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 |
|---|---|---|---|---|---|---|---|---|
21022
|
Matsudaira Yasuhide was born to the high-ranking hatamoto Matsudaira Yasuzumi, a samurai in the direct service of the Tokugawa shogunate, and later took the name of Daimyō.
|
{
"evidence": [
"Matsudaira Yasuhide_sentence_3",
"Hatamoto_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:\nMatsudaira Yasuhide was born to the high-ranking hatamoto Matsudaira Yasuzumi, a samurai in the direct service of the Tokugawa shogunate, and later took the name of Daimyō.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Matsudaira Yasuhide_sentence_3",
"Hatamoto_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4884,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.597",
"value": "?search="
},
{
"operation": "search",
"time": "6.051",
"value": "Matsudaira Yasuhide"
},
{
"operation": "Now on",
"time": "6.722",
"value": "Matsudaira Yasuhide"
},
{
"operation": "Page search",
"time": "15.258",
"value": "Matsudaira Yasuzumi"
},
{
"operation": "Highlighting",
"time": "19.446",
"value": "Matsudaira Yasuhide_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "36.88",
"value": "Matsudaira Yasuhide_sentence_3"
},
{
"operation": "Highlighting",
"time": "37.348",
"value": "Matsudaira Yasuhide_sentence_3"
},
{
"operation": "Page search",
"time": "41.177",
"value": "samurai"
},
{
"operation": "Highlighting deleted",
"time": "43.275",
"value": "Matsudaira Yasuhide_sentence_3"
},
{
"operation": "Highlighting",
"time": "44.028",
"value": "Matsudaira Yasuhide_sentence_3"
},
{
"operation": "hyperlink",
"time": "55.354",
"value": "hatamoto"
},
{
"operation": "Now on",
"time": "56.132",
"value": "Hatamoto"
},
{
"operation": "Page search",
"time": "59.597",
"value": "Matsudaira Yasuzumi"
},
{
"operation": "Page search",
"time": "64.876",
"value": "samurai"
},
{
"operation": "Highlighting",
"time": "67.783",
"value": "Hatamoto_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "73.285",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "73.373",
"value": "Matsudaira Yasuhide"
},
{
"operation": "Page search",
"time": "92.11",
"value": "Daimyō"
},
{
"operation": "hyperlink",
"time": "114.036",
"value": "hatamoto"
},
{
"operation": "Now on",
"time": "114.446",
"value": "Hatamoto"
},
{
"operation": "Page search",
"time": "118.214",
"value": "Daimyō"
},
{
"operation": "back-button-clicked",
"time": "130.679",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "130.759",
"value": "Matsudaira Yasuhide"
},
{
"operation": "Page search",
"time": "135.886",
"value": "Daimyō"
},
{
"operation": "Page search",
"time": "144.337",
"value": "name"
},
{
"operation": "finish",
"time": "191",
"value": "finish"
}
]
}
|
exfever_5184
|
Chris Hemsworth is an Australian actor who appears in the action film, a 2009 American science fiction action film directed by J. J. Abrams.
|
{
"evidence": [
"Chris Hemsworth",
"Star Trek (film)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nChris Hemsworth is an Australian actor who appears in the action film, a 2009 American science fiction action film directed by J. J. Abrams.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Chris Hemsworth",
"Star Trek (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4885,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2508
|
Deadpool is a 2016 superhero film with a screenplay by Rhett Reese and a Canadian television/movie screenwriter and producer.
|
{
"evidence": [
"Deadpool (film)",
"Paul Wernick"
],
"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:\nDeadpool is a 2016 superhero film with a screenplay by Rhett Reese and a Canadian television/movie screenwriter and producer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Deadpool (film)",
"Paul Wernick"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4886,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
32355
|
"Trains and Winter Rains" is a single by Irish musician Sinead O'Connor that ranked 15th and 39th in the Belgian chart and 15th in the Italian chart in 2008.
|
{
"evidence": [
"Trains and Winter Rains_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:\n\"Trains and Winter Rains\" is a single by Irish musician Sinead O'Connor that ranked 15th and 39th in the Belgian chart and 15th in the Italian chart in 2008.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Trains and Winter Rains_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4887,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.328",
"value": "?search="
},
{
"operation": "Now on",
"time": "2.123",
"value": "?search="
},
{
"operation": "search",
"time": "7.727",
"value": "Trains and Winter Rains"
},
{
"operation": "Now on",
"time": "8.264",
"value": "Trains and Winter Rains"
},
{
"operation": "Highlighting",
"time": "10.726",
"value": "Trains and Winter Rains_sentence_0"
},
{
"operation": "finish",
"time": "26.7",
"value": "finish"
}
]
}
|
22889
|
Westbeach Recorders was established in 1985 by Bad Religion guitarist Brett Gurewitz in Culver City, California, then Donnell Cameron (a record producer who worked with Sublime and Blink-182) became a partner in 1988.
|
{
"evidence": [
"Westbeach Recorders_sentence_1",
"Westbeach Recorders_sentence_4",
"Donnell Cameron_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:\nWestbeach Recorders was established in 1985 by Bad Religion guitarist Brett Gurewitz in Culver City, California, then Donnell Cameron (a record producer who worked with Sublime and Blink-182) became a partner in 1988.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Westbeach Recorders_sentence_1",
"Westbeach Recorders_sentence_4",
"Donnell Cameron_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4888,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.844",
"value": "?search="
},
{
"operation": "search",
"time": "8.34",
"value": "Westbeach Recorders"
},
{
"operation": "Now on",
"time": "9.482",
"value": "Westbeach Recorders"
},
{
"operation": "Highlighting",
"time": "11.579",
"value": "Westbeach Recorders_cell_0_3_1"
},
{
"operation": "Highlighting",
"time": "11.582",
"value": "Westbeach Recorders_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "15.583",
"value": "Westbeach Recorders_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "17.375",
"value": "Westbeach Recorders_cell_0_3_1"
},
{
"operation": "Highlighting deleted",
"time": "17.379",
"value": "Westbeach Recorders_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "24.148",
"value": "Westbeach Recorders_sentence_4"
},
{
"operation": "hyperlink",
"time": "32.392",
"value": "Donnell Cameron"
},
{
"operation": "Now on",
"time": "33.958",
"value": "Donnell Cameron"
},
{
"operation": "Highlighting",
"time": "35.172",
"value": "Donnell Cameron_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "36.214",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "36.305",
"value": "Westbeach Recorders"
},
{
"operation": "finish",
"time": "40.604",
"value": "finish"
}
]
}
|
exfever_8051
|
Split is a 2016 American film which has a combining thriller and psychological fiction genres.
|
{
"evidence": [
"Split (2016 American film)",
"Psychological thriller"
],
"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:\nSplit is a 2016 American film which has a combining thriller and psychological fiction genres.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Split (2016 American film)",
"Psychological thriller"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4889,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
40362
|
Nezamysl was one of seven mythical princes whose names were recorded in Cosmas chronicle and then transmitted into historical works, but they are not connected to the Ducal Rotunda of the Virgin Mary dating back to the 11th or 12th century.
|
{
"evidence": [
"Nezamysl_sentence_0",
"Nezamysl_sentence_1",
"Nezamysl_sentence_2",
"Nezamysl_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:\nNezamysl was one of seven mythical princes whose names were recorded in Cosmas chronicle and then transmitted into historical works, but they are not connected to the Ducal Rotunda of the Virgin Mary dating back to the 11th or 12th century.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Nezamysl_sentence_0",
"Nezamysl_sentence_1",
"Nezamysl_sentence_2",
"Nezamysl_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4890,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.676",
"value": "?search="
},
{
"operation": "search",
"time": "25.585",
"value": "Nezamysi"
},
{
"operation": "Now on",
"time": "26.4",
"value": "undefined"
},
{
"operation": "search",
"time": "45.632",
"value": "Cosmas chronicle"
},
{
"operation": "Now on",
"time": "46.836",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "50.173",
"value": "Přemysl the Ploughman"
},
{
"operation": "Now on",
"time": "50.687",
"value": "Přemysl the Ploughman"
},
{
"operation": "back-button-clicked",
"time": "54.54",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "55.127",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "57.069",
"value": "Nezamysl"
},
{
"operation": "Now on",
"time": "57.581",
"value": "Nezamysl"
},
{
"operation": "Highlighting",
"time": "60.861",
"value": "Nezamysl_sentence_0"
},
{
"operation": "Highlighting",
"time": "66.857",
"value": "Nezamysl_sentence_1"
},
{
"operation": "Highlighting",
"time": "81.426",
"value": "Nezamysl_sentence_2"
},
{
"operation": "Highlighting",
"time": "98.954",
"value": "Nezamysl_sentence_3"
},
{
"operation": "finish",
"time": "132.839",
"value": "finish"
}
]
}
|
64954
|
The Dwarf killer whale dwarfism is caused by achondroplasia which is a genetic disorder whose primary feature is dwarfism.
|
{
"evidence": [
"Dwarf killer whale_sentence_10",
"Achondroplasia_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Dwarf killer whale dwarfism is caused by achondroplasia which is a genetic disorder whose primary feature is dwarfism.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Dwarf killer whale_sentence_10",
"Achondroplasia_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4891,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.321",
"value": "?search="
},
{
"operation": "search",
"time": "21.095",
"value": "Dwarf killer whale"
},
{
"operation": "Now on",
"time": "22.275",
"value": "Dwarf killer whale"
},
{
"operation": "Highlighting",
"time": "147.281",
"value": "Dwarf killer whale_sentence_10"
},
{
"operation": "hyperlink",
"time": "201.164",
"value": "achondroplasia"
},
{
"operation": "Now on",
"time": "202.22",
"value": "Achondroplasia"
},
{
"operation": "Highlighting",
"time": "205.631",
"value": "Achondroplasia_sentence_0"
},
{
"operation": "finish",
"time": "234.796",
"value": "finish"
}
]
}
|
exfever_4095
|
Al Capone attained notoriety during the nationwide constitutional ban on the production, importation, transportation, and sale of alcoholic beverages.
|
{
"evidence": [
"Al Capone",
"Prohibition in the United States"
],
"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:\nAl Capone attained notoriety during the nationwide constitutional ban on the production, importation, transportation, and sale of alcoholic beverages.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Al Capone",
"Prohibition in the United States"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4892,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
88229
|
Sani Monastery is a monastery of the Kagyu sect of Tibetan Buddhism and oldest religious site in the whole region of Ladakh and Zanskar, dedicated to the deity Avalokiteshvara that is located in northern India
|
{
"evidence": [
"Sani Monastery_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:\nSani Monastery is a monastery of the Kagyu sect of Tibetan Buddhism and oldest religious site in the whole region of Ladakh and Zanskar, dedicated to the deity Avalokiteshvara that is located in northern India\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sani Monastery_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4893,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.094",
"value": "?search="
},
{
"operation": "search",
"time": "19.112",
"value": "Sani Monastery"
},
{
"operation": "Now on",
"time": "20.036",
"value": "Sani Monastery"
},
{
"operation": "Highlighting",
"time": "68.48",
"value": "Sani Monastery_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "69.566",
"value": "Sani Monastery_cell_0_3_1"
},
{
"operation": "Highlighting",
"time": "69.586",
"value": "Sani Monastery_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "69.587",
"value": "Sani Monastery_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "69.588",
"value": "Sani Monastery_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "81.823",
"value": "Sani Monastery_sentence_2"
},
{
"operation": "Highlighting",
"time": "91.37",
"value": "Sani Monastery_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "101.39",
"value": "Sani Monastery_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "104.107",
"value": "Sani Monastery_cell_0_3_1"
},
{
"operation": "Highlighting deleted",
"time": "104.112",
"value": "Sani Monastery_header_cell_0_3_0"
},
{
"operation": "Highlighting deleted",
"time": "104.115",
"value": "Sani Monastery_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "104.116",
"value": "Sani Monastery_header_cell_0_0_0"
},
{
"operation": "finish",
"time": "134.545",
"value": "finish"
}
]
}
|
7358
|
Construction for the tower started in 2007 with some design changes from Sheffield-based architects AXIS Architecture, but the original plans for the site was for a single tower named Eclipse which was later abandoned by Cala Developments due to unfavourable economic conditions.
|
{
"evidence": [
"Velocity Tower_sentence_8",
"Velocity Tower_sentence_5",
"Velocity Tower_sentence_2",
"Velocity Tower_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:\nConstruction for the tower started in 2007 with some design changes from Sheffield-based architects AXIS Architecture, but the original plans for the site was for a single tower named Eclipse which was later abandoned by Cala Developments due to unfavourable economic conditions.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Velocity Tower_sentence_8",
"Velocity Tower_sentence_5",
"Velocity Tower_sentence_2",
"Velocity Tower_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4894,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.265",
"value": "?search="
},
{
"operation": "search",
"time": "24.238",
"value": "Contains...tower named Eclipse"
},
{
"operation": "Now on",
"time": "25.912",
"value": "undefined"
},
{
"operation": "search",
"time": "53.565",
"value": "Contains...AXIS Architecture"
},
{
"operation": "Now on",
"time": "55.139",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "64.674",
"value": "Kelham Island Quarter"
},
{
"operation": "Now on",
"time": "65.583",
"value": "Kelham Island Quarter"
},
{
"operation": "search",
"time": "93.875",
"value": "Contains...single tower named Eclipse"
},
{
"operation": "Now on",
"time": "94.983",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "133.739",
"value": "Cala Eclipse Velocity Tower_section_0"
},
{
"operation": "Now on",
"time": "134.477",
"value": "Velocity Tower#Cala Eclipse Velocity Tower section 0"
},
{
"operation": "Highlighting",
"time": "172.046",
"value": "Velocity Tower_sentence_2"
},
{
"operation": "Highlighting",
"time": "185.798",
"value": "Velocity Tower_header_cell_0_6_0"
},
{
"operation": "Highlighting",
"time": "187.747",
"value": "Velocity Tower_cell_0_6_1"
},
{
"operation": "Highlighting",
"time": "187.802",
"value": "Velocity Tower_header_cell_0_6_0"
},
{
"operation": "Highlighting",
"time": "187.808",
"value": "Velocity Tower_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "187.813",
"value": "Velocity Tower_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "213.828",
"value": "Velocity Tower_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "220.134",
"value": "Velocity Tower_cell_0_6_1"
},
{
"operation": "Highlighting deleted",
"time": "220.183",
"value": "Velocity Tower_header_cell_0_6_0"
},
{
"operation": "Highlighting deleted",
"time": "220.187",
"value": "Velocity Tower_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "220.191",
"value": "Velocity Tower_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "268.838",
"value": "Velocity Tower_sentence_8"
},
{
"operation": "Highlighting",
"time": "292.999",
"value": "Velocity Tower_sentence_5"
},
{
"operation": "Highlighting",
"time": "302.871",
"value": "Velocity Tower_sentence_2"
},
{
"operation": "Highlighting",
"time": "336.24",
"value": "Velocity Tower_sentence_4"
},
{
"operation": "finish",
"time": "384.114",
"value": "finish"
}
]
}
|
34421
|
Construction started only ten years after the Stockton & Darlington Railway commenced operations in the United States and the public passenger railway service in Jamaica, which was closed in October 1992, had a brief revival in 2011 only to be closed once again in August 2012.
|
{
"evidence": [
"Rail transport in Jamaica_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:\nConstruction started only ten years after the Stockton & Darlington Railway commenced operations in the United States and the public passenger railway service in Jamaica, which was closed in October 1992, had a brief revival in 2011 only to be closed once again in August 2012.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Rail transport in Jamaica_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4895,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.539",
"value": "?search="
},
{
"operation": "Now on",
"time": "10.179",
"value": "?search="
},
{
"operation": "search",
"time": "30.429",
"value": "Stockton & Darlington Railway"
},
{
"operation": "Now on",
"time": "31.731",
"value": "Stockton and Darlington Railway"
},
{
"operation": "Highlighting",
"time": "80.563",
"value": "Stockton and Darlington Railway_sentence_2"
},
{
"operation": "Highlighting",
"time": "80.566",
"value": "Stockton and Darlington Railway_title"
},
{
"operation": "search",
"time": "87.947",
"value": "public passenger railway service in Jamaica"
},
{
"operation": "Now on",
"time": "90.577",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "92.663",
"value": "Rail transport in Jamaica"
},
{
"operation": "Now on",
"time": "93.509",
"value": "Rail transport in Jamaica"
},
{
"operation": "Highlighting",
"time": "96.651",
"value": "Rail transport in Jamaica_sentence_1"
},
{
"operation": "finish",
"time": "109.041",
"value": "finish"
}
]
}
|
84703
|
Its forewing has postdiscal series of black-centered, and outward turned slender white lunules, while the hindwing has broad black costal and terminal margins with black markings.
|
{
"evidence": [
"Cethosia nietneri_sentence_20",
"Cethosia nietneri_sentence_21"
],
"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:\nIts forewing has postdiscal series of black-centered, and outward turned slender white lunules, while the hindwing has broad black costal and terminal margins with black markings.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Cethosia nietneri_sentence_20",
"Cethosia nietneri_sentence_21"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4896,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.262",
"value": "?search="
},
{
"operation": "search",
"time": "22.082",
"value": "Contains...Its forewing has postdiscal series of black-centered, and outward turned slender white lunules,"
},
{
"operation": "Now on",
"time": "23.41",
"value": "undefined"
},
{
"operation": "search",
"time": "80.935",
"value": "Cethosia nietneri"
},
{
"operation": "Now on",
"time": "81.792",
"value": "Cethosia nietneri"
},
{
"operation": "Page search",
"time": "93.864",
"value": "followed by a postdiscal series of black-centred, outward turned, slender white lunules"
},
{
"operation": "Highlighting",
"time": "249.813",
"value": "Cethosia nietneri_sentence_20"
},
{
"operation": "Highlighting",
"time": "274.3",
"value": "Cethosia nietneri_sentence_21"
},
{
"operation": "Page search",
"time": "289.336",
"value": "postdiscal"
},
{
"operation": "finish",
"time": "329.87",
"value": "finish"
}
]
}
|
93855
|
Manchester One, at 260 feet, remains to this day one of Manchester's tallest buildings.
|
{
"evidence": [
"Manchester One_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:\nManchester One, at 260 feet, remains to this day one of Manchester's tallest buildings.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Manchester One_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4897,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.15",
"value": "?search="
},
{
"operation": "search",
"time": "10.273",
"value": "Manchester One"
},
{
"operation": "Now on",
"time": "11.113",
"value": "Manchester One"
},
{
"operation": "Highlighting",
"time": "20.552",
"value": "Manchester One_sentence_6"
},
{
"operation": "finish",
"time": "47.63",
"value": "finish"
}
]
}
|
93106
|
Their first single, "That's What Boys Do", was a hit and was also used in The Powerpuff Girls Series, and later that year, was used in The Hot Chick and in 2002 after much success with the group, they toured alongside Aaron Carter on his "Oh Aaron tour" as a part of a Jive Records collaboration.
|
{
"evidence": [
"Nancarrow_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:\nTheir first single, \"That's What Boys Do\", was a hit and was also used in The Powerpuff Girls Series, and later that year, was used in The Hot Chick and in 2002 after much success with the group, they toured alongside Aaron Carter on his \"Oh Aaron tour\" as a part of a Jive Records collaboration.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Nancarrow_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4898,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.783",
"value": "?search="
},
{
"operation": "search",
"time": "18.921",
"value": "Contains...That's What Boys Do"
},
{
"operation": "Now on",
"time": "24.004",
"value": "undefined"
},
{
"operation": "search",
"time": "119.932",
"value": "Nancarrow"
},
{
"operation": "Now on",
"time": "121.432",
"value": "Nancarrow"
},
{
"operation": "search",
"time": "185.613",
"value": "Nancarrow"
},
{
"operation": "Now on",
"time": "186.322",
"value": "Nancarrow"
},
{
"operation": "Highlighting",
"time": "204.852",
"value": "Nancarrow_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "229.43",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "229.498",
"value": "Nancarrow"
},
{
"operation": "Now on",
"time": "241.779",
"value": "Nancarrow"
},
{
"operation": "finish",
"time": "248.033",
"value": "finish"
}
]
}
|
87926
|
Junior Eurovision Song Contest 2011 was held in the previous year's winning country and Public Television of Armenia ARMTV was the main organizer and Bulgaria (first appearance 2007) returned.
|
{
"evidence": [
"Junior Eurovision Song Contest 2011_sentence_1",
"Junior Eurovision Song Contest 2011_sentence_2",
"Junior Eurovision Song Contest 2011_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:\nJunior Eurovision Song Contest 2011 was held in the previous year's winning country and Public Television of Armenia ARMTV was the main organizer and Bulgaria (first appearance 2007) returned.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Junior Eurovision Song Contest 2011_sentence_1",
"Junior Eurovision Song Contest 2011_sentence_2",
"Junior Eurovision Song Contest 2011_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4899,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.072",
"value": "?search="
},
{
"operation": "search",
"time": "5.969",
"value": "Junior Eurovision Song Contest 2011"
},
{
"operation": "Now on",
"time": "6.782",
"value": "Junior Eurovision Song Contest 2011"
},
{
"operation": "Highlighting",
"time": "19.053",
"value": "Junior Eurovision Song Contest 2011_sentence_1"
},
{
"operation": "Highlighting",
"time": "26.183",
"value": "Junior Eurovision Song Contest 2011_sentence_2"
},
{
"operation": "Highlighting",
"time": "172.808",
"value": "Junior Eurovision Song Contest 2011_sentence_3"
},
{
"operation": "finish",
"time": "199.658",
"value": "finish"
}
]
}
|
exfever_5294
|
In Dubious Battle is a 2016 American drama film starring an American actor and filmmaker whose career spans more than nine decades.
|
{
"evidence": [
"In Dubious Battle (film)",
"Robert Duvall"
],
"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 Dubious Battle is a 2016 American drama film starring an American actor and filmmaker whose career spans more than nine decades.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"In Dubious Battle (film)",
"Robert Duvall"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4900,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
7325
|
His Excellency, The Most Reverend Fernando Isern is a Cuban-born bishop of the Roman Catholic Church in the Archdiocese of California from December 10, 2009 to June 13, 2013.
|
{
"evidence": [
"Fernando Isern_sentence_0",
"Fernando Isern_sentence_15",
"Fernando Isern_sentence_18"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHis Excellency, The Most Reverend Fernando Isern is a Cuban-born bishop of the Roman Catholic Church in the Archdiocese of California from December 10, 2009 to June 13, 2013.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Fernando Isern_sentence_0",
"Fernando Isern_sentence_15",
"Fernando Isern_sentence_18"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4901,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.623",
"value": "?search="
},
{
"operation": "search",
"time": "17.415",
"value": "Most Reverend Fernando"
},
{
"operation": "Now on",
"time": "18.7",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "27.73",
"value": "Fernando Vellosillo Barrio"
},
{
"operation": "Now on",
"time": "28.741",
"value": "Fernando Vellosillo Barrio"
},
{
"operation": "back-button-clicked",
"time": "35.123",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "35.843",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "60.715",
"value": "Fernando Vellosillo Barrio"
},
{
"operation": "Now on",
"time": "61.15",
"value": "Fernando Vellosillo Barrio"
},
{
"operation": "back-button-clicked",
"time": "67.215",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "67.96",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "76.418",
"value": "Fernando Isern"
},
{
"operation": "Now on",
"time": "77.086",
"value": "Fernando Isern"
},
{
"operation": "Highlighting",
"time": "82.188",
"value": "Fernando Isern_sentence_0"
},
{
"operation": "Highlighting",
"time": "121.928",
"value": "Fernando Isern_sentence_14"
},
{
"operation": "Highlighting",
"time": "131.296",
"value": "Fernando Isern_sentence_15"
},
{
"operation": "Highlighting deleted",
"time": "135.014",
"value": "Fernando Isern_sentence_14"
},
{
"operation": "Highlighting",
"time": "145.12",
"value": "Fernando Isern_sentence_18"
},
{
"operation": "finish",
"time": "183.113",
"value": "finish"
}
]
}
|
85882
|
1711 Sandrine was discovered in 1935 by E. Delporte in an observatory in South Africa; it's orbital period if five point twenty three years.
|
{
"evidence": [
"1711 Sandrine_sentence_1",
"Royal Observatory of Belgium_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:\n1711 Sandrine was discovered in 1935 by E. Delporte in an observatory in South Africa; it's orbital period if five point twenty three years.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"1711 Sandrine_sentence_1",
"Royal Observatory of Belgium_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4902,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.729",
"value": "?search="
},
{
"operation": "search",
"time": "25.23",
"value": "1711 Sandrine"
},
{
"operation": "Now on",
"time": "25.81",
"value": "1711 Sandrine"
},
{
"operation": "Highlighting",
"time": "41.809",
"value": "1711 Sandrine_sentence_1"
},
{
"operation": "hyperlink",
"time": "56.138",
"value": "Royal Observatory of Belgium"
},
{
"operation": "Now on",
"time": "56.725",
"value": "Royal Observatory of Belgium"
},
{
"operation": "Highlighting",
"time": "61.736",
"value": "Royal Observatory of Belgium_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "63.213",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "63.337",
"value": "1711 Sandrine"
},
{
"operation": "finish",
"time": "71.453",
"value": "finish"
}
]
}
|
93694
|
Dave Hodgson is a Mayor.
|
{
"evidence": [
"Dave Hodgson_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:\nDave Hodgson is a Mayor.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Dave Hodgson_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4903,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.968",
"value": "?search="
},
{
"operation": "search",
"time": "103.441",
"value": "David Hodgson"
},
{
"operation": "Now on",
"time": "104.501",
"value": "David Hodgson"
},
{
"operation": "hyperlink",
"time": "112.207",
"value": "David Hodgson (judge)"
},
{
"operation": "Now on",
"time": "112.738",
"value": "David Hodgson (judge)"
},
{
"operation": "back-button-clicked",
"time": "123.874",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "123.918",
"value": "David Hodgson"
},
{
"operation": "hyperlink",
"time": "130.859",
"value": "Dave Hodgson"
},
{
"operation": "Now on",
"time": "131.358",
"value": "Dave Hodgson"
},
{
"operation": "Highlighting",
"time": "138.577",
"value": "Dave Hodgson_sentence_0"
},
{
"operation": "finish",
"time": "145.89",
"value": "finish"
}
]
}
|
48022
|
The Songs About Jane Tour was the debut concert tour of Twenty One Pilots where they visited Asia and performed two shows in a year.
|
{
"evidence": [
"Songs About Jane Tour_sentence_0"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Songs About Jane Tour was the debut concert tour of Twenty One Pilots where they visited Asia and performed two shows in a year.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Songs About Jane Tour_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4904,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.623",
"value": "?search="
},
{
"operation": "search",
"time": "18.109",
"value": "Songs About Jane Tour"
},
{
"operation": "Now on",
"time": "18.63",
"value": "Songs About Jane Tour"
},
{
"operation": "Highlighting",
"time": "67.581",
"value": "Songs About Jane Tour_sentence_0"
},
{
"operation": "hyperlink",
"time": "136.572",
"value": "Maroon 5"
},
{
"operation": "Now on",
"time": "137.327",
"value": "Maroon 5"
},
{
"operation": "finish",
"time": "187.436",
"value": "finish"
}
]
}
|
exfever_2692
|
Jackie is a biographical film directed by Pablo Larraín and characterized as a category of narrative fiction intended to be more serious than humorous in tone.
|
{
"evidence": [
"Jackie (2016 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:\nJackie is a biographical film directed by Pablo Larraín and characterized as a category of narrative fiction intended to be more serious than humorous in tone.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jackie (2016 film)",
"Drama (film and television)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4905,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1651
|
The Contender is a 2000 American sci-fi drama film starring an American actress who won an award only given to actresses for quality leading roles in a Broadway play.
|
{
"evidence": [
"The Contender (2000 film)",
"Joan Allen",
"Tony Award for Best Actress in a Play"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Contender is a 2000 American sci-fi drama film starring an American actress who won an award only given to actresses for quality leading roles in a Broadway play.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Contender (2000 film)",
"Joan Allen",
"Tony Award for Best Actress in a Play"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4906,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6659
|
Dwight D.Eisenhower was a German military officer and statesman who served as Supreme Commander of the Allied Expeditionary Force in Europe during a global war that lasted from 1939 to 1945.
|
{
"evidence": [
"Dwight D. Eisenhower",
"World War II"
],
"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:\nDwight D.Eisenhower was a German military officer and statesman who served as Supreme Commander of the Allied Expeditionary Force in Europe during a global war that lasted from 1939 to 1945.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dwight D. Eisenhower",
"World War II"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4907,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1889
|
The Semitic languages, which are spoken by more than 400 million people across West Asia, are a branch of a language family of about 300 languages that are spoken on a large peninsula in East Africa, located on the easternmost part of the African mainland.
|
{
"evidence": [
"Semitic languages",
"Afroasiatic languages",
"Horn of Africa"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Semitic languages, which are spoken by more than 400 million people across West Asia, are a branch of a language family of about 300 languages that are spoken on a large peninsula in East Africa, located on the easternmost part of the African mainland.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Semitic languages",
"Afroasiatic languages",
"Horn of Africa"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4908,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7911
|
China is the largest country in Asia.
|
{
"evidence": [
"K2",
"Taxkorgan Tajik Autonomous County"
],
"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:\nChina is the largest country in Asia.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"K2",
"Taxkorgan Tajik Autonomous County"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4909,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1125
|
Choirs may sing without instrumental accompaniment or a large instrumental ensemble typical of classical music that combines instruments from different families, including double bass woodwinds such as an aerophone or reedless wind instrument that produces its sound from the flow of air across an opening.
|
{
"evidence": [
"Choir",
"Orchestra",
"Flute"
],
"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:\nChoirs may sing without instrumental accompaniment or a large instrumental ensemble typical of classical music that combines instruments from different families, including double bass woodwinds such as an aerophone or reedless wind instrument that produces its sound from the flow of air across an opening.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Choir",
"Orchestra",
"Flute"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4910,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
51095
|
The Great Gambini is a 1937 American mystery film which is a genre of film that revolves around the solution of a problem or a crime.
|
{
"evidence": [
"The Great Gambini_sentence_0",
"Mystery film_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Great Gambini is a 1937 American mystery film which is a genre of film that revolves around the solution of a problem or a crime.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"The Great Gambini_sentence_0",
"Mystery film_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4911,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.172",
"value": "?search="
},
{
"operation": "search",
"time": "16.829",
"value": "Great Gambini"
},
{
"operation": "Now on",
"time": "17.952",
"value": "undefined"
},
{
"operation": "search",
"time": "33.808",
"value": "The Great Gambini"
},
{
"operation": "Now on",
"time": "34.986",
"value": "The Great Gambini"
},
{
"operation": "Highlighting",
"time": "78.687",
"value": "The Great Gambini_sentence_0"
},
{
"operation": "hyperlink",
"time": "79.634",
"value": "mystery film"
},
{
"operation": "Now on",
"time": "84.746",
"value": "Mystery film"
},
{
"operation": "Highlighting",
"time": "85.676",
"value": "Mystery film_sentence_0"
},
{
"operation": "finish",
"time": "91.698",
"value": "finish"
}
]
}
|
58653
|
Horticultural societies typically engage in small-scale gardening with simple tools, these methods allow for higher population densities and the common type of horticulture is slash-and-burn cultivation.
|
{
"evidence": [
"Horticulture_sentence_16",
"Horticulture_sentence_22",
"Horticulture_sentence_19",
"Horticulture_sentence_23",
"Royal Horticultural Society_sentence_0",
"Horticulture_sentence_3",
"Horticulture_sentence_8",
"Horticulture_sentence_14"
],
"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:\nHorticultural societies typically engage in small-scale gardening with simple tools, these methods allow for higher population densities and the common type of horticulture is slash-and-burn cultivation.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Horticulture_sentence_16",
"Horticulture_sentence_22",
"Horticulture_sentence_19",
"Horticulture_sentence_23",
"Royal Horticultural Society_sentence_0",
"Horticulture_sentence_3",
"Horticulture_sentence_8",
"Horticulture_sentence_14"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4912,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.169",
"value": "?search="
},
{
"operation": "search",
"time": "18.693",
"value": "horticultural societies"
},
{
"operation": "Now on",
"time": "20.176",
"value": "undefined"
},
{
"operation": "search",
"time": "30.572",
"value": "Horticultural Society of London"
},
{
"operation": "Now on",
"time": "32.756",
"value": "Royal Horticultural Society"
},
{
"operation": "search",
"time": "89.502",
"value": "Horticulture"
},
{
"operation": "Now on",
"time": "91.7",
"value": "Horticulture"
},
{
"operation": "hyperlink",
"time": "100.325",
"value": "4 Organizations Horticulture_section_4"
},
{
"operation": "Page search",
"time": "119.744",
"value": "small"
},
{
"operation": "Highlighting",
"time": "139.081",
"value": "Horticulture_sentence_16"
},
{
"operation": "Highlighting",
"time": "156.562",
"value": "Horticulture_sentence_20"
},
{
"operation": "Highlighting deleted",
"time": "161.138",
"value": "Horticulture_sentence_20"
},
{
"operation": "Highlighting",
"time": "162.365",
"value": "Horticulture_sentence_21"
},
{
"operation": "Highlighting",
"time": "163.314",
"value": "Horticulture_sentence_22"
},
{
"operation": "Highlighting",
"time": "180.767",
"value": "Horticulture_sentence_19"
},
{
"operation": "Highlighting",
"time": "192.615",
"value": "Horticulture_sentence_23"
},
{
"operation": "Page search",
"time": "205.569",
"value": "slash"
},
{
"operation": "Page search",
"time": "227.889",
"value": "tool"
},
{
"operation": "Page search",
"time": "249.949",
"value": "densi*"
},
{
"operation": "Page search",
"time": "273.825",
"value": "densit"
},
{
"operation": "Page search",
"time": "276.7",
"value": "dens"
},
{
"operation": "search",
"time": "290.509",
"value": "slash-and-burn"
},
{
"operation": "Now on",
"time": "292.923",
"value": "undefined"
},
{
"operation": "search",
"time": "314.19",
"value": "horticultural society"
},
{
"operation": "Now on",
"time": "316.761",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "323.202",
"value": "Singapore Botanic Gardens"
},
{
"operation": "Now on",
"time": "328.765",
"value": "Singapore Botanic Gardens"
},
{
"operation": "back-button-clicked",
"time": "333.445",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "334.578",
"value": "undefined"
},
{
"operation": "search",
"time": "340.536",
"value": "horticultural societ"
},
{
"operation": "Now on",
"time": "341.289",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "344.761",
"value": "horticultural society"
},
{
"operation": "Now on",
"time": "346.758",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "349.608",
"value": "Francis Walker (entomologist)"
},
{
"operation": "Now on",
"time": "351.055",
"value": "Francis Walker (entomologist)"
},
{
"operation": "Page search",
"time": "362.131",
"value": "horticul"
},
{
"operation": "hyperlink",
"time": "368.6",
"value": "Royal Horticultural Society"
},
{
"operation": "Now on",
"time": "370.129",
"value": "Royal Horticultural Society"
},
{
"operation": "Highlighting",
"time": "372.002",
"value": "Royal Horticultural Society_sentence_0"
},
{
"operation": "search",
"time": "379.181",
"value": "horticulture"
},
{
"operation": "Now on",
"time": "380.828",
"value": "Horticulture"
},
{
"operation": "Highlighting",
"time": "389.835",
"value": "Horticulture_sentence_3"
},
{
"operation": "Highlighting",
"time": "401.363",
"value": "Horticulture_sentence_8"
},
{
"operation": "Highlighting",
"time": "418.38",
"value": "Horticulture_sentence_14"
},
{
"operation": "Highlighting deleted",
"time": "432.569",
"value": "Horticulture_sentence_21"
},
{
"operation": "finish",
"time": "460.781",
"value": "finish"
}
]
}
|
41959
|
Ascum first appears in 528, his title in Latin would be magister militum andd the Byzantine army defeated one group of invaders.
|
{
"evidence": [
"Ascum_sentence_7",
"Ascum_sentence_8",
"Ascum_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:\nAscum first appears in 528, his title in Latin would be magister militum andd the Byzantine army defeated one group of invaders.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Ascum_sentence_7",
"Ascum_sentence_8",
"Ascum_sentence_11"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4913,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.942",
"value": "?search="
},
{
"operation": "search",
"time": "11.741",
"value": "Ascum"
},
{
"operation": "Now on",
"time": "12.833",
"value": "Ascum"
},
{
"operation": "Highlighting",
"time": "16.78",
"value": "Ascum_sentence_7"
},
{
"operation": "Highlighting",
"time": "24.724",
"value": "Ascum_sentence_8"
},
{
"operation": "Highlighting",
"time": "38.516",
"value": "Ascum_sentence_11"
},
{
"operation": "finish",
"time": "65.36",
"value": "finish"
}
]
}
|
93651
|
Calum French received his first major championship medal (gold) at the 2017 European Amateur Boxing Championship, then qualified in the 2017 Commonwealth Games.
|
{
"evidence": [
"Calum French_sentence_7"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nCalum French received his first major championship medal (gold) at the 2017 European Amateur Boxing Championship, then qualified in the 2017 Commonwealth Games.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Calum French_sentence_7"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4914,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.199",
"value": "?search="
},
{
"operation": "search",
"time": "17.559",
"value": "Calum French"
},
{
"operation": "Now on",
"time": "18.177",
"value": "Calum French"
},
{
"operation": "Highlighting",
"time": "23.4",
"value": "Calum French_sentence_7"
},
{
"operation": "finish",
"time": "28.913",
"value": "finish"
}
]
}
|
exfever_5742
|
In 2016, the Government of India honoured the Indian actress, model, and singer, Priyanka Chopra, with the fourth-highest civilian award in the Republic of America.
|
{
"evidence": [
"Priyanka Chopra",
"Padma Shri"
],
"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 2016, the Government of India honoured the Indian actress, model, and singer, Priyanka Chopra, with the fourth-highest civilian award in the Republic of America.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Priyanka Chopra",
"Padma Shri"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4915,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4379
|
Stagira is the most populous city Greek province.
|
{
"evidence": [
"Stagira (ancient city)",
"Central Macedonia"
],
"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:\nStagira is the most populous city Greek province.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Stagira (ancient city)",
"Central Macedonia"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4916,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7491
|
Diane Ruggiero is an African television writer and producer whose credits include an American supernatural procedural crime drama television series co-developed by Rob Thomas for an American English-language commercial broadcast television network.
|
{
"evidence": [
"Diane Ruggiero",
"IZombie (TV series)",
"The CW"
],
"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:\nDiane Ruggiero is an African television writer and producer whose credits include an American supernatural procedural crime drama television series co-developed by Rob Thomas for an American English-language commercial broadcast television network.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Diane Ruggiero",
"IZombie (TV series)",
"The CW"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4917,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
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": 4918,
"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_1970
|
Anaconda is a song by Trinidadian American rapper and singer Nicki Minaj which was released on August 4, 2014 by an American record label founded by rapper Lil Wayne although the president is his lifelong friend and fellow rapper, an American record executive, rapper, and songwriter from New Orleans, Louisiana.
|
{
"evidence": [
"Anaconda (Nicki Minaj song)",
"Young Money Entertainment",
"Mack Maine"
],
"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:\nAnaconda is a song by Trinidadian American rapper and singer Nicki Minaj which was released on August 4, 2014 by an American record label founded by rapper Lil Wayne although the president is his lifelong friend and fellow rapper, an American record executive, rapper, and songwriter from New Orleans, Louisiana.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Anaconda (Nicki Minaj song)",
"Young Money Entertainment",
"Mack Maine"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4919,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7906
|
Sharon Tate was an American actress and a person with a role neither to promote, display or advertise commercial products.
|
{
"evidence": [
"Sharon Tate",
"Model (person)"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSharon Tate was an American actress and a person with a role neither to promote, display or advertise commercial products.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sharon Tate",
"Model (person)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4920,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_697
|
Stanley Park is a private park that borders the neighborhoods of Coal Harbour and a neighbourhood located in the financial and central business districts of Downtown Vancouver to the east, Stanley Park to the northwest, the English Bay to the west, and to the southwest, a neighborhood named after Squamish Chief August Jack Khatsahlano.
|
{
"evidence": [
"Stanley Park",
"West End, Vancouver",
"Kitsilano"
],
"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:\nStanley Park is a private park that borders the neighborhoods of Coal Harbour and a neighbourhood located in the financial and central business districts of Downtown Vancouver to the east, Stanley Park to the northwest, the English Bay to the west, and to the southwest, a neighborhood named after Squamish Chief August Jack Khatsahlano.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Stanley Park",
"West End, Vancouver",
"Kitsilano"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4921,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
73654
|
The one who was riding for the women's UCI team NÖ Radunion Vitalogic after moving to Austria at the age of six was Alisa Van Oijen (born 1 October 1992)
|
{
"evidence": [
"Alisa Van Oijen_sentence_0",
"Alisa Van Oijen_sentence_1",
"Alisa Van Oijen_sentence_2"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe one who was riding for the women's UCI team NÖ Radunion Vitalogic after moving to Austria at the age of six was Alisa Van Oijen (born 1 October 1992)\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Alisa Van Oijen_sentence_0",
"Alisa Van Oijen_sentence_1",
"Alisa Van Oijen_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4922,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1",
"value": "?search="
},
{
"operation": "search",
"time": "13.648",
"value": "Alisa Van Oijen"
},
{
"operation": "Now on",
"time": "15.353",
"value": "Alisa Van Oijen"
},
{
"operation": "Highlighting",
"time": "28.499",
"value": "Alisa Van Oijen_sentence_0"
},
{
"operation": "Highlighting",
"time": "29.49",
"value": "Alisa Van Oijen_sentence_1"
},
{
"operation": "Highlighting",
"time": "30.41",
"value": "Alisa Van Oijen_sentence_2"
},
{
"operation": "finish",
"time": "57.055",
"value": "finish"
}
]
}
|
exfever_7548
|
George Dewey Cukor was an American film director and film producer who directed the 1933 American pre-Code drama film based on the 1868-69 two-volume novel by Louisa May Alcott.
|
{
"evidence": [
"George Cukor",
"Little Women (1933 film)"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nGeorge Dewey Cukor was an American film director and film producer who directed the 1933 American pre-Code drama film based on the 1868-69 two-volume novel by Louisa May Alcott.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"George Cukor",
"Little Women (1933 film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4923,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
15740
|
The winner of the 2004 Vojvodina assembly election Vrbas was Igor Bečić of the Serbian Radical Party, far-right ultranationalist political party.
|
{
"evidence": [
"Igor Bečić_sentence_2",
"Igor Bečić_sentence_16",
"Serbian Radical Party_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:\nThe winner of the 2004 Vojvodina assembly election Vrbas was Igor Bečić of the Serbian Radical Party, far-right ultranationalist political party.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Igor Bečić_sentence_2",
"Igor Bečić_sentence_16",
"Serbian Radical Party_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4924,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.661",
"value": "?search="
},
{
"operation": "search",
"time": "18.839",
"value": "2004 Vojvodina provincial election"
},
{
"operation": "Now on",
"time": "20.285",
"value": "2004 Vojvodina provincial election"
},
{
"operation": "Page search",
"time": "32.228",
"value": "igor"
},
{
"operation": "search",
"time": "56.586",
"value": "2004 vojvodina assembly election"
},
{
"operation": "Now on",
"time": "59.261",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "64.644",
"value": "Igor Bečić"
},
{
"operation": "Now on",
"time": "66.036",
"value": "Igor Bečić"
},
{
"operation": "Page search",
"time": "70.737",
"value": "2004"
},
{
"operation": "Page search",
"time": "71.816",
"value": "2004"
},
{
"operation": "Highlighting",
"time": "79.2",
"value": "Igor Bečić_sentence_2"
},
{
"operation": "Highlighting",
"time": "85.429",
"value": "Igor Bečić_sentence_16"
},
{
"operation": "hyperlink",
"time": "104.61",
"value": "Serbian Radical Party"
},
{
"operation": "Now on",
"time": "107.058",
"value": "Serbian Radical Party"
},
{
"operation": "Highlighting",
"time": "108.279",
"value": "Serbian Radical Party_sentence_1"
},
{
"operation": "finish",
"time": "124.498",
"value": "finish"
}
]
}
|
exfever_2115
|
Caitlyn Jenner, an American media personality and retired Olympic gold medal-winning decathlete had a six-year career in a sport that is a combined event in athletics consisting of nine track and field events.
|
{
"evidence": [
"Caitlyn Jenner",
"Decathlon"
],
"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:\nCaitlyn Jenner, an American media personality and retired Olympic gold medal-winning decathlete had a six-year career in a sport that is a combined event in athletics consisting of nine track and field events.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Caitlyn Jenner",
"Decathlon"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4925,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7581
|
The Aegean Sea is an elongated embayment of the Mediterranean Sea between Europe and Earth's largest and most populous continent, located primarily in the Northern and the half of Earth, which is the prime meridian west.
|
{
"evidence": [
"Aegean Sea",
"Asia",
"Eastern Hemisphere"
],
"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 Aegean Sea is an elongated embayment of the Mediterranean Sea between Europe and Earth's largest and most populous continent, located primarily in the Northern and the half of Earth, which is the prime meridian west.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Aegean Sea",
"Asia",
"Eastern Hemisphere"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4926,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
39270
|
Mirza Varesanovic had signed a six month contract to manage Tuzla City in February 2019, but after their last game in May 2019, he decided to leave, three months into the contract.
|
{
"evidence": [
"Mirza Varešanović_sentence_20",
"Mirza Varešanović_sentence_22"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMirza Varesanovic had signed a six month contract to manage Tuzla City in February 2019, but after their last game in May 2019, he decided to leave, three months into the contract.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Mirza Varešanović_sentence_20",
"Mirza Varešanović_sentence_22"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4927,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.971",
"value": "?search="
},
{
"operation": "search",
"time": "9.84",
"value": "Mirza Varešanović"
},
{
"operation": "Now on",
"time": "10.948",
"value": "Mirza Varešanović"
},
{
"operation": "Page search",
"time": "34.095",
"value": "contract"
},
{
"operation": "hyperlink",
"time": "47.444",
"value": "3.5 Tuzla City Mirza Varešanović_section_7"
},
{
"operation": "Highlighting",
"time": "49.459",
"value": "Mirza Varešanović_sentence_20"
},
{
"operation": "Highlighting",
"time": "62.066",
"value": "Mirza Varešanović_sentence_22"
},
{
"operation": "finish",
"time": "146.453",
"value": "finish"
}
]
}
|
exfever_4609
|
Steel is a 1997 African superhero film loosely based on the American comic book character of the same name a subsidiary of Warner Bros.
|
{
"evidence": [
"Steel (1997 film)",
"DC Comics"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nSteel is a 1997 African superhero film loosely based on the American comic book character of the same name a subsidiary of Warner Bros.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Steel (1997 film)",
"DC Comics"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4928,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
39515
|
Barbara Stauffacher Solomon was born in 1928 as a third-generation San Franciscan and studied painting and sculpture at San Francisco Art Institute, a private college of contemporary art in San Francisco, California.
|
{
"evidence": [
"Barbara Stauffacher Solomon_sentence_2",
"Barbara Stauffacher Solomon_sentence_3",
"San Francisco Art Institute_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:\nBarbara Stauffacher Solomon was born in 1928 as a third-generation San Franciscan and studied painting and sculpture at San Francisco Art Institute, a private college of contemporary art in San Francisco, California.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Barbara Stauffacher Solomon_sentence_2",
"Barbara Stauffacher Solomon_sentence_3",
"San Francisco Art Institute_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4929,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.734",
"value": "?search="
},
{
"operation": "search",
"time": "9.044",
"value": "Barbara Stauffacher Solomon"
},
{
"operation": "Now on",
"time": "10.471",
"value": "Barbara Stauffacher Solomon"
},
{
"operation": "Highlighting",
"time": "13.582",
"value": "Barbara Stauffacher Solomon_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "16.136",
"value": "Barbara Stauffacher Solomon_sentence_0"
},
{
"operation": "Highlighting",
"time": "17.051",
"value": "Barbara Stauffacher Solomon_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "17.065",
"value": "Barbara Stauffacher Solomon_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "17.066",
"value": "Barbara Stauffacher Solomon_header_cell_0_0_0"
},
{
"operation": "Highlighting deleted",
"time": "25.001",
"value": "Barbara Stauffacher Solomon_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "25.025",
"value": "Barbara Stauffacher Solomon_header_cell_0_1_0"
},
{
"operation": "Highlighting deleted",
"time": "25.03",
"value": "Barbara Stauffacher Solomon_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "26.272",
"value": "Barbara Stauffacher Solomon_sentence_0"
},
{
"operation": "Highlighting",
"time": "422.432",
"value": "Barbara Stauffacher Solomon_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "429.912",
"value": "Barbara Stauffacher Solomon_sentence_0"
},
{
"operation": "Highlighting",
"time": "442.865",
"value": "Barbara Stauffacher Solomon_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "444.182",
"value": "Barbara Stauffacher Solomon_sentence_4"
},
{
"operation": "Highlighting",
"time": "446.072",
"value": "Barbara Stauffacher Solomon_sentence_3"
},
{
"operation": "hyperlink",
"time": "461.117",
"value": "San Francisco Art Institute"
},
{
"operation": "Now on",
"time": "462.297",
"value": "San Francisco Art Institute"
},
{
"operation": "Highlighting",
"time": "464.475",
"value": "San Francisco Art Institute_sentence_1"
},
{
"operation": "finish",
"time": "477.495",
"value": "finish"
}
]
}
|
30580
|
In 1904, Henry James Bruce entered the Foreign Office, and in 1905 was sent to Vienna as a junior diplomat in the British mission to the Austrian Empire, a Central European multinational great power from 1804 to 1867, and was appointed Third Secretary there in 1906.
|
{
"evidence": [
"Henry James Bruce_sentence_8",
"Austrian Empire_sentence_3",
"Henry James Bruce_sentence_9"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn 1904, Henry James Bruce entered the Foreign Office, and in 1905 was sent to Vienna as a junior diplomat in the British mission to the Austrian Empire, a Central European multinational great power from 1804 to 1867, and was appointed Third Secretary there in 1906.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Henry James Bruce_sentence_8",
"Austrian Empire_sentence_3",
"Henry James Bruce_sentence_9"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4930,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.672",
"value": "?search="
},
{
"operation": "search",
"time": "7.809",
"value": "Henry James Bruce"
},
{
"operation": "Now on",
"time": "8.414",
"value": "Henry James Bruce"
},
{
"operation": "Highlighting",
"time": "12.018",
"value": "Henry James Bruce_sentence_8"
},
{
"operation": "hyperlink",
"time": "21.058",
"value": "Austrian Empire"
},
{
"operation": "Now on",
"time": "22.128",
"value": "Austrian Empire"
},
{
"operation": "Highlighting",
"time": "24.135",
"value": "Austrian Empire_sentence_3"
},
{
"operation": "back-button-clicked",
"time": "29.407",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "29.501",
"value": "Henry James Bruce"
},
{
"operation": "Highlighting",
"time": "30.555",
"value": "Henry James Bruce_sentence_9"
},
{
"operation": "finish",
"time": "34.884",
"value": "finish"
}
]
}
|
exfever_8463
|
KFC is the most successful fast-food restaurant chain in Louisville, Kentucky.
|
{
"evidence": [
"KFC",
"Fried chicken"
],
"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:\nKFC is the most successful fast-food restaurant chain in Louisville, Kentucky.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"KFC",
"Fried chicken"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4931,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
58448
|
Movement of the National Left, a left-wing political party (which typically involve a concern for those in society whom its adherents perceive as disadvantaged relative to others as well as a belief that there are unjustified inequalities that need to be reduced or abolished) in Bolivia was founded by Luis Sandoval Moron in 1978 when he split from the Revolutionary Nationalist Movement.
|
{
"evidence": [
"Movement of the National Left_sentence_0",
"Movement of the National Left_sentence_1",
"Left-wing politics_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:\nMovement of the National Left, a left-wing political party (which typically involve a concern for those in society whom its adherents perceive as disadvantaged relative to others as well as a belief that there are unjustified inequalities that need to be reduced or abolished) in Bolivia was founded by Luis Sandoval Moron in 1978 when he split from the Revolutionary Nationalist Movement.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Movement of the National Left_sentence_0",
"Movement of the National Left_sentence_1",
"Left-wing politics_sentence_6"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4932,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.221",
"value": "?search="
},
{
"operation": "search",
"time": "22.084",
"value": "Movement of the National Left"
},
{
"operation": "Now on",
"time": "24.109",
"value": "Movement of the National Left"
},
{
"operation": "Highlighting",
"time": "35.922",
"value": "Movement of the National Left_sentence_0"
},
{
"operation": "Highlighting",
"time": "48.462",
"value": "Movement of the National Left_sentence_1"
},
{
"operation": "hyperlink",
"time": "55.299",
"value": "left-wing"
},
{
"operation": "Now on",
"time": "57.324",
"value": "Left-wing politics"
},
{
"operation": "Highlighting",
"time": "67.3",
"value": "Left-wing politics_sentence_6"
},
{
"operation": "finish",
"time": "86.248",
"value": "finish"
}
]
}
|
exfever_337
|
Christina Applegate is an American actress who established a successful film and television career in her adult years, winning seven nominations for an award bestowed by ATAS in recognition of excellence in Australian primetime television programming.
|
{
"evidence": [
"Christina Applegate",
"Primetime Emmy Award"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nChristina Applegate is an American actress who established a successful film and television career in her adult years, winning seven nominations for an award bestowed by ATAS in recognition of excellence in Australian primetime television programming.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Christina Applegate",
"Primetime Emmy Award"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4933,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
4509
|
The Makerere University School of Public Health was established in 2001 in Uganda.
|
{
"evidence": [
"Makerere University School of Public Health_sentence_6",
"Makerere University School of Public Health_sentence_8"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Makerere University School of Public Health was established in 2001 in Uganda.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Makerere University School of Public Health_sentence_6",
"Makerere University School of Public Health_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4934,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "34.71",
"value": "Makerere University School of Public Health"
},
{
"operation": "Highlighting",
"time": "48.857",
"value": "Makerere University School of Public Health_sentence_6"
},
{
"operation": "Highlighting",
"time": "61.31",
"value": "Makerere University School of Public Health_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "61.367",
"value": "Makerere University School of Public Health_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "67.656",
"value": "Makerere University School of Public Health_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "70.958",
"value": "Makerere University School of Public Health_cell_0_1_1"
},
{
"operation": "Highlighting deleted",
"time": "71.03",
"value": "Makerere University School of Public Health_header_cell_0_1_0"
},
{
"operation": "finish",
"time": "77.073",
"value": "finish"
}
]
}
|
48737
|
South Korea won the most number of medals in the 2010 World Junior Short Track Speed Skating Championships, which took place between January 8 to 10, 2010 neither in Taipei nor in Taiwan at the Taipei Arena, with eleven gold, six silver, and five bronze medals.
|
{
"evidence": [
"2010 World Junior Short Track Speed Skating Championships_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:\nSouth Korea won the most number of medals in the 2010 World Junior Short Track Speed Skating Championships, which took place between January 8 to 10, 2010 neither in Taipei nor in Taiwan at the Taipei Arena, with eleven gold, six silver, and five bronze medals.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"2010 World Junior Short Track Speed Skating Championships_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4935,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.818",
"value": "?search="
},
{
"operation": "search",
"time": "54.002",
"value": "Contains...South Korea in the 2010 World Junior Short Track Speed Skating Championship"
},
{
"operation": "Now on",
"time": "56.779",
"value": "undefined"
},
{
"operation": "search",
"time": "96.765",
"value": "2010 World Junior Short Track Speed Skating Championships"
},
{
"operation": "Now on",
"time": "97.789",
"value": "2010 World Junior Short Track Speed Skating Championships"
},
{
"operation": "Highlighting",
"time": "102.479",
"value": "2010 World Junior Short Track Speed Skating Championships_sentence_0"
},
{
"operation": "finish",
"time": "120.331",
"value": "finish"
}
]
}
|
exfever_15
|
Richard Nixon was the president of the United States who was involved in the first arrival of a spacecraft on the surface of the Moon with a spacecraft landing that does not result in significant damage to or destruction of the vehicle or its payload.
|
{
"evidence": [
"Richard Nixon",
"Moon landing",
"Soft landing (rocketry)"
],
"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:\nRichard Nixon was the president of the United States who was involved in the first arrival of a spacecraft on the surface of the Moon with a spacecraft landing that does not result in significant damage to or destruction of the vehicle or its payload.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Richard Nixon",
"Moon landing",
"Soft landing (rocketry)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4936,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
27062
|
The Belloa plant belongs to the sunflower family.
|
{
"evidence": [
"Belloa_sentence_0"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Belloa plant belongs to the sunflower family.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Belloa_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4937,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.619",
"value": "?search="
},
{
"operation": "search",
"time": "12.519",
"value": "Belloa"
},
{
"operation": "Now on",
"time": "13.151",
"value": "Belloa"
},
{
"operation": "Highlighting",
"time": "17.119",
"value": "Belloa_sentence_0"
},
{
"operation": "finish",
"time": "41.5",
"value": "finish"
}
]
}
|
8598
|
The acronyms in healthcare are considered as common terminologies and is criticized for potentially leading to incorrect assumptions and medication errors.
|
{
"evidence": [
"Jargon_sentence_4",
"Acronyms in healthcare_sentence_8"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe acronyms in healthcare are considered as common terminologies and is criticized for potentially leading to incorrect assumptions and medication errors.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Jargon_sentence_4",
"Acronyms in healthcare_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4938,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.611",
"value": "?search="
},
{
"operation": "search",
"time": "17.496",
"value": "Acronyms in healthcare"
},
{
"operation": "Now on",
"time": "18.746",
"value": "Acronyms in healthcare"
},
{
"operation": "Highlighting",
"time": "29.818",
"value": "Acronyms in healthcare_sentence_4"
},
{
"operation": "Highlighting",
"time": "36.963",
"value": "Acronyms in healthcare_sentence_3"
},
{
"operation": "hyperlink",
"time": "38.253",
"value": "jargon"
},
{
"operation": "Now on",
"time": "39.334",
"value": "Jargon"
},
{
"operation": "Highlighting",
"time": "50.011",
"value": "Jargon_sentence_4"
},
{
"operation": "back-button-clicked",
"time": "63.614",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "63.705",
"value": "Acronyms in healthcare"
},
{
"operation": "Highlighting deleted",
"time": "75.437",
"value": "Acronyms in healthcare_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "76.374",
"value": "Acronyms in healthcare_sentence_3"
},
{
"operation": "Highlighting",
"time": "85.287",
"value": "Acronyms in healthcare_sentence_8"
},
{
"operation": "finish",
"time": "121.225",
"value": "finish"
}
]
}
|
31095
|
Woodruff, a quarterback and punter at Kent State University, joined the University of Washington staff in Seattle as a graduate assistant under head coach Don James in 1980.
|
{
"evidence": [
"Jeff Woodruff_sentence_2",
"Jeff Woodruff_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:\nWoodruff, a quarterback and punter at Kent State University, joined the University of Washington staff in Seattle as a graduate assistant under head coach Don James in 1980.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Jeff Woodruff_sentence_2",
"Jeff Woodruff_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4939,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.685",
"value": "?search="
},
{
"operation": "search",
"time": "32.198",
"value": "Woodruff quarterback kent state"
},
{
"operation": "Now on",
"time": "33.727",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "38.804",
"value": "Jeff Woodruff"
},
{
"operation": "Now on",
"time": "39.514",
"value": "Jeff Woodruff"
},
{
"operation": "Highlighting",
"time": "43.948",
"value": "Jeff Woodruff_sentence_2"
},
{
"operation": "Highlighting",
"time": "59.508",
"value": "Jeff Woodruff_cell_0_5_1"
},
{
"operation": "Highlighting",
"time": "59.519",
"value": "Jeff Woodruff_header_cell_0_5_0"
},
{
"operation": "Highlighting",
"time": "59.519",
"value": "Jeff Woodruff_header_cell_0_3_0"
},
{
"operation": "Highlighting",
"time": "71.66",
"value": "Jeff Woodruff_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "86.108",
"value": "Jeff Woodruff_cell_0_5_1"
},
{
"operation": "Highlighting deleted",
"time": "86.124",
"value": "Jeff Woodruff_header_cell_0_5_0"
},
{
"operation": "Highlighting deleted",
"time": "86.125",
"value": "Jeff Woodruff_header_cell_0_3_0"
},
{
"operation": "finish",
"time": "102.82",
"value": "finish"
}
]
}
|
exfever_7700
|
Zootopia is a film produced by a Japanese animation studio that creates animated short films and is headquartered in a city in the southeastern end of the San Fernando Valley in Los Angeles County, California.
|
{
"evidence": [
"Zootopia",
"Walt Disney Animation Studios",
"Burbank, California"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nZootopia is a film produced by a Japanese animation studio that creates animated short films and is headquartered in a city in the southeastern end of the San Fernando Valley in Los Angeles County, California.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Zootopia",
"Walt Disney Animation Studios",
"Burbank, California"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4940,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_45
|
Paul Feig is an actor and collaborator of a fashion designer.
|
{
"evidence": [
"Paul Feig",
"Melissa McCarthy"
],
"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 Feig is an actor and collaborator of a fashion designer.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Paul Feig",
"Melissa McCarthy"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4941,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6676
|
Jerry Goldsmith was an American composer who composed scores for a 1976 American film starring a British actress born on December 20, 1952.
|
{
"evidence": [
"Jerry Goldsmith",
"Logan's Run (film)",
"Jenny Agutter"
],
"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:\nJerry Goldsmith was an American composer who composed scores for a 1976 American film starring a British actress born on December 20, 1952.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jerry Goldsmith",
"Logan's Run (film)",
"Jenny Agutter"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4942,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1195
|
Tunisia is the northernmost country in Africa, bordered by a country in the Maghreb region of South Africa.
|
{
"evidence": [
"Tunisia",
"Libya"
],
"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:\nTunisia is the northernmost country in Africa, bordered by a country in the Maghreb region of South Africa.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tunisia",
"Libya"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4943,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
53507
|
Hunter Region is a very small area.
|
{
"evidence": [
"Hunter Region_sentence_2",
"Hunter Region_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:\nHunter Region is a very small area.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Hunter Region_sentence_2",
"Hunter Region_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4944,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.796",
"value": "?search="
},
{
"operation": "search",
"time": "10.303",
"value": "Hunter Region"
},
{
"operation": "Now on",
"time": "11.493",
"value": "Hunter Region"
},
{
"operation": "Highlighting",
"time": "111.124",
"value": "Hunter Region_sentence_2"
},
{
"operation": "Highlighting",
"time": "129.101",
"value": "Hunter Region_sentence_4"
},
{
"operation": "finish",
"time": "135.514",
"value": "finish"
}
]
}
|
24281
|
Tridrepana acuta's forewing is orange and costa is reddish and have an light spot on discocellulars and the outer margin is red brown.
|
{
"evidence": [
"Tridrepana acuta_sentence_9",
"Tridrepana acuta_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:\nTridrepana acuta's forewing is orange and costa is reddish and have an light spot on discocellulars and the outer margin is red brown.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Tridrepana acuta_sentence_9",
"Tridrepana acuta_sentence_11"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4945,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.868",
"value": "?search="
},
{
"operation": "search",
"time": "10.59",
"value": "Tridrepana acuta"
},
{
"operation": "Now on",
"time": "11.422",
"value": "Tridrepana acuta"
},
{
"operation": "Highlighting",
"time": "19.849",
"value": "Tridrepana acuta_sentence_9"
},
{
"operation": "Highlighting",
"time": "34.038",
"value": "Tridrepana acuta_sentence_11"
},
{
"operation": "finish",
"time": "59.757",
"value": "finish"
}
]
}
|
exfever_4875
|
Red is a 2010 American action comedy film produced by Di Bonaventura Pictures and distributed by an American film production and distribution company headquartered in Santa Monica, in a state in the Western United Kingdom.
|
{
"evidence": [
"Red (2010 film)",
"Summit Entertainment",
"California"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nRed is a 2010 American action comedy film produced by Di Bonaventura Pictures and distributed by an American film production and distribution company headquartered in Santa Monica, in a state in the Western United Kingdom.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Red (2010 film)",
"Summit Entertainment",
"California"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4946,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_197
|
Year Zero is the third studio album by American industrial rock band Nine Inch Nails, released by an American record label owned by a Dutch–American multinational music corporation.
|
{
"evidence": [
"Year Zero (album)",
"Interscope Records",
"Universal Music Group"
],
"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:\nYear Zero is the third studio album by American industrial rock band Nine Inch Nails, released by an American record label owned by a Dutch–American multinational music corporation.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Year Zero (album)",
"Interscope Records",
"Universal Music Group"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4947,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
29188
|
Lonely Accordion is a song whose text was written in a poem by Mikhail Isakovsky and composed by Boris Mokrousov that was performed by the Pyatnitsky Choir (but did not receive significant fame).
|
{
"evidence": [
"Lonely Accordion_sentence_0",
"Lonely Accordion_sentence_1",
"Lonely Accordion_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:\nLonely Accordion is a song whose text was written in a poem by Mikhail Isakovsky and composed by Boris Mokrousov that was performed by the Pyatnitsky Choir (but did not receive significant fame).\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Lonely Accordion_sentence_0",
"Lonely Accordion_sentence_1",
"Lonely Accordion_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4948,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "8.408",
"value": "?search="
},
{
"operation": "search",
"time": "32.924",
"value": "Lonely Accordion"
},
{
"operation": "Now on",
"time": "34.207",
"value": "Lonely Accordion"
},
{
"operation": "Highlighting",
"time": "38.315",
"value": "Lonely Accordion_sentence_0"
},
{
"operation": "Highlighting",
"time": "41.388",
"value": "Lonely Accordion_sentence_1"
},
{
"operation": "Highlighting",
"time": "50.057",
"value": "Lonely Accordion_sentence_4"
},
{
"operation": "finish",
"time": "69.891",
"value": "finish"
}
]
}
|
exfever_6028
|
The writer of The Affair ghostwrote the book with O. J. Simpson,?an English screenwriter and ghostwriter based in Los Angeles, California.
|
{
"evidence": [
"Pablo Fenjves",
"If I Did It",
"Pablo Fenjves"
],
"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 writer of The Affair ghostwrote the book with O. J. Simpson,?an English screenwriter and ghostwriter based in Los Angeles, California.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Pablo Fenjves",
"If I Did It",
"Pablo Fenjves"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4949,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
92676
|
House of Bluebird is a 50- episode television series from South Korea.
|
{
"evidence": [
"House of Bluebird_sentence_1",
"House of Bluebird_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:\nHouse of Bluebird is a 50- episode television series from South Korea.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"House of Bluebird_sentence_1",
"House of Bluebird_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4950,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.173",
"value": "?search="
},
{
"operation": "search",
"time": "12.218",
"value": "House of Bluebird"
},
{
"operation": "Now on",
"time": "13.842",
"value": "House of Bluebird"
},
{
"operation": "Highlighting",
"time": "21.625",
"value": "House of Bluebird_sentence_1"
},
{
"operation": "Highlighting",
"time": "23.393",
"value": "House of Bluebird_sentence_2"
},
{
"operation": "finish",
"time": "36.476",
"value": "finish"
}
]
}
|
exfever_6952
|
Hotel Transylvania is a Canadian animated media franchise that has an ensemble cast, usually led by the voices of an American comedian who is a cast member on an American late-night live television sketch comedy and variety show created by Lorne Michaels.
|
{
"evidence": [
"Hotel Transylvania",
"Andy Samberg",
"Saturday Night Live"
],
"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:\nHotel Transylvania is a Canadian animated media franchise that has an ensemble cast, usually led by the voices of an American comedian who is a cast member on an American late-night live television sketch comedy and variety show created by Lorne Michaels.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hotel Transylvania",
"Andy Samberg",
"Saturday Night Live"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4951,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7786
|
The Spanish language is credited as the least divergent from it, together with a Romance language, considered as the language that is closest to Latin among all its genealogical descendants.
|
{
"evidence": [
"Italian language",
"Sardinian language"
],
"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 Spanish language is credited as the least divergent from it, together with a Romance language, considered as the language that is closest to Latin among all its genealogical descendants.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Italian language",
"Sardinian language"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4952,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
3526
|
Sveti Primoz nad Muto is located in Bosnia, has a total area of 3.20 square miles, and in 2002 is had a population of 304 people.
|
{
"evidence": [
"Sveti Primož nad Muto_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:\nSveti Primoz nad Muto is located in Bosnia, has a total area of 3.20 square miles, and in 2002 is had a population of 304 people.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Sveti Primož nad Muto_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4953,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "7.208",
"value": "Sveti Primož nad Muto"
},
{
"operation": "Highlighting",
"time": "14.412",
"value": "Sveti Primož nad Muto_sentence_0"
},
{
"operation": "finish",
"time": "33.162",
"value": "finish"
}
]
}
|
93807
|
The first known species of the Cordylanthus was Cordylanthus rigidus described in 1801 by the English plant taxonomist George Bentham who also found other species.
|
{
"evidence": [
"Cordylanthus_sentence_5"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe first known species of the Cordylanthus was Cordylanthus rigidus described in 1801 by the English plant taxonomist George Bentham who also found other species.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Cordylanthus_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4954,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.129",
"value": "undefined"
},
{
"operation": "search",
"time": "10.56",
"value": "Cordylanthus"
},
{
"operation": "Now on",
"time": "11.751",
"value": "Cordylanthus"
},
{
"operation": "Highlighting",
"time": "24.175",
"value": "Cordylanthus_sentence_5"
},
{
"operation": "finish",
"time": "33.899",
"value": "finish"
}
]
}
|
exfever_7747
|
Jeff Goldblum is the youngest actor who has starred in the highest-grossing films of his era.
|
{
"evidence": [
"Jeff Goldblum",
"Jurassic World: Fallen Kingdom"
],
"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:\nJeff Goldblum is the youngest actor who has starred in the highest-grossing films of his era.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jeff Goldblum",
"Jurassic World: Fallen Kingdom"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4955,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8982
|
Pi is a mathematical constant that is the ratio of a circle's circumference to any straight line segment that passes through the center of the circle and whose endpoints lie in the square.
|
{
"evidence": [
"Pi",
"Diameter"
],
"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:\nPi is a mathematical constant that is the ratio of a circle's circumference to any straight line segment that passes through the center of the circle and whose endpoints lie in the square.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Pi",
"Diameter"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4956,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2204
|
Richard Nixon served five years in the White House, which witnessed the first arrival of a spacecraft on the surface of the Moon.
|
{
"evidence": [
"Richard Nixon",
"Moon landing"
],
"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:\nRichard Nixon served five years in the White House, which witnessed the first arrival of a spacecraft on the surface of the Moon.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Richard Nixon",
"Moon landing"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4957,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
67770
|
Jessica Mauboy has never won Female Artist of the Year at the Deadly Awards, despite two nominations.
|
{
"evidence": [
"Jessica Mauboy_sentence_69"
],
"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:\nJessica Mauboy has never won Female Artist of the Year at the Deadly Awards, despite two nominations.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Jessica Mauboy_sentence_69"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4958,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.82",
"value": "?search="
},
{
"operation": "search",
"time": "16.055",
"value": "Jessica Mauboy"
},
{
"operation": "Now on",
"time": "17.649",
"value": "Jessica Mauboy"
},
{
"operation": "Page search",
"time": "45.261",
"value": "DEADLY"
},
{
"operation": "Highlighting",
"time": "61.971",
"value": "Jessica Mauboy_sentence_69"
},
{
"operation": "finish",
"time": "77.645",
"value": "finish"
}
]
}
|
exfever_4458
|
Michael Keaton is an American actor best known for portraying Bruce Wayne, a superhero who appeared in comic books published by one of the largest and oldest American comic book companies.
|
{
"evidence": [
"Michael Keaton",
"DC Comics"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMichael Keaton is an American actor best known for portraying Bruce Wayne, a superhero who appeared in comic books published by one of the largest and oldest American comic book companies.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Michael Keaton",
"DC Comics"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4959,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7521
|
Las Vegas is the city that anchors the major metropolitan area in the constituent political entity, which is currently 50.
|
{
"evidence": [
"Las Vegas",
"Las Vegas Valley",
"U.S. state"
],
"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:\nLas Vegas is the city that anchors the major metropolitan area in the constituent political entity, which is currently 50.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Las Vegas",
"Las Vegas Valley",
"U.S. state"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4960,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_409
|
Kingdom Hearts is a crossover of various Disney properties based on an original self-consistent setting with events that differ from the real world.
|
{
"evidence": [
"Kingdom Hearts",
"Fictional universe"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nKingdom Hearts is a crossover of various Disney properties based on an original self-consistent setting with events that differ from the real world.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Kingdom Hearts",
"Fictional universe"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4961,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
33375
|
On July 29, 1900, Gaetano Bresci an Italian-American Anarchist, assassinated Victor Emmanuel III of Italy.
|
{
"evidence": [
"Gaetano Bresci_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:\nOn July 29, 1900, Gaetano Bresci an Italian-American Anarchist, assassinated Victor Emmanuel III of Italy.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Gaetano Bresci_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4962,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "2.099",
"value": "?search="
},
{
"operation": "search",
"time": "12.842",
"value": "Gaetano Bresci"
},
{
"operation": "Now on",
"time": "14.377",
"value": "Gaetano Bresci"
},
{
"operation": "Highlighting",
"time": "19.522",
"value": "Gaetano Bresci_sentence_0"
},
{
"operation": "finish",
"time": "69.519",
"value": "finish"
}
]
}
|
exfever_6343
|
Women Talking 2022 was the latest movie starring Ben Whishaw.
|
{
"evidence": [
"Ben Whishaw",
"Hamlet",
"Prince Hamlet"
],
"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:\nWomen Talking 2022 was the latest movie starring Ben Whishaw.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ben Whishaw",
"Hamlet",
"Prince Hamlet"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4963,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3323
|
Harvey Kurtzman began to work on the New Trend line of comic books at an American publisher of comic books notably the American bi-monthly drama comic anthology series.
|
{
"evidence": [
"Harvey Kurtzman",
"EC Comics",
"Tales from the Crypt (comics)"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHarvey Kurtzman began to work on the New Trend line of comic books at an American publisher of comic books notably the American bi-monthly drama comic anthology series.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Harvey Kurtzman",
"EC Comics",
"Tales from the Crypt (comics)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4964,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_891
|
The Northeast megalopolis is located primarily on the Atlantic Coast of a geographical region of the United States.
|
{
"evidence": [
"Northeast megalopolis",
"Northeastern United States"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Northeast megalopolis is located primarily on the Atlantic Coast of a geographical region of the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Northeast megalopolis",
"Northeastern United States"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4965,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2229
|
Miley Cyrus is an American singer, songwriter, actress, and television personality known for her music incorporates elements of varied styles and genres, including a fusion genre of country music and pop known originally as the sound originated during the mid-1950s as a subgenre of American country music, replacing the chart dominance of the rough honky tonk music.
|
{
"evidence": [
"Miley Cyrus",
"Country pop",
"Nashville sound"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMiley Cyrus is an American singer, songwriter, actress, and television personality known for her music incorporates elements of varied styles and genres, including a fusion genre of country music and pop known originally as the sound originated during the mid-1950s as a subgenre of American country music, replacing the chart dominance of the rough honky tonk music.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Miley Cyrus",
"Country pop",
"Nashville sound"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4966,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_938
|
Stadium Arcadium was the most famous studio album by the American rock band Red Hot Chili Peppers.
|
{
"evidence": [
"Stadium Arcadium",
"Dani California"
],
"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:\nStadium Arcadium was the most famous studio album by the American rock band Red Hot Chili Peppers.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Stadium Arcadium",
"Dani California"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4967,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_3941
|
Julius Caesar is a 1955 American film starring Marlon Brando as a Roman general who played a critical role in transforming the Roman Republic from a constitutional republic into the autocratic Roman Empire.
|
{
"evidence": [
"Julius Caesar (1953 film)",
"Mark Antony"
],
"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:\nJulius Caesar is a 1955 American film starring Marlon Brando as a Roman general who played a critical role in transforming the Roman Republic from a constitutional republic into the autocratic Roman Empire.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Julius Caesar (1953 film)",
"Mark Antony"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4968,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7912
|
The Bible is the most famous book in the world.
|
{
"evidence": [
"Bible",
"Judaism"
],
"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 Bible is the most famous book in the world.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Bible",
"Judaism"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4969,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_6551
|
Daft Punk were a French electronic music duo whose genre combined elements of house music with funk, techno, disco, pop, indie rock, and a popular music developed in Canada by inner-city African Americans.
|
{
"evidence": [
"Daft Punk",
"Hip hop music"
],
"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:\nDaft Punk were a French electronic music duo whose genre combined elements of house music with funk, techno, disco, pop, indie rock, and a popular music developed in Canada by inner-city African Americans.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Daft Punk",
"Hip hop music"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4970,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
16684
|
Efim Shifrin would not start acting until he graduated from the State University of Circus and Stage Art in 1978.
|
{
"evidence": [
"Efim Shifrin_sentence_3",
"Efim Shifrin_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:\nEfim Shifrin would not start acting until he graduated from the State University of Circus and Stage Art in 1978.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Efim Shifrin_sentence_3",
"Efim Shifrin_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4971,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.367",
"value": "?search="
},
{
"operation": "search",
"time": "14.618",
"value": "Efim Shifrin"
},
{
"operation": "Now on",
"time": "15.699",
"value": "Efim Shifrin"
},
{
"operation": "Page search",
"time": "41.848",
"value": "until"
},
{
"operation": "Page search",
"time": "50.304",
"value": "grad"
},
{
"operation": "Highlighting",
"time": "82.756",
"value": "Efim Shifrin_sentence_3"
},
{
"operation": "Highlighting",
"time": "235.757",
"value": "Efim Shifrin_sentence_4"
},
{
"operation": "finish",
"time": "261.196",
"value": "finish"
}
]
}
|
exfever_4459
|
Leopold II was the second King of the Belgians born in a region of Belgium comprising 20 municipalities, including the City of Brussels, which is the capital of Belgium.
|
{
"evidence": [
"Leopold II of Belgium",
"Brussels"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLeopold II was the second King of the Belgians born in a region of Belgium comprising 20 municipalities, including the City of Brussels, which is the capital of Belgium.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Leopold II of Belgium",
"Brussels"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4972,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_408
|
The Hobbit is set within Tolkien's fictional universe that follows the quest of the title character, the hobbit, who is a supporting character in an epic high-fantasy novel and one of the best-selling books ever written.
|
{
"evidence": [
"The Hobbit",
"Bilbo Baggins",
"The Lord of the Rings"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Hobbit is set within Tolkien's fictional universe that follows the quest of the title character, the hobbit, who is a supporting character in an epic high-fantasy novel and one of the best-selling books ever written.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Hobbit",
"Bilbo Baggins",
"The Lord of the Rings"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 4973,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4010
|
Charles Robert Redford Jr. is the first celebrity who won the Cecil B. DeMille Award.
|
{
"evidence": [
"Robert Redford",
"Presidential Medal of Freedom"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nCharles Robert Redford Jr. is the first celebrity who won the Cecil B. DeMille Award.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Robert Redford",
"Presidential Medal of Freedom"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4974,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
34338
|
In 2011, KKOV (1550Hz), a station owned by Intelli LLC, had changed its format from "talk" to adult standards, while in the following year, it was switched back to talk, then branded as "Talk 1550."
|
{
"evidence": [
"KKOV_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:\nIn 2011, KKOV (1550Hz), a station owned by Intelli LLC, had changed its format from \"talk\" to adult standards, while in the following year, it was switched back to talk, then branded as \"Talk 1550.\"\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"KKOV_sentence_11"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4975,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.397",
"value": "?search="
},
{
"operation": "search",
"time": "39.737",
"value": "KKOV"
},
{
"operation": "Now on",
"time": "40.986",
"value": "KKOV"
},
{
"operation": "Highlighting",
"time": "77.708",
"value": "KKOV_sentence_11"
},
{
"operation": "finish",
"time": "121.458",
"value": "finish"
}
]
}
|
exfever_2280
|
A Few Good Men was the twelfth film released by Columbia Pictures.
|
{
"evidence": [
"A Few Good Men",
"Andrew Scheinman",
"A Few Good Men"
],
"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:\nA Few Good Men was the twelfth film released by Columbia Pictures.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"A Few Good Men",
"Andrew Scheinman",
"A Few Good Men"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4976,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_703
|
Harry Styles is a French singer-songwriter who released his self-titled debut solo album through an American record label owned by Sony Music Entertainment.
|
{
"evidence": [
"Harry Styles",
"Columbia Records"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHarry Styles is a French singer-songwriter who released his self-titled debut solo album through an American record label owned by Sony Music Entertainment.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Harry Styles",
"Columbia Records"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4977,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4599
|
Philip V was the youngest king of Spain.
|
{
"evidence": [
"Philip V of Spain",
"Louis, Grand Dauphin"
],
"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:\nPhilip V was the youngest king of Spain.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Philip V of Spain",
"Louis, Grand Dauphin"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4978,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
34228
|
Rokach's first public position was council member in Jaffa, then in 1922, he was elected to the city council of Tel Aviv, representing neither the United Centrist List party nor the Jewish neighborhoods of Neve Tzedek.
|
{
"evidence": [
"Israel Rokach_sentence_12"
],
"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:\nRokach's first public position was council member in Jaffa, then in 1922, he was elected to the city council of Tel Aviv, representing neither the United Centrist List party nor the Jewish neighborhoods of Neve Tzedek.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Israel Rokach_sentence_12"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4979,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.236",
"value": "?search="
},
{
"operation": "search",
"time": "17.996",
"value": "Rokach"
},
{
"operation": "Now on",
"time": "18.903",
"value": "Rokeach"
},
{
"operation": "hyperlink",
"time": "33.788",
"value": "Israel Rokach"
},
{
"operation": "Now on",
"time": "34.359",
"value": "Israel Rokach"
},
{
"operation": "Highlighting",
"time": "79.412",
"value": "Israel Rokach_sentence_11"
},
{
"operation": "Highlighting",
"time": "94.932",
"value": "Israel Rokach_sentence_12"
},
{
"operation": "Highlighting deleted",
"time": "100.644",
"value": "Israel Rokach_sentence_11"
},
{
"operation": "finish",
"time": "107.028",
"value": "finish"
}
]
}
|
exfever_1003
|
Leslie Howard directed his last film, The Gentle Sex, in 1943.
|
{
"evidence": [
"Leslie Howard (actor)",
"Gone with the Wind (film)"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLeslie Howard directed his last film, The Gentle Sex, in 1943.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Leslie Howard (actor)",
"Gone with the Wind (film)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4980,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
9544
|
Give a Little was released a year after it was announced.
|
{
"evidence": [
"Give a Little_sentence_2",
"Give a Little_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:\nGive a Little was released a year after it was announced.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Give a Little_sentence_2",
"Give a Little_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4981,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.88",
"value": "?search="
},
{
"operation": "search",
"time": "9.866",
"value": "Give a Little"
},
{
"operation": "Now on",
"time": "10.748",
"value": "Give a Little"
},
{
"operation": "Highlighting",
"time": "19.275",
"value": "Give a Little_sentence_2"
},
{
"operation": "Highlighting",
"time": "19.862",
"value": "Give a Little_sentence_3"
},
{
"operation": "finish",
"time": "37.548",
"value": "finish"
}
]
}
|
exfever_2682
|
The phalanx was composed of heavy infantry armed with a pole weapon consisting of shaft wood or may be made of durable material, such as a chemical element with the symbol Fe and atomic number 24.
|
{
"evidence": [
"Phalanx",
"Spear",
"Iron"
],
"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 phalanx was composed of heavy infantry armed with a pole weapon consisting of shaft wood or may be made of durable material, such as a chemical element with the symbol Fe and atomic number 24.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Phalanx",
"Spear",
"Iron"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 4982,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
28410
|
Cases of small intestine neuroendocrine tumor can be diagnosed by urine sampling.
|
{
"evidence": [
"Small intestine neuroendocrine tumor_sentence_5",
"Small intestine neuroendocrine tumor_sentence_6",
"Small intestine neuroendocrine tumor_sentence_7"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nCases of small intestine neuroendocrine tumor can be diagnosed by urine sampling.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Small intestine neuroendocrine tumor_sentence_5",
"Small intestine neuroendocrine tumor_sentence_6",
"Small intestine neuroendocrine tumor_sentence_7"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 4983,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.218",
"value": "?search="
},
{
"operation": "search",
"time": "15.076",
"value": "Small intestine neuroendocrine tumor"
},
{
"operation": "Now on",
"time": "16.365",
"value": "Small intestine neuroendocrine tumor"
},
{
"operation": "Page search",
"time": "47.539",
"value": "urine"
},
{
"operation": "Highlighting",
"time": "59.157",
"value": "Small intestine neuroendocrine tumor_sentence_5"
},
{
"operation": "Highlighting",
"time": "64.191",
"value": "Small intestine neuroendocrine tumor_sentence_6"
},
{
"operation": "hyperlink",
"time": "127.147",
"value": "Tachykinin peptides"
},
{
"operation": "Now on",
"time": "128.384",
"value": "Tachykinin peptides"
},
{
"operation": "Page search",
"time": "134.984",
"value": "urine"
},
{
"operation": "back-button-clicked",
"time": "148.9",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "149.078",
"value": "Small intestine neuroendocrine tumor"
},
{
"operation": "Highlighting",
"time": "179.69",
"value": "Small intestine neuroendocrine tumor_sentence_7"
},
{
"operation": "finish",
"time": "184.929",
"value": "finish"
}
]
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.