "Highlighting deleted",
"time": "47.107",
"value": "Castle Lake loop_sentence_0"
},
{
"operation": "Page search",
"time": "53.06",
"value": "limerick"
},
{
"operation": "finish",
"time": "68.297",
"value": "finish"
}
]
}
|
exfever_2096
|
Thomas William Hiddleston is the richest English actor.
|
{
"evidence": [
"Tom Hiddleston",
"Thor (film)",
"Superhero 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:\nThomas William Hiddleston is the richest English actor.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tom Hiddleston",
"Thor (film)",
"Superhero film"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1362,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7006
|
The Wolfman in 2010 is a remake of a horror film written by a Filipino-American novelist and screenwriter.
|
{
"evidence": [
"The Wolfman (2010 film)",
"The Wolf Man (1941 film)",
"Curt Siodmak"
],
"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 Wolfman in 2010 is a remake of a horror film written by a Filipino-American novelist and screenwriter.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Wolfman (2010 film)",
"The Wolf Man (1941 film)",
"Curt Siodmak"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1363,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6771
|
Kristofferson joined fellow country artist, an American musician, actor, and activist, in forming the country music supergroup, The Highwaymen.
|
{
"evidence": [
"Kris Kristofferson",
"Willie Nelson"
],
"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:\nKristofferson joined fellow country artist, an American musician, actor, and activist, in forming the country music supergroup, The Highwaymen.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Kris Kristofferson",
"Willie Nelson"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1364,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
1793
|
Zhang Wenli was the first son of Wang Rong who prevented a mutiny against the Wang clan, and was eventually killed by Li Cunxu.
|
{
"evidence": [
"Zhang Wenli_sentence_10",
"Zhang Wenli_sentence_3",
"Zhang Wenli_sentence_45",
"Zhang Wenli_sentence_46"
],
"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:\nZhang Wenli was the first son of Wang Rong who prevented a mutiny against the Wang clan, and was eventually killed by Li Cunxu.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Zhang Wenli_sentence_10",
"Zhang Wenli_sentence_3",
"Zhang Wenli_sentence_45",
"Zhang Wenli_sentence_46"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1365,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "11.563",
"value": "Zhang Wenli"
},
{
"operation": "Highlighting",
"time": "63.512",
"value": "Zhang Wenli_sentence_10"
},
{
"operation": "Highlighting",
"time": "114.669",
"value": "Zhang Wenli_sentence_3"
},
{
"operation": "Highlighting",
"time": "192.988",
"value": "Zhang Wenli_sentence_45"
},
{
"operation": "Highlighting",
"time": "202.382",
"value": "Zhang Wenli_sentence_46"
},
{
"operation": "finish",
"time": "250.575",
"value": "finish"
}
]
}
|
exfever_1628
|
Lebanon is a country in the westernmost subregion of the larger geographical region of Asia, almost entirely a part of the Middle East, which includes a country in Western Asia bordered to the west by Iraq and Turkey.
|
{
"evidence": [
"Lebanon",
"Western Asia",
"Iran"
],
"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:\nLebanon is a country in the westernmost subregion of the larger geographical region of Asia, almost entirely a part of the Middle East, which includes a country in Western Asia bordered to the west by Iraq and Turkey.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lebanon",
"Western Asia",
"Iran"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1366,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
33179
|
The court martial with the remaining accused began on 12 February 2018, but the hearing which was expected to last four weeks was soon terminated after the judge ruled that Royal Military Police (RMP) had abused the investigatory process and that a fair trial would therefore not be possible.
|
{
"evidence": [
"Army Foundation College recruit abuse investigation 2014–18_sentence_5",
"Army Foundation College recruit abuse investigation 2014–18_sentence_6"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe court martial with the remaining accused began on 12 February 2018, but the hearing which was expected to last four weeks was soon terminated after the judge ruled that Royal Military Police (RMP) had abused the investigatory process and that a fair trial would therefore not be possible.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Army Foundation College recruit abuse investigation 2014–18_sentence_5",
"Army Foundation College recruit abuse investigation 2014–18_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1367,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0",
"value": "?search="
},
{
"operation": "search",
"time": "0.029",
"value": "court martial with the remaining accused began"
},
{
"operation": "Now on",
"time": "0.031",
"value": "undefined"
},
{
"operation": "search",
"time": "0.08",
"value": "judge ruled that Royal Military Police"
},
{
"operation": "Now on",
"time": "0.082",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "0.086",
"value": "Army Foundation College recruit abuse investigation 2014–18"
},
{
"operation": "Now on",
"time": "0.086",
"value": "Army Foundation College recruit abuse investigation 2014–18"
},
{
"operation": "Highlighting",
"time": "0.102",
"value": "Army Foundation College recruit abuse investigation 2014–18_sentence_5"
},
{
"operation": "Highlighting",
"time": "0.113",
"value": "Army Foundation College recruit abuse investigation 2014–18_sentence_6"
},
{
"operation": "finish",
"time": "0.179",
"value": "finish"
},
{
"operation": "Now on",
"time": "1620325091.25",
"value": "Army Foundation College recruit abuse investigation 2014–18"
},
{
"operation": "finish",
"time": "1620325094.612",
"value": "finish"
}
]
}
|
exfever_6634
|
Salman Khan is an Indian actor who has received numerous awards as a film producer, including two prestigious awards at the most prestigious film award ceremony that is given by the government of India.
|
{
"evidence": [
"Salman Khan",
"National 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:\nSalman Khan is an Indian actor who has received numerous awards as a film producer, including two prestigious awards at the most prestigious film award ceremony that is given by the government of India.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Salman Khan",
"National Film Awards"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1368,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
77470
|
South Hadley Canal (operation commencing in 1795) funds were raised locally after John Hancock signed the charter.
|
{
"evidence": [
"South Hadley Canal_sentence_1",
"South Hadley Canal_sentence_5",
"South Hadley Canal_sentence_7"
],
"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:\nSouth Hadley Canal (operation commencing in 1795) funds were raised locally after John Hancock signed the charter.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"South Hadley Canal_sentence_1",
"South Hadley Canal_sentence_5",
"South Hadley Canal_sentence_7"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1369,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.2",
"value": "?search="
},
{
"operation": "search",
"time": "15.248",
"value": "South Hadley Canal"
},
{
"operation": "Now on",
"time": "16.082",
"value": "South Hadley Canal"
},
{
"operation": "Highlighting",
"time": "26.535",
"value": "South Hadley Canal_sentence_1"
},
{
"operation": "Page search",
"time": "38.751",
"value": "John Hancock"
},
{
"operation": "Highlighting",
"time": "42.581",
"value": "South Hadley Canal_sentence_5"
},
{
"operation": "Highlighting",
"time": "48.405",
"value": "South Hadley Canal_sentence_7"
},
{
"operation": "finish",
"time": "76.822",
"value": "finish"
}
]
}
|
exfever_2557
|
John Cowan Hartford was an American folk, country, and bluegrass composer and musician known for his mastery of a musical instrument that produces sound from a vibrating cord.
|
{
"evidence": [
"John Hartford",
"Fiddle",
"String instrument"
],
"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:\nJohn Cowan Hartford was an American folk, country, and bluegrass composer and musician known for his mastery of a musical instrument that produces sound from a vibrating cord.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"John Hartford",
"Fiddle",
"String instrument"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1370,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
32940
|
East Greenland Orogen ( also known as East Greenland mountain resort, is the circular mountain resort along the eastern Greenland coast, from 70 to 82 degrees north Katieude, Geologically, the mountain chain consists of Silurian to early Devonian (490 to 390 million years ago) fold and thrust belts.
|
{
"evidence": [
"East Greenland Orogen_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:\nEast Greenland Orogen ( also known as East Greenland mountain resort, is the circular mountain resort along the eastern Greenland coast, from 70 to 82 degrees north Katieude, Geologically, the mountain chain consists of Silurian to early Devonian (490 to 390 million years ago) fold and thrust belts.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"East Greenland Orogen_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1371,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.644",
"value": "?search="
},
{
"operation": "search",
"time": "7.183",
"value": "East Greenland Orogen"
},
{
"operation": "Now on",
"time": "7.81",
"value": "East Greenland Orogen"
},
{
"operation": "Highlighting",
"time": "10.926",
"value": "East Greenland Orogen_sentence_0"
},
{
"operation": "Page search",
"time": "18.586",
"value": "resort"
},
{
"operation": "Page search",
"time": "18.941",
"value": "resort"
},
{
"operation": "Page search",
"time": "19.105",
"value": "resort"
},
{
"operation": "Page search",
"time": "19.26",
"value": "resort"
},
{
"operation": "finish",
"time": "31.49",
"value": "finish"
}
]
}
|
90947
|
Saw pit is a pit where a lumber is positioned to be sawed by the sawyer (a person who cuts woods for a living).
|
{
"evidence": [
"Saw pit_sentence_1",
"Saw pit_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:\nSaw pit is a pit where a lumber is positioned to be sawed by the sawyer (a person who cuts woods for a living).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Saw pit_sentence_1",
"Saw pit_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1372,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.392",
"value": "?search="
},
{
"operation": "search",
"time": "26.131",
"value": "Saw pit"
},
{
"operation": "Now on",
"time": "27.067",
"value": "Saw pit"
},
{
"operation": "Highlighting",
"time": "96.338",
"value": "Saw pit_sentence_1"
},
{
"operation": "Highlighting",
"time": "117.772",
"value": "Saw pit_sentence_5"
},
{
"operation": "finish",
"time": "168.269",
"value": "finish"
}
]
}
|
48194
|
Jean Giraudeau took part in the premiere of Martine by Henri Rabaud (10 November 1873 – 11 September 1949) and made his debut at the Opéra-Comique where he later sang in Madame Bovary
|
{
"evidence": [
"Jean Giraudeau_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:\nJean Giraudeau took part in the premiere of Martine by Henri Rabaud (10 November 1873 – 11 September 1949) and made his debut at the Opéra-Comique where he later sang in Madame Bovary\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Jean Giraudeau_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1373,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.836",
"value": "?search="
},
{
"operation": "search",
"time": "48.132",
"value": "Contains...John Giraudeau"
},
{
"operation": "Now on",
"time": "48.816",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "50.897",
"value": "john girardeau"
},
{
"operation": "Now on",
"time": "51.733",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "52.745",
"value": "John Girardeau"
},
{
"operation": "Now on",
"time": "53.288",
"value": "John Girardeau"
},
{
"operation": "hyperlink",
"time": "78.688",
"value": "libertarianism"
},
{
"operation": "Now on",
"time": "79.595",
"value": "Libertarianism (metaphysics)"
},
{
"operation": "search",
"time": "88.679",
"value": "Jean Giraudeau"
},
{
"operation": "Now on",
"time": "89.597",
"value": "Jean Giraudeau"
},
{
"operation": "Page search",
"time": "106.465",
"value": "Martine"
},
{
"operation": "Highlighting",
"time": "108.89",
"value": "Jean Giraudeau_sentence_5"
},
{
"operation": "Page search",
"time": "119.913",
"value": "debut"
},
{
"operation": "Page search",
"time": "147.309",
"value": "Madame Bovary"
},
{
"operation": "Highlighting",
"time": "161.865",
"value": "Jean Giraudeau_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "162.825",
"value": "Jean Giraudeau_sentence_4"
},
{
"operation": "Highlighting",
"time": "163.761",
"value": "Jean Giraudeau_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "164.874",
"value": "Jean Giraudeau_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "165.753",
"value": "Jean Giraudeau_sentence_7"
},
{
"operation": "Highlighting",
"time": "166.665",
"value": "Jean Giraudeau_sentence_4"
},
{
"operation": "finish",
"time": "185.305",
"value": "finish"
}
]
}
|
31541
|
Iowa State Capitol, a 275 ft. high building, was the tallest building in Iowa from 1884 to 1924; then from 1991 up to the present, 801 Grand in Des Moines, which is 630 ft. high with 45 floors, holds the title.
|
{
"evidence": [
"Iowa State Capitol_sentence_57"
],
"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:\nIowa State Capitol, a 275 ft. high building, was the tallest building in Iowa from 1884 to 1924; then from 1991 up to the present, 801 Grand in Des Moines, which is 630 ft. high with 45 floors, holds the title.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Iowa State Capitol_sentence_57"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1374,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.063",
"value": "undefined"
},
{
"operation": "Now on",
"time": "24.613",
"value": "undefined"
},
{
"operation": "search",
"time": "125.226",
"value": "iowa state capitol"
},
{
"operation": "Now on",
"time": "126.384",
"value": "Iowa State Capitol"
},
{
"operation": "Highlighting",
"time": "258.848",
"value": "Iowa State Capitol_sentence_57"
},
{
"operation": "search",
"time": "376.425",
"value": "801 grand"
},
{
"operation": "Now on",
"time": "377.571",
"value": "801 Grand"
},
{
"operation": "finish",
"time": "481.613",
"value": "finish"
}
]
}
|
66122
|
Pennock Huey, was the son of Jacob Huey, a farmer in Chester County, Pennsylvania (one of the three original Pennsylvania counties created by William Penn in 1682), and pursued a career as a merchant before war broke out in 1861.
|
{
"evidence": [
"Pennock Huey_sentence_1",
"Chester County, Pennsylvania_sentence_5",
"Pennock Huey_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:\nPennock Huey, was the son of Jacob Huey, a farmer in Chester County, Pennsylvania (one of the three original Pennsylvania counties created by William Penn in 1682), and pursued a career as a merchant before war broke out in 1861.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Pennock Huey_sentence_1",
"Chester County, Pennsylvania_sentence_5",
"Pennock Huey_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1375,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "6.837",
"value": "?search="
},
{
"operation": "search",
"time": "216.668",
"value": "Pennock Huey"
},
{
"operation": "Now on",
"time": "218.382",
"value": "Pennock Huey"
},
{
"operation": "Highlighting",
"time": "226.865",
"value": "Pennock Huey_sentence_1"
},
{
"operation": "hyperlink",
"time": "229.689",
"value": "Chester County, Pennsylvania"
},
{
"operation": "Now on",
"time": "249.697",
"value": "Pennock Huey"
},
{
"operation": "hyperlink",
"time": "251.608",
"value": "Chester County, Pennsylvania"
},
{
"operation": "Now on",
"time": "255.084",
"value": "Chester County, Pennsylvania"
},
{
"operation": "Highlighting",
"time": "258.385",
"value": "Chester County, Pennsylvania_sentence_5"
},
{
"operation": "back-button-clicked",
"time": "259.778",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "259.862",
"value": "Pennock Huey"
},
{
"operation": "Highlighting",
"time": "263.825",
"value": "Pennock Huey_sentence_2"
},
{
"operation": "finish",
"time": "270.279",
"value": "finish"
}
]
}
|
22352
|
Portuguese sports club União Desportiva Oliveirense won the Segunda Divisão – Zona Centro ahead of 18 other teams, promoting them to Liga de Honra.
|
{
"evidence": [
"U.D. Oliveirense_sentence_18",
"U.D. Oliveirense_sentence_23",
"U.D. Oliveirense (basketball)_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:\nPortuguese sports club União Desportiva Oliveirense won the Segunda Divisão – Zona Centro ahead of 18 other teams, promoting them to Liga de Honra.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"U.D. Oliveirense_sentence_18",
"U.D. Oliveirense_sentence_23",
"U.D. Oliveirense (basketball)_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1376,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.63",
"value": "undefined"
},
{
"operation": "search",
"time": "32.528",
"value": "Segunda Divisão – Zona Centro"
},
{
"operation": "Now on",
"time": "34.031",
"value": "undefined"
},
{
"operation": "search",
"time": "83.315",
"value": "U.D. Oliveirense"
},
{
"operation": "Now on",
"time": "84.211",
"value": "U.D. Oliveirense"
},
{
"operation": "Page search",
"time": "87.543",
"value": "promote"
},
{
"operation": "Highlighting",
"time": "93.512",
"value": "U.D. Oliveirense_sentence_18"
},
{
"operation": "Page search",
"time": "197.431",
"value": "honra"
},
{
"operation": "Highlighting",
"time": "202.612",
"value": "U.D. Oliveirense_sentence_23"
},
{
"operation": "hyperlink",
"time": "421.216",
"value": "U.D. Oliveirense (basketball)"
},
{
"operation": "Now on",
"time": "422.226",
"value": "Oliveirense Basquetebol"
},
{
"operation": "Page search",
"time": "426.918",
"value": "honra"
},
{
"operation": "Highlighting",
"time": "439.6",
"value": "U.D. Oliveirense (basketball)_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "443.956",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "443.993",
"value": "U.D. Oliveirense"
},
{
"operation": "search",
"time": "465.42",
"value": "U. D. Oliveirense"
},
{
"operation": "Now on",
"time": "466.132",
"value": "U.D. Oliveirense"
},
{
"operation": "search",
"time": "479.471",
"value": "u. d. oliveirense football"
},
{
"operation": "Now on",
"time": "480.954",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "488.735",
"value": "U.D. Oliveirense (basketball)"
},
{
"operation": "Now on",
"time": "489.26",
"value": "U.D. Oliveirense (basketball)"
},
{
"operation": "hyperlink",
"time": "493.836",
"value": "football"
},
{
"operation": "Now on",
"time": "494.698",
"value": "Association football"
},
{
"operation": "finish",
"time": "537.506",
"value": "finish"
}
]
}
|
exfever_6642
|
Leonard Norman Cohen was a Canadian singer-songwriter, poet and novelist who was inducted into the Canadian Music Hall of Fame and in the Canadian non-profit organization founded in 1998 by Frank Davies.
|
{
"evidence": [
"Leonard Cohen",
"Canadian Songwriters Hall of Fame"
],
"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:\nLeonard Norman Cohen was a Canadian singer-songwriter, poet and novelist who was inducted into the Canadian Music Hall of Fame and in the Canadian non-profit organization founded in 1998 by Frank Davies.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Leonard Cohen",
"Canadian Songwriters Hall of Fame"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1377,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7530
|
"Me Before You" is a romantic drama film starring an Asian actress who had her breakthrough role as a fictional character in the series of epic fantasy novels, A Song of Ice and Fire.
|
{
"evidence": [
"Me Before You (film)",
"Emilia Clarke",
"Daenerys Targaryen"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\n\"Me Before You\" is a romantic drama film starring an Asian actress who had her breakthrough role as a fictional character in the series of epic fantasy novels, A Song of Ice and Fire.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Me Before You (film)",
"Emilia Clarke",
"Daenerys Targaryen"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1378,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4327
|
Solving the optically inactive form of tartaric acid was the greatest discovery of Louis Pasteur.
|
{
"evidence": [
"Adderall",
"Racemic mixture",
"Racemic acid"
],
"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:\nSolving the optically inactive form of tartaric acid was the greatest discovery of Louis Pasteur.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Adderall",
"Racemic mixture",
"Racemic acid"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1379,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1783
|
Jemaine Atea Mahana Clement created a comedy series for Home Box Office.
|
{
"evidence": [
"Jemaine Clement",
"HBO"
],
"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:\nJemaine Atea Mahana Clement created a comedy series for Home Box Office.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jemaine Clement",
"HBO"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1380,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3458
|
Jennifer Grey is an American actress who made her breakthrough with the 1986 American children's comedy film written, co-produced, and directed by John Hughes.
|
{
"evidence": [
"Jennifer Grey",
"Ferris Bueller's Day Off"
],
"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:\nJennifer Grey is an American actress who made her breakthrough with the 1986 American children's comedy film written, co-produced, and directed by John Hughes.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jennifer Grey",
"Ferris Bueller's Day Off"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1381,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7942
|
Michael Kelland John Hutchence recorded some solo material and acted in feature films, including a 1990 science fiction horror film based on Brian Aldiss' 1973 novel of the same name, starring an English actor whose career spanned over five decades.
|
{
"evidence": [
"Michael Hutchence",
"Frankenstein Unbound",
"John Hurt"
],
"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:\nMichael Kelland John Hutchence recorded some solo material and acted in feature films, including a 1990 science fiction horror film based on Brian Aldiss' 1973 novel of the same name, starring an English actor whose career spanned over five decades.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Michael Hutchence",
"Frankenstein Unbound",
"John Hurt"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1382,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
25344
|
BraunAbility established two new divisions in 1977 before a fire swept through their headquarters in 1979.
|
{
"evidence": [
"BraunAbility_sentence_21",
"BraunAbility_sentence_22"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBraunAbility established two new divisions in 1977 before a fire swept through their headquarters in 1979.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"BraunAbility_sentence_21",
"BraunAbility_sentence_22"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1383,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.183",
"value": "?search="
},
{
"operation": "search",
"time": "21.125",
"value": "BraunAbility"
},
{
"operation": "Now on",
"time": "22.356",
"value": "BraunAbility"
},
{
"operation": "Highlighting",
"time": "48.385",
"value": "BraunAbility_sentence_21"
},
{
"operation": "Highlighting",
"time": "51.353",
"value": "BraunAbility_sentence_22"
},
{
"operation": "finish",
"time": "77.455",
"value": "finish"
}
]
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.