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 |
|---|---|---|---|---|---|---|---|---|
13738
|
For Malaysiakini first story, posted a report on 20 November criticising the practices of Sin Chew Daily and made news again when it discovered and reported the secret detention of 10 political activists.
|
{
"evidence": [
"Malaysiakini_sentence_10",
"Malaysiakini_sentence_13"
],
"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:\nFor Malaysiakini first story, posted a report on 20 November criticising the practices of Sin Chew Daily and made news again when it discovered and reported the secret detention of 10 political activists.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Malaysiakini_sentence_10",
"Malaysiakini_sentence_13"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6384,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.587",
"value": "?search="
},
{
"operation": "search",
"time": "36.368",
"value": "Malaysiakini"
},
{
"operation": "Now on",
"time": "37.242",
"value": "Malaysiakini"
},
{
"operation": "Highlighting",
"time": "51.939",
"value": "Malaysiakini_sentence_10"
},
{
"operation": "Highlighting",
"time": "66.801",
"value": "Malaysiakini_sentence_13"
},
{
"operation": "finish",
"time": "75.197",
"value": "finish"
}
]
}
|
exfever_823
|
Lost Highway is a 1997 neo-noir film, starring an American actor who also starred in a 1987 American space film directed by Mel Gibson.
|
{
"evidence": [
"Lost Highway (film)",
"Bill Pullman",
"Spaceballs"
],
"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:\nLost Highway is a 1997 neo-noir film, starring an American actor who also starred in a 1987 American space film directed by Mel Gibson.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lost Highway (film)",
"Bill Pullman",
"Spaceballs"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6385,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
79657
|
Barbara Corlett was a staunch supporter of women's rights and education, establishing a Society for Promoting the Employment of Women to further her main goal of bringing attention to the urgent need to provide training to women.
|
{
"evidence": [
"Barbara Corlett_sentence_22",
"Barbara Corlett_sentence_3",
"Barbara Corlett_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:\nBarbara Corlett was a staunch supporter of women's rights and education, establishing a Society for Promoting the Employment of Women to further her main goal of bringing attention to the urgent need to provide training to women.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Barbara Corlett_sentence_22",
"Barbara Corlett_sentence_3",
"Barbara Corlett_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6386,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.903",
"value": "?search="
},
{
"operation": "search",
"time": "11.889",
"value": "Barbara Corlett"
},
{
"operation": "Now on",
"time": "12.871",
"value": "Barbara Corlett"
},
{
"operation": "Page search",
"time": "29.273",
"value": "staunch"
},
{
"operation": "Highlighting",
"time": "32.946",
"value": "Barbara Corlett_sentence_22"
},
{
"operation": "Highlighting",
"time": "45.898",
"value": "Barbara Corlett_sentence_3"
},
{
"operation": "Highlighting",
"time": "54.804",
"value": "Barbara Corlett_sentence_5"
},
{
"operation": "finish",
"time": "58.489",
"value": "finish"
}
]
}
|
exfever_8080
|
To Kill a Mockingbird is a novel by the American author Harper Lee, loosely based on events that occurred in the county seat of Monroe County.
|
{
"evidence": [
"To Kill a Mockingbird",
"Monroeville, Alabama"
],
"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:\nTo Kill a Mockingbird is a novel by the American author Harper Lee, loosely based on events that occurred in the county seat of Monroe County.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"To Kill a Mockingbird",
"Monroeville, Alabama"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6387,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8344
|
The Martian is a 2015 fictional science-based stylistic category for motion pictures.
|
{
"evidence": [
"The Martian (film)",
"Science fiction film",
"Film 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:\nThe Martian is a 2015 fictional science-based stylistic category for motion pictures.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Martian (film)",
"Science fiction film",
"Film genre"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6388,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2678
|
Woody Harrelson is a British actor known for his role in the NBC sitcom created by American screenwriters best known for the movie Taxi.
|
{
"evidence": [
"Woody Harrelson",
"Cheers",
"Glen and Les Charles"
],
"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:\nWoody Harrelson is a British actor known for his role in the NBC sitcom created by American screenwriters best known for the movie Taxi.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Woody Harrelson",
"Cheers",
"Glen and Les Charles"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6389,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4739
|
Coldplay is the tenth best-selling British band worldwide.
|
{
"evidence": [
"Coldplay",
"Rock music"
],
"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:\nColdplay is the tenth best-selling British band worldwide.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Coldplay",
"Rock music"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6390,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
27603
|
Dipeptidases dehydrate (to break a bond apart with water) bound pairs of amino acids called dipeptides and can be found within the enterocytes themselves.
|
{
"evidence": [
"Dipeptidase_sentence_1",
"Dipeptidase_sentence_4",
"Hydrolysis_sentence_7",
"Hydrolysis_sentence_10"
],
"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:\nDipeptidases dehydrate (to break a bond apart with water) bound pairs of amino acids called dipeptides and can be found within the enterocytes themselves.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Dipeptidase_sentence_1",
"Dipeptidase_sentence_4",
"Hydrolysis_sentence_7",
"Hydrolysis_sentence_10"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6391,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.665",
"value": "?search="
},
{
"operation": "search",
"time": "11.606",
"value": "Contains...Dipeptidases dehydrate"
},
{
"operation": "Now on",
"time": "12.076",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "15.662",
"value": "dipeptidase dehydratase"
},
{
"operation": "Now on",
"time": "16.238",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "19.637",
"value": "List of MeSH codes (D08)"
},
{
"operation": "Now on",
"time": "21.944",
"value": "List of MeSH codes (D08)"
},
{
"operation": "search",
"time": "51.854",
"value": "Dipeptidases"
},
{
"operation": "Now on",
"time": "52.503",
"value": "Dipeptidase"
},
{
"operation": "Highlighting",
"time": "63.238",
"value": "Dipeptidase_sentence_1"
},
{
"operation": "Highlighting",
"time": "76.438",
"value": "Dipeptidase_sentence_4"
},
{
"operation": "search",
"time": "86.264",
"value": "Hydrolyzed"
},
{
"operation": "Now on",
"time": "86.968",
"value": "Hydrolysis"
},
{
"operation": "Highlighting",
"time": "122.589",
"value": "Hydrolysis_sentence_7"
},
{
"operation": "Highlighting",
"time": "123.174",
"value": "Hydrolysis_sentence_10"
},
{
"operation": "search",
"time": "131.193",
"value": "Dehydration"
},
{
"operation": "Now on",
"time": "132.104",
"value": "Dehydration"
},
{
"operation": "finish",
"time": "170.799",
"value": "finish"
}
]
}
|
44931
|
Ursula Holliger was an important harpist who always dreamed of being able to teach at Hochschule für Musik Freiburg.
|
{
"evidence": [
"Ursula Holliger_sentence_0",
"Ursula Holliger_sentence_3",
"Ursula Holliger_sentence_6"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nUrsula Holliger was an important harpist who always dreamed of being able to teach at Hochschule für Musik Freiburg.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ursula Holliger_sentence_0",
"Ursula Holliger_sentence_3",
"Ursula Holliger_sentence_6"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6392,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.363",
"value": "?search="
},
{
"operation": "search",
"time": "17.904",
"value": "Ursula Holliger"
},
{
"operation": "Now on",
"time": "18.624",
"value": "Ursula Holliger"
},
{
"operation": "Highlighting",
"time": "46.008",
"value": "Ursula Holliger_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "46.444",
"value": "Ursula Holliger_sentence_0"
},
{
"operation": "Highlighting",
"time": "49.821",
"value": "Ursula Holliger_sentence_0"
},
{
"operation": "Highlighting",
"time": "142.947",
"value": "Ursula Holliger_sentence_3"
},
{
"operation": "Page search",
"time": "177.275",
"value": "dream"
},
{
"operation": "Highlighting",
"time": "185.096",
"value": "Ursula Holliger_sentence_6"
},
{
"operation": "finish",
"time": "229.609",
"value": "finish"
}
]
}
|
33083
|
Julia Zorn was selected for the Germany women's national ice hockey team in the 2014 Winter Olympics and played in all five games scoring no goal.
|
{
"evidence": [
"Julia Zorn_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nJulia Zorn was selected for the Germany women's national ice hockey team in the 2014 Winter Olympics and played in all five games scoring no goal.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Julia Zorn_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6393,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.176",
"value": "?search="
},
{
"operation": "search",
"time": "11.419",
"value": "Julia Zorn"
},
{
"operation": "Now on",
"time": "13.066",
"value": "Julia Zorn"
},
{
"operation": "hyperlink",
"time": "22.818",
"value": "2014 Winter Olympics"
},
{
"operation": "Now on",
"time": "25.623",
"value": "Julia Zorn"
},
{
"operation": "Highlighting",
"time": "26.959",
"value": "Julia Zorn_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "31.117",
"value": "Julia Zorn_sentence_5"
},
{
"operation": "hyperlink",
"time": "32.149",
"value": "2014 Winter Olympics"
},
{
"operation": "Now on",
"time": "34.393",
"value": "Julia Zorn"
},
{
"operation": "Highlighting",
"time": "37.012",
"value": "Julia Zorn_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "39.626",
"value": "Julia Zorn_sentence_4"
},
{
"operation": "Highlighting",
"time": "40.461",
"value": "Julia Zorn_sentence_5"
},
{
"operation": "finish",
"time": "56.831",
"value": "finish"
}
]
}
|
exfever_7340
|
Silicon Valley is a region in a northern state comprising the easternmost states.
|
{
"evidence": [
"Silicon Valley",
"California",
"Western United States"
],
"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:\nSilicon Valley is a region in a northern state comprising the easternmost states.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Silicon Valley",
"California",
"Western United States"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6394,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5513
|
Alison Brie Schermerhorn is an African actress whose breakthrough role as Trudy Campbell in the drama series on American television produced by Lionsgate Television and created by best known as a writer and executive producer of The Sopranos.
|
{
"evidence": [
"Alison Brie",
"Mad Men",
"Matthew Weiner"
],
"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:\nAlison Brie Schermerhorn is an African actress whose breakthrough role as Trudy Campbell in the drama series on American television produced by Lionsgate Television and created by best known as a writer and executive producer of The Sopranos.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Alison Brie",
"Mad Men",
"Matthew Weiner"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6395,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4781
|
Chris Evans was the youngest cast member of the television show Opposite Sex.
|
{
"evidence": [
"Chris Evans (actor)",
"Comic book"
],
"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:\nChris Evans was the youngest cast member of the television show Opposite Sex.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Chris Evans (actor)",
"Comic book"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6396,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
36812
|
Petter Eliassen skied professionally throughout his 20s.
|
{
"evidence": [
"Petter Eliassen_sentence_0",
"Petter Eliassen_sentence_1",
"Petter Eliassen_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:\nPetter Eliassen skied professionally throughout his 20s.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Petter Eliassen_sentence_0",
"Petter Eliassen_sentence_1",
"Petter Eliassen_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6397,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.383",
"value": "?search="
},
{
"operation": "search",
"time": "45.795",
"value": "Petter Eliassen"
},
{
"operation": "Now on",
"time": "46.948",
"value": "Petter Eliassen"
},
{
"operation": "Highlighting",
"time": "71.843",
"value": "Petter Eliassen_sentence_0"
},
{
"operation": "Highlighting",
"time": "76.938",
"value": "Petter Eliassen_sentence_1"
},
{
"operation": "Highlighting",
"time": "77.522",
"value": "Petter Eliassen_sentence_2"
},
{
"operation": "finish",
"time": "128.541",
"value": "finish"
}
]
}
|
53332
|
Jeff and Marty meet with the nun to obtain evidence after she informs Jeff that her convent’s funds are being diverted by their accountant.
|
{
"evidence": [
"You Can Always Find a Fall Guy_sentence_4",
"You Can Always Find a Fall Guy_sentence_5",
"You Can Always Find a Fall Guy_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:\nJeff and Marty meet with the nun to obtain evidence after she informs Jeff that her convent’s funds are being diverted by their accountant.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"You Can Always Find a Fall Guy_sentence_4",
"You Can Always Find a Fall Guy_sentence_5",
"You Can Always Find a Fall Guy_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6398,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.699",
"value": "undefined"
},
{
"operation": "search",
"time": "20.827",
"value": "jeff and marty nun"
},
{
"operation": "Now on",
"time": "22.694",
"value": "undefined"
},
{
"operation": "search",
"time": "64.1",
"value": "insource"
},
{
"operation": "Now on",
"time": "65.475",
"value": "undefined"
},
{
"operation": "search",
"time": "122.983",
"value": "jeff and marty"
},
{
"operation": "Now on",
"time": "125.447",
"value": "undefined"
},
{
"operation": "search",
"time": "174.255",
"value": "insource"
},
{
"operation": "Now on",
"time": "176.078",
"value": "undefined"
},
{
"operation": "search",
"time": "205.643",
"value": "you can always find a fall guy"
},
{
"operation": "Now on",
"time": "207.449",
"value": "You Can Always Find a Fall Guy"
},
{
"operation": "Highlighting",
"time": "239.405",
"value": "You Can Always Find a Fall Guy_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "251.924",
"value": "You Can Always Find a Fall Guy_sentence_4"
},
{
"operation": "Highlighting",
"time": "252.494",
"value": "You Can Always Find a Fall Guy_sentence_3"
},
{
"operation": "Highlighting",
"time": "253.2",
"value": "You Can Always Find a Fall Guy_sentence_4"
},
{
"operation": "Highlighting",
"time": "255.666",
"value": "You Can Always Find a Fall Guy_sentence_5"
},
{
"operation": "Highlighting",
"time": "261.23",
"value": "You Can Always Find a Fall Guy_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "264.368",
"value": "You Can Always Find a Fall Guy_sentence_3"
},
{
"operation": "finish",
"time": "319.702",
"value": "finish"
}
]
}
|
84024
|
The Eastern Michigan Hurons, under head coach Dan Boisture, compiled a 6-4 record in the 1973 NCAA Division II football season.
|
{
"evidence": [
"1973 Eastern Michigan Hurons football team_sentence_1",
"1973 Eastern Michigan Hurons football team_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Eastern Michigan Hurons, under head coach Dan Boisture, compiled a 6-4 record in the 1973 NCAA Division II football season.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"1973 Eastern Michigan Hurons football team_sentence_1",
"1973 Eastern Michigan Hurons football team_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6399,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.834",
"value": "?search="
},
{
"operation": "search",
"time": "25.547",
"value": "1973 Eastern Michigan Hurons football team"
},
{
"operation": "Now on",
"time": "26.312",
"value": "1973 Eastern Michigan Hurons football team"
},
{
"operation": "Highlighting",
"time": "50.054",
"value": "1973 Eastern Michigan Hurons football team_sentence_1"
},
{
"operation": "Highlighting",
"time": "59.194",
"value": "1973 Eastern Michigan Hurons football team_sentence_0"
},
{
"operation": "finish",
"time": "74.37",
"value": "finish"
}
]
}
|
51425
|
The story written by Cho Jeongrae (a South Korean film director born October 15, 1973) tries to describe and reveal a few generations-lasting conflicts between the proprietors and peasants and provided further insight into the politically controversial ideological issue over the '40s in South Korea.
|
{
"evidence": [
"The Taebaek Mountains_sentence_1",
"The Taebaek Mountains_sentence_2",
"The Taebaek Mountains_sentence_4"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe story written by Cho Jeongrae (a South Korean film director born October 15, 1973) tries to describe and reveal a few generations-lasting conflicts between the proprietors and peasants and provided further insight into the politically controversial ideological issue over the '40s in South Korea.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"The Taebaek Mountains_sentence_1",
"The Taebaek Mountains_sentence_2",
"The Taebaek Mountains_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6400,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.934",
"value": "?search="
},
{
"operation": "search",
"time": "55.704",
"value": "Cho Jeongrae"
},
{
"operation": "Now on",
"time": "56.836",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "73.862",
"value": "The Taebaek Mountains"
},
{
"operation": "Now on",
"time": "74.893",
"value": "The Taebaek Mountains"
},
{
"operation": "Highlighting",
"time": "125.483",
"value": "The Taebaek Mountains_sentence_1"
},
{
"operation": "Highlighting",
"time": "139.45",
"value": "The Taebaek Mountains_sentence_2"
},
{
"operation": "Highlighting",
"time": "203.175",
"value": "The Taebaek Mountains_sentence_4"
},
{
"operation": "Highlighting",
"time": "222.413",
"value": "The Taebaek Mountains_header_cell_0_7_0"
},
{
"operation": "Highlighting",
"time": "225.944",
"value": "The Taebaek Mountains_cell_0_7_1"
},
{
"operation": "Highlighting",
"time": "225.968",
"value": "The Taebaek Mountains_header_cell_0_7_0"
},
{
"operation": "Highlighting",
"time": "225.968",
"value": "The Taebaek Mountains_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "229.331",
"value": "The Taebaek Mountains_cell_0_7_1"
},
{
"operation": "Highlighting deleted",
"time": "229.34",
"value": "The Taebaek Mountains_header_cell_0_7_0"
},
{
"operation": "Highlighting deleted",
"time": "229.342",
"value": "The Taebaek Mountains_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "232.17",
"value": "The Taebaek Mountains_cell_0_7_1"
},
{
"operation": "Highlighting",
"time": "232.181",
"value": "The Taebaek Mountains_header_cell_0_7_0"
},
{
"operation": "Highlighting",
"time": "232.181",
"value": "The Taebaek Mountains_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "240.999",
"value": "The Taebaek Mountains_cell_0_7_1"
},
{
"operation": "Highlighting deleted",
"time": "241.016",
"value": "The Taebaek Mountains_header_cell_0_7_0"
},
{
"operation": "Highlighting deleted",
"time": "241.018",
"value": "The Taebaek Mountains_header_cell_0_0_0"
},
{
"operation": "search",
"time": "469.256",
"value": "Cho Jeongrae"
},
{
"operation": "Now on",
"time": "470.473",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "491.715",
"value": "cho jeong jae"
},
{
"operation": "Now on",
"time": "493.454",
"value": "undefined"
},
{
"operation": "search",
"time": "529.743",
"value": "Cho Jeong-rae"
},
{
"operation": "Now on",
"time": "530.797",
"value": "undefined"
},
{
"operation": "search",
"time": "741.159",
"value": "Cho Jeong-rae"
},
{
"operation": "Now on",
"time": "741.913",
"value": "undefined"
},
{
"operation": "finish",
"time": "777.391",
"value": "finish"
}
]
}
|
exfever_3875
|
Black Panther is a 2018 American superhero film distributed by the American film distribution company that promotes films by the American computer animation studio based in Oakland, California.
|
{
"evidence": [
"Black Panther (film)",
"Walt Disney Studios Motion Pictures",
"Pixar"
],
"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:\nBlack Panther is a 2018 American superhero film distributed by the American film distribution company that promotes films by the American computer animation studio based in Oakland, California.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Black Panther (film)",
"Walt Disney Studios Motion Pictures",
"Pixar"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6401,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
51769
|
Stephen Blair Hedges is the director of the Center for Biodiversity at Temple University where he researches the tree of life, or universal tree of life which is a model and research tool used to explore the evolution of life and describe the relationships between organisms, both living and extinct, as described in a famous passage in Charles Darwin's On the Origin of Species (1859).
|
{
"evidence": [
"Stephen Blair Hedges_sentence_0",
"Tree of life (biology)_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:\nStephen Blair Hedges is the director of the Center for Biodiversity at Temple University where he researches the tree of life, or universal tree of life which is a model and research tool used to explore the evolution of life and describe the relationships between organisms, both living and extinct, as described in a famous passage in Charles Darwin's On the Origin of Species (1859).\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Stephen Blair Hedges_sentence_0",
"Tree of life (biology)_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6402,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.339",
"value": "?search="
},
{
"operation": "search",
"time": "13.62",
"value": "Stephen Blair Hedges"
},
{
"operation": "Now on",
"time": "14.604",
"value": "Stephen Blair Hedges"
},
{
"operation": "Highlighting",
"time": "28.396",
"value": "Stephen Blair Hedges_sentence_0"
},
{
"operation": "hyperlink",
"time": "30.55",
"value": "tree of life"
},
{
"operation": "Now on",
"time": "31.219",
"value": "Tree of life (biology)"
},
{
"operation": "Highlighting",
"time": "48.135",
"value": "Tree of life (biology)_sentence_1"
},
{
"operation": "finish",
"time": "57.05",
"value": "finish"
}
]
}
|
exfever_7467
|
To Die For is a 2005 film directed by Gus Van Sant based on the novel of the same name by an American novelist who began her career in journalism in the 1970s, writing for several publications.
|
{
"evidence": [
"To Die For",
"Joyce Maynard"
],
"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:\nTo Die For is a 2005 film directed by Gus Van Sant based on the novel of the same name by an American novelist who began her career in journalism in the 1970s, writing for several publications.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"To Die For",
"Joyce Maynard"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6403,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
7443
|
"Jerry Was a Race Car Driver" is the 1991 fourth single by the American funk metal band Primus form the album Sailing the Seas of Cheese, that features Bill Moseley's character Chop Top and received heavy airplay on rock radio, and peaked at number 23 on the Modern Rock Tracks in 1991.
|
{
"evidence": [
"Jerry Was a Race Car Driver_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:\n\"Jerry Was a Race Car Driver\" is the 1991 fourth single by the American funk metal band Primus form the album Sailing the Seas of Cheese, that features Bill Moseley's character Chop Top and received heavy airplay on rock radio, and peaked at number 23 on the Modern Rock Tracks in 1991.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Jerry Was a Race Car Driver_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6404,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Highlighting",
"time": "0",
"value": "Coroner (TV series)_sentence_5"
},
{
"operation": "Highlighting",
"time": "0",
"value": "Coroner (TV series)_title"
},
{
"operation": "Now on",
"time": "0",
"value": "?search="
},
{
"operation": "search",
"time": "0",
"value": "Jerry Was a Race Car Driver"
},
{
"operation": "Now on",
"time": "0",
"value": "Jerry Was a Race Car Driver"
},
{
"operation": "Highlighting",
"time": "0",
"value": "Jerry Was a Race Car Driver_sentence_1"
},
{
"operation": "finish",
"time": "0",
"value": "finish"
},
{
"operation": "Now on",
"time": "1619029.56",
"value": "Marie Dompnier"
},
{
"operation": "finish",
"time": "1619029.564",
"value": "finish"
},
{
"operation": "Now on",
"time": "1619029594.071",
"value": "Marie Dompnier"
},
{
"operation": "finish",
"time": "1619029601.036",
"value": "finish"
}
]
}
|
4024
|
Chi2 Orionis (right ascension 06 03 55.18482) is a star in the Little Dipper, with an visual magnitude of 4.63 and the star serves as an anchor point that stars are classified.
|
{
"evidence": [
"Little Dipper (disambiguation)_sentence_0",
"Chi2 Orionis_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:\nChi2 Orionis (right ascension 06 03 55.18482) is a star in the Little Dipper, with an visual magnitude of 4.63 and the star serves as an anchor point that stars are classified.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Little Dipper (disambiguation)_sentence_0",
"Chi2 Orionis_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6405,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "14.708",
"value": "Chi2 Orionis"
},
{
"operation": "hyperlink",
"time": "33.589",
"value": "Orion"
},
{
"operation": "search",
"time": "57.413",
"value": "Little Dipper"
},
{
"operation": "search",
"time": "72.403",
"value": "little dipper"
},
{
"operation": "search",
"time": "80.574",
"value": "Little Dipper (disambiguation)"
},
{
"operation": "Highlighting",
"time": "92.484",
"value": "Little Dipper (disambiguation)_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "99.031",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "100.993",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "100.998",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "102.364",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "102.37",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "105.054",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "105.058",
"value": "back-button-clicked"
},
{
"operation": "Highlighting",
"time": "108.236",
"value": "Chi2 Orionis_sentence_1"
},
{
"operation": "finish",
"time": "125.353",
"value": "finish"
}
]
}
|
exfever_7284
|
Cersei Lannister is a fictional character in the series of epic fantasy novels by the American novelist and screenwriter George R. R. Martin, which has a fourth of seven planned novels in the epic fantasy series, A Song of Ice and Fire.
|
{
"evidence": [
"Cersei Lannister",
"A Song of Ice and Fire",
"A Feast for Crows"
],
"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:\nCersei Lannister is a fictional character in the series of epic fantasy novels by the American novelist and screenwriter George R. R. Martin, which has a fourth of seven planned novels in the epic fantasy series, A Song of Ice and Fire.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Cersei Lannister",
"A Song of Ice and Fire",
"A Feast for Crows"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6406,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8722
|
The Prestige is a 2006 thriller film directed by Christopher Nolan whose genre revolves around the solution of a problem or a crime.
|
{
"evidence": [
"The Prestige (film)",
"Mystery film"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Prestige is a 2006 thriller film directed by Christopher Nolan whose genre revolves around the solution of a problem or a crime.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Prestige (film)",
"Mystery film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6407,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
49304
|
Saint Teresa of Avila School, neither a private, nor a co-educational school established in 1960, was accredited by the Middle States Association of Colleges and Schools.
|
{
"evidence": [
"Saint Teresa of Avila School_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:\nSaint Teresa of Avila School, neither a private, nor a co-educational school established in 1960, was accredited by the Middle States Association of Colleges and Schools.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Saint Teresa of Avila School_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6408,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.776",
"value": "?search="
},
{
"operation": "search",
"time": "12.032",
"value": "Saint Teresa of Avila School"
},
{
"operation": "Now on",
"time": "12.819",
"value": "Saint Teresa of Avila School"
},
{
"operation": "Highlighting",
"time": "25.729",
"value": "Saint Teresa of Avila School_sentence_1"
},
{
"operation": "finish",
"time": "98.548",
"value": "finish"
}
]
}
|
exfever_3496
|
Attack on Titan is a Japanese comic that originated in Japan that was written and illustrated by Hajime Isayama.
|
{
"evidence": [
"Attack on Titan",
"Manga"
],
"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:\nAttack on Titan is a Japanese comic that originated in Japan that was written and illustrated by Hajime Isayama.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Attack on Titan",
"Manga"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6409,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
64272
|
Francisco Camilo (Madrid, 1615 - ibídem, 1673) was the son of Domenico Camilo and a painter who produced a series of paintings of Kings of Spain for the theater of Buenretiro and adorn the western gallery of the palace with 14 frescoes from Ovid's Metamorphoses.
|
{
"evidence": [
"Francisco Camilo_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:\nFrancisco Camilo (Madrid, 1615 - ibídem, 1673) was the son of Domenico Camilo and a painter who produced a series of paintings of Kings of Spain for the theater of Buenretiro and adorn the western gallery of the palace with 14 frescoes from Ovid's Metamorphoses.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Francisco Camilo_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6410,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.758",
"value": "?search="
},
{
"operation": "search",
"time": "22.196",
"value": "Francisco Camilo"
},
{
"operation": "Now on",
"time": "23.302",
"value": "Francisco Camilo"
},
{
"operation": "Highlighting",
"time": "226.807",
"value": "Francisco Camilo_sentence_0"
},
{
"operation": "finish",
"time": "324.73",
"value": "finish"
}
]
}
|
45219
|
Though being rejected as an entry of the National Register of Historic Places, Pine Crest Inn is a historic restort inn complex that was built in 1906 where inn and three of the ten cottages were for the Thermal Belt Sanatorium for tuberculosis care, then the remaining cottages were built after the conversion of the property to an inn in 1917.
|
{
"evidence": [
"Pine Crest Inn_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThough being rejected as an entry of the National Register of Historic Places, Pine Crest Inn is a historic restort inn complex that was built in 1906 where inn and three of the ten cottages were for the Thermal Belt Sanatorium for tuberculosis care, then the remaining cottages were built after the conversion of the property to an inn in 1917.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Pine Crest Inn_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6411,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.831",
"value": "?search="
},
{
"operation": "search",
"time": "10.976",
"value": "Pine Crest Inn"
},
{
"operation": "Now on",
"time": "11.852",
"value": "Pine Crest Inn"
},
{
"operation": "Highlighting",
"time": "14.913",
"value": "Pine Crest Inn_cell_0_6_1"
},
{
"operation": "Highlighting",
"time": "14.917",
"value": "Pine Crest Inn_header_cell_0_6_0"
},
{
"operation": "Highlighting",
"time": "16.848",
"value": "Pine Crest Inn_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "41.162",
"value": "Pine Crest Inn_cell_0_6_1"
},
{
"operation": "Highlighting deleted",
"time": "41.168",
"value": "Pine Crest Inn_header_cell_0_6_0"
},
{
"operation": "finish",
"time": "50.509",
"value": "finish"
}
]
}
|
75384
|
Fettiplace included John Fettiplace who was a London draper who possessed the manors of East Shefford and had two sons.
|
{
"evidence": [
"Fettiplace_sentence_10"
],
"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:\nFettiplace included John Fettiplace who was a London draper who possessed the manors of East Shefford and had two sons.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Fettiplace_sentence_10"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6412,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.419",
"value": "?search="
},
{
"operation": "search",
"time": "10.885",
"value": "Fettiplace"
},
{
"operation": "Now on",
"time": "12.334",
"value": "Fettiplace"
},
{
"operation": "Page search",
"time": "36.378",
"value": "John Fettiplace"
},
{
"operation": "Page search",
"time": "75.108",
"value": "John"
},
{
"operation": "Highlighting",
"time": "94.454",
"value": "Fettiplace_sentence_0"
},
{
"operation": "Highlighting",
"time": "96.861",
"value": "Fettiplace_sentence_5"
},
{
"operation": "Page search",
"time": "107.806",
"value": "draper"
},
{
"operation": "Highlighting",
"time": "111.707",
"value": "Fettiplace_sentence_8"
},
{
"operation": "Highlighting",
"time": "129.763",
"value": "Fettiplace_sentence_9"
},
{
"operation": "Highlighting",
"time": "136.54",
"value": "Fettiplace_sentence_10"
},
{
"operation": "finish",
"time": "141.691",
"value": "finish"
}
]
}
|
exfever_7157
|
WALL-E is the first computer-animated science fiction film released by Walt Disney Pictures.
|
{
"evidence": [
"WALL-E",
"Andrew Stanton",
"A Bug's Life"
],
"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:\nWALL-E is the first computer-animated science fiction film released by Walt Disney Pictures.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"WALL-E",
"Andrew Stanton",
"A Bug's Life"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6413,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
57738
|
Zun flanges appear on the vessels body and neck while the same decoration is used and the shoulders consist of small animal heads.
|
{
"evidence": [
"Zun_sentence_44"
],
"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:\nZun flanges appear on the vessels body and neck while the same decoration is used and the shoulders consist of small animal heads.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Zun_sentence_44"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6414,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.631",
"value": "?search="
},
{
"operation": "search",
"time": "11.691",
"value": "Contains...zun flanges"
},
{
"operation": "Now on",
"time": "12.623",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "24.649",
"value": "Guang (vessel)"
},
{
"operation": "Now on",
"time": "25.659",
"value": "Guang (vessel)"
},
{
"operation": "Highlighting",
"time": "127.783",
"value": "Guang (vessel)_sentence_2"
},
{
"operation": "Highlighting",
"time": "127.784",
"value": "Guang (vessel)_title"
},
{
"operation": "hyperlink",
"time": "134.566",
"value": "zun"
},
{
"operation": "Now on",
"time": "136.031",
"value": "Zun"
},
{
"operation": "Highlighting",
"time": "178.714",
"value": "Zun_sentence_44"
},
{
"operation": "finish",
"time": "262.712",
"value": "finish"
}
]
}
|
exfever_8967
|
Vivah is Shahid Kapoor's third film.
|
{
"evidence": [
"Shahid Kapoor",
"Filmfare Awards"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nVivah is Shahid Kapoor's third film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Shahid Kapoor",
"Filmfare Awards"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6415,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_174
|
Angelina Jolie is an American actress, filmmaker, and humanitarian who made her screen debut as a child in a 1982 American comedy film directed by Hal Ashby.
|
{
"evidence": [
"Angelina Jolie",
"Lookin' to Get Out"
],
"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:\nAngelina Jolie is an American actress, filmmaker, and humanitarian who made her screen debut as a child in a 1982 American comedy film directed by Hal Ashby.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Angelina Jolie",
"Lookin' to Get Out"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6416,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_796
|
Ballspielverein Borussia 09 e. V. Dortmund is the winningest professional sports club in Germany.
|
{
"evidence": [
"Borussia Dortmund",
"North Rhine-Westphalia"
],
"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:\nBallspielverein Borussia 09 e. V. Dortmund is the winningest professional sports club in Germany.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Borussia Dortmund",
"North Rhine-Westphalia"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6417,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
49134
|
Charlemagne, the monarch responsible for uniting western and central Europe in the middle ages, has a connection to the vita of Amalberga according to her cult.
|
{
"evidence": [
"Charlemagne_sentence_1",
"Charlemagne_sentence_2",
"Charlemagne_sentence_508"
],
"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:\nCharlemagne, the monarch responsible for uniting western and central Europe in the middle ages, has a connection to the vita of Amalberga according to her cult.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Charlemagne_sentence_1",
"Charlemagne_sentence_2",
"Charlemagne_sentence_508"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6418,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.114",
"value": "?search="
},
{
"operation": "search",
"time": "8.832",
"value": "Charlemagne"
},
{
"operation": "Now on",
"time": "10.878",
"value": "Charlemagne"
},
{
"operation": "Highlighting",
"time": "45.373",
"value": "Charlemagne_sentence_1"
},
{
"operation": "Highlighting",
"time": "47.029",
"value": "Charlemagne_sentence_2"
},
{
"operation": "Page search",
"time": "53.562",
"value": "vita"
},
{
"operation": "Page search",
"time": "112.829",
"value": "amalberga"
},
{
"operation": "Page search",
"time": "137.045",
"value": "vita of"
},
{
"operation": "Page search",
"time": "156.369",
"value": "vita"
},
{
"operation": "Highlighting",
"time": "182.735",
"value": "Charlemagne_sentence_508"
},
{
"operation": "search",
"time": "210.416",
"value": "vita of amalberga"
},
{
"operation": "Now on",
"time": "211.123",
"value": "undefined"
},
{
"operation": "finish",
"time": "233.857",
"value": "finish"
}
]
}
|
exfever_3428
|
Claire Danes is the youngest American actress who was awarded a star on the Hollywood Walk of Fame in 2015.
|
{
"evidence": [
"Claire Danes",
"Golden Globe Award"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nClaire Danes is the youngest American actress who was awarded a star on the Hollywood Walk of Fame in 2015.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Claire Danes",
"Golden Globe Award"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6419,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
94167
|
Quinalow was affected by the 2010–2011 Queensland floods, including the Quinalow Library opened in March 1984.
|
{
"evidence": [
"Quinalow_sentence_7",
"Quinalow_sentence_8",
"Quinalow_sentence_6"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nQuinalow was affected by the 2010–2011 Queensland floods, including the Quinalow Library opened in March 1984.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Quinalow_sentence_7",
"Quinalow_sentence_8",
"Quinalow_sentence_6"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6420,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.425",
"value": "?search="
},
{
"operation": "search",
"time": "8.966",
"value": "Quinalow"
},
{
"operation": "Now on",
"time": "9.675",
"value": "Quinalow"
},
{
"operation": "Highlighting",
"time": "26.214",
"value": "Quinalow_sentence_7"
},
{
"operation": "Highlighting",
"time": "33.102",
"value": "Quinalow_sentence_8"
},
{
"operation": "Highlighting",
"time": "53.582",
"value": "Quinalow_sentence_6"
},
{
"operation": "search",
"time": "63.925",
"value": "Quinalow Library"
},
{
"operation": "Now on",
"time": "64.588",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "68.502",
"value": "Library Quinalow_section_2"
},
{
"operation": "Now on",
"time": "68.911",
"value": "Quinalow#Library Quinalow section 2"
},
{
"operation": "hyperlink",
"time": "74.286",
"value": "Toowoomba Regional Council"
},
{
"operation": "Now on",
"time": "75.157",
"value": "Toowoomba Region"
},
{
"operation": "Page search",
"time": "92.984",
"value": "library"
},
{
"operation": "hyperlink",
"time": "98.918",
"value": "7 Toowoomba Regional Library Services Toowoomba Region_section_9"
},
{
"operation": "hyperlink",
"time": "104.31",
"value": "Quinalow"
},
{
"operation": "Now on",
"time": "104.825",
"value": "Quinalow"
},
{
"operation": "finish",
"time": "134.414",
"value": "finish"
}
]
}
|
56634
|
Cave of Dogs is a small lake on the eastern side of the Phlegraean Fields which is a large volcano situated to the west of Naples, Italy.
|
{
"evidence": [
"Cave of Dogs_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:\nCave of Dogs is a small lake on the eastern side of the Phlegraean Fields which is a large volcano situated to the west of Naples, Italy.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Cave of Dogs_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6421,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.126",
"value": "?search="
},
{
"operation": "search",
"time": "12.849",
"value": "Cave of Dogs"
},
{
"operation": "Now on",
"time": "14.11",
"value": "Cave of Dogs"
},
{
"operation": "Highlighting",
"time": "20.446",
"value": "Cave of Dogs_sentence_0"
},
{
"operation": "finish",
"time": "40.311",
"value": "finish"
}
]
}
|
exfever_4867
|
Election is a 1999 British film directed by Alexander Payne starring an actor known for his musical The Producers, a 2005 film written by Mel Brooks.
|
{
"evidence": [
"Election (1999 film)",
"Matthew Broderick",
"The Producers (2005 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:\nElection is a 1999 British film directed by Alexander Payne starring an actor known for his musical The Producers, a 2005 film written by Mel Brooks.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Election (1999 film)",
"Matthew Broderick",
"The Producers (2005 film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6422,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8385
|
Pulp Fiction is a 1994 American black comedy crime film that starred a Canadian actor and singer.
|
{
"evidence": [
"Pulp Fiction",
"John Travolta"
],
"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:\nPulp Fiction is a 1994 American black comedy crime film that starred a Canadian actor and singer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Pulp Fiction",
"John Travolta"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6423,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
85496
|
Diplocystidae, which was created by Bhatia in 1938, are a family of species which infect insects and are spread by the orofaecal route.
|
{
"evidence": [
"Diplocystidae_sentence_3",
"Diplocystidae_sentence_5",
"Diplocystidae_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:\nDiplocystidae, which was created by Bhatia in 1938, are a family of species which infect insects and are spread by the orofaecal route.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Diplocystidae_sentence_3",
"Diplocystidae_sentence_5",
"Diplocystidae_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6424,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.358",
"value": "?search="
},
{
"operation": "search",
"time": "7.609",
"value": "Diplocystidae"
},
{
"operation": "Now on",
"time": "8.347",
"value": "Diplocystidae"
},
{
"operation": "Highlighting",
"time": "101.215",
"value": "Diplocystidae_sentence_3"
},
{
"operation": "Highlighting",
"time": "104.825",
"value": "Diplocystidae_sentence_0"
},
{
"operation": "Highlighting",
"time": "111.227",
"value": "Diplocystidae_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "113.013",
"value": "Diplocystidae_sentence_5"
},
{
"operation": "Highlighting",
"time": "145.786",
"value": "Diplocystidae_sentence_5"
},
{
"operation": "Highlighting",
"time": "156.417",
"value": "Diplocystidae_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "158.914",
"value": "Diplocystidae_sentence_0"
},
{
"operation": "finish",
"time": "169.384",
"value": "finish"
}
]
}
|
exfever_8501
|
At 8,611 meters, K2 is the world's third-highest mountain, after Mount Everest, and is located in the Gilgit-Baltistan region of Western Asia's fifth-most populous country.
|
{
"evidence": [
"K2",
"Pakistan"
],
"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:\nAt 8,611 meters, K2 is the world's third-highest mountain, after Mount Everest, and is located in the Gilgit-Baltistan region of Western Asia's fifth-most populous country.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"K2",
"Pakistan"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6425,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5197
|
Merlin is a fantasy-adventure drama television programme created for the national broadcaster of the United States that is funded by the television licence fee, which is charged for using any type of equipment to receive its video-on-demand service.
|
{
"evidence": [
"Merlin (2008 TV series)",
"BBC",
"BBC iPlayer"
],
"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:\nMerlin is a fantasy-adventure drama television programme created for the national broadcaster of the United States that is funded by the television licence fee, which is charged for using any type of equipment to receive its video-on-demand service.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Merlin (2008 TV series)",
"BBC",
"BBC iPlayer"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6426,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2205
|
The film Mrigayaa is the directorial debut of Mrinal Sen.
|
{
"evidence": [
"Mrigayaa",
"Film score"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe film Mrigayaa is the directorial debut of Mrinal Sen.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mrigayaa",
"Film score"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6427,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8074
|
Chris Hemsworth is an actor and also a musician.
|
{
"evidence": [
"Chris Hemsworth",
"Home and Away"
],
"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:\nChris Hemsworth is an actor and also a musician.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Chris Hemsworth",
"Home and Away"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6428,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
75385
|
The Historia Scientiarum is a bilingual academic journal headed by editor-in-chief Takehiko Hashimoto.
|
{
"evidence": [
"Historia Scientiarum_sentence_4"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Historia Scientiarum is a bilingual academic journal headed by editor-in-chief Takehiko Hashimoto.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Historia Scientiarum_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6429,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.862",
"value": "?search="
},
{
"operation": "search",
"time": "18.202",
"value": "Historia Scientiarum"
},
{
"operation": "Now on",
"time": "19.377",
"value": "Historia Scientiarum"
},
{
"operation": "Highlighting",
"time": "75.334",
"value": "Historia Scientiarum_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "84.464",
"value": "Historia Scientiarum_sentence_3"
},
{
"operation": "Highlighting",
"time": "86.489",
"value": "Historia Scientiarum_sentence_4"
},
{
"operation": "finish",
"time": "93.861",
"value": "finish"
}
]
}
|
exfever_1747
|
Wallander is a movie series work by a British cinematographer and photographer.
|
{
"evidence": [
"Wallander (UK TV series)",
"Anthony Dod Mantle"
],
"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:\nWallander is a movie series work by a British cinematographer and photographer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Wallander (UK TV series)",
"Anthony Dod Mantle"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6430,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
47815
|
Vladimir Sukachev found the Department of Dendrology and Systematics of Plants at the Imperial Forestry Institute in 1919 and organized the Forestry Institute of the USSR Academy of Sciences in 1944.
|
{
"evidence": [
"Vladimir Sukachev_sentence_4",
"Vladimir Sukachev_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:\nVladimir Sukachev found the Department of Dendrology and Systematics of Plants at the Imperial Forestry Institute in 1919 and organized the Forestry Institute of the USSR Academy of Sciences in 1944.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Vladimir Sukachev_sentence_4",
"Vladimir Sukachev_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6431,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.693",
"value": "?search="
},
{
"operation": "search",
"time": "169.034",
"value": "Vladimir Sukachev"
},
{
"operation": "Now on",
"time": "169.971",
"value": "Vladimir Sukachev"
},
{
"operation": "Highlighting",
"time": "182.16",
"value": "Vladimir Sukachev_sentence_4"
},
{
"operation": "Highlighting",
"time": "191.741",
"value": "Vladimir Sukachev_sentence_6"
},
{
"operation": "finish",
"time": "220.72",
"value": "finish"
}
]
}
|
37153
|
The 1969 winner of the New York City-based Overseas Press Club award was Peter Leibing for his well-known Cold War photo of Schumann jumping off the Berlin Wall's barbed wire fence.
|
{
"evidence": [
"Peter Leibing_sentence_6"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe 1969 winner of the New York City-based Overseas Press Club award was Peter Leibing for his well-known Cold War photo of Schumann jumping off the Berlin Wall's barbed wire fence.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Peter Leibing_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6432,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.868",
"value": "?search="
},
{
"operation": "search",
"time": "28.708",
"value": "Overseas Press Club"
},
{
"operation": "Now on",
"time": "29.568",
"value": "Overseas Press Club"
},
{
"operation": "Highlighting",
"time": "38.603",
"value": "Overseas Press Club_sentence_1"
},
{
"operation": "Highlighting",
"time": "38.605",
"value": "Overseas Press Club_title"
},
{
"operation": "search",
"time": "108.712",
"value": "Peter Leibing"
},
{
"operation": "Now on",
"time": "109.381",
"value": "Peter Leibing"
},
{
"operation": "Highlighting",
"time": "120.166",
"value": "Peter Leibing_sentence_0"
},
{
"operation": "search",
"time": "189.97",
"value": "Contains...1969 Overseas Press Club award"
},
{
"operation": "Now on",
"time": "191.822",
"value": "undefined"
},
{
"operation": "search",
"time": "206.594",
"value": "Overseas Press Club Award"
},
{
"operation": "Now on",
"time": "207.425",
"value": "Overseas Press Club"
},
{
"operation": "hyperlink",
"time": "252.366",
"value": "Robert Capa Gold Medal Award"
},
{
"operation": "Now on",
"time": "253.246",
"value": "Robert Capa Gold Medal"
},
{
"operation": "back-button-clicked",
"time": "288.292",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "288.398",
"value": "Overseas Press Club"
},
{
"operation": "hyperlink",
"time": "318.112",
"value": "Foreign Correspondents' Club"
},
{
"operation": "Now on",
"time": "318.92",
"value": "Foreign Correspondents' Club"
},
{
"operation": "search",
"time": "328.026",
"value": "Peter Leibing"
},
{
"operation": "Now on",
"time": "328.984",
"value": "Peter Leibing"
},
{
"operation": "Highlighting",
"time": "345.434",
"value": "Peter Leibing_sentence_6"
},
{
"operation": "finish",
"time": "368.205",
"value": "finish"
}
]
}
|
21645
|
Love at First Sting is the ninth studio album by German hard rock band Rammstein, which received a rating of 7/10 from Collector's Guide to Heavy Metal and 7.5/10 from Rock Hard.
|
{
"evidence": [
"Love at First Sting_sentence_0",
"Rammstein discography_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:\nLove at First Sting is the ninth studio album by German hard rock band Rammstein, which received a rating of 7/10 from Collector's Guide to Heavy Metal and 7.5/10 from Rock Hard.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Love at First Sting_sentence_0",
"Rammstein discography_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6433,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.572",
"value": "?search="
},
{
"operation": "search",
"time": "25.985",
"value": "Love at First Sting"
},
{
"operation": "Now on",
"time": "26.707",
"value": "Love at First Sting"
},
{
"operation": "Highlighting",
"time": "36.894",
"value": "Love at First Sting_sentence_0"
},
{
"operation": "hyperlink",
"time": "46.286",
"value": "Scorpions"
},
{
"operation": "Now on",
"time": "47.572",
"value": "Scorpions (band)"
},
{
"operation": "search",
"time": "55.176",
"value": "Rammstein"
},
{
"operation": "Now on",
"time": "56.804",
"value": "Rammstein"
},
{
"operation": "hyperlink",
"time": "75.593",
"value": "Rammstein discography"
},
{
"operation": "Now on",
"time": "76.669",
"value": "Rammstein discography"
},
{
"operation": "Highlighting",
"time": "88.936",
"value": "Rammstein discography_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "89.77",
"value": "Rammstein discography_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "90.37",
"value": "Rammstein discography_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "90.38",
"value": "Rammstein discography_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "90.383",
"value": "Rammstein discography_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "101.521",
"value": "Rammstein discography_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "101.525",
"value": "Rammstein discography_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "101.525",
"value": "Rammstein discography_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "104.048",
"value": "Rammstein discography_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "112.706",
"value": "Rammstein discography_sentence_0"
},
{
"operation": "Highlighting",
"time": "113.819",
"value": "Rammstein discography_sentence_0"
},
{
"operation": "finish",
"time": "141.334",
"value": "finish"
}
]
}
|
exfever_3363
|
Lauren Graham is an American actress best known for her role as Lorelai Gilmore on an American comedy-drama television series created by Amy Sherman-Palladino.
|
{
"evidence": [
"Lauren Graham",
"Gilmore Girls"
],
"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:\nLauren Graham is an American actress best known for her role as Lorelai Gilmore on an American comedy-drama television series created by Amy Sherman-Palladino.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lauren Graham",
"Gilmore Girls"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6434,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
17116
|
Charles Le Brun was a pupil of Simon Vouet, his only teacher, and received commissions at age fifteen from Cardinal Richelieu.
|
{
"evidence": [
"Charles Le Brun_sentence_3",
"Charles Le Brun_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:\nCharles Le Brun was a pupil of Simon Vouet, his only teacher, and received commissions at age fifteen from Cardinal Richelieu.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Charles Le Brun_sentence_3",
"Charles Le Brun_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6435,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.748",
"value": "?search="
},
{
"operation": "search",
"time": "10.971",
"value": "Charles Le Brun"
},
{
"operation": "Now on",
"time": "11.929",
"value": "Charles Le Brun"
},
{
"operation": "Page search",
"time": "53.848",
"value": "simon vouet"
},
{
"operation": "Highlighting",
"time": "63.748",
"value": "Charles Le Brun_sentence_3"
},
{
"operation": "Highlighting",
"time": "76.276",
"value": "Charles Le Brun_sentence_4"
},
{
"operation": "finish",
"time": "94.944",
"value": "finish"
}
]
}
|
90050
|
Gerhard Berger, has a record ten Grands Prix, achieved 48 podiums, 12 poles and 21 fastest laps, from Germany won the 1992 Australian Grand Prix, driving a McLaren-Honda.
|
{
"evidence": [
"Gerhard Berger_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:\nGerhard Berger, has a record ten Grands Prix, achieved 48 podiums, 12 poles and 21 fastest laps, from Germany won the 1992 Australian Grand Prix, driving a \tMcLaren-Honda.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Gerhard Berger_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6436,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.978",
"value": "?search="
},
{
"operation": "search",
"time": "11.37",
"value": "Gerhard Berger"
},
{
"operation": "Now on",
"time": "12.622",
"value": "Gerhard Berger"
},
{
"operation": "Page search",
"time": "40.686",
"value": "grand prix"
},
{
"operation": "Page search",
"time": "65.177",
"value": "record"
},
{
"operation": "Page search",
"time": "104.573",
"value": "ten"
},
{
"operation": "page-search-reset",
"time": "114.991",
"value": "page-search-reset"
},
{
"operation": "Highlighting",
"time": "119.512",
"value": "Gerhard Berger_sentence_2"
},
{
"operation": "Page search",
"time": "144.354",
"value": "germany"
},
{
"operation": "search",
"time": "177.148",
"value": "1992 Australian Grand Prix"
},
{
"operation": "Now on",
"time": "178.286",
"value": "1992 Australian Grand Prix"
},
{
"operation": "Highlighting",
"time": "187.387",
"value": "1992 Australian Grand Prix_sentence_0"
},
{
"operation": "Highlighting",
"time": "187.388",
"value": "1992 Australian Grand Prix_title"
},
{
"operation": "hyperlink",
"time": "188.873",
"value": "Adelaide"
},
{
"operation": "Now on",
"time": "189.734",
"value": "Adelaide Street Circuit"
},
{
"operation": "Highlighting",
"time": "200.52",
"value": "Adelaide Street Circuit_sentence_0"
},
{
"operation": "Highlighting",
"time": "200.521",
"value": "Adelaide Street Circuit_title"
},
{
"operation": "back-button-clicked",
"time": "207.433",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "207.594",
"value": "1992 Australian Grand Prix"
},
{
"operation": "back-button-clicked",
"time": "209.402",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "209.406",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "209.724",
"value": "Gerhard Berger"
},
{
"operation": "Highlighting deleted",
"time": "221.273",
"value": "Gerhard Berger_sentence_2"
},
{
"operation": "Highlighting",
"time": "223.704",
"value": "Gerhard Berger_sentence_0"
},
{
"operation": "Highlighting",
"time": "232.904",
"value": "Gerhard Berger_sentence_0"
},
{
"operation": "finish",
"time": "247.927",
"value": "finish"
}
]
}
|
exfever_8973
|
Apocalypse Now is a 1979 American epic psychological war film starring an American actor who first became known for his role in a 1973 American neo-noir period crime drama film directed by Terrence Malick.
|
{
"evidence": [
"Apocalypse Now",
"Martin Sheen",
"Badlands (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:\nApocalypse Now is a 1979 American epic psychological war film starring an American actor who first became known for his role in a 1973 American neo-noir period crime drama film directed by Terrence Malick.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Apocalypse Now",
"Martin Sheen",
"Badlands (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6437,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8246
|
The Fate of the Furious is written by an American screenwriter with multiple screenplays in the Fox crime drama series that aired on Fox from May 22 to August 14, 2014.
|
{
"evidence": [
"The Fate of the Furious",
"Chris Morgan (writer)",
"Gang Related (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:\nThe Fate of the Furious is written by an American screenwriter with multiple screenplays in the Fox crime drama series that aired on Fox from May 22 to August 14, 2014.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Fate of the Furious",
"Chris Morgan (writer)",
"Gang Related (TV series)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6438,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
17195
|
From 1948 to 1975, Dave Fuller was the head coach of the Florida Gators baseball team, and eight of his athletes went on to play in Major League Baseball.
|
{
"evidence": [
"Dave Fuller_sentence_8",
"Dave Fuller_sentence_11"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nFrom 1948 to 1975, Dave Fuller was the head coach of the Florida Gators baseball team, and eight of his athletes went on to play in Major League Baseball.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Dave Fuller_sentence_8",
"Dave Fuller_sentence_11"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6439,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.041",
"value": "?search="
},
{
"operation": "search",
"time": "13.21",
"value": "Dave fuller"
},
{
"operation": "Now on",
"time": "14.329",
"value": "Dave Fuller"
},
{
"operation": "Highlighting",
"time": "25.145",
"value": "Dave Fuller_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "30.217",
"value": "Dave Fuller_sentence_0"
},
{
"operation": "Highlighting",
"time": "49.097",
"value": "Dave Fuller_sentence_8"
},
{
"operation": "Highlighting",
"time": "58.857",
"value": "Dave Fuller_sentence_11"
},
{
"operation": "finish",
"time": "66.537",
"value": "finish"
}
]
}
|
60325
|
Greg Romeus sat out the 2006 season as on redshirt status, but in 2007, he saw action in all 12 games and recorded 41 tackles
|
{
"evidence": [
"Greg Romeus_sentence_11",
"Greg Romeus_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:\nGreg Romeus sat out the 2006 season as on redshirt status, but in 2007, he saw action in all 12 games and recorded 41 tackles\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Greg Romeus_sentence_11",
"Greg Romeus_sentence_12"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6440,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.098",
"value": "?search="
},
{
"operation": "search",
"time": "12.408",
"value": "Greg Romeus"
},
{
"operation": "Now on",
"time": "13.398",
"value": "Greg Romeus"
},
{
"operation": "Highlighting",
"time": "90.849",
"value": "Greg Romeus_sentence_11"
},
{
"operation": "Highlighting",
"time": "93.236",
"value": "Greg Romeus_sentence_12"
},
{
"operation": "finish",
"time": "156.02",
"value": "finish"
}
]
}
|
exfever_5087
|
Prominent examples of countercultures in the Western world include non-traditional lifestyles often were expressed through free love, frugality, and—in some case— a lifestyle of living in a vehicle.
|
{
"evidence": [
"Counterculture",
"Vandwelling"
],
"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:\nProminent examples of countercultures in the Western world include non-traditional lifestyles often were expressed through free love, frugality, and—in some case— a lifestyle of living in a vehicle.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Counterculture",
"Vandwelling"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6441,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
36316
|
Stamps that were issued include the Malayan Union and Singapore Peace Commemorative stamps of 1946, which were issued for local political reasons, and essays for four values (30c, 50c, 1s.30, 2s.50) of the 1960-62 Kenya, Uganda & Tanganyika definitive series overprinted to mark the 60th session of the International Olympic Committee in Nairobi, 1963.
|
{
"evidence": [
"Crown Agents Philatelic and Security Printing Archive_sentence_6"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nStamps that were issued include the Malayan Union and Singapore Peace Commemorative stamps of 1946, which were issued for local political reasons, and essays for four values (30c, 50c, 1s.30, 2s.50) of the 1960-62 Kenya, Uganda & Tanganyika definitive series overprinted to mark the 60th session of the International Olympic Committee in Nairobi, 1963.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Crown Agents Philatelic and Security Printing Archive_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6442,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.777",
"value": "?search="
},
{
"operation": "search",
"time": "50.087",
"value": "singapore stamps"
},
{
"operation": "Now on",
"time": "51.182",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "59.308",
"value": "Postage stamps and postal history of Singapore"
},
{
"operation": "Now on",
"time": "60.023",
"value": "Postage stamps and postal history of Singapore"
},
{
"operation": "search",
"time": "155.199",
"value": "malayan union"
},
{
"operation": "Now on",
"time": "155.928",
"value": "Malayan Union"
},
{
"operation": "search",
"time": "164.395",
"value": "singapore postage stamps"
},
{
"operation": "Now on",
"time": "165.928",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "167.785",
"value": "Postage stamps and postal history of Singapore"
},
{
"operation": "Now on",
"time": "168.219",
"value": "Postage stamps and postal history of Singapore"
},
{
"operation": "Page search",
"time": "172.443",
"value": "malayan union"
},
{
"operation": "search",
"time": "257.452",
"value": "60th session of the International Olympic Committee"
},
{
"operation": "Now on",
"time": "259.418",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "262.898",
"value": "Crown Agents Philatelic and Security Printing Archive"
},
{
"operation": "Now on",
"time": "263.749",
"value": "Crown Agents Philatelic and Security Printing Archive"
},
{
"operation": "Highlighting",
"time": "284.958",
"value": "Crown Agents Philatelic and Security Printing Archive_sentence_6"
},
{
"operation": "finish",
"time": "312.015",
"value": "finish"
}
]
}
|
65278
|
In 2012, Maria Kong performed at the Dancing Poznan Festival in Poznan, a city on the Warta River in west-central Poland, and at the Drupa Exhibition in Dusseldorf, the capital city of North Rhine-Westphalia in Germany.
|
{
"evidence": [
"Maria Kong_sentence_4",
"Poznań_sentence_4",
"Maria Kong_sentence_5",
"Düsseldorf_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:\nIn 2012, Maria Kong performed at the Dancing Poznan Festival in Poznan, a city on the Warta River in west-central Poland, and at the Drupa Exhibition in Dusseldorf, the capital city of North Rhine-Westphalia in Germany.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Maria Kong_sentence_4",
"Poznań_sentence_4",
"Maria Kong_sentence_5",
"Düsseldorf_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6443,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.125",
"value": "?search="
},
{
"operation": "search",
"time": "12.276",
"value": "Maria Kong"
},
{
"operation": "Now on",
"time": "13.339",
"value": "Maria Kong"
},
{
"operation": "Highlighting",
"time": "20.495",
"value": "Maria Kong_sentence_4"
},
{
"operation": "hyperlink",
"time": "36.896",
"value": "Poznan, Poland"
},
{
"operation": "Now on",
"time": "38.665",
"value": "Poznań"
},
{
"operation": "Page search",
"time": "57.369",
"value": "Warta"
},
{
"operation": "Highlighting",
"time": "60.659",
"value": "Poznań_sentence_4"
},
{
"operation": "back-button-clicked",
"time": "66.672",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "67.016",
"value": "Maria Kong"
},
{
"operation": "Highlighting",
"time": "69.907",
"value": "Maria Kong_sentence_5"
},
{
"operation": "hyperlink",
"time": "75.643",
"value": "Düsseldorf, Germany"
},
{
"operation": "Now on",
"time": "83.783",
"value": "Maria Kong"
},
{
"operation": "hyperlink",
"time": "85.411",
"value": "Düsseldorf, Germany"
},
{
"operation": "Now on",
"time": "93.166",
"value": "Maria Kong"
},
{
"operation": "search",
"time": "98.072",
"value": "Düsseldorf"
},
{
"operation": "Now on",
"time": "99.889",
"value": "Düsseldorf"
},
{
"operation": "Highlighting",
"time": "103.247",
"value": "Düsseldorf_sentence_2"
},
{
"operation": "finish",
"time": "138.332",
"value": "finish"
}
]
}
|
exfever_8924
|
Elizabeth Meriwether is a producer known for creating the sitcom New Girl for an American commercial broadcast television network.
|
{
"evidence": [
"Elizabeth Meriwether",
"Fox Broadcasting Company"
],
"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:\nElizabeth Meriwether is a producer known for creating the sitcom New Girl for an American commercial broadcast television network.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Elizabeth Meriwether",
"Fox Broadcasting Company"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6444,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3053
|
The Netherlands' official language is Dutch, and secondary in the province of Friesland, is a language spoken by about 500,000 people who live on the southern fringes of the North Sea in the Netherlands and Germany.
|
{
"evidence": [
"Netherlands",
"Frisian languages"
],
"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 Netherlands' official language is Dutch, and secondary in the province of Friesland, is a language spoken by about 500,000 people who live on the southern fringes of the North Sea in the Netherlands and Germany.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Netherlands",
"Frisian languages"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6445,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8102
|
Céline Marie Claudette Dion is a Canadian dancer born into a large family from an off-island suburb of Montreal, Quebec, Canada.
|
{
"evidence": [
"Celine Dion",
"Charlemagne, Quebec"
],
"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:\nCéline Marie Claudette Dion is a Canadian dancer born into a large family from an off-island suburb of Montreal, Quebec, Canada.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Celine Dion",
"Charlemagne, Quebec"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6446,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
52603
|
Roland Carter (who was born in 1892 in Raukkan, South Australia) was the first of the Ngarrindjeri people to enlist in the First Australian Imperial Force to fight in World War I, where was wounded and captured during combat at Noreuil in northern France on 2 April 1917.
|
{
"evidence": [
"Roland Carter_sentence_0",
"Roland Carter_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:\nRoland Carter (who was born in 1892 in Raukkan, South Australia) was the first of the Ngarrindjeri people to enlist in the First Australian Imperial Force to fight in World War I, where was wounded and captured during combat at Noreuil in northern France on 2 April 1917.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Roland Carter_sentence_0",
"Roland Carter_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6447,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.383",
"value": "?search="
},
{
"operation": "search",
"time": "9.345",
"value": "Roland Carter"
},
{
"operation": "Now on",
"time": "10.126",
"value": "Roland Carter"
},
{
"operation": "Highlighting",
"time": "16.859",
"value": "Roland Carter_sentence_0"
},
{
"operation": "Highlighting",
"time": "31.049",
"value": "Roland Carter_sentence_1"
},
{
"operation": "finish",
"time": "59.154",
"value": "finish"
}
]
}
|
exfever_5214
|
PageRank is an algorithm used by Google Search to rank web pages in their search engine results, which is named after its co-founder, an Australian computer scientist and internet entrepreneur.
|
{
"evidence": [
"PageRank",
"Larry Page"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPageRank is an algorithm used by Google Search to rank web pages in their search engine results, which is named after its co-founder, an Australian computer scientist and internet entrepreneur.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"PageRank",
"Larry Page"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6448,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
27571
|
Jack Wolf received his undergraduate degree from the University of Pennsylvania, he held faculty appointments at New York University and was appointed inn any position by Qualcomm headquartered in San Diego.
|
{
"evidence": [
"Jack Wolf_sentence_2",
"Jack Wolf_sentence_3",
"Jack Wolf_sentence_5",
"Qualcomm_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:\nJack Wolf received his undergraduate degree from the University of Pennsylvania, he held faculty appointments at New York University and was appointed inn any position by Qualcomm headquartered in San Diego.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Jack Wolf_sentence_2",
"Jack Wolf_sentence_3",
"Jack Wolf_sentence_5",
"Qualcomm_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6449,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.085",
"value": "?search="
},
{
"operation": "search",
"time": "17.391",
"value": "Jack Wolf"
},
{
"operation": "Now on",
"time": "18.825",
"value": "Jack Wolf"
},
{
"operation": "Highlighting",
"time": "24.532",
"value": "Jack Wolf_sentence_2"
},
{
"operation": "Highlighting",
"time": "30.434",
"value": "Jack Wolf_sentence_3"
},
{
"operation": "Highlighting",
"time": "42.685",
"value": "Jack Wolf_sentence_5"
},
{
"operation": "hyperlink",
"time": "51.608",
"value": "Qualcomm"
},
{
"operation": "Now on",
"time": "53.786",
"value": "Qualcomm"
},
{
"operation": "Highlighting",
"time": "55.534",
"value": "Qualcomm_sentence_0"
},
{
"operation": "finish",
"time": "76.738",
"value": "finish"
}
]
}
|
5536
|
Sola, a Spanish footballer born in Pamplona, Navarre, made his senior debuts with UDC Chantrea in 2011; then made a deal with CA Osasuna on February 16, 2013.
|
{
"evidence": [
"Yoel Sola_sentence_2",
"Yoel Sola_sentence_4",
"Yoel Sola_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:\nSola, a Spanish footballer born in Pamplona, Navarre, made his senior debuts with UDC Chantrea in 2011; then made a deal with CA Osasuna on February 16, 2013.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Yoel Sola_sentence_2",
"Yoel Sola_sentence_4",
"Yoel Sola_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6450,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.563",
"value": "?search="
},
{
"operation": "search",
"time": "14.386",
"value": "sola"
},
{
"operation": "Now on",
"time": "15.261",
"value": "Sola"
},
{
"operation": "search",
"time": "42.25",
"value": "UDC chantrea"
},
{
"operation": "Now on",
"time": "43.011",
"value": "UDC Txantrea"
},
{
"operation": "Page search",
"time": "56.142",
"value": "sola"
},
{
"operation": "Highlighting",
"time": "66.766",
"value": "UDC Txantrea_item_1_2"
},
{
"operation": "Highlighting",
"time": "66.769",
"value": "UDC Txantrea_title"
},
{
"operation": "Highlighting",
"time": "66.771",
"value": "UDC Txantrea_section_1"
},
{
"operation": "hyperlink",
"time": "68.186",
"value": "Miguel Ángel Sola"
},
{
"operation": "Now on",
"time": "68.932",
"value": "Miguel Sola"
},
{
"operation": "Highlighting",
"time": "90.597",
"value": "Miguel Sola_table_caption_0"
},
{
"operation": "Highlighting",
"time": "90.603",
"value": "Miguel Sola_title"
},
{
"operation": "Highlighting",
"time": "94.712",
"value": "Miguel Sola_cell_0_3_1"
},
{
"operation": "Highlighting",
"time": "94.773",
"value": "Miguel Sola_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "94.781",
"value": "Miguel Sola_header_cell_0_0_0"
},
{
"operation": "Page search",
"time": "198.181",
"value": "chantrea"
},
{
"operation": "Page search",
"time": "215.412",
"value": "osasuna"
},
{
"operation": "search",
"time": "264.776",
"value": "CA Osasuna on February 16, 2013"
},
{
"operation": "Now on",
"time": "267.203",
"value": "undefined"
},
{
"operation": "search",
"time": "296.917",
"value": "CA Osasuna on February 16, 2013 sola"
},
{
"operation": "Now on",
"time": "297.594",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "317.451",
"value": "Kike Sola"
},
{
"operation": "Now on",
"time": "318.31",
"value": "Kike Sola"
},
{
"operation": "Page search",
"time": "332.931",
"value": "CA Osasuna on February 16, 2013"
},
{
"operation": "Page search",
"time": "339.842",
"value": "CA Osasuna"
},
{
"operation": "search",
"time": "408.13",
"value": "CA Osasuna February 16, 2013 sola"
},
{
"operation": "Now on",
"time": "408.652",
"value": "undefined"
},
{
"operation": "search",
"time": "426.497",
"value": "sola UDC chantrea"
},
{
"operation": "Now on",
"time": "427.148",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "430.871",
"value": "Club career Yoel Sola_section_0"
},
{
"operation": "Now on",
"time": "431.628",
"value": "Yoel Sola#Club career Yoel Sola section 0"
},
{
"operation": "Highlighting",
"time": "479.975",
"value": "Yoel Sola_cell_0_3_1"
},
{
"operation": "Highlighting",
"time": "480.003",
"value": "Yoel Sola_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "480.005",
"value": "Yoel Sola_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "485.502",
"value": "Yoel Sola_sentence_2"
},
{
"operation": "Highlighting",
"time": "496.496",
"value": "Yoel Sola_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "513.493",
"value": "Yoel Sola_cell_0_3_1"
},
{
"operation": "Highlighting deleted",
"time": "513.52",
"value": "Yoel Sola_header_cell_0_3_0"
},
{
"operation": "Highlighting deleted",
"time": "513.523",
"value": "Yoel Sola_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "527.578",
"value": "Yoel Sola_sentence_1"
},
{
"operation": "finish",
"time": "659.695",
"value": "finish"
}
]
}
|
37498
|
Henry Seymour Portman was a Member (MP) for St Mawes and was also commissioner for assessment for London.
|
{
"evidence": [
"Henry Seymour Portman_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHenry Seymour Portman was a Member (MP) for St Mawes and was also commissioner for assessment for London.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Henry Seymour Portman_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6451,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.907",
"value": "?search="
},
{
"operation": "search",
"time": "10.963",
"value": "Henry Seymour Portman"
},
{
"operation": "Now on",
"time": "11.864",
"value": "Henry Seymour Portman"
},
{
"operation": "Highlighting",
"time": "23.962",
"value": "Henry Seymour Portman_sentence_4"
},
{
"operation": "Page search",
"time": "38.776",
"value": "commissioner"
},
{
"operation": "Highlighting",
"time": "41.215",
"value": "Henry Seymour Portman_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "45.255",
"value": "Henry Seymour Portman_sentence_4"
},
{
"operation": "finish",
"time": "60.976",
"value": "finish"
}
]
}
|
exfever_5243
|
Ricky Martin's last album was A Quien Quiera Escuchar.
|
{
"evidence": [
"Ricky Martin",
"San Juan, Puerto Rico",
"Municipalities of Puerto Rico"
],
"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:\nRicky Martin's last album was A Quien Quiera Escuchar.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ricky Martin",
"San Juan, Puerto Rico",
"Municipalities of Puerto Rico"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6452,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_492
|
Samantha Robson was born on March 22, 1966, in a district of South London.
|
{
"evidence": [
"Samantha Robson",
"Wandsworth"
],
"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:\nSamantha Robson was born on March 22, 1966, in a district of South London.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Samantha Robson",
"Wandsworth"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6453,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
65894
|
Gymnopilus subfulgens is a fungus.
|
{
"evidence": [
"Gymnopilus subfulgens_sentence_0",
"Mushroom_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:\nGymnopilus subfulgens is a fungus.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Gymnopilus subfulgens_sentence_0",
"Mushroom_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6454,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.064",
"value": "?search="
},
{
"operation": "search",
"time": "17.236",
"value": "Gymnopilus subfulgens"
},
{
"operation": "Now on",
"time": "18.651",
"value": "Gymnopilus subfulgens"
},
{
"operation": "Highlighting",
"time": "32.714",
"value": "Gymnopilus subfulgens_sentence_0"
},
{
"operation": "hyperlink",
"time": "34.306",
"value": "mushroom"
},
{
"operation": "Now on",
"time": "48.756",
"value": "Gymnopilus subfulgens"
},
{
"operation": "search",
"time": "53.45",
"value": "Mushroom"
},
{
"operation": "Now on",
"time": "65.304",
"value": "Mushroom"
},
{
"operation": "Highlighting",
"time": "68.561",
"value": "Mushroom_sentence_5"
},
{
"operation": "finish",
"time": "84.858",
"value": "finish"
}
]
}
|
37473
|
Stuart Gill began his career in the UK Government’s Department for Trade and Industry, he was later appointed as Her Majesty's Ambassador to the Republic of Iceland but never worked in the legal department.
|
{
"evidence": [
"Stuart Gill_sentence_2",
"Stuart Gill_sentence_4"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nStuart Gill began his career in the UK Government’s Department for Trade and Industry, he was later appointed as Her Majesty's Ambassador to the Republic of Iceland but never worked in the legal department.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Stuart Gill_sentence_2",
"Stuart Gill_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6455,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.145",
"value": "?search="
},
{
"operation": "search",
"time": "15.157",
"value": "Stuart Gill"
},
{
"operation": "Now on",
"time": "17.843",
"value": "Stuart Gill"
},
{
"operation": "Highlighting",
"time": "20.276",
"value": "Stuart Gill_sentence_2"
},
{
"operation": "Highlighting",
"time": "27.82",
"value": "Stuart Gill_sentence_4"
},
{
"operation": "finish",
"time": "79.076",
"value": "finish"
}
]
}
|
82596
|
The Iranian Cyber Army, an Iranian white hacker group which was directly hired by the Iranian government, pledged loyalty to the Supreme Leader of Iran and was assumed to have been managed by Mohamad Hussein Tajik.
|
{
"evidence": [
"Iranian Cyber Army_sentence_0",
"Iranian Cyber Army_sentence_1",
"Iranian Cyber Army_sentence_2",
"Iranian Cyber Army_sentence_4"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Iranian Cyber Army, an Iranian white hacker group which was directly hired by the Iranian government, pledged loyalty to the Supreme Leader of Iran and was assumed to have been managed by Mohamad Hussein Tajik.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Iranian Cyber Army_sentence_0",
"Iranian Cyber Army_sentence_1",
"Iranian Cyber Army_sentence_2",
"Iranian Cyber Army_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6456,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.004",
"value": "?search="
},
{
"operation": "search",
"time": "16.239",
"value": "Iranian Cyber Army"
},
{
"operation": "Now on",
"time": "16.949",
"value": "Iranian Cyber Army"
},
{
"operation": "Highlighting",
"time": "82.221",
"value": "Iranian Cyber Army_sentence_0"
},
{
"operation": "Highlighting",
"time": "86.432",
"value": "Iranian Cyber Army_sentence_1"
},
{
"operation": "Highlighting",
"time": "98.028",
"value": "Iranian Cyber Army_sentence_2"
},
{
"operation": "Highlighting",
"time": "117.009",
"value": "Iranian Cyber Army_sentence_4"
},
{
"operation": "hyperlink",
"time": "189.06",
"value": "Iranian"
},
{
"operation": "Now on",
"time": "190.445",
"value": "Iran"
},
{
"operation": "finish",
"time": "200.45",
"value": "finish"
}
]
}
|
exfever_2905
|
The album's cover designer, David Juniper, won a Grammy Award for Best Recording Package in 1970.
|
{
"evidence": [
"Led Zeppelin II",
"Jimmy Page"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe album's cover designer, David Juniper, won a Grammy Award for Best Recording Package in 1970.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Led Zeppelin II",
"Jimmy Page"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6457,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
6422
|
Ernie Machin, who finished his playing career with Brighton & Hove Albion, became the second English football player to successfully challenge a fine and suspension by the Football Association in the English courts.
|
{
"evidence": [
"Ernie Machin_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:\nErnie Machin, who finished his playing career with Brighton & Hove Albion, became the second English football player to successfully challenge a fine and suspension by the Football Association in the English courts.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ernie Machin_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6458,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.55",
"value": "?search="
},
{
"operation": "search",
"time": "4.488",
"value": "ernie machin"
},
{
"operation": "Now on",
"time": "5.131",
"value": "Ernie Machin"
},
{
"operation": "Highlighting",
"time": "173.259",
"value": "Ernie Machin_sentence_5"
},
{
"operation": "Page search",
"time": "180.548",
"value": "second"
},
{
"operation": "Page search",
"time": "188.348",
"value": "challenge"
},
{
"operation": "Highlighting",
"time": "191.376",
"value": "Ernie Machin_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "192.244",
"value": "Ernie Machin_sentence_5"
},
{
"operation": "finish",
"time": "201.117",
"value": "finish"
}
]
}
|
exfever_214
|
Underworld is a series of dark fantasy action horror films starring an Asian actress and model who appeared in a 1994 drama adventure film co-written and directed by Gabriel Axel.
|
{
"evidence": [
"Underworld (film series)",
"Kate Beckinsale",
"Prince of Jutland"
],
"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:\nUnderworld is a series of dark fantasy action horror films starring an Asian actress and model who appeared in a 1994 drama adventure film co-written and directed by Gabriel Axel.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Underworld (film series)",
"Kate Beckinsale",
"Prince of Jutland"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6459,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4096
|
Stefanie Maria "Steffi" Graf is a German former professional tennis player who was ranked third-most of all time that won 22 major singles titles behind an American professional tennis player who won 23 Grand Slam singles titles.
|
{
"evidence": [
"Steffi Graf",
"Serena Williams"
],
"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:\nStefanie Maria \"Steffi\" Graf is a German former professional tennis player who was ranked third-most of all time that won 22 major singles titles behind an American professional tennis player who won 23 Grand Slam singles titles.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Steffi Graf",
"Serena Williams"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6460,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_129
|
Sarah Silverman is an American comedian who has won two awards bestowed by the Academy of Television Arts & Sciences to recognize excellence in American primetime television programming.
|
{
"evidence": [
"Sarah Silverman",
"Primetime Emmy Award"
],
"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:\nSarah Silverman is an American comedian who has won two awards bestowed by the Academy of Television Arts & Sciences to recognize excellence in American primetime television programming.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sarah Silverman",
"Primetime Emmy Award"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6461,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2052
|
Pompeii was an ancient city located in what is now the comune of Pompei, near Naples in region of Italy, which has a population of around 5,820,000 people making it Italy's third most populous region.
|
{
"evidence": [
"Pompeii",
"Campania"
],
"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:\nPompeii was an ancient city located in what is now the comune of Pompei, near Naples in region of Italy, which has a population of around 5,820,000 people making it Italy's third most populous region.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Pompeii",
"Campania"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6462,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1838
|
BYD Auto is the subsidiary of the BYD Company that produces a public transport road vehicle designed to carry significantly more passengers than the average car or van.?
|
{
"evidence": [
"BYD Auto",
"Bus"
],
"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:\nBYD Auto is the subsidiary of the BYD Company that produces a public transport road vehicle designed to carry significantly more passengers than the average car or van.?\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"BYD Auto",
"Bus"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6463,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
26387
|
Macau University of Science and Technology Hospital, which opened in 2006, is one of three hospitals in Macau, and it focuses on both Chinese and Western medicine.
|
{
"evidence": [
"Macau University of Science and Technology Hospital_sentence_2",
"Macau University of Science and Technology Hospital_sentence_0",
"Macau University of Science and Technology Hospital_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:\nMacau University of Science and Technology Hospital, which opened in 2006, is one of three hospitals in Macau, and it focuses on both Chinese and Western medicine.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Macau University of Science and Technology Hospital_sentence_2",
"Macau University of Science and Technology Hospital_sentence_0",
"Macau University of Science and Technology Hospital_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6464,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.235",
"value": "?search="
},
{
"operation": "search",
"time": "21.718",
"value": "Macau University of Science and Technology Hospital"
},
{
"operation": "Now on",
"time": "22.432",
"value": "Macau University of Science and Technology Hospital"
},
{
"operation": "Page search",
"time": "101.019",
"value": "2006"
},
{
"operation": "Highlighting",
"time": "104.476",
"value": "Macau University of Science and Technology Hospital_sentence_2"
},
{
"operation": "Page search",
"time": "108.594",
"value": "three"
},
{
"operation": "Highlighting",
"time": "110.301",
"value": "Macau University of Science and Technology Hospital_sentence_0"
},
{
"operation": "Highlighting",
"time": "117.883",
"value": "Macau University of Science and Technology Hospital_sentence_3"
},
{
"operation": "finish",
"time": "157.263",
"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": 6465,
"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"
}
]
}
|
94453
|
Lansing School District covers 52 square miles in Lansing, the capital of the state of Michigan.
|
{
"evidence": [
"Lansing School District_sentence_0",
"Lansing, Michigan_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:\nLansing School District covers 52 square miles in Lansing, the capital of the state of Michigan.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Lansing School District_sentence_0",
"Lansing, Michigan_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6466,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.423",
"value": "?search="
},
{
"operation": "search",
"time": "9.662",
"value": "Lansing School District"
},
{
"operation": "Now on",
"time": "10.625",
"value": "Lansing School District"
},
{
"operation": "Highlighting",
"time": "27.707",
"value": "Lansing School District_sentence_0"
},
{
"operation": "hyperlink",
"time": "31.256",
"value": "Lansing"
},
{
"operation": "Now on",
"time": "33.363",
"value": "Lansing, Michigan"
},
{
"operation": "Highlighting",
"time": "34.118",
"value": "Lansing, Michigan_sentence_2"
},
{
"operation": "finish",
"time": "277.304",
"value": "finish"
}
]
}
|
exfever_7002
|
Minions is a 2015 American computer-animated drama film that stars the voice of an American actress who made her acting debut in the 1987 American action thriller film written and directed by J. Christian Ingvordsen.
|
{
"evidence": [
"Minions (film)",
"Sandra Bullock",
"Hangmen (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:\nMinions is a 2015 American computer-animated drama film that stars the voice of an American actress who made her acting debut in the 1987 American action thriller film written and directed by J. Christian Ingvordsen.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Minions (film)",
"Sandra Bullock",
"Hangmen (film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6467,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1239
|
Carol Susan Jane Danvers is a fictional character appearing in American comic books, who first appeared as a colleague of a Kree superhero who is the original bearer of the name "Captain Marvel" within the Marvel brand.
|
{
"evidence": [
"Carol Danvers",
"Captain Marvel (Mar-Vell)"
],
"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:\nCarol Susan Jane Danvers is a fictional character appearing in American comic books, who first appeared as a colleague of a Kree superhero who is the original bearer of the name \"Captain Marvel\" within the Marvel brand.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Carol Danvers",
"Captain Marvel (Mar-Vell)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6468,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4299
|
The Georgia Institute of Technology has a satellite campus in the oldest city in the U.S. state and was a strategic port city during the ideological and political revolution that occurred in British America between 1765 and 1791.
|
{
"evidence": [
"Georgia Institute of Technology",
"Savannah, Georgia",
"American Revolution"
],
"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 Georgia Institute of Technology has a satellite campus in the oldest city in the U.S. state and was a strategic port city during the ideological and political revolution that occurred in British America between 1765 and 1791.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Georgia Institute of Technology",
"Savannah, Georgia",
"American Revolution"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6469,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4858
|
Tennis is the only sports that can be plyaed in different courts.
|
{
"evidence": [
"Timea Bacsinszky",
"2013 French Open"
],
"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:\nTennis is the only sports that can be plyaed in different courts.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Timea Bacsinszky",
"2013 French Open"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6470,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
62329
|
Xuzhou Metro Line 1 is the first rapid transit system to open in Xuzhou, Jiangsu Province in China, which began construction in February 2014 to run from Luwo to Zushoudong Raily Station.
|
{
"evidence": [
"Xuzhou Metro_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:\nXuzhou Metro Line 1 is the first rapid transit system to open in Xuzhou, Jiangsu Province in China, which began construction in February 2014 to run from Luwo to Zushoudong Raily Station.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Xuzhou Metro_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6471,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.158",
"value": "?search="
},
{
"operation": "Now on",
"time": "9.531",
"value": "?search="
},
{
"operation": "search",
"time": "90.826",
"value": "Xuzhou Metro"
},
{
"operation": "Now on",
"time": "91.777",
"value": "Xuzhou Metro"
},
{
"operation": "Highlighting",
"time": "264.625",
"value": "Xuzhou Metro_sentence_0"
},
{
"operation": "finish",
"time": "355.942",
"value": "finish"
}
]
}
|
11096
|
During D-Day, the 3rd Canadian Division was ashore and leading elements had pushed inland to seize bridges over the Seulles by noon and had captured Saint-Aubin-sur-Mer protected by the German WN27 Resistance Nest by 6:00 pm.
|
{
"evidence": [
"3rd Canadian Division_sentence_34",
"3rd Canadian Division_sentence_35"
],
"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:\nDuring D-Day, the 3rd Canadian Division was ashore and leading elements had pushed inland to seize bridges over the Seulles by noon and had captured Saint-Aubin-sur-Mer protected by the German WN27 Resistance Nest by 6:00 pm.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"3rd Canadian Division_sentence_34",
"3rd Canadian Division_sentence_35"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6472,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.525",
"value": "?search="
},
{
"operation": "search",
"time": "25.131",
"value": "3rd Canadian Division"
},
{
"operation": "Now on",
"time": "26.285",
"value": "3rd Canadian Division"
},
{
"operation": "hyperlink",
"time": "41.951",
"value": "1.2.1 Juno Beach, D Day 3rd Canadian Division_section_4"
},
{
"operation": "Page search",
"time": "101.219",
"value": "leading elements had pushed inland"
},
{
"operation": "Highlighting",
"time": "138.988",
"value": "3rd Canadian Division_sentence_34"
},
{
"operation": "Highlighting",
"time": "155.482",
"value": "3rd Canadian Division_sentence_35"
},
{
"operation": "Page search",
"time": "207.076",
"value": "WN27 Resistance"
},
{
"operation": "hyperlink",
"time": "235.851",
"value": "Saint-Aubin-sur-Mer"
},
{
"operation": "Now on",
"time": "236.718",
"value": "Saint-Aubin-sur-Mer, Calvados"
},
{
"operation": "Page search",
"time": "242.761",
"value": "WN27 Resistance"
},
{
"operation": "back-button-clicked",
"time": "250.466",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "250.727",
"value": "3rd Canadian Division#Juno Beach.2C D Day 3rd Canadian Division section 4"
},
{
"operation": "Page search",
"time": "268.574",
"value": "wn27"
},
{
"operation": "hyperlink",
"time": "351.751",
"value": "27th Armoured Regiment (The Sherbrooke Fusiliers Regiment)"
},
{
"operation": "Now on",
"time": "352.609",
"value": "Les Fusiliers de Sherbrooke"
},
{
"operation": "finish",
"time": "368.396",
"value": "finish"
}
]
}
|
20899
|
Intrepid Pictures went out of business in 2016.
|
{
"evidence": [
"Intrepid Pictures_sentence_15",
"Intrepid Pictures_sentence_17",
"Intrepid Pictures_sentence_18"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIntrepid Pictures went out of business in 2016.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Intrepid Pictures_sentence_15",
"Intrepid Pictures_sentence_17",
"Intrepid Pictures_sentence_18"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6473,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.971",
"value": "?search="
},
{
"operation": "search",
"time": "21.99",
"value": "Intrepid Pictures"
},
{
"operation": "Now on",
"time": "23.163",
"value": "Intrepid Pictures"
},
{
"operation": "Highlighting",
"time": "122.912",
"value": "Intrepid Pictures_sentence_15"
},
{
"operation": "Highlighting",
"time": "124.141",
"value": "Intrepid Pictures_sentence_16"
},
{
"operation": "Highlighting deleted",
"time": "125.025",
"value": "Intrepid Pictures_sentence_16"
},
{
"operation": "Highlighting",
"time": "126.49",
"value": "Intrepid Pictures_sentence_17"
},
{
"operation": "Highlighting",
"time": "128.633",
"value": "Intrepid Pictures_sentence_18"
},
{
"operation": "finish",
"time": "134.695",
"value": "finish"
}
]
}
|
79217
|
Islington-City Centre West is a commercial and residential neighbourhood in Etobicoke, Toronto, Ontario, Canada and it exits in Brown's Line.
|
{
"evidence": [
"Islington-City Centre West_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:\nIslington-City Centre West is a commercial and residential neighbourhood in Etobicoke, Toronto, Ontario, Canada and it exits in Brown's Line.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Islington-City Centre West_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6474,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.795",
"value": "?search="
},
{
"operation": "search",
"time": "12.725",
"value": "Islington-City Centre West"
},
{
"operation": "Now on",
"time": "13.396",
"value": "Islington-City Centre West"
},
{
"operation": "Highlighting",
"time": "16.94",
"value": "Islington-City Centre West_sentence_0"
},
{
"operation": "Page search",
"time": "32.356",
"value": "brown"
},
{
"operation": "Page search",
"time": "40.404",
"value": "line"
},
{
"operation": "Page search",
"time": "74.643",
"value": "exit"
},
{
"operation": "Page search",
"time": "102.236",
"value": "Brown's"
},
{
"operation": "hyperlink",
"time": "198.388",
"value": "Bloor–Danforth line"
},
{
"operation": "Now on",
"time": "199.074",
"value": "Bloor–Danforth line"
},
{
"operation": "Page search",
"time": "207.247",
"value": "brown"
},
{
"operation": "search",
"time": "225.982",
"value": "islington"
},
{
"operation": "Now on",
"time": "226.716",
"value": "Islington"
},
{
"operation": "search",
"time": "234.335",
"value": "Islington-City Centre West"
},
{
"operation": "Now on",
"time": "234.819",
"value": "Islington-City Centre West"
},
{
"operation": "search",
"time": "283.836",
"value": "Brown's Line"
},
{
"operation": "Now on",
"time": "285.034",
"value": "List of north–south roads in Toronto"
},
{
"operation": "Page search",
"time": "297.827",
"value": "islington"
},
{
"operation": "finish",
"time": "358.531",
"value": "finish"
}
]
}
|
exfever_4165
|
Lauren Graham is an American actress and author best known for her roles as Lorelai Gilmore in the television series Gilmore Girls that aired on the American television network first launched on broadcast television on January 11, 1995.
|
{
"evidence": [
"Lauren Graham",
"The WB"
],
"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:\nLauren Graham is an American actress and author best known for her roles as Lorelai Gilmore in the television series Gilmore Girls that aired on the American television network first launched on broadcast television on January 11, 1995.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lauren Graham",
"The WB"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6475,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8220
|
Ready Player One is a 2002 science fiction novel of American author Ernest Cline, which received an award given annually to recognize "ten books written for adults that have special appeal to young adults ages 12 through 18".
|
{
"evidence": [
"Ready Player One",
"Alex Awards"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nReady Player One is a 2002 science fiction novel of American author Ernest Cline, which received an award given annually to recognize \"ten books written for adults that have special appeal to young adults ages 12 through 18\".\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ready Player One",
"Alex Awards"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6476,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
94475
|
In the Summer Games Tennis Tournament, Chung Hyeon ranked first in Group A while Gianluigi Quinzi ranked last in the group.
|
{
"evidence": [
"The Championships, Wimbledon_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:\nIn the Summer Games Tennis Tournament, Chung Hyeon ranked first in Group A while Gianluigi Quinzi ranked last in the group.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"The Championships, Wimbledon_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6477,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.402",
"value": "?search="
},
{
"operation": "search",
"time": "16.51",
"value": "Contains...Summer Games Tennis Tournament"
},
{
"operation": "Now on",
"time": "18.352",
"value": "undefined"
},
{
"operation": "search",
"time": "38.626",
"value": "Contains...Chung Hyeon ranked first in Group A"
},
{
"operation": "Now on",
"time": "41.49",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "46.566",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "47.616",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "53.041",
"value": "The Championships, Wimbledon"
},
{
"operation": "Now on",
"time": "55.546",
"value": "The Championships, Wimbledon"
},
{
"operation": "Highlighting",
"time": "63.56",
"value": "The Championships, Wimbledon_sentence_0"
},
{
"operation": "finish",
"time": "69.107",
"value": "finish"
}
]
}
|
exfever_8277
|
Jermaine Lamarr Cole is an American rapper and singer-songwriter who launched his debut studio album, released on September 27, 2011, by Roc Nation and Columbia Records.
|
{
"evidence": [
"J. Cole",
"Cole World: The Sideline Story"
],
"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:\nJermaine Lamarr Cole is an American rapper and singer-songwriter who launched his debut studio album, released on September 27, 2011, by Roc Nation and Columbia Records.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"J. Cole",
"Cole World: The Sideline Story"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6478,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
27374
|
The President of Canada awarded the Alberta Order of Excellence as a Presidential Decree, eligible only to American residents living in a selected state in Canada.
|
{
"evidence": [
"Alberta Order of Excellence_sentence_3",
"Alberta Order of Excellence_sentence_4"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe President of Canada awarded the Alberta Order of Excellence as a Presidential Decree, eligible only to American residents living in a selected state in Canada.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Alberta Order of Excellence_sentence_3",
"Alberta Order of Excellence_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6479,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.275",
"value": "?search="
},
{
"operation": "search",
"time": "23.742",
"value": "Alberta Order of Excellence"
},
{
"operation": "Now on",
"time": "24.859",
"value": "Alberta Order of Excellence"
},
{
"operation": "Page search",
"time": "56.444",
"value": "American"
},
{
"operation": "Page search",
"time": "75.32",
"value": "eligible"
},
{
"operation": "Highlighting",
"time": "78.677",
"value": "Alberta Order of Excellence_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "94.066",
"value": "Alberta Order of Excellence_sentence_4"
},
{
"operation": "Highlighting",
"time": "94.739",
"value": "Alberta Order of Excellence_sentence_3"
},
{
"operation": "Highlighting",
"time": "95.546",
"value": "Alberta Order of Excellence_sentence_4"
},
{
"operation": "finish",
"time": "121.907",
"value": "finish"
}
]
}
|
68989
|
XDAIS algorithms are a standard for algorithm development for the TMS320 DSP family, first introduced in 1999.
|
{
"evidence": [
"XDAIS algorithms_sentence_0",
"XDAIS algorithms_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:\nXDAIS algorithms are a standard for algorithm development for the TMS320 DSP family, first introduced in 1999.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"XDAIS algorithms_sentence_0",
"XDAIS algorithms_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6480,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.079",
"value": "?search="
},
{
"operation": "search",
"time": "13.329",
"value": "XDAIS algorithms"
},
{
"operation": "Now on",
"time": "13.955",
"value": "XDAIS algorithms"
},
{
"operation": "Highlighting",
"time": "17.863",
"value": "XDAIS algorithms_sentence_0"
},
{
"operation": "Highlighting",
"time": "26.399",
"value": "XDAIS algorithms_sentence_1"
},
{
"operation": "finish",
"time": "196.537",
"value": "finish"
}
]
}
|
43335
|
Jane Slowey was Vice Chair of the National Council for Voluntary Organizations from 2003 to 2008 and was previously chief executive of Birmingham Voluntary Youth Service Council.
|
{
"evidence": [
"Jane Slowey_sentence_3",
"Jane Slowey_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:\nJane Slowey was Vice Chair of the National Council for Voluntary Organizations from 2003 to 2008 and was previously chief executive of Birmingham Voluntary Youth Service Council.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Jane Slowey_sentence_3",
"Jane Slowey_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 6481,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.565",
"value": "?search="
},
{
"operation": "search",
"time": "23.283",
"value": "Jane Slowey"
},
{
"operation": "Now on",
"time": "24.093",
"value": "Jane Slowey"
},
{
"operation": "Highlighting",
"time": "29.962",
"value": "Jane Slowey_sentence_3"
},
{
"operation": "Highlighting",
"time": "33.346",
"value": "Jane Slowey_sentence_4"
},
{
"operation": "finish",
"time": "42.082",
"value": "finish"
}
]
}
|
14359
|
Hittite phonology has ten consonant pronouns and m (The bilabial nasal occurs in English) is a Labial Nasal pronoun.
|
{
"evidence": [
"Voiced bilabial nasal_sentence_2",
"Voiced bilabial nasal_sentence_3"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHittite phonology has ten consonant pronouns and m (The bilabial nasal occurs in English) is a Labial Nasal pronoun.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Voiced bilabial nasal_sentence_2",
"Voiced bilabial nasal_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 6482,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.466",
"value": "?search="
},
{
"operation": "search",
"time": "15.111",
"value": "Hittite"
},
{
"operation": "Now on",
"time": "16.449",
"value": "Hittite"
},
{
"operation": "hyperlink",
"time": "20.998",
"value": "Hittite language"
},
{
"operation": "Now on",
"time": "22.035",
"value": "Hittite language"
},
{
"operation": "search",
"time": "64.026",
"value": "\"Hittite\" \"consonant\" \"pronoun\" \"nasal\""
},
{
"operation": "Now on",
"time": "64.927",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "85.723",
"value": "Hittite Indo-European copula_section_27"
},
{
"operation": "Now on",
"time": "87.693",
"value": "Indo-European copula#Hittite Indo-European copula section 27"
},
{
"operation": "Page search",
"time": "129.765",
"value": "consonant"
},
{
"operation": "search",
"time": "324.68",
"value": "Contains...bilabial nasal occurs in English"
},
{
"operation": "Now on",
"time": "326.595",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "351.76",
"value": "Voiced bilabial nasal"
},
{
"operation": "Now on",
"time": "353.179",
"value": "Voiced bilabial nasal"
},
{
"operation": "search",
"time": "400.788",
"value": "Hittite phonology has ten consonant pronouns"
},
{
"operation": "Now on",
"time": "402.113",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "421.552",
"value": "Anatolian languages"
},
{
"operation": "Now on",
"time": "422.648",
"value": "Anatolian languages"
},
{
"operation": "Page search",
"time": "429.609",
"value": "pronoun"
},
{
"operation": "search",
"time": "512.447",
"value": "Labial Nasal pronoun"
},
{
"operation": "Now on",
"time": "513.577",
"value": "undefined"
},
{
"operation": "search",
"time": "534.973",
"value": "\"m\" \"nasal\" \"bilabial\""
},
{
"operation": "Now on",
"time": "535.806",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "554.752",
"value": "Sumerian language"
},
{
"operation": "Now on",
"time": "556.377",
"value": "Sumerian language"
},
{
"operation": "back-button-clicked",
"time": "557.123",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "557.853",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "586.805",
"value": "Voiced bilabial nasal"
},
{
"operation": "Now on",
"time": "587.52",
"value": "Voiced bilabial nasal"
},
{
"operation": "Highlighting",
"time": "625.437",
"value": "Voiced bilabial nasal_sentence_2"
},
{
"operation": "Highlighting",
"time": "627.043",
"value": "Voiced bilabial nasal_sentence_3"
},
{
"operation": "finish",
"time": "635.624",
"value": "finish"
}
]
}
|
exfever_1143
|
The third season of Games of Thrones was aired on HBO from March 31, 2013 to June 9, 2013 and was adapted from a novel published in 1995 by an American novelist, screenwriter, tv producer and short story writer.
|
{
"evidence": [
"Game of Thrones (season 3)",
"George R. R. Martin",
"A Song of Ice and Fire"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe third season of Games of Thrones was aired on HBO from March 31, 2013 to June 9, 2013 and was adapted from a novel published in 1995 by an American novelist, screenwriter, tv producer and short story writer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Game of Thrones (season 3)",
"George R. R. Martin",
"A Song of Ice and Fire"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 6483,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.