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 |
|---|---|---|---|---|---|---|---|---|
91900
|
Albert Fishlow (economist and professor) studied and published papers regarding Brazil and Latin America's development strategies, which later helped him to receive the National Order of the Southern Cross award of the Brazilian government.
|
{
"evidence": [
"Albert Fishlow_sentence_0",
"Albert Fishlow_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:\nAlbert Fishlow (economist and professor) studied and published papers regarding Brazil and Latin America's development strategies, which later helped him to receive the National Order of the Southern Cross award of the Brazilian government.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Albert Fishlow_sentence_0",
"Albert Fishlow_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5184,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.578",
"value": "?search="
},
{
"operation": "search",
"time": "8.795",
"value": "Albert Fishlow"
},
{
"operation": "Now on",
"time": "9.713",
"value": "Albert Fishlow"
},
{
"operation": "Highlighting",
"time": "15.051",
"value": "Albert Fishlow_sentence_0"
},
{
"operation": "Highlighting",
"time": "99.877",
"value": "Albert Fishlow_sentence_3"
},
{
"operation": "finish",
"time": "363.965",
"value": "finish"
}
]
}
|
55589
|
Guidano, neither an Italian neuropsychiatrist nor the first president of the Italian Society of Behavioural and Cognitive Therapy (SITCC), created the cognitive post-rationalist model based on the attachment theory, evolutionary epistemology, and the complex systems theory.
|
{
"evidence": [
"Vittorio Guidano_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:\nGuidano, neither an Italian neuropsychiatrist nor the first president of the Italian Society of Behavioural and Cognitive Therapy (SITCC), created the cognitive post-rationalist model based on the attachment theory, evolutionary epistemology, and the complex systems theory.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Vittorio Guidano_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5185,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.359",
"value": "?search="
},
{
"operation": "search",
"time": "23.273",
"value": "Contains...cognitive post-rationalist model"
},
{
"operation": "Now on",
"time": "24.933",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "28.99",
"value": "Vittorio Guidano"
},
{
"operation": "Now on",
"time": "30.223",
"value": "Vittorio Guidano"
},
{
"operation": "Highlighting",
"time": "33.928",
"value": "Vittorio Guidano_sentence_0"
},
{
"operation": "finish",
"time": "50.7",
"value": "finish"
}
]
}
|
53967
|
Hassanat Akinwande started her career in the 1980s and eventually starred in over 50 videos but was arrested in 2006 for possession of cocaine.
|
{
"evidence": [
"Hassanat Akinwande_sentence_2",
"Hassanat Akinwande_sentence_3",
"Hassanat Akinwande_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:\nHassanat Akinwande started her career in the 1980s and eventually starred in over 50 videos but was arrested in 2006 for possession of cocaine.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Hassanat Akinwande_sentence_2",
"Hassanat Akinwande_sentence_3",
"Hassanat Akinwande_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5186,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.299",
"value": "?search="
},
{
"operation": "search",
"time": "11.692",
"value": "Hassanat Akinwande"
},
{
"operation": "Now on",
"time": "12.584",
"value": "Hassanat Akinwande"
},
{
"operation": "Highlighting",
"time": "22.536",
"value": "Hassanat Akinwande_sentence_2"
},
{
"operation": "Highlighting",
"time": "25.455",
"value": "Hassanat Akinwande_sentence_3"
},
{
"operation": "Highlighting",
"time": "28.083",
"value": "Hassanat Akinwande_sentence_4"
},
{
"operation": "finish",
"time": "156.317",
"value": "finish"
}
]
}
|
exfever_7294
|
Beasts of the Southern Wild is a 2012 fantasy-drama film directed by Benh Zeitlin that was adapted by an American screenwriter and playwright from Alibar's one-act play Juicy and Delicious.
|
{
"evidence": [
"Beasts of the Southern Wild",
"Lucy Alibar"
],
"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:\nBeasts of the Southern Wild is a 2012 fantasy-drama film directed by Benh Zeitlin that was adapted by an American screenwriter and playwright from Alibar's one-act play Juicy and Delicious.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Beasts of the Southern Wild",
"Lucy Alibar"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5187,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7195
|
Paul Labile Pogba is a French professional footballer who plays in the EPL, the top level of the English football league system, for Manchester United and the French national team.
|
{
"evidence": [
"Paul Pogba",
"Premier League"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPaul Labile Pogba is a French professional footballer who plays in the EPL, the top level of the English football league system, for Manchester United and the French national team.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Paul Pogba",
"Premier League"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5188,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
73887
|
3962 Valyaev was discovered by Smirnova, a Soviet/Russian astronomer and a discoverer of minor planets and comets, on February 8, 1967 and was named after Russian astronomer Valerij Valyaev.
|
{
"evidence": [
"3962 Valyaev_sentence_3",
"Tamara Smirnova_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:\n3962 Valyaev was discovered by Smirnova, a Soviet/Russian astronomer and a discoverer of minor planets and comets, on February 8, 1967 and was named after Russian astronomer Valerij Valyaev.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"3962 Valyaev_sentence_3",
"Tamara Smirnova_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5189,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.245",
"value": "?search="
},
{
"operation": "search",
"time": "18.799",
"value": "3962 Valyaev"
},
{
"operation": "Now on",
"time": "19.726",
"value": "3962 Valyaev"
},
{
"operation": "Page search",
"time": "70.863",
"value": "smirnova"
},
{
"operation": "Highlighting",
"time": "73.258",
"value": "3962 Valyaev_sentence_3"
},
{
"operation": "hyperlink",
"time": "87.761",
"value": "Tamara Smirnova"
},
{
"operation": "Now on",
"time": "88.723",
"value": "Tamara Smirnova"
},
{
"operation": "Highlighting",
"time": "91.105",
"value": "Tamara Smirnova_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "92.553",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "92.712",
"value": "3962 Valyaev"
},
{
"operation": "finish",
"time": "391.495",
"value": "finish"
}
]
}
|
19343
|
Anastacia was often nominated in European ceremonies, and she has won 78% of her nominations.
|
{
"evidence": [
"List of awards and nominations received by Anastacia_sentence_4",
"List of awards and nominations received by Anastacia_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:\nAnastacia was often nominated in European ceremonies, and she has won 78% of her nominations.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"List of awards and nominations received by Anastacia_sentence_4",
"List of awards and nominations received by Anastacia_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5190,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.881",
"value": "?search="
},
{
"operation": "search",
"time": "10.515",
"value": "Anastacia"
},
{
"operation": "Now on",
"time": "13.356",
"value": "Anastacia"
},
{
"operation": "Page search",
"time": "19.826",
"value": "nominated"
},
{
"operation": "hyperlink",
"time": "31.482",
"value": "8 Awards Anastacia_section_14"
},
{
"operation": "hyperlink",
"time": "33.379",
"value": "List of awards and nominations received by Anastacia"
},
{
"operation": "Now on",
"time": "35.156",
"value": "List of awards and nominations received by Anastacia"
},
{
"operation": "Highlighting",
"time": "42.858",
"value": "List of awards and nominations received by Anastacia_sentence_4"
},
{
"operation": "Highlighting",
"time": "49.645",
"value": "List of awards and nominations received by Anastacia_sentence_3"
},
{
"operation": "Now on",
"time": "62.622",
"value": "List of awards and nominations received by Anastacia"
},
{
"operation": "finish",
"time": "77.925",
"value": "finish"
}
]
}
|
9647
|
Art Dubai Residents (a programme-cum-gallery section in which multidisciplinary gallery artists from all over the world are invited to the UAE to create new works) presented a four-week residency programme for first edition residents in March 2018; and the 2019 edition of residents focused excclusively on Asia.
|
{
"evidence": [
"Art Dubai_sentence_22"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nArt Dubai Residents (a programme-cum-gallery section in which multidisciplinary gallery artists from all over the world are invited to the UAE to create new works) presented a four-week residency programme for first edition residents in March 2018; and the 2019 edition of residents focused excclusively on Asia.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Art Dubai_sentence_22"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5191,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.829",
"value": "?search="
},
{
"operation": "search",
"time": "20.4",
"value": "Art Dubai"
},
{
"operation": "Now on",
"time": "21.342",
"value": "Art Dubai"
},
{
"operation": "search",
"time": "44.728",
"value": "Contains...Art Dubai Residents"
},
{
"operation": "Now on",
"time": "46.506",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "53.339",
"value": "Art Dubai"
},
{
"operation": "Now on",
"time": "53.639",
"value": "Art Dubai"
},
{
"operation": "Page search",
"time": "63.14",
"value": "programme-cum-gallery"
},
{
"operation": "Highlighting",
"time": "70.206",
"value": "Art Dubai_sentence_20"
},
{
"operation": "Highlighting",
"time": "81.535",
"value": "Art Dubai_sentence_19"
},
{
"operation": "Highlighting",
"time": "89.023",
"value": "Art Dubai_sentence_22"
},
{
"operation": "Highlighting deleted",
"time": "93.173",
"value": "Art Dubai_sentence_20"
},
{
"operation": "Highlighting deleted",
"time": "93.723",
"value": "Art Dubai_sentence_19"
},
{
"operation": "finish",
"time": "114.139",
"value": "finish"
}
]
}
|
1323
|
METAP2 is wrongly believed that the exact mechanism of catalysis is not known and is studied via x-ray crystallography.
|
{
"evidence": [
"METAP2_sentence_23"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMETAP2 is wrongly believed that the exact mechanism of catalysis is not known and is studied via x-ray crystallography.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"METAP2_sentence_23"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5192,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "15.7",
"value": "METAP2"
},
{
"operation": "Page search",
"time": "29.553",
"value": "catalysis"
},
{
"operation": "Highlighting",
"time": "74.71",
"value": "METAP2_sentence_23"
},
{
"operation": "finish",
"time": "90.685",
"value": "finish"
}
]
}
|
50548
|
Ed Rose got a Fostex X-15 recorder for Christmas as a teenager and a Fostex 250 the next year.
|
{
"evidence": [
"Ed Rose_sentence_5",
"Ed Rose_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:\nEd Rose got a Fostex X-15 recorder for Christmas as a teenager and a Fostex 250 the next year.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ed Rose_sentence_5",
"Ed Rose_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5193,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.725",
"value": "?search="
},
{
"operation": "search",
"time": "11.671",
"value": "Ed Rose"
},
{
"operation": "Now on",
"time": "12.388",
"value": "Ed Rose"
},
{
"operation": "Highlighting",
"time": "220.383",
"value": "Ed Rose_sentence_5"
},
{
"operation": "Highlighting",
"time": "231.662",
"value": "Ed Rose_sentence_6"
},
{
"operation": "finish",
"time": "239.742",
"value": "finish"
}
]
}
|
29480
|
The Specials had a successful first album, after which member Dammers wanted to expand their sound into other genres of music; the second More Specials album was probably the beginning of the fracturing of the band, because Dammers had a heated row with Bradbury, who wanted to do a "2 Tone review, mixing ska, soul, and reggae."
|
{
"evidence": [
"The Specials_sentence_30",
"The Specials_sentence_35",
"The Specials_sentence_39",
"The Specials_sentence_64",
"The Specials_sentence_40",
"The Specials_sentence_41",
"The Specials_sentence_42",
"The Specials_sentence_43",
"The Specials_sentence_4",
"The Specials_sentence_6",
"The Specials_sentence_47",
"The Specials (album)_sentence_2",
"The Specials (album)_sentence_47"
],
"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 Specials had a successful first album, after which member Dammers wanted to expand their sound into other genres of music; the second More Specials album was probably the beginning of the fracturing of the band, because Dammers had a heated row with Bradbury, who wanted to do a \"2 Tone review, mixing ska, soul, and reggae.\"\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"The Specials_sentence_30",
"The Specials_sentence_35",
"The Specials_sentence_39",
"The Specials_sentence_64",
"The Specials_sentence_40",
"The Specials_sentence_41",
"The Specials_sentence_42",
"The Specials_sentence_43",
"The Specials_sentence_4",
"The Specials_sentence_6",
"The Specials_sentence_47",
"The Specials (album)_sentence_2",
"The Specials (album)_sentence_47"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5194,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.61",
"value": "?search="
},
{
"operation": "search",
"time": "14.694",
"value": "The Specials"
},
{
"operation": "Now on",
"time": "16.174",
"value": "The Specials"
},
{
"operation": "Page search",
"time": "25.382",
"value": "first album"
},
{
"operation": "Highlighting",
"time": "91.662",
"value": "The Specials_sentence_30"
},
{
"operation": "Highlighting",
"time": "112.254",
"value": "The Specials_sentence_35"
},
{
"operation": "Highlighting",
"time": "133.739",
"value": "The Specials_sentence_39"
},
{
"operation": "Highlighting",
"time": "157.329",
"value": "The Specials_sentence_64"
},
{
"operation": "Page search",
"time": "193.669",
"value": "heated row"
},
{
"operation": "Page search",
"time": "237.13",
"value": "expand"
},
{
"operation": "Highlighting",
"time": "268.316",
"value": "The Specials_sentence_27"
},
{
"operation": "Highlighting deleted",
"time": "270.418",
"value": "The Specials_sentence_27"
},
{
"operation": "Highlighting",
"time": "311.776",
"value": "The Specials_sentence_40"
},
{
"operation": "Highlighting",
"time": "314.231",
"value": "The Specials_sentence_41"
},
{
"operation": "Highlighting",
"time": "324.859",
"value": "The Specials_sentence_42"
},
{
"operation": "Highlighting",
"time": "325.64",
"value": "The Specials_sentence_43"
},
{
"operation": "Highlighting",
"time": "354.786",
"value": "The Specials_sentence_4"
},
{
"operation": "Highlighting",
"time": "358.855",
"value": "The Specials_sentence_6"
},
{
"operation": "Highlighting",
"time": "389.421",
"value": "The Specials_sentence_47"
},
{
"operation": "Page search",
"time": "403.772",
"value": "bradbury"
},
{
"operation": "hyperlink",
"time": "432.887",
"value": "eponymous debut album"
},
{
"operation": "Now on",
"time": "433.929",
"value": "The Specials (album)"
},
{
"operation": "Highlighting",
"time": "481.662",
"value": "The Specials (album)_sentence_2"
},
{
"operation": "Highlighting",
"time": "554.568",
"value": "The Specials (album)_sentence_47"
},
{
"operation": "finish",
"time": "566.869",
"value": "finish"
}
]
}
|
84994
|
Top of the Rockies is the name of a National Scenic Byway in Colorado and most of the byway is above 9,000 ft and thus named Top of the Rockies.
|
{
"evidence": [
"Top of the Rockies_sentence_0",
"Top of the Rockies_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:\nTop of the Rockies is the name of a National Scenic Byway in Colorado and most of the byway is above 9,000 ft and thus named Top of the Rockies.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Top of the Rockies_sentence_0",
"Top of the Rockies_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5195,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.436",
"value": "?search="
},
{
"operation": "search",
"time": "12.029",
"value": "Top of the Rockies"
},
{
"operation": "Now on",
"time": "13.45",
"value": "Top of the Rockies"
},
{
"operation": "Highlighting",
"time": "16.361",
"value": "Top of the Rockies_sentence_0"
},
{
"operation": "Highlighting",
"time": "22.806",
"value": "Top of the Rockies_sentence_1"
},
{
"operation": "finish",
"time": "45.549",
"value": "finish"
}
]
}
|
40374
|
Christian Niccum, a four time Junior World Champion (with teammate McClain), and was the youngest double rider to win a World Cup in 1996 (he won two more two years later), had his best doubles finish in 2008 with sixth place in Oberhof, Germany.
|
{
"evidence": [
"Christian Niccum_sentence_6",
"Christian Niccum_sentence_7",
"Christian Niccum_sentence_8",
"Christian Niccum_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:\nChristian Niccum, a four time Junior World Champion (with teammate McClain), and was the youngest double rider to win a World Cup in 1996 (he won two more two years later), had his best doubles finish in 2008 with sixth place in Oberhof, Germany.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Christian Niccum_sentence_6",
"Christian Niccum_sentence_7",
"Christian Niccum_sentence_8",
"Christian Niccum_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5196,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.117",
"value": "?search="
},
{
"operation": "search",
"time": "19.898",
"value": "Christian Niccum"
},
{
"operation": "Now on",
"time": "21.361",
"value": "Christian Niccum"
},
{
"operation": "Highlighting",
"time": "40.966",
"value": "Christian Niccum_sentence_6"
},
{
"operation": "Highlighting",
"time": "51.757",
"value": "Christian Niccum_sentence_7"
},
{
"operation": "Highlighting",
"time": "62.424",
"value": "Christian Niccum_sentence_8"
},
{
"operation": "Highlighting",
"time": "85.386",
"value": "Christian Niccum_sentence_5"
},
{
"operation": "finish",
"time": "136.279",
"value": "finish"
}
]
}
|
81505
|
Wedson Nyirenda was involved in the management of the Zambian Under-20 team and led Zambia at the African Nations Championships (a football tournament which was first announced in September 11, 2007 and administered by the Confederation of African Football) finals.
|
{
"evidence": [
"Wedson Nyirenda_sentence_6",
"Wedson Nyirenda_sentence_8",
"African Nations Championship_sentence_1",
"African Nations Championship_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:\nWedson Nyirenda was involved in the management of the Zambian Under-20 team and led Zambia at the African Nations Championships (a football tournament which was first announced in September 11, 2007 and administered by the Confederation of African Football) finals.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Wedson Nyirenda_sentence_6",
"Wedson Nyirenda_sentence_8",
"African Nations Championship_sentence_1",
"African Nations Championship_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5197,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.02",
"value": "?search="
},
{
"operation": "search",
"time": "16.815",
"value": "Wedson Nyirenda"
},
{
"operation": "Now on",
"time": "17.495",
"value": "Wedson Nyirenda"
},
{
"operation": "Highlighting",
"time": "35.822",
"value": "Wedson Nyirenda_sentence_6"
},
{
"operation": "Highlighting",
"time": "52.021",
"value": "Wedson Nyirenda_sentence_8"
},
{
"operation": "hyperlink",
"time": "60.725",
"value": "African Nations Championships"
},
{
"operation": "Now on",
"time": "61.464",
"value": "African Nations Championship"
},
{
"operation": "Highlighting",
"time": "68.784",
"value": "African Nations Championship_sentence_1"
},
{
"operation": "Highlighting",
"time": "76.279",
"value": "African Nations Championship_sentence_2"
},
{
"operation": "finish",
"time": "196.66",
"value": "finish"
}
]
}
|
90204
|
Käferberg is in the north of the Limmat Valley, a river valley and a region in Zürich and Aargau in Switzerland.
|
{
"evidence": [
"Käferberg_sentence_4",
"Limmat Valley_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:\nKäferberg is in the north of the Limmat Valley, a river valley and a region in Zürich and Aargau in Switzerland.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Käferberg_sentence_4",
"Limmat Valley_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5198,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.462",
"value": "?search="
},
{
"operation": "search",
"time": "11.129",
"value": "Käferberg"
},
{
"operation": "Now on",
"time": "12.138",
"value": "Käferberg"
},
{
"operation": "Highlighting",
"time": "44.089",
"value": "Käferberg_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "60.517",
"value": "Käferberg_sentence_2"
},
{
"operation": "search",
"time": "70.555",
"value": "Limmat Valley"
},
{
"operation": "Now on",
"time": "71.618",
"value": "Limmat Valley"
},
{
"operation": "Highlighting",
"time": "85.826",
"value": "Limmat Valley_sentence_0"
},
{
"operation": "Page search",
"time": "95.201",
"value": "Käferberg"
},
{
"operation": "back-button-clicked",
"time": "115.608",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "115.911",
"value": "Käferberg"
},
{
"operation": "Page search",
"time": "130.282",
"value": "Limmat Valley"
},
{
"operation": "Highlighting",
"time": "162.689",
"value": "Käferberg_sentence_4"
},
{
"operation": "Highlighting",
"time": "179.353",
"value": "Käferberg_sentence_4"
},
{
"operation": "hyperlink",
"time": "200.539",
"value": "Limmat valley"
},
{
"operation": "Now on",
"time": "201.576",
"value": "Limmat valley"
},
{
"operation": "Highlighting",
"time": "203.892",
"value": "Limmat Valley_sentence_0"
},
{
"operation": "finish",
"time": "220.275",
"value": "finish"
}
]
}
|
exfever_1926
|
Yango, also known as Rangoon, is the capital of the Yangon Region and the largest city in Burma, a country in Southeast Asia.
|
{
"evidence": [
"Yangon",
"Myanmar"
],
"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:\nYango, also known as Rangoon, is the capital of the Yangon Region and the largest city in Burma, a country in Southeast Asia.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Yangon",
"Myanmar"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5199,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5890
|
Touch My Body is a song by an American singer and songwriter who released a debut album under the guidance of an American music executive, producer, and author.
|
{
"evidence": [
"Touch My Body",
"Tommy Mottola"
],
"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:\nTouch My Body is a song by an American singer and songwriter who released a debut album under the guidance of an American music executive, producer, and author.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Touch My Body",
"Tommy Mottola"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5200,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
61541
|
Eudaronia aperta is a four-legged species in the kingdom Animalia.
|
{
"evidence": [
"Eudaronia aperta_sentence_0",
"Mollusca_sentence_17"
],
"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:\nEudaronia aperta is a four-legged species in the kingdom Animalia.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Eudaronia aperta_sentence_0",
"Mollusca_sentence_17"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5201,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.966",
"value": "?search="
},
{
"operation": "search",
"time": "24.328",
"value": "Eudaronia aperta"
},
{
"operation": "Now on",
"time": "25.586",
"value": "Eudaronia aperta"
},
{
"operation": "Highlighting",
"time": "39.277",
"value": "Eudaronia aperta_sentence_0"
},
{
"operation": "hyperlink",
"time": "46.891",
"value": "mollusk"
},
{
"operation": "Now on",
"time": "49.561",
"value": "Mollusca"
},
{
"operation": "Highlighting",
"time": "79.155",
"value": "Mollusca_sentence_17"
},
{
"operation": "finish",
"time": "98.705",
"value": "finish"
}
]
}
|
exfever_3793
|
James Eugene Carrey gained further notice for his portrayal of the fictional character created by Dr. Seuss, who has also been portrayed and voiced by many different actors, including an American actor known for providing the voice of Walt Disney's George Darling.
|
{
"evidence": [
"Jim Carrey",
"Grinch",
"Hans Conried"
],
"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:\nJames Eugene Carrey gained further notice for his portrayal of the fictional character created by Dr. Seuss, who has also been portrayed and voiced by many different actors, including an American actor known for providing the voice of Walt Disney's George Darling.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jim Carrey",
"Grinch",
"Hans Conried"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5202,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
2349
|
Named after Anabaptist movement leader Michael Sattler, the Sattler College in Boston has an annual tuition rate of $9,270 in 2016.
|
{
"evidence": [
"Sattler College_sentence_0",
"Sattler College_sentence_11"
],
"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:\nNamed after Anabaptist movement leader Michael Sattler, the Sattler College in Boston has an annual tuition rate of $9,270 in 2016.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Sattler College_sentence_0",
"Sattler College_sentence_11"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5203,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "48.358",
"value": "Sattler College"
},
{
"operation": "Highlighting",
"time": "101.812",
"value": "Sattler College_sentence_5"
},
{
"operation": "Highlighting",
"time": "101.84",
"value": "Sattler College_section_0"
},
{
"operation": "Highlighting",
"time": "153.837",
"value": "Sattler College_sentence_0"
},
{
"operation": "Highlighting",
"time": "157.313",
"value": "Sattler College_sentence_11"
},
{
"operation": "hyperlink",
"time": "158.986",
"value": "Anabaptist"
},
{
"operation": "back-button-clicked",
"time": "160.496",
"value": "back-button-clicked"
},
{
"operation": "Highlighting deleted",
"time": "161.775",
"value": "Sattler College_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "161.842",
"value": "Sattler College_section_0"
},
{
"operation": "finish",
"time": "184.648",
"value": "finish"
}
]
}
|
36065
|
The Ammons haunting case occurred in Gary, Indiana where numerous movies have been filmed; and the Department of Child Services was alerted to the Ammon family whose story was sensationalized by news outlets.
|
{
"evidence": [
"Ammons haunting case_sentence_0",
"Ammons haunting case_sentence_41",
"Ammons haunting case_sentence_17",
"Ammons haunting case_sentence_19"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Ammons haunting case occurred in Gary, Indiana where numerous movies have been filmed; and the Department of Child Services was alerted to the Ammon family whose story was sensationalized by news outlets.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Ammons haunting case_sentence_0",
"Ammons haunting case_sentence_41",
"Ammons haunting case_sentence_17",
"Ammons haunting case_sentence_19"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5204,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.091",
"value": "?search="
},
{
"operation": "search",
"time": "24.299",
"value": "Ammons haunting case"
},
{
"operation": "Now on",
"time": "25.292",
"value": "Ammons haunting case"
},
{
"operation": "Highlighting",
"time": "31.508",
"value": "Ammons haunting case_sentence_0"
},
{
"operation": "Page search",
"time": "39.32",
"value": "movies"
},
{
"operation": "Page search",
"time": "65.641",
"value": "movie"
},
{
"operation": "Page search",
"time": "74.106",
"value": "films"
},
{
"operation": "Highlighting",
"time": "149.577",
"value": "Ammons haunting case_sentence_41"
},
{
"operation": "Page search",
"time": "171.873",
"value": "child"
},
{
"operation": "Highlighting",
"time": "197.667",
"value": "Ammons haunting case_sentence_17"
},
{
"operation": "Highlighting",
"time": "228.49",
"value": "Ammons haunting case_sentence_19"
},
{
"operation": "finish",
"time": "279.186",
"value": "finish"
}
]
}
|
67593
|
Human rights in Estonia has faced areas of concern such as detention conditions, police use of force, and child abuse; which is the physical, sexual, and /or psychological maltreatment or neglect of a child or children, especially by a parent or a caregiver.
|
{
"evidence": [
"Human rights in Estonia_sentence_1",
"Child abuse_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:\nHuman rights in Estonia has faced areas of concern such as detention conditions, police use of force, and child abuse; which is the physical, sexual, and /or psychological maltreatment or neglect of a child or children, especially by a parent or a caregiver.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Human rights in Estonia_sentence_1",
"Child abuse_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5205,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.095",
"value": "?search="
},
{
"operation": "search",
"time": "9.487",
"value": "Human rights in Estonia"
},
{
"operation": "Now on",
"time": "10.808",
"value": "Human rights in Estonia"
},
{
"operation": "Highlighting",
"time": "18.584",
"value": "Human rights in Estonia_sentence_1"
},
{
"operation": "hyperlink",
"time": "34.589",
"value": "child abuse"
},
{
"operation": "Now on",
"time": "36.562",
"value": "Child abuse"
},
{
"operation": "Highlighting",
"time": "39.053",
"value": "Child abuse_sentence_2"
},
{
"operation": "finish",
"time": "54.388",
"value": "finish"
}
]
}
|
exfever_6898
|
Suzanne Brigit Bird is the first American-Israeli professional basketball player for Seattle Storm of the Women's National Basketball Association.
|
{
"evidence": [
"Sue Bird",
"Denver Nuggets",
"Northwest Division (NBA)"
],
"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:\nSuzanne Brigit Bird is the first American-Israeli professional basketball player for Seattle Storm of the Women's National Basketball Association.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sue Bird",
"Denver Nuggets",
"Northwest Division (NBA)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5206,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8957
|
Harold Ramis was an American actor, comedian, director, and writer, best known for his film acting roles in a 1981 American war comedy film directed by Norman Jewison.
|
{
"evidence": [
"Harold Ramis",
"Stripes (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:\nHarold Ramis was an American actor, comedian, director, and writer, best known for his film acting roles in a 1981 American war comedy film directed by Norman Jewison.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Harold Ramis",
"Stripes (film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5207,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1939
|
The Shape of Water is a 2017 British film directed by Guillermo def toro is a subgenre of fantasy fiction, that describes a fantasy story using many of the elements and conventions of the chivalric romance genre.
|
{
"evidence": [
"The Shape of Water (film)",
"Romantic fantasy"
],
"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 Shape of Water is a 2017 British film directed by Guillermo def toro is a subgenre of fantasy fiction, that describes a fantasy story using many of the elements and conventions of the chivalric romance genre.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Shape of Water (film)",
"Romantic fantasy"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5208,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
48755
|
John Lindsay set the world record for the Men's 200m TW3 event at the ninth Summer Paralympics event in 1992.
|
{
"evidence": [
"John Lindsay (Paralympian)_sentence_0",
"John Lindsay (Paralympian)_sentence_4",
"John Lindsay (Paralympian)_sentence_2",
"1992 Summer Paralympics_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:\nJohn Lindsay set the world record for the Men's 200m TW3 event at the ninth Summer Paralympics event in 1992.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"John Lindsay (Paralympian)_sentence_0",
"John Lindsay (Paralympian)_sentence_4",
"John Lindsay (Paralympian)_sentence_2",
"1992 Summer Paralympics_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5209,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.272",
"value": "?search="
},
{
"operation": "search",
"time": "8.586",
"value": "John Lindsay"
},
{
"operation": "Now on",
"time": "9.606",
"value": "John Lindsay"
},
{
"operation": "hyperlink",
"time": "13.565",
"value": "John Lindsay (disambiguation)"
},
{
"operation": "Now on",
"time": "14.186",
"value": "John Lindsay (disambiguation)"
},
{
"operation": "hyperlink",
"time": "19.006",
"value": "3 Sports John Lindsay (disambiguation)_section_2"
},
{
"operation": "hyperlink",
"time": "26.537",
"value": "John Lindsay (Paralympian)"
},
{
"operation": "Now on",
"time": "27.458",
"value": "John Lindsay (Paralympian)"
},
{
"operation": "Highlighting",
"time": "32.22",
"value": "John Lindsay (Paralympian)_sentence_0"
},
{
"operation": "Highlighting",
"time": "67.67",
"value": "John Lindsay (Paralympian)_sentence_4"
},
{
"operation": "Highlighting",
"time": "99.82",
"value": "John Lindsay (Paralympian)_sentence_2"
},
{
"operation": "hyperlink",
"time": "111.866",
"value": "1992 Barcelona Games"
},
{
"operation": "Now on",
"time": "112.67",
"value": "1992 Summer Paralympics"
},
{
"operation": "Highlighting",
"time": "115.12",
"value": "1992 Summer Paralympics_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "140.111",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "140.459",
"value": "John Lindsay (Paralympian)"
},
{
"operation": "finish",
"time": "169.858",
"value": "finish"
}
]
}
|
76291
|
A Meeting by the River received professional ratings from three musci critics.
|
{
"evidence": [
"A Meeting by the River_sentence_27",
"A Meeting by the River_sentence_29"
],
"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:\nA Meeting by the River received professional ratings from three musci critics.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"A Meeting by the River_sentence_27",
"A Meeting by the River_sentence_29"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5210,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.758",
"value": "?search="
},
{
"operation": "search",
"time": "14.332",
"value": "A Meeting by the River"
},
{
"operation": "Now on",
"time": "15.309",
"value": "A Meeting by the River"
},
{
"operation": "Highlighting",
"time": "45.328",
"value": "A Meeting by the River_sentence_27"
},
{
"operation": "Highlighting",
"time": "49.445",
"value": "A Meeting by the River_sentence_29"
},
{
"operation": "finish",
"time": "99.712",
"value": "finish"
}
]
}
|
83249
|
Wetlet Township is a former township of the Shwebo district of the Sagaing region of Burma, a country which is also known as Myanmar.
|
{
"evidence": [
"Wetlet Township_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:\nWetlet Township is a former township of the Shwebo district of the Sagaing region of Burma, a country which is also known as Myanmar.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Wetlet Township_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5211,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.935",
"value": "?search="
},
{
"operation": "search",
"time": "8.409",
"value": "Wetlet Township"
},
{
"operation": "Now on",
"time": "9.265",
"value": "Wetlet Township"
},
{
"operation": "Highlighting",
"time": "25.791",
"value": "Wetlet Township_sentence_0"
},
{
"operation": "finish",
"time": "61.082",
"value": "finish"
}
]
}
|
33352
|
Polypodium is derived from Ancient Greek, they are commonly called polypodies and are terrestrial.
|
{
"evidence": [
"Polypodium_sentence_3",
"Polypodium_sentence_4",
"Polypodium_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:\nPolypodium is derived from Ancient Greek, they are commonly called polypodies and are terrestrial.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Polypodium_sentence_3",
"Polypodium_sentence_4",
"Polypodium_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5212,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.887",
"value": "?search="
},
{
"operation": "search",
"time": "10.596",
"value": "Polypodium"
},
{
"operation": "Now on",
"time": "11.389",
"value": "Polypodium"
},
{
"operation": "Highlighting",
"time": "14.323",
"value": "Polypodium_sentence_3"
},
{
"operation": "Highlighting",
"time": "20.791",
"value": "Polypodium_sentence_4"
},
{
"operation": "Highlighting",
"time": "23.873",
"value": "Polypodium_sentence_5"
},
{
"operation": "finish",
"time": "29.028",
"value": "finish"
}
]
}
|
13166
|
Medallists in synchronised swimming at the Commonwealth Games include winners such as Sylvie Fréchette, and Lisa Alexander for gold, and Kerry Shacklock for bronze.
|
{
"evidence": [
"Kerry Shacklock_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMedallists in synchronised swimming at the Commonwealth Games include winners such as Sylvie Fréchette, and Lisa Alexander for gold, and Kerry Shacklock for bronze.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Kerry Shacklock_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5213,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.436",
"value": "undefined"
},
{
"operation": "search",
"time": "15.653",
"value": "Medallists in synchronised swimming at the Commonwealth Games"
},
{
"operation": "Now on",
"time": "16.755",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "30.162",
"value": "2010 Commonwealth Games"
},
{
"operation": "Now on",
"time": "31.43",
"value": "2010 Commonwealth Games"
},
{
"operation": "Page search",
"time": "45.199",
"value": "Lisa Alexander"
},
{
"operation": "Page search",
"time": "46.304",
"value": "Lisa Alexander"
},
{
"operation": "search",
"time": "48.467",
"value": "Lisa Alexander f"
},
{
"operation": "Now on",
"time": "49.922",
"value": "undefined"
},
{
"operation": "search",
"time": "52.218",
"value": "Lisa Alexander"
},
{
"operation": "Now on",
"time": "53.15",
"value": "Lisa Alexander"
},
{
"operation": "hyperlink",
"time": "67.492",
"value": "1994 Commonwealth Games"
},
{
"operation": "Now on",
"time": "69.438",
"value": "1994 Commonwealth Games"
},
{
"operation": "back-button-clicked",
"time": "73.807",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "73.854",
"value": "Lisa Alexander"
},
{
"operation": "Highlighting",
"time": "75.352",
"value": "Lisa Alexander_sentence_5"
},
{
"operation": "Highlighting",
"time": "75.354",
"value": "Lisa Alexander_title"
},
{
"operation": "Highlighting",
"time": "75.354",
"value": "Lisa Alexander_section_0"
},
{
"operation": "search",
"time": "87.809",
"value": "Kerry Shacklock"
},
{
"operation": "Now on",
"time": "88.559",
"value": "Kerry Shacklock"
},
{
"operation": "Highlighting",
"time": "114.967",
"value": "Kerry Shacklock_cell_0_15_0"
},
{
"operation": "Highlighting",
"time": "114.969",
"value": "Kerry Shacklock_section_0"
},
{
"operation": "Highlighting",
"time": "114.974",
"value": "Kerry Shacklock_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "115.505",
"value": "Kerry Shacklock_cell_0_15_1"
},
{
"operation": "Highlighting",
"time": "115.509",
"value": "Kerry Shacklock_header_cell_0_0_1"
},
{
"operation": "Highlighting",
"time": "116.155",
"value": "Kerry Shacklock_cell_0_16_0"
},
{
"operation": "Highlighting",
"time": "116.16",
"value": "Kerry Shacklock_header_cell_0_0_3"
},
{
"operation": "Highlighting",
"time": "116.676",
"value": "Kerry Shacklock_cell_0_15_3"
},
{
"operation": "Highlighting",
"time": "119.445",
"value": "Kerry Shacklock_cell_0_5_0"
},
{
"operation": "Highlighting",
"time": "119.831",
"value": "Kerry Shacklock_cell_0_5_1"
},
{
"operation": "Highlighting",
"time": "120.849",
"value": "Kerry Shacklock_cell_0_6_0"
},
{
"operation": "Highlighting",
"time": "121.299",
"value": "Kerry Shacklock_cell_0_5_3"
},
{
"operation": "Highlighting",
"time": "129.897",
"value": "Kerry Shacklock_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "149.902",
"value": "Kerry Shacklock_cell_0_5_1"
},
{
"operation": "Highlighting deleted",
"time": "151.734",
"value": "Kerry Shacklock_cell_0_5_3"
},
{
"operation": "Highlighting deleted",
"time": "152.104",
"value": "Kerry Shacklock_cell_0_6_0"
},
{
"operation": "Highlighting deleted",
"time": "153.082",
"value": "Kerry Shacklock_cell_0_5_0"
},
{
"operation": "Highlighting deleted",
"time": "153.988",
"value": "Kerry Shacklock_cell_0_15_0"
},
{
"operation": "Highlighting deleted",
"time": "153.993",
"value": "Kerry Shacklock_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "154.396",
"value": "Kerry Shacklock_cell_0_15_1"
},
{
"operation": "Highlighting deleted",
"time": "154.401",
"value": "Kerry Shacklock_header_cell_0_0_1"
},
{
"operation": "Highlighting deleted",
"time": "155.226",
"value": "Kerry Shacklock_cell_0_16_0"
},
{
"operation": "Highlighting deleted",
"time": "155.619",
"value": "Kerry Shacklock_cell_0_15_3"
},
{
"operation": "Highlighting deleted",
"time": "155.621",
"value": "Kerry Shacklock_section_0"
},
{
"operation": "Highlighting deleted",
"time": "155.624",
"value": "Kerry Shacklock_header_cell_0_0_3"
},
{
"operation": "finish",
"time": "165.084",
"value": "finish"
}
]
}
|
48030
|
A portable electronic voting authentication device, the INEC card reader uses cryptographic technology and was first used in Zambia's fifth presidential election in 2015, since the end of military rule in 1999.
|
{
"evidence": [
"INEC card reader_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:\nA portable electronic voting authentication device, the INEC card reader uses cryptographic technology and was first used in Zambia's fifth presidential election in 2015, since the end of military rule in 1999.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"INEC card reader_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5214,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.965",
"value": "?search="
},
{
"operation": "search",
"time": "11.623",
"value": "INEC card reader"
},
{
"operation": "Now on",
"time": "12.682",
"value": "INEC card reader"
},
{
"operation": "Highlighting",
"time": "18.599",
"value": "INEC card reader_sentence_0"
},
{
"operation": "Highlighting",
"time": "26.879",
"value": "INEC card reader_sentence_3"
},
{
"operation": "Highlighting",
"time": "40.379",
"value": "INEC card reader_sentence_6"
},
{
"operation": "hyperlink",
"time": "45.475",
"value": "Nigeria's presidential election"
},
{
"operation": "Now on",
"time": "46.674",
"value": "2015 Nigerian general election"
},
{
"operation": "Highlighting",
"time": "52.113",
"value": "2015 Nigerian general election_sentence_0"
},
{
"operation": "Highlighting",
"time": "52.114",
"value": "2015 Nigerian general election_title"
},
{
"operation": "hyperlink",
"time": "85.289",
"value": "Nigeria"
},
{
"operation": "Now on",
"time": "86.949",
"value": "Nigeria"
},
{
"operation": "back-button-clicked",
"time": "98.363",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "99.211",
"value": "2015 Nigerian general election"
},
{
"operation": "back-button-clicked",
"time": "101.119",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "101.123",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "101.464",
"value": "INEC card reader"
},
{
"operation": "Highlighting deleted",
"time": "103.256",
"value": "INEC card reader_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "105.044",
"value": "INEC card reader_sentence_3"
},
{
"operation": "search",
"time": "121.334",
"value": "Zambia"
},
{
"operation": "Now on",
"time": "122.471",
"value": "Zambia"
},
{
"operation": "back-button-clicked",
"time": "174.301",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "174.557",
"value": "INEC card reader"
},
{
"operation": "hyperlink",
"time": "191.029",
"value": "Nigeria's presidential election"
},
{
"operation": "Now on",
"time": "192.291",
"value": "Nigerian general election, 2015"
},
{
"operation": "finish",
"time": "246.446",
"value": "finish"
}
]
}
|
exfever_8229
|
30 Seconds to Mars had been in the works for a couple of years, with the majority of the songs written by an American actor known for his method of acting in a variety of roles.
|
{
"evidence": [
"30 Seconds to Mars (album)",
"Jared Leto"
],
"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:\n30 Seconds to Mars had been in the works for a couple of years, with the majority of the songs written by an American actor known for his method of acting in a variety of roles.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"30 Seconds to Mars (album)",
"Jared Leto"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5215,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5918
|
Dangerous Beauty is a 1998 American biographical drama film starring an English actress on stage and screen who appeared in several films, including Braveheart and The Land Girls.
|
{
"evidence": [
"Dangerous Beauty",
"Catherine McCormack"
],
"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:\nDangerous Beauty is a 1998 American biographical drama film starring an English actress on stage and screen who appeared in several films, including Braveheart and The Land Girls.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dangerous Beauty",
"Catherine McCormack"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5216,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
74955
|
Methacton High School (believed to be an American Indian term for "windy hill.") was falsely believed to be renamed from the Lower Providence-Worcester Joint Junior-Senior High School with the first class graduating in 1964
|
{
"evidence": [
"Methacton High School_sentence_7",
"Methacton High School_sentence_5",
"Methacton High School_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:\nMethacton High School (believed to be an American Indian term for \"windy hill.\") was falsely believed to be renamed from the Lower Providence-Worcester Joint Junior-Senior High School with the first class graduating in 1964\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Methacton High School_sentence_7",
"Methacton High School_sentence_5",
"Methacton High School_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5217,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.945",
"value": "?search="
},
{
"operation": "search",
"time": "25.756",
"value": "Methacton High School"
},
{
"operation": "Now on",
"time": "26.6",
"value": "Methacton High School"
},
{
"operation": "Page search",
"time": "45.485",
"value": "windy"
},
{
"operation": "Highlighting",
"time": "49.459",
"value": "Methacton High School_sentence_7"
},
{
"operation": "Highlighting",
"time": "102.318",
"value": "Methacton High School_sentence_5"
},
{
"operation": "Highlighting",
"time": "103.384",
"value": "Methacton High School_sentence_6"
},
{
"operation": "finish",
"time": "117.594",
"value": "finish"
}
]
}
|
90476
|
According to Benin's 2013 census, the total population of the Borgou department was 1,214,249, with 607,013 males and 607,236 females and the proportion of women was 50.00%.
|
{
"evidence": [
"Borgou Department_sentence_16",
"Borgou Department_sentence_17"
],
"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:\nAccording to Benin's 2013 census, the total population of the Borgou department was 1,214,249, with 607,013 males and 607,236 females and the proportion of women was 50.00%.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Borgou Department_sentence_16",
"Borgou Department_sentence_17"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5218,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.026",
"value": "?search="
},
{
"operation": "search",
"time": "64.985",
"value": "Contains...Benin's 2013 census"
},
{
"operation": "Now on",
"time": "66.514",
"value": "undefined"
},
{
"operation": "search",
"time": "90.194",
"value": "Borgou Department"
},
{
"operation": "Now on",
"time": "90.904",
"value": "Borgou Department"
},
{
"operation": "Highlighting",
"time": "113.457",
"value": "Borgou Department_sentence_16"
},
{
"operation": "Highlighting",
"time": "125.648",
"value": "Borgou Department_sentence_17"
},
{
"operation": "finish",
"time": "142.788",
"value": "finish"
}
]
}
|
61300
|
Claude Sheppard was in the Royal Artillary, colloquially known as the Gunners, which is the artillary arm of the British Army.
|
{
"evidence": [
"Claude Sheppard_sentence_4",
"Royal Artillery_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:\nClaude Sheppard was in the Royal Artillary, colloquially known as the Gunners, which is the artillary arm of the British Army.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Claude Sheppard_sentence_4",
"Royal Artillery_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5219,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.269",
"value": "?search="
},
{
"operation": "search",
"time": "8.482",
"value": "Claude Sheppard"
},
{
"operation": "Now on",
"time": "9.526",
"value": "Claude Sheppard"
},
{
"operation": "Highlighting",
"time": "52.019",
"value": "Claude Sheppard_sentence_4"
},
{
"operation": "hyperlink",
"time": "55.144",
"value": "Royal Artillery"
},
{
"operation": "Now on",
"time": "56.254",
"value": "Royal Artillery"
},
{
"operation": "Highlighting",
"time": "57.384",
"value": "Royal Artillery_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "59.134",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "59.2",
"value": "Claude Sheppard"
},
{
"operation": "hyperlink",
"time": "62.515",
"value": "Royal Artillery"
},
{
"operation": "Now on",
"time": "63.276",
"value": "Royal Artillery"
},
{
"operation": "finish",
"time": "68.611",
"value": "finish"
}
]
}
|
exfever_2934
|
Before being accepted at the Bristol Old Vic Theatre School, Daniel Day-Lewis excelled on the stage of a youth theatre and a registered charity in the capital and largest city in England and the United Kingdom.
|
{
"evidence": [
"Daniel Day-Lewis",
"National Youth Theatre",
"London"
],
"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:\nBefore being accepted at the Bristol Old Vic Theatre School, Daniel Day-Lewis excelled on the stage of a youth theatre and a registered charity in the capital and largest city in England and the United Kingdom.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Daniel Day-Lewis",
"National Youth Theatre",
"London"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5220,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1976
|
One Piece is a Japanese manga series written and illustrated by Masashi Kishimoto, who received an honorable mention in a Japanese firm founded in 1925's monthly Hop Step Award.
|
{
"evidence": [
"Naruto",
"Shueisha"
],
"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:\nOne Piece is a Japanese manga series written and illustrated by Masashi Kishimoto, who received an honorable mention in a Japanese firm founded in 1925's monthly Hop Step Award.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Naruto",
"Shueisha"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5221,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2326
|
Vantage Point is a political action thriller film that focuses on an assassination attempt on the President of the United States in Salamanca, in a country in southwestern Asia.
|
{
"evidence": [
"Vantage Point (film)",
"Spain"
],
"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:\nVantage Point is a political action thriller film that focuses on an assassination attempt on the President of the United States in Salamanca, in a country in southwestern Asia.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Vantage Point (film)",
"Spain"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5222,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
44325
|
Live à Paris is the third live album featuring primarily French-language songs by Canadian singer Celine Dion, released on 21 October 1996 by Columbia Records (a British record label owned by Sony Music Entertainment).
|
{
"evidence": [
"Live à Paris_sentence_1",
"Columbia Records_sentence_3"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLive à Paris is the third live album featuring primarily French-language songs by Canadian singer Celine Dion, released on 21 October 1996 by Columbia Records (a British record label owned by Sony Music Entertainment).\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Live à Paris_sentence_1",
"Columbia Records_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5223,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.153",
"value": "?search="
},
{
"operation": "search",
"time": "12.843",
"value": "Live à Paris"
},
{
"operation": "Now on",
"time": "13.707",
"value": "Live à Paris"
},
{
"operation": "Highlighting",
"time": "33.946",
"value": "Live à Paris_sentence_1"
},
{
"operation": "Highlighting",
"time": "41.012",
"value": "Live à Paris_sentence_2"
},
{
"operation": "hyperlink",
"time": "50.044",
"value": "Columbia Records"
},
{
"operation": "Now on",
"time": "51.713",
"value": "Columbia Records"
},
{
"operation": "back-button-clicked",
"time": "56.609",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "56.679",
"value": "Live à Paris"
},
{
"operation": "Highlighting",
"time": "57.538",
"value": "Live à Paris_sentence_1"
},
{
"operation": "hyperlink",
"time": "58.313",
"value": "Columbia Records"
},
{
"operation": "Now on",
"time": "58.729",
"value": "Columbia Records"
},
{
"operation": "Highlighting",
"time": "62.172",
"value": "Columbia Records_sentence_3"
},
{
"operation": "finish",
"time": "74.671",
"value": "finish"
}
]
}
|
59704
|
Jhujhar Singh was a raja of the Orchha region (a kingdom situated in the Bundelkhand region and later a princely state in British India) during the 17th century, and the first-born son of Vir Singh Deo and the senior of his three queens.
|
{
"evidence": [
"Jhujhar Singh_sentence_2",
"Jhujhar Singh_sentence_3",
"Orchha State_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:\nJhujhar Singh was a raja of the Orchha region (a kingdom situated in the Bundelkhand region and later a princely state in British India) during the 17th century, and the first-born son of Vir Singh Deo and the senior of his three queens.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Jhujhar Singh_sentence_2",
"Jhujhar Singh_sentence_3",
"Orchha State_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5224,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.289",
"value": "?search="
},
{
"operation": "search",
"time": "11.745",
"value": "Jhujhar Singh"
},
{
"operation": "Now on",
"time": "12.562",
"value": "Jhujhar Singh"
},
{
"operation": "Highlighting",
"time": "17.768",
"value": "Jhujhar Singh_sentence_2"
},
{
"operation": "Highlighting",
"time": "28.712",
"value": "Jhujhar Singh_sentence_3"
},
{
"operation": "hyperlink",
"time": "37.928",
"value": "Orchha region"
},
{
"operation": "Now on",
"time": "38.882",
"value": "Orchha State"
},
{
"operation": "Highlighting",
"time": "49.492",
"value": "Orchha State_sentence_0"
},
{
"operation": "finish",
"time": "57.874",
"value": "finish"
}
]
}
|
exfever_4490
|
Colin Edward Quinn's notable film work includes his role as Dooey in a 1999 American comedy film.
|
{
"evidence": [
"Colin Quinn",
"A Night at the Roxbury"
],
"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:\nColin Edward Quinn's notable film work includes his role as Dooey in a 1999 American comedy film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Colin Quinn",
"A Night at the Roxbury"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5225,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
87522
|
Tortoise Beats Hare is only a fairy tale that is still in the process of becoming a film.
|
{
"evidence": [
"Tortoise Beats Hare_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:\nTortoise Beats Hare is only a fairy tale that is still in the process of becoming a film.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Tortoise Beats Hare_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5226,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.125",
"value": "?search="
},
{
"operation": "search",
"time": "124.584",
"value": "Tortoise Beats Hare"
},
{
"operation": "Now on",
"time": "126.573",
"value": "Tortoise Beats Hare"
},
{
"operation": "Highlighting",
"time": "289.526",
"value": "Tortoise Beats Hare_sentence_1"
},
{
"operation": "finish",
"time": "299.359",
"value": "finish"
}
]
}
|
76622
|
Norwegian rider Kurt Asle Arvesen of Team CSC lost to Italian rider Enrico Gasparotto of Team Liquigas in the 2007 Danmark Rundt.
|
{
"evidence": [
"2007 Danmark Rundt_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:\nNorwegian rider Kurt Asle Arvesen of Team CSC lost to Italian rider Enrico Gasparotto of Team Liquigas in the 2007 Danmark Rundt.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"2007 Danmark Rundt_sentence_6"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5227,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.057",
"value": "?search="
},
{
"operation": "search",
"time": "23.169",
"value": "2007 Danmark Rundt"
},
{
"operation": "Now on",
"time": "23.771",
"value": "2007 Danmark Rundt"
},
{
"operation": "Highlighting",
"time": "34.287",
"value": "2007 Danmark Rundt_sentence_6"
},
{
"operation": "finish",
"time": "60.367",
"value": "finish"
}
]
}
|
exfever_497
|
James Blake Litherland is an English multi-instrumentalist and record producer who collaborated with an American indie-folk band founded in 2009 by singer-songwriter Justin Vernon.
|
{
"evidence": [
"James Blake (musician)",
"Bon Iver"
],
"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:\nJames Blake Litherland is an English multi-instrumentalist and record producer who collaborated with an American indie-folk band founded in 2009 by singer-songwriter Justin Vernon.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"James Blake (musician)",
"Bon Iver"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5228,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
15515
|
Nikon D3S (first full-frame camera to feature HD (720p/30) video recording) is a Digital single-lens reflex camera released on 14 October 2009.
|
{
"evidence": [
"Nikon D3S_sentence_2",
"Nikon D3S_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:\nNikon D3S (first full-frame camera to feature HD (720p/30) video recording) is a Digital single-lens reflex camera released on 14 October 2009.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Nikon D3S_sentence_2",
"Nikon D3S_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5229,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.566",
"value": "?search="
},
{
"operation": "search",
"time": "7.638",
"value": "Nikon D3S"
},
{
"operation": "Now on",
"time": "8.39",
"value": "Nikon D3S"
},
{
"operation": "Highlighting",
"time": "27.542",
"value": "Nikon D3S_sentence_2"
},
{
"operation": "Highlighting",
"time": "37.495",
"value": "Nikon D3S_sentence_0"
},
{
"operation": "finish",
"time": "44.162",
"value": "finish"
}
]
}
|
56694
|
Devan Ramachandran has been working on the conditions of the bad roads along with the flooding in Kochi City which has been vastly improved ever since.
|
{
"evidence": [
"Devan Ramachandran_sentence_18",
"Devan Ramachandran_sentence_20"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDevan Ramachandran has been working on the conditions of the bad roads along with the flooding in Kochi City which has been vastly improved ever since.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Devan Ramachandran_sentence_18",
"Devan Ramachandran_sentence_20"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5230,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.844",
"value": "?search="
},
{
"operation": "search",
"time": "25.186",
"value": "Devan Ramachandran"
},
{
"operation": "Now on",
"time": "25.838",
"value": "Devan Ramachandran"
},
{
"operation": "Page search",
"time": "31.941",
"value": "Kochi city"
},
{
"operation": "Highlighting",
"time": "44.194",
"value": "Devan Ramachandran_sentence_18"
},
{
"operation": "Highlighting",
"time": "49.276",
"value": "Devan Ramachandran_sentence_20"
},
{
"operation": "finish",
"time": "65.875",
"value": "finish"
}
]
}
|
exfever_5844
|
Clinton Eastwood Jr. is a British actor who rose to international fame in a film series by an Italian film director, producer, and screenwriter who released an Italian film series consisting of three Spaghetti Western films.
|
{
"evidence": [
"Clint Eastwood",
"Sergio Leone",
"Dollars Trilogy"
],
"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:\nClinton Eastwood Jr. is a British actor who rose to international fame in a film series by an Italian film director, producer, and screenwriter who released an Italian film series consisting of three Spaghetti Western films.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Clint Eastwood",
"Sergio Leone",
"Dollars Trilogy"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5231,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1052
|
Hamlet is a 1990 film that has a category of narrative fiction intended to be more serious than humorous in tone, based on the Shakespearean tragedy of the same name, directed by Franco Zeffirelli.
|
{
"evidence": [
"Hamlet (1990 film)",
"Drama (film and television)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHamlet is a 1990 film that has a category of narrative fiction intended to be more serious than humorous in tone, based on the Shakespearean tragedy of the same name, directed by Franco Zeffirelli.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hamlet (1990 film)",
"Drama (film and television)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5232,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
59932
|
Huka Falls has a set of small waterfalls at the top with a final six meter drop at the end of the waterfall.
|
{
"evidence": [
"Huka Falls_sentence_6",
"Huka Falls_sentence_7"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHuka Falls has a set of small waterfalls at the top with a final six meter drop at the end of the waterfall.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Huka Falls_sentence_6",
"Huka Falls_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5233,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.049",
"value": "?search="
},
{
"operation": "search",
"time": "11.583",
"value": "Huka Falls"
},
{
"operation": "Now on",
"time": "12.626",
"value": "Huka Falls"
},
{
"operation": "Highlighting",
"time": "27.02",
"value": "Huka Falls_sentence_6"
},
{
"operation": "Highlighting",
"time": "33.97",
"value": "Huka Falls_sentence_7"
},
{
"operation": "finish",
"time": "60.947",
"value": "finish"
}
]
}
|
85622
|
Antipathy, observed by Sophie Bryant (an Anglo-Irish mathematician, educator, feminist and activist), is based on the positive character of others.
|
{
"evidence": [
"Sophie Bryant_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:\nAntipathy, observed by Sophie Bryant (an Anglo-Irish mathematician, educator, feminist and activist), is based on the positive character of others.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Sophie Bryant_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5234,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.231",
"value": "?search="
},
{
"operation": "search",
"time": "58.866",
"value": "sophie bryant"
},
{
"operation": "Now on",
"time": "59.956",
"value": "Sophie Bryant"
},
{
"operation": "Page search",
"time": "65.994",
"value": "obser"
},
{
"operation": "Highlighting",
"time": "70.404",
"value": "Sophie Bryant_sentence_0"
},
{
"operation": "Page search",
"time": "76.338",
"value": "charac"
},
{
"operation": "search",
"time": "150.16",
"value": "Antipathy"
},
{
"operation": "Now on",
"time": "150.841",
"value": "Antipathy"
},
{
"operation": "finish",
"time": "166.463",
"value": "finish"
}
]
}
|
54284
|
The motto Per aspera in astera was used by Étienne Jérôme Rouchouze who resided in Valparaíso, Chile
|
{
"evidence": [
"Étienne Jérôme Rouchouze_sentence_5",
"Étienne Jérôme Rouchouze_sentence_4"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe motto Per aspera in astera was used by Étienne Jérôme Rouchouze who resided in Valparaíso, Chile\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Étienne Jérôme Rouchouze_sentence_5",
"Étienne Jérôme Rouchouze_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5235,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.805",
"value": "?search="
},
{
"operation": "search",
"time": "18.316",
"value": "per aspera in astera"
},
{
"operation": "Now on",
"time": "18.974",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "22.011",
"value": "Étienne Jérôme Rouchouze"
},
{
"operation": "Now on",
"time": "22.641",
"value": "Étienne Jérôme Rouchouze"
},
{
"operation": "Page search",
"time": "27.139",
"value": "per aspera"
},
{
"operation": "Highlighting",
"time": "29.987",
"value": "Étienne Jérôme Rouchouze_sentence_5"
},
{
"operation": "Highlighting",
"time": "40.747",
"value": "Étienne Jérôme Rouchouze_sentence_4"
},
{
"operation": "finish",
"time": "53.345",
"value": "finish"
}
]
}
|
42855
|
Paint seller, Rodda Paint, was founded in 1932 in Seattle, Washington.
|
{
"evidence": [
"Rodda Paint_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:\nPaint seller, Rodda Paint, was founded in 1932 in Seattle, Washington.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Rodda Paint_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5236,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.099",
"value": "?search="
},
{
"operation": "search",
"time": "12.201",
"value": "Rodda Paint"
},
{
"operation": "Now on",
"time": "12.762",
"value": "Rodda Paint"
},
{
"operation": "Highlighting",
"time": "14.901",
"value": "Rodda Paint_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "15.89",
"value": "Rodda Paint_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "15.9",
"value": "Rodda Paint_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "24.824",
"value": "Rodda Paint_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "24.83",
"value": "Rodda Paint_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "26.575",
"value": "Rodda Paint_sentence_0"
},
{
"operation": "finish",
"time": "41.089",
"value": "finish"
}
]
}
|
72064
|
A Jain temple at Moodabidri in Karnataka, southern India, had stored for centuries its manuscript of palm-leaves, a 9th-century commentary in Prakrit and Sanskrit.
|
{
"evidence": [
"Bharatiya Jnanpith_sentence_2",
"Bharatiya Jnanpith_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:\nA Jain temple at Moodabidri in Karnataka, southern India, had stored for centuries its manuscript of palm-leaves, a 9th-century commentary in Prakrit and Sanskrit.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Bharatiya Jnanpith_sentence_2",
"Bharatiya Jnanpith_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5237,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.759",
"value": "?search="
},
{
"operation": "search",
"time": "42.207",
"value": "Bharatiya Jnanpith"
},
{
"operation": "Now on",
"time": "42.941",
"value": "Bharatiya Jnanpith"
},
{
"operation": "Highlighting",
"time": "48.948",
"value": "Bharatiya Jnanpith_sentence_2"
},
{
"operation": "Highlighting",
"time": "73.351",
"value": "Bharatiya Jnanpith_sentence_3"
},
{
"operation": "finish",
"time": "83.736",
"value": "finish"
}
]
}
|
57176
|
Slovakia's national football team won every time against Estadio Nacional de Costa Rica.
|
{
"evidence": [
"Estadio Nacional de Costa Rica (2011)_sentence_1",
"Estadio Nacional de Costa Rica (1924)_sentence_0",
"Slovakia national football team_sentence_2"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSlovakia's national football team won every time against Estadio Nacional de Costa Rica.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Estadio Nacional de Costa Rica (2011)_sentence_1",
"Estadio Nacional de Costa Rica (1924)_sentence_0",
"Slovakia national football team_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5238,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.455",
"value": "?search="
},
{
"operation": "search",
"time": "20.665",
"value": "Slovakia national football team"
},
{
"operation": "Now on",
"time": "24.528",
"value": "Slovakia national football team"
},
{
"operation": "Page search",
"time": "69.602",
"value": "estadio nacional"
},
{
"operation": "search",
"time": "110.454",
"value": "Estadio Nacional de Costa Rica"
},
{
"operation": "Now on",
"time": "111.811",
"value": "Estadio Nacional de Costa Rica"
},
{
"operation": "Highlighting",
"time": "116.182",
"value": "Estadio Nacional de Costa Rica_sentence_0"
},
{
"operation": "Highlighting",
"time": "116.184",
"value": "Estadio Nacional de Costa Rica_title"
},
{
"operation": "Highlighting",
"time": "117.138",
"value": "Estadio Nacional de Costa Rica_item_0_0"
},
{
"operation": "Highlighting deleted",
"time": "120.785",
"value": "Estadio Nacional de Costa Rica_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "121.834",
"value": "Estadio Nacional de Costa Rica_item_0_0"
},
{
"operation": "Highlighting deleted",
"time": "121.835",
"value": "Estadio Nacional de Costa Rica_title"
},
{
"operation": "hyperlink",
"time": "122.691",
"value": "Estadio Nacional de Costa Rica (1924)"
},
{
"operation": "Now on",
"time": "123.384",
"value": "Estadio Nacional de Costa Rica (1924)"
},
{
"operation": "back-button-clicked",
"time": "131.311",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "131.389",
"value": "Estadio Nacional de Costa Rica"
},
{
"operation": "hyperlink",
"time": "133.223",
"value": "Estadio Nacional de Costa Rica (2011)"
},
{
"operation": "Now on",
"time": "135.13",
"value": "Estadio Nacional de Costa Rica (2011)"
},
{
"operation": "Highlighting",
"time": "145.932",
"value": "Estadio Nacional de Costa Rica (2011)_sentence_1"
},
{
"operation": "back-button-clicked",
"time": "148.115",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "148.196",
"value": "Estadio Nacional de Costa Rica"
},
{
"operation": "hyperlink",
"time": "150.039",
"value": "Estadio Nacional de Costa Rica (1924)"
},
{
"operation": "Now on",
"time": "150.942",
"value": "Estadio Nacional de Costa Rica (1924)"
},
{
"operation": "Highlighting",
"time": "152.476",
"value": "Estadio Nacional de Costa Rica (1924)_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "165.778",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "165.86",
"value": "Estadio Nacional de Costa Rica"
},
{
"operation": "back-button-clicked",
"time": "166.296",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "166.301",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "166.309",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "166.316",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "166.805",
"value": "Slovakia national football team"
},
{
"operation": "Highlighting",
"time": "174.254",
"value": "Slovakia national football team_sentence_2"
},
{
"operation": "search",
"time": "193.771",
"value": "Estadio Nacional de Costa Rica"
},
{
"operation": "Now on",
"time": "195.056",
"value": "Estadio Nacional de Costa Rica"
},
{
"operation": "hyperlink",
"time": "197.293",
"value": "Estadio Nacional de Costa Rica (1924)"
},
{
"operation": "Now on",
"time": "197.612",
"value": "Estadio Nacional de Costa Rica (1924)"
},
{
"operation": "back-button-clicked",
"time": "210.364",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "210.456",
"value": "Estadio Nacional de Costa Rica"
},
{
"operation": "finish",
"time": "218.225",
"value": "finish"
}
]
}
|
8717
|
JD Natasha's debut album was never nominated in any grand awards.
|
{
"evidence": [
"JD Natasha_sentence_8",
"Imperfecta-Imperfect_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:\nJD Natasha's debut album was never nominated in any grand awards.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"JD Natasha_sentence_8",
"Imperfecta-Imperfect_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5239,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.567",
"value": "?search="
},
{
"operation": "search",
"time": "11.073",
"value": "JD Natasha"
},
{
"operation": "Now on",
"time": "11.979",
"value": "JD Natasha"
},
{
"operation": "Highlighting",
"time": "44.836",
"value": "JD Natasha_sentence_8"
},
{
"operation": "hyperlink",
"time": "48",
"value": "Imperfecta-Imperfect"
},
{
"operation": "Now on",
"time": "48.521",
"value": "Imperfecta-Imperfect"
},
{
"operation": "Highlighting",
"time": "62.535",
"value": "Imperfecta-Imperfect_sentence_4"
},
{
"operation": "finish",
"time": "79.238",
"value": "finish"
}
]
}
|
45500
|
Otis Turner, an American director, screenwriter and producer born on November 29, 1862 in Fairfield, Indiana, directed 133 films and wrote 40 scenarios.
|
{
"evidence": [
"Otis Turner_sentence_0",
"Otis Turner_sentence_2",
"Otis Turner_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:\nOtis Turner, an American director, screenwriter and producer born on November 29, 1862 in Fairfield, Indiana, directed 133 films and wrote 40 scenarios.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Otis Turner_sentence_0",
"Otis Turner_sentence_2",
"Otis Turner_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5240,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.226",
"value": "?search="
},
{
"operation": "search",
"time": "6.81",
"value": "Otis Turner"
},
{
"operation": "Now on",
"time": "7.828",
"value": "Otis Turner"
},
{
"operation": "Highlighting",
"time": "11.413",
"value": "Otis Turner_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "13.932",
"value": "Otis Turner_sentence_0"
},
{
"operation": "Highlighting",
"time": "14.446",
"value": "Otis Turner_sentence_0"
},
{
"operation": "Highlighting",
"time": "44.655",
"value": "Otis Turner_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "44.66",
"value": "Otis Turner_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "44.66",
"value": "Otis Turner_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "56.116",
"value": "Otis Turner_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "67.203",
"value": "Otis Turner_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "67.206",
"value": "Otis Turner_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "67.207",
"value": "Otis Turner_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "68.157",
"value": "Otis Turner_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "73.735",
"value": "Otis Turner_sentence_1"
},
{
"operation": "Highlighting",
"time": "73.944",
"value": "Otis Turner_sentence_1"
},
{
"operation": "finish",
"time": "94.088",
"value": "finish"
}
]
}
|
82191
|
Ralph Lewis was the youngest of seven son, hosted radio shows and moved to Niagara Falls.
|
{
"evidence": [
"Sons of Ralph_sentence_5",
"Sons of Ralph_sentence_6",
"Sons of Ralph_sentence_7"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRalph Lewis was the youngest of seven son, hosted radio shows and moved to Niagara Falls.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Sons of Ralph_sentence_5",
"Sons of Ralph_sentence_6",
"Sons of Ralph_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5241,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.127",
"value": "?search="
},
{
"operation": "search",
"time": "34.488",
"value": "Ralph Lewis"
},
{
"operation": "Now on",
"time": "35.557",
"value": "Ralph Lewis"
},
{
"operation": "search",
"time": "111.648",
"value": "Sons of Ralph"
},
{
"operation": "Now on",
"time": "112.327",
"value": "Sons of Ralph"
},
{
"operation": "Highlighting",
"time": "142.456",
"value": "Sons of Ralph_sentence_5"
},
{
"operation": "Highlighting",
"time": "160.582",
"value": "Sons of Ralph_sentence_6"
},
{
"operation": "Highlighting",
"time": "174.213",
"value": "Sons of Ralph_sentence_7"
},
{
"operation": "finish",
"time": "210.942",
"value": "finish"
}
]
}
|
exfever_353
|
That's So Raven is an American teen sitcom that aired on Netflix and centers on a teenager with hidden psychic abilities who shared with her best friend acted by a Dutch-born American singer who recorded several songs for an American multinational entertainment and media conglomerate headquartered at the Walt Disney Studios complex in Burbank, California.
|
{
"evidence": [
"That's So Raven",
"Anneliese van der Pol",
"The Walt Disney Company"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThat's So Raven is an American teen sitcom that aired on Netflix and centers on a teenager with hidden psychic abilities who shared with her best friend acted by a Dutch-born American singer who recorded several songs for an American multinational entertainment and media conglomerate headquartered at the Walt Disney Studios complex in Burbank, California.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"That's So Raven",
"Anneliese van der Pol",
"The Walt Disney Company"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5242,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
2898
|
There were two churches that got closed in 2013, the first one was St. Andrew in Colwyn Bay, which was founded in 1908, and St. Catherine in Old Colwyn, which was founded in 1837.
|
{
"evidence": [
"St. Andrew's Church_sentence_0",
"St. Catherine's Church_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:\nThere were two churches that got closed in 2013, the first one was St. Andrew in Colwyn Bay, which was founded in 1908, and St. Catherine in Old Colwyn, which was founded in 1837.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"St. Andrew's Church_sentence_0",
"St. Catherine's Church_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5243,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "24.517",
"value": "st. andrew colwyn bay"
},
{
"operation": "search",
"time": "37.766",
"value": "St. Andrew's Church"
},
{
"operation": "Page search",
"time": "47.838",
"value": "closed"
},
{
"operation": "Page search",
"time": "60.218",
"value": "colwyn bay"
},
{
"operation": "search",
"time": "93.641",
"value": "Colwyn Bay"
},
{
"operation": "Page search",
"time": "104.521",
"value": "st andrew"
},
{
"operation": "search",
"time": "118.827",
"value": "church closures 2013"
},
{
"operation": "hyperlink",
"time": "122.383",
"value": "church closure 2013"
},
{
"operation": "Page search",
"time": "130.15",
"value": "st. andrew"
},
{
"operation": "search",
"time": "146.38",
"value": "old colwyn"
},
{
"operation": "hyperlink",
"time": "173.19",
"value": "2 History Old Colwyn_section_1"
},
{
"operation": "Page search",
"time": "181.367",
"value": "2013"
},
{
"operation": "search",
"time": "191.315",
"value": "church closure"
},
{
"operation": "Page search",
"time": "197.362",
"value": "2013"
},
{
"operation": "hyperlink",
"time": "204.775",
"value": "Church closure Alan Campbell (pastor)_section_1"
},
{
"operation": "Highlighting",
"time": "210.668",
"value": "Alan Campbell (pastor)_sentence_18"
},
{
"operation": "Highlighting",
"time": "210.672",
"value": "Alan Campbell (pastor)_title"
},
{
"operation": "Highlighting",
"time": "210.674",
"value": "Alan Campbell (pastor)_section_1"
},
{
"operation": "back-button-clicked",
"time": "236.222",
"value": "back-button-clicked"
},
{
"operation": "hyperlink",
"time": "245.707",
"value": "St Mary's Church, Rogiet"
},
{
"operation": "Highlighting",
"time": "250.463",
"value": "St Mary's Church, Rogiet_sentence_10"
},
{
"operation": "Highlighting",
"time": "250.467",
"value": "St Mary's Church, Rogiet_title"
},
{
"operation": "Highlighting",
"time": "250.47",
"value": "St Mary's Church, Rogiet_section_0"
},
{
"operation": "back-button-clicked",
"time": "252.818",
"value": "back-button-clicked"
},
{
"operation": "hyperlink",
"time": "258.894",
"value": "St. George's (Round) Church, Halifax, Nova Scotia"
},
{
"operation": "Page search",
"time": "271.097",
"value": "closure"
},
{
"operation": "back-button-clicked",
"time": "281.019",
"value": "back-button-clicked"
},
{
"operation": "Page search",
"time": "287.947",
"value": "2013"
},
{
"operation": "hyperlink",
"time": "290.589",
"value": "Holy Trinity Church, Brighton"
},
{
"operation": "back-button-clicked",
"time": "305.705",
"value": "back-button-clicked"
},
{
"operation": "search",
"time": "327.866",
"value": "St. Andrew's Church"
},
{
"operation": "Page search",
"time": "333.485",
"value": "closure"
},
{
"operation": "hyperlink",
"time": "357.007",
"value": "25.4 Wales St. Andrew's Church_section_63"
},
{
"operation": "Highlighting",
"time": "371.196",
"value": "St. Andrew's Church_sentence_0"
},
{
"operation": "search",
"time": "402.904",
"value": "St. Catherine Church, Birmingham"
},
{
"operation": "back-button-clicked",
"time": "416.374",
"value": "back-button-clicked"
},
{
"operation": "Page search",
"time": "423.595",
"value": "1908"
},
{
"operation": "search",
"time": "444.417",
"value": "st. catherine's church"
},
{
"operation": "Highlighting",
"time": "450.114",
"value": "St. Catherine's Church_sentence_0"
},
{
"operation": "finish",
"time": "483.321",
"value": "finish"
}
]
}
|
84449
|
The longest-reigning in the list of premiers or the first minister for Alberta was Ernest Manning with seven mandates counting 25 years and 195 days in office, followed by Peter Lougheed with four mandates serving 14 years and 52 days.
|
{
"evidence": [
"List of premiers of Alberta_sentence_0",
"Ernest Manning_sentence_1",
"Ernest Manning_sentence_0"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe longest-reigning in the list of premiers or the first minister for Alberta was Ernest Manning with seven mandates counting 25 years and 195 days in office, followed by Peter Lougheed with four mandates serving 14 years and 52 days.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"List of premiers of Alberta_sentence_0",
"Ernest Manning_sentence_1",
"Ernest Manning_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5244,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.469",
"value": "?search="
},
{
"operation": "Now on",
"time": "13.521",
"value": "?search="
},
{
"operation": "search",
"time": "214.313",
"value": "List of premiers of Alberta"
},
{
"operation": "Now on",
"time": "216.607",
"value": "List of premiers of Alberta"
},
{
"operation": "Highlighting",
"time": "218.76",
"value": "List of premiers of Alberta_sentence_0"
},
{
"operation": "search",
"time": "462.689",
"value": "Contains...first minister for Alberta"
},
{
"operation": "Now on",
"time": "464.711",
"value": "undefined"
},
{
"operation": "search",
"time": "477.235",
"value": "Ernest Manning"
},
{
"operation": "Now on",
"time": "478.424",
"value": "Ernest Manning"
},
{
"operation": "Highlighting",
"time": "486.609",
"value": "Ernest Manning_sentence_1"
},
{
"operation": "Highlighting",
"time": "493.013",
"value": "Ernest Manning_sentence_0"
},
{
"operation": "finish",
"time": "548.006",
"value": "finish"
}
]
}
|
27127
|
Singapore had the most number of athletes participating in cue sports at the Asian Games in 2010.
|
{
"evidence": [
"Cue sports at the 2010 Asian Games_sentence_1"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSingapore had the most number of athletes participating in cue sports at the Asian Games in 2010.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Cue sports at the 2010 Asian Games_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5245,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.836",
"value": "?search="
},
{
"operation": "search",
"time": "35.805",
"value": "Contains...cue sports at the Asian Games in 2010"
},
{
"operation": "Now on",
"time": "38.305",
"value": "undefined"
},
{
"operation": "search",
"time": "76.756",
"value": "Contains...Asian Games in 2010"
},
{
"operation": "Now on",
"time": "79.26",
"value": "undefined"
},
{
"operation": "search",
"time": "106.859",
"value": "Cue sports at the 2010 Asian Games"
},
{
"operation": "Now on",
"time": "108.419",
"value": "Cue sports at the 2010 Asian Games"
},
{
"operation": "hyperlink",
"time": "303.722",
"value": "Singapore"
},
{
"operation": "Now on",
"time": "304.932",
"value": "Singapore at the 2010 Asian Games"
},
{
"operation": "back-button-clicked",
"time": "315.939",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "316.841",
"value": "Cue sports at the 2010 Asian Games"
},
{
"operation": "hyperlink",
"time": "323.019",
"value": "4 Participating nations Cue sports at the 2010 Asian Games_section_5"
},
{
"operation": "Highlighting",
"time": "325.156",
"value": "Cue sports at the 2010 Asian Games_sentence_1"
},
{
"operation": "finish",
"time": "333.21",
"value": "finish"
}
]
}
|
exfever_6137
|
The capital of Indonesia is the world's second-most populous urban area and the largest city in Indonesia and Southeastern Asia, the geographical southeastern subregion of Asia.
|
{
"evidence": [
"Indonesia",
"Jakarta",
"Southeast Asia"
],
"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 capital of Indonesia is the world's second-most populous urban area and the largest city in Indonesia and Southeastern Asia, the geographical southeastern subregion of Asia.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Indonesia",
"Jakarta",
"Southeast Asia"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5246,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6150
|
A seminary is an educational institution for educating students generally to prepare them for ordination to serve as clergy in an activity carried out by Christians to express or spread their faith, the prototype being the Vision Statements.
|
{
"evidence": [
"Seminary",
"Christian ministry"
],
"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:\nA seminary is an educational institution for educating students generally to prepare them for ordination to serve as clergy in an activity carried out by Christians to express or spread their faith, the prototype being the Vision Statements.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Seminary",
"Christian ministry"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5247,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7358
|
N.W.A. was the most successful rap group ever.
|
{
"evidence": [
"Dr. Dre",
"Chief executive officer"
],
"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:\nN.W.A. was the most successful rap group ever.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dr. Dre",
"Chief executive officer"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5248,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6399
|
In 1972, Jackie Robinson was inducted into the Museum that was operated by private interests in Cooperstown, New York.
|
{
"evidence": [
"Jackie Robinson",
"National Baseball Hall of Fame and Museum"
],
"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:\nIn 1972, Jackie Robinson was inducted into the Museum that was operated by private interests in Cooperstown, New York.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jackie Robinson",
"National Baseball Hall of Fame and Museum"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5249,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
32291
|
Mirosław Wodzyński failed to place in the top ten in 1973, but placed his personal best in Warsaw.
|
{
"evidence": [
"Mirosław Wodzyński_sentence_3"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMirosław Wodzyński failed to place in the top ten in 1973, but placed his personal best in Warsaw.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Mirosław Wodzyński_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5250,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.767",
"value": "?search="
},
{
"operation": "search",
"time": "15.27",
"value": "Mirosław Wodzyński"
},
{
"operation": "Now on",
"time": "15.968",
"value": "Mirosław Wodzyński"
},
{
"operation": "Highlighting",
"time": "38.601",
"value": "Mirosław Wodzyński_sentence_3"
},
{
"operation": "finish",
"time": "47.859",
"value": "finish"
}
]
}
|
exfever_8372
|
Ella Purnell is a French actress best known for her performance as Gwyn in the animated series Star Trek: Prodigy on an American pay television channel that launched on April 1, 1979, as the first cable channel for children.
|
{
"evidence": [
"Ella Purnell",
"Nickelodeon"
],
"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:\nElla Purnell is a French actress best known for her performance as Gwyn in the animated series Star Trek: Prodigy on an American pay television channel that launched on April 1, 1979, as the first cable channel for children.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ella Purnell",
"Nickelodeon"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5251,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
57723
|
Chafik Tigroudja joined Ligue 2 side Chamois Niortais on a two-year contract, then made his debut for the club in the 2–0 win at Genêts Anglet, a French football team founded in 1910 that play at the Stade Choisy in Anglet.
|
{
"evidence": [
"Chafik Tigroudja_sentence_3",
"Chafik Tigroudja_sentence_4",
"Genêts Anglet_sentence_0",
"Genêts Anglet_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:\nChafik Tigroudja joined Ligue 2 side Chamois Niortais on a two-year contract, then made his debut for the club in the 2–0 win at Genêts Anglet, a French football team founded in 1910 that play at the Stade Choisy in Anglet.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Chafik Tigroudja_sentence_3",
"Chafik Tigroudja_sentence_4",
"Genêts Anglet_sentence_0",
"Genêts Anglet_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5252,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.303",
"value": "?search="
},
{
"operation": "search",
"time": "8.55",
"value": "Chafik Tigroudja"
},
{
"operation": "Now on",
"time": "9.258",
"value": "Chafik Tigroudja"
},
{
"operation": "Page search",
"time": "18.667",
"value": "ligue"
},
{
"operation": "Highlighting",
"time": "39.493",
"value": "Chafik Tigroudja_sentence_3"
},
{
"operation": "Highlighting",
"time": "57.311",
"value": "Chafik Tigroudja_sentence_4"
},
{
"operation": "hyperlink",
"time": "114.766",
"value": "Genêts Anglet"
},
{
"operation": "Now on",
"time": "115.147",
"value": "Genêts Anglet"
},
{
"operation": "Highlighting",
"time": "123.143",
"value": "Genêts Anglet_sentence_0"
},
{
"operation": "Highlighting",
"time": "130.65",
"value": "Genêts Anglet_sentence_2"
},
{
"operation": "finish",
"time": "177.708",
"value": "finish"
}
]
}
|
78768
|
McFarland won the election in a tight race, only winning by 292 votes over the liberal candidate (the oldest active political party in Alberta and was the dominant political party until the 1921 election), which was the closest the liberal had come to winning in 70 years.
|
{
"evidence": [
"Liberal Party of Canada_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:\nMcFarland won the election in a tight race, only winning by 292 votes over the liberal candidate (the oldest active political party in Alberta and was the dominant political party until the 1921 election), which was the closest the liberal had come to winning in 70 years.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Liberal Party of Canada_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5253,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.663",
"value": "?search="
},
{
"operation": "search",
"time": "171.884",
"value": "Contains...McFarland won the election in a tight race"
},
{
"operation": "Now on",
"time": "173.58",
"value": "undefined"
},
{
"operation": "search",
"time": "242.329",
"value": "Contains...closest the liberal had come to winning in 70 years"
},
{
"operation": "Now on",
"time": "245.244",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "257.138",
"value": "2008 Canadian federal election"
},
{
"operation": "Now on",
"time": "258.071",
"value": "2008 Canadian federal election"
},
{
"operation": "Page search",
"time": "264.326",
"value": "McFarland"
},
{
"operation": "search",
"time": "314.101",
"value": "winning by 292"
},
{
"operation": "Now on",
"time": "316.103",
"value": "undefined"
},
{
"operation": "search",
"time": "371.829",
"value": "Liberal Party Canada"
},
{
"operation": "Now on",
"time": "373.088",
"value": "Liberal Party of Canada"
},
{
"operation": "Highlighting",
"time": "377.569",
"value": "Liberal Party of Canada_sentence_0"
},
{
"operation": "Page search",
"time": "404.814",
"value": "1921"
},
{
"operation": "finish",
"time": "463.027",
"value": "finish"
}
]
}
|
28267
|
Rosy was an Urdu adaptation of the Hollywood movie "Devil Wears Prada", and Moin Akhter was the one who directed it and Saira Kazmi was the one who wrote the script for it.
|
{
"evidence": [
"Moin Akhter_sentence_15"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRosy was an Urdu adaptation of the Hollywood movie \"Devil Wears Prada\", and Moin Akhter was the one who directed it and Saira Kazmi was the one who wrote the script for it.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Moin Akhter_sentence_15"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5254,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.615",
"value": "?search="
},
{
"operation": "search",
"time": "21.015",
"value": "Rosy"
},
{
"operation": "Now on",
"time": "21.667",
"value": "Rosy"
},
{
"operation": "search",
"time": "64.572",
"value": "Moin Akhter"
},
{
"operation": "Now on",
"time": "65.564",
"value": "Moin Akhter"
},
{
"operation": "Highlighting",
"time": "85.876",
"value": "Moin Akhter_sentence_15"
},
{
"operation": "finish",
"time": "99.603",
"value": "finish"
}
]
}
|
exfever_4477
|
Travis Scott signed his first major-label contract in 2012 with Epic Records and a publishing deal with an American rapper widely regarded as one of the greatest and most influential hip hop musicians of all time.
|
{
"evidence": [
"Travis Scott",
"Kanye West"
],
"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:\nTravis Scott signed his first major-label contract in 2012 with Epic Records and a publishing deal with an American rapper widely regarded as one of the greatest and most influential hip hop musicians of all time.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Travis Scott",
"Kanye West"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5255,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
23416
|
Vallée graduated with a bachelor's degree in Management Information Systems from the University of Ottawa; then received the Ottawa Forty Under 40 award from the Ottawa Business Journal in 2011.
|
{
"evidence": [
"Paul Vallée_sentence_2",
"Paul Vallée_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:\nVallée graduated with a bachelor's degree in Management Information Systems from the University of Ottawa; then received the Ottawa Forty Under 40 award from the Ottawa Business Journal in 2011.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Paul Vallée_sentence_2",
"Paul Vallée_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5256,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.7",
"value": "?search="
},
{
"operation": "search",
"time": "19.884",
"value": "Ottawa Forty Under 40"
},
{
"operation": "Now on",
"time": "21.745",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "24.085",
"value": "Paul Vallée"
},
{
"operation": "Now on",
"time": "24.901",
"value": "Paul Vallée"
},
{
"operation": "Highlighting",
"time": "38.185",
"value": "Paul Vallée_sentence_2"
},
{
"operation": "Highlighting",
"time": "43.642",
"value": "Paul Vallée_sentence_4"
},
{
"operation": "finish",
"time": "157.15",
"value": "finish"
}
]
}
|
exfever_3739
|
Edmund North was a screenwriter who shared an award for artistic and technical merit in the film industry where the statuette sketch by a British art director.
|
{
"evidence": [
"Edmund H. North",
"Academy Awards",
"Cedric Gibbons"
],
"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:\nEdmund North was a screenwriter who shared an award for artistic and technical merit in the film industry where the statuette sketch by a British art director.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Edmund H. North",
"Academy Awards",
"Cedric Gibbons"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5257,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3469
|
Kirk Douglas was an Indian actor and filmmaker, who became an international star for his role as an unscrupulous boxing hero in a 1949 American drama film noir sport film which pictures the struggles of "Midge" Kelly, an athlete in combat sport in which two people, usually wearing protective gloves and other protective equipment.
|
{
"evidence": [
"Kirk Douglas",
"Champion (1949 film)",
"Boxing"
],
"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:\nKirk Douglas was an Indian actor and filmmaker, who became an international star for his role as an unscrupulous boxing hero in a 1949 American drama film noir sport film which pictures the struggles of \"Midge\" Kelly, an athlete in combat sport in which two people, usually wearing protective gloves and other protective equipment.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Kirk Douglas",
"Champion (1949 film)",
"Boxing"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5258,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2315
|
Zhou Enlai was the second Premier of the People's Republic of China who served under Chairman Mao, a Chinese communist revolutionary who was the founder of the People's Republic of China.
|
{
"evidence": [
"Zhou Enlai",
"Mao Zedong"
],
"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:\nZhou Enlai was the second Premier of the People's Republic of China who served under Chairman Mao, a Chinese communist revolutionary who was the founder of the People's Republic of China.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Zhou Enlai",
"Mao Zedong"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5259,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
73667
|
Texas department of corrections used to have prisoners pretend to escape and hunted them with dogs for entertainment, the inmates who handled the dogs were called dawg boys and a movie was made about it.
|
{
"evidence": [
"Dog boy_sentence_0",
"Dog boy_sentence_3",
"Dog boy_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:\nTexas department of corrections used to have prisoners pretend to escape and hunted them with dogs for entertainment, the inmates who handled the dogs were called dawg boys and a movie was made about it.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Dog boy_sentence_0",
"Dog boy_sentence_3",
"Dog boy_sentence_11"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5260,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.093",
"value": "?search="
},
{
"operation": "search",
"time": "33.43",
"value": "Contains...dawg boys"
},
{
"operation": "Now on",
"time": "35.115",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "42.556",
"value": "Dog boy"
},
{
"operation": "Now on",
"time": "43.249",
"value": "Dog boy"
},
{
"operation": "Highlighting",
"time": "46.416",
"value": "Dog boy_sentence_0"
},
{
"operation": "Highlighting",
"time": "54.396",
"value": "Dog boy_sentence_3"
},
{
"operation": "Highlighting",
"time": "94.968",
"value": "Dog boy_sentence_11"
},
{
"operation": "finish",
"time": "144.368",
"value": "finish"
}
]
}
|
exfever_1155
|
The film Oz the Great and Powerful was based on early 19th century Oz novels and set 20 years before the events of the children's novel written by author L. Frank Baum.
|
{
"evidence": [
"Oz the Great and Powerful",
"The Wonderful Wizard of Oz"
],
"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 film Oz the Great and Powerful was based on early 19th century Oz novels and set 20 years before the events of the children's novel written by author L. Frank Baum.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Oz the Great and Powerful",
"The Wonderful Wizard of Oz"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5261,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8169
|
My Summer of Love is a 2004 British drama film filmed in a metropolitan and ceremonial county in Switzerland.
|
{
"evidence": [
"My Summer of Love",
"West Yorkshire"
],
"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:\nMy Summer of Love is a 2004 British drama film filmed in a metropolitan and ceremonial county in Switzerland.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"My Summer of Love",
"West Yorkshire"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5262,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7324
|
The Muppet Christmas Carol is a 1992 Christmas film directed by the chairman of Muppets Inc., an Asian entertainment company in Los Angeles, California.
|
{
"evidence": [
"The Muppet Christmas Carol",
"Brian Henson",
"The Jim Henson Company"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Muppet Christmas Carol is a 1992 Christmas film directed by the chairman of Muppets Inc., an Asian entertainment company in Los Angeles, California.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Muppet Christmas Carol",
"Brian Henson",
"The Jim Henson Company"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5263,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
33001
|
More Democratic than Republican candidates won in the 1968 United States gubernatorial special elections.
|
{
"evidence": [
"By-election_sentence_0",
"1968 United States gubernatorial elections_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:\nMore Democratic than Republican candidates won in the 1968 United States gubernatorial special elections.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"By-election_sentence_0",
"1968 United States gubernatorial elections_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5264,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.632",
"value": "?search="
},
{
"operation": "search",
"time": "18.865",
"value": "Contains...1968 United States gubernatorial special elections"
},
{
"operation": "Now on",
"time": "21.568",
"value": "undefined"
},
{
"operation": "search",
"time": "46.017",
"value": "1968 gubernatorial"
},
{
"operation": "Now on",
"time": "47.23",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "50.899",
"value": "1968 United States gubernatorial elections"
},
{
"operation": "Now on",
"time": "51.604",
"value": "1968 United States gubernatorial elections"
},
{
"operation": "search",
"time": "79.935",
"value": "gubernatorial special elections"
},
{
"operation": "Now on",
"time": "80.695",
"value": "undefined"
},
{
"operation": "search",
"time": "106.448",
"value": "1968 gubernatorial special"
},
{
"operation": "Now on",
"time": "107.07",
"value": "undefined"
},
{
"operation": "search",
"time": "134.042",
"value": "Contains...1968 United States gubernatorial special elections"
},
{
"operation": "Now on",
"time": "134.878",
"value": "undefined"
},
{
"operation": "search",
"time": "178.392",
"value": "1968 United States gubernatorial elections"
},
{
"operation": "Now on",
"time": "179.019",
"value": "1968 United States gubernatorial elections"
},
{
"operation": "hyperlink",
"time": "212.087",
"value": "1968 United States elections"
},
{
"operation": "Now on",
"time": "212.841",
"value": "1968 United States elections"
},
{
"operation": "search",
"time": "237.886",
"value": "Special elections"
},
{
"operation": "Now on",
"time": "238.887",
"value": "By-election"
},
{
"operation": "Highlighting",
"time": "283.498",
"value": "By-election_sentence_0"
},
{
"operation": "search",
"time": "295.655",
"value": "1968 gubernatorial"
},
{
"operation": "Now on",
"time": "296.546",
"value": "undefined"
},
{
"operation": "search",
"time": "326.544",
"value": "1968 gubernatorial elections"
},
{
"operation": "Now on",
"time": "327.041",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "332.205",
"value": "1968 United States gubernatorial elections"
},
{
"operation": "Now on",
"time": "332.673",
"value": "1968 United States gubernatorial elections"
},
{
"operation": "Highlighting",
"time": "339.759",
"value": "1968 United States gubernatorial elections_sentence_0"
},
{
"operation": "finish",
"time": "355.509",
"value": "finish"
}
]
}
|
6838
|
Blatna is a name shared by a village in Republika Srpska, Bosnia and Herzegovina and a town in the Czech Republic.
|
{
"evidence": [
"Blatna_sentence_1",
"Blatná_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:\nBlatna is a name shared by a village in Republika Srpska, Bosnia and Herzegovina and a town in the Czech Republic.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Blatna_sentence_1",
"Blatná_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5265,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.641",
"value": "?search="
},
{
"operation": "search",
"time": "95.027",
"value": "blatna"
},
{
"operation": "Now on",
"time": "95.816",
"value": "Blatna"
},
{
"operation": "Highlighting",
"time": "101.726",
"value": "Blatna_sentence_1"
},
{
"operation": "hyperlink",
"time": "125.294",
"value": "Czech Republic"
},
{
"operation": "Now on",
"time": "127.61",
"value": "Czech Republic"
},
{
"operation": "Highlighting",
"time": "127.802",
"value": "Czech Republic_sentence_1"
},
{
"operation": "Highlighting",
"time": "127.828",
"value": "Czech Republic_title"
},
{
"operation": "Highlighting deleted",
"time": "128.868",
"value": "Czech Republic_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "128.888",
"value": "Czech Republic_title"
},
{
"operation": "Highlighting",
"time": "130.55",
"value": "Czech Republic_sentence_2"
},
{
"operation": "Highlighting",
"time": "130.57",
"value": "Czech Republic_title"
},
{
"operation": "Highlighting deleted",
"time": "137.959",
"value": "Czech Republic_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "137.985",
"value": "Czech Republic_title"
},
{
"operation": "back-button-clicked",
"time": "139.605",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "139.695",
"value": "Blatna"
},
{
"operation": "hyperlink",
"time": "141.6",
"value": "Blatná"
},
{
"operation": "Now on",
"time": "142.524",
"value": "Blatná"
},
{
"operation": "Highlighting",
"time": "143.442",
"value": "Blatná_sentence_0"
},
{
"operation": "finish",
"time": "152.715",
"value": "finish"
}
]
}
|
34685
|
Frederick Marrable (the first Chief Architect for the Metropolitan Board of Works) was beginning to feel that he was grossly underpaid and overworked by the end of 1860, and demanded an increase in his £800 annual salary.
|
{
"evidence": [
"Frederick Marrable_sentence_0",
"Frederick Marrable_sentence_11",
"Frederick Marrable_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:\nFrederick Marrable (the first Chief Architect for the Metropolitan Board of Works) was beginning to feel that he was grossly underpaid and overworked by the end of 1860, and demanded an increase in his £800 annual salary.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Frederick Marrable_sentence_0",
"Frederick Marrable_sentence_11",
"Frederick Marrable_sentence_12"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5266,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.281",
"value": "?search="
},
{
"operation": "search",
"time": "11.15",
"value": "Frederick Marrable"
},
{
"operation": "Now on",
"time": "12.184",
"value": "Frederick Marrable"
},
{
"operation": "Highlighting",
"time": "17.043",
"value": "Frederick Marrable_sentence_0"
},
{
"operation": "Highlighting",
"time": "38.38",
"value": "Frederick Marrable_sentence_11"
},
{
"operation": "Highlighting",
"time": "43.052",
"value": "Frederick Marrable_sentence_12"
},
{
"operation": "finish",
"time": "83.021",
"value": "finish"
}
]
}
|
73505
|
The Australia at the 1984 Summer Paralympics results show that Australia won 55 gold medals in Athletics (a group of sporting events that involves competitive running, jumping, throwing, and walking), ten in Lawn Bowls and five in shooting.
|
{
"evidence": [
"Australia at the 1984 Summer Paralympics_sentence_49"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Australia at the 1984 Summer Paralympics results show that Australia won 55 gold medals in Athletics (a group of sporting events that involves competitive running, jumping, throwing, and walking), ten in Lawn Bowls and five in shooting.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Australia at the 1984 Summer Paralympics_sentence_49"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5267,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.498",
"value": "?search="
},
{
"operation": "search",
"time": "17.446",
"value": "Australia at the 1984 Summer Paralympics"
},
{
"operation": "Now on",
"time": "19.576",
"value": "Australia at the 1984 Summer Paralympics"
},
{
"operation": "hyperlink",
"time": "29.26",
"value": "7 Results Australia at the 1984 Summer Paralympics_section_10"
},
{
"operation": "Highlighting",
"time": "37.51",
"value": "Australia at the 1984 Summer Paralympics_cell_1_2_0"
},
{
"operation": "Highlighting",
"time": "37.516",
"value": "Australia at the 1984 Summer Paralympics_section_11"
},
{
"operation": "Highlighting",
"time": "37.52",
"value": "Australia at the 1984 Summer Paralympics_header_cell_1_0_0"
},
{
"operation": "Highlighting deleted",
"time": "69.043",
"value": "Australia at the 1984 Summer Paralympics_cell_1_2_0"
},
{
"operation": "Highlighting deleted",
"time": "69.049",
"value": "Australia at the 1984 Summer Paralympics_section_11"
},
{
"operation": "Highlighting deleted",
"time": "69.055",
"value": "Australia at the 1984 Summer Paralympics_header_cell_1_0_0"
},
{
"operation": "Highlighting",
"time": "76.8",
"value": "Australia at the 1984 Summer Paralympics_sentence_49"
},
{
"operation": "finish",
"time": "116.714",
"value": "finish"
}
]
}
|
exfever_2685
|
J. Cole is an American rapper, singer, songwriter, and record producer who released an album release by Roc Nation.
|
{
"evidence": [
"J. Cole",
"Born Sinner"
],
"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:\nJ. Cole is an American rapper, singer, songwriter, and record producer who released an album release by Roc Nation.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"J. Cole",
"Born Sinner"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5268,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5960
|
Drop Dead Gorgeous is a 2000 American satirical black comedy mockumentary film starring talented artists, including an American actress and singer born in Atlanta.
|
{
"evidence": [
"Drop Dead Gorgeous (film)",
"Brittany Murphy"
],
"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:\nDrop Dead Gorgeous is a 2000 American satirical black comedy mockumentary film starring talented artists, including an American actress and singer born in Atlanta.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Drop Dead Gorgeous (film)",
"Brittany Murphy"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5269,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5258
|
Mark Bryan is an American singer who gained fame as the lead vocalist of Hootie & the Blowfish, who released a solo album in 2002 on an American independent record label that was formed in 1998 by Steven McKeever.
|
{
"evidence": [
"Darius Rucker",
"Hidden Beach Recordings"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMark Bryan is an American singer who gained fame as the lead vocalist of Hootie & the Blowfish, who released a solo album in 2002 on an American independent record label that was formed in 1998 by Steven McKeever.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Darius Rucker",
"Hidden Beach Recordings"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5270,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
73719
|
Brewer, born neither in 1810, nor in Norwich to a Baptist schoolmaster, was matriculated at Queen's College, Oxford in 1827, before being ordained in the Church of England in 1837.
|
{
"evidence": [
"John Sherren Brewer_sentence_0",
"John Sherren Brewer_sentence_3"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nBrewer, born neither in 1810, nor in Norwich to a Baptist schoolmaster, was matriculated at Queen's College, Oxford in 1827, before being ordained in the Church of England in 1837.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"John Sherren Brewer_sentence_0",
"John Sherren Brewer_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5271,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "7.273",
"value": "?search="
},
{
"operation": "search",
"time": "41.882",
"value": "\"matriculated at Queen's College, Oxford\""
},
{
"operation": "Now on",
"time": "43.26",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "54",
"value": "John Sherren Brewer"
},
{
"operation": "Now on",
"time": "55.663",
"value": "John Sherren Brewer"
},
{
"operation": "Highlighting",
"time": "56.857",
"value": "John Sherren Brewer_sentence_0"
},
{
"operation": "Highlighting",
"time": "60.932",
"value": "John Sherren Brewer_sentence_3"
},
{
"operation": "finish",
"time": "75.298",
"value": "finish"
}
]
}
|
exfever_4340
|
The bonobo is an invulnerable great ape that is one of the two species making up the taxonomic rank Pan that is used in the biological classification of living and fossil organisms as well as viruses.
|
{
"evidence": [
"Bonobo",
"Genus"
],
"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 bonobo is an invulnerable great ape that is one of the two species making up the taxonomic rank Pan that is used in the biological classification of living and fossil organisms as well as viruses.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Bonobo",
"Genus"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5272,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1449
|
Southampton F.C. was the first ever club that played in the Premier League.
|
{
"evidence": [
"Southampton F.C.",
"Association football"
],
"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:\nSouthampton F.C. was the first ever club that played in the Premier League.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Southampton F.C.",
"Association football"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 5273,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7384
|
The Caracazo is the name given to the wave of protests, riots, and looting that started on February 27, 1979, in a city in Miranda, Venezuela that was established in 1621 as Nuestra Senora de Copacabana de Los Guarenas.
|
{
"evidence": [
"Caracazo",
"Guarenas"
],
"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 Caracazo is the name given to the wave of protests, riots, and looting that started on February 27, 1979, in a city in Miranda, Venezuela that was established in 1621 as Nuestra Senora de Copacabana de Los Guarenas.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Caracazo",
"Guarenas"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5274,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8085
|
The Golden Girls is an American sitcom aired on CBS, starring Bea Arthur, Betty White, Estelle Getty, and an American actress and comedian best known for her role as Vivian Harmon on Maude.
|
{
"evidence": [
"The Golden Girls",
"Rue McClanahan"
],
"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 Golden Girls is an American sitcom aired on CBS, starring Bea Arthur, Betty White, Estelle Getty, and an American actress and comedian best known for her role as Vivian Harmon on Maude.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Golden Girls",
"Rue McClanahan"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5275,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
32505
|
Teldenia sparsata is a moth in the family Drepanidae found in New Guinea.
|
{
"evidence": [
"Teldenia sparsata_sentence_0",
"Teldenia sparsata_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:\nTeldenia sparsata is a moth in the family Drepanidae found in New Guinea.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Teldenia sparsata_sentence_0",
"Teldenia sparsata_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5276,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.935",
"value": "?search="
},
{
"operation": "search",
"time": "10.726",
"value": "Teldenia sparsata"
},
{
"operation": "Now on",
"time": "11.527",
"value": "Teldenia sparsata"
},
{
"operation": "Highlighting",
"time": "14.001",
"value": "Teldenia sparsata_sentence_0"
},
{
"operation": "Highlighting",
"time": "18.936",
"value": "Teldenia sparsata_sentence_2"
},
{
"operation": "finish",
"time": "31.759",
"value": "finish"
}
]
}
|
23403
|
Released March 13, 2018, Flower Road (a Japanese term, new glorious journey) was created by Big Bang, and makes a reference to their 2008 single.
|
{
"evidence": [
"Flower Road_sentence_14"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nReleased March 13, 2018, Flower Road (a Japanese term, new glorious journey) was created by Big Bang, and makes a reference to their 2008 single.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Flower Road_sentence_14"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5277,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.641",
"value": "?search="
},
{
"operation": "search",
"time": "11.711",
"value": "Flower Road"
},
{
"operation": "Now on",
"time": "12.4",
"value": "Flower Road"
},
{
"operation": "Highlighting",
"time": "34.717",
"value": "Flower Road_sentence_0"
},
{
"operation": "hyperlink",
"time": "53.036",
"value": "Burning Sun scandal"
},
{
"operation": "Now on",
"time": "54.42",
"value": "Burning Sun scandal"
},
{
"operation": "back-button-clicked",
"time": "141.67",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "141.774",
"value": "Flower Road"
},
{
"operation": "Page search",
"time": "160.129",
"value": "2008"
},
{
"operation": "Highlighting",
"time": "167.331",
"value": "Flower Road_sentence_15"
},
{
"operation": "Highlighting",
"time": "175.375",
"value": "Flower Road_sentence_14"
},
{
"operation": "Highlighting deleted",
"time": "177.446",
"value": "Flower Road_sentence_15"
},
{
"operation": "Highlighting deleted",
"time": "179.716",
"value": "Flower Road_sentence_0"
},
{
"operation": "finish",
"time": "186.765",
"value": "finish"
}
]
}
|
46156
|
Dion Lewis played in three teams, one of which was named TEN.
|
{
"evidence": [
"Dion Lewis_sentence_0",
"Dion Lewis_sentence_1",
"Dion Lewis_sentence_2",
"Dion Lewis_sentence_3"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDion Lewis played in three teams, one of which was named TEN.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Dion Lewis_sentence_0",
"Dion Lewis_sentence_1",
"Dion Lewis_sentence_2",
"Dion Lewis_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5278,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.234",
"value": "undefined"
},
{
"operation": "search",
"time": "13.743",
"value": "Dion Lewis"
},
{
"operation": "Now on",
"time": "15.306",
"value": "Dion Lewis"
},
{
"operation": "Highlighting",
"time": "25.941",
"value": "Dion Lewis_sentence_0"
},
{
"operation": "Highlighting",
"time": "29.896",
"value": "Dion Lewis_sentence_1"
},
{
"operation": "Highlighting",
"time": "38.712",
"value": "Dion Lewis_sentence_2"
},
{
"operation": "Highlighting",
"time": "41.791",
"value": "Dion Lewis_sentence_3"
},
{
"operation": "finish",
"time": "78.259",
"value": "finish"
}
]
}
|
exfever_520
|
The Hit is a British road crime film starring an English actor who collaborated with Quentin Tarantino on several films, such as an American anthology black comedy film co-written and co-directed by Theodore Roosevelt.
|
{
"evidence": [
"The Hit (1984 film)",
"Tim Roth",
"Four Rooms"
],
"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 Hit is a British road crime film starring an English actor who collaborated with Quentin Tarantino on several films, such as an American anthology black comedy film co-written and co-directed by Theodore Roosevelt.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Hit (1984 film)",
"Tim Roth",
"Four Rooms"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5279,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
87883
|
Safe Road Trains for the Environment were testing at the proving grounds of Volvo, a manufacturing company.
|
{
"evidence": [
"Safe Road Trains for the Environment_sentence_15",
"Volvo_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:\nSafe Road Trains for the Environment were testing at the proving grounds of Volvo, a manufacturing company.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Safe Road Trains for the Environment_sentence_15",
"Volvo_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5280,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.273",
"value": "?search="
},
{
"operation": "search",
"time": "12.787",
"value": "Safe Road Trains for the Environment"
},
{
"operation": "Now on",
"time": "13.773",
"value": "Safe Road Trains for the Environment"
},
{
"operation": "Highlighting",
"time": "133.43",
"value": "Safe Road Trains for the Environment_sentence_15"
},
{
"operation": "hyperlink",
"time": "136.061",
"value": "Volvo"
},
{
"operation": "Now on",
"time": "136.877",
"value": "Volvo"
},
{
"operation": "Highlighting",
"time": "138.984",
"value": "Volvo_sentence_3"
},
{
"operation": "finish",
"time": "151.034",
"value": "finish"
}
]
}
|
exfever_4364
|
Wi-Fi is a family of wireless network protocols commonly used for local area networking of devices and a global system of interconnected mobile networks that uses the Internet protocol suite to communicate between networks and devices.
|
{
"evidence": [
"Wi-Fi",
"Internet"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWi-Fi is a family of wireless network protocols commonly used for local area networking of devices and a global system of interconnected mobile networks that uses the Internet protocol suite to communicate between networks and devices.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Wi-Fi",
"Internet"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5281,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5038
|
Russell Crowe is an actor from Australia who came to international attention for his role as Roman General Maximus Decimus Meridius in a 1996 epic historical drama film directed by Ridley Scott.
|
{
"evidence": [
"Russell Crowe",
"Gladiator (2000 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:\nRussell Crowe is an actor from Australia who came to international attention for his role as Roman General Maximus Decimus Meridius in a 1996 epic historical drama film directed by Ridley Scott.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Russell Crowe",
"Gladiator (2000 film)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 5282,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
84977
|
Chaotic systems generate large numbers of patterns of behavior and exhibit sensitivity to initial conditions which, in practice, means that it can switch between patterns extremely fast.
|
{
"evidence": [
"Chaos computing_sentence_2",
"Chaos computing_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:\nChaotic systems generate large numbers of patterns of behavior and exhibit sensitivity to initial conditions which, in practice, means that it can switch between patterns extremely fast.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Chaos computing_sentence_2",
"Chaos computing_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 5283,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.145",
"value": "?search="
},
{
"operation": "search",
"time": "14.895",
"value": "Chaotic systems generate large numbers of patterns of behavior"
},
{
"operation": "Now on",
"time": "17.372",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "19.529",
"value": "Chaos computing"
},
{
"operation": "Now on",
"time": "19.951",
"value": "Chaos computing"
},
{
"operation": "Page search",
"time": "26.708",
"value": "Chaotic systems generate large numbers of patterns of behavior"
},
{
"operation": "Highlighting",
"time": "29.477",
"value": "Chaos computing_sentence_2"
},
{
"operation": "Highlighting",
"time": "36.072",
"value": "Chaos computing_sentence_3"
},
{
"operation": "finish",
"time": "49.536",
"value": "finish"
}
]
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.