id
stringlengths 1
12
| question
stringlengths 22
566
| golden_answers
dict | data_source
stringclasses 2
values | prompt
listlengths 1
1
| ability
stringclasses 6
values | reward_model
dict | extra_info
dict | metadata
dict |
|---|---|---|---|---|---|---|---|---|
exfever_2188
|
Cheyenne Jackson was also nominated for the Outer Critics Circle Award - Outstanding Performance by a Lead Actor.
|
{
"evidence": [
"Cheyenne Jackson",
"Thoroughly Modern Millie (musical)",
"Thoroughly Modern Millie"
],
"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:\nCheyenne Jackson was also nominated for the Outer Critics Circle Award - Outstanding Performance by a Lead Actor.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Cheyenne Jackson",
"Thoroughly Modern Millie (musical)",
"Thoroughly Modern Millie"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 984,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
47535
|
Richard Wallace "Rick" Bloomingdale is a labor union activist from Pittsburg, Pennsylvania who has served as Secretary-Treasurer and President of the Pennsylvania AFL-CIO since 2010
|
{
"evidence": [
"Richard Bloomingdale_sentence_0",
"Richard Bloomingdale_sentence_2",
"Richard Bloomingdale_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:\nRichard Wallace \"Rick\" Bloomingdale is a labor union activist from Pittsburg, Pennsylvania who has served as Secretary-Treasurer and President of the Pennsylvania AFL-CIO since 2010\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Richard Bloomingdale_sentence_0",
"Richard Bloomingdale_sentence_2",
"Richard Bloomingdale_sentence_3"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 985,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.776",
"value": "?search="
},
{
"operation": "search",
"time": "40.331",
"value": "Contains...Richard Wallace \"Rick\" Bloomingdale"
},
{
"operation": "Now on",
"time": "41.468",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "43.981",
"value": "Richard Bloomingdale"
},
{
"operation": "Now on",
"time": "44.573",
"value": "Richard Bloomingdale"
},
{
"operation": "Highlighting",
"time": "75.706",
"value": "Richard Bloomingdale_sentence_0"
},
{
"operation": "Highlighting",
"time": "156.879",
"value": "Richard Bloomingdale_sentence_2"
},
{
"operation": "Highlighting",
"time": "162.529",
"value": "Richard Bloomingdale_sentence_3"
},
{
"operation": "finish",
"time": "175.862",
"value": "finish"
}
]
}
|
30832
|
Above the tree line the open slopes are used as grazing grounds for sheep, goats, cows and horses; then in summer, flocks of sheep are especially common around the Steinbergstein (a 7,267 ft. high mountain in the Kitzbühel Alps in Austria).
|
{
"evidence": [
"Steinbergstein_sentence_0",
"Steinbergstein_sentence_4",
"Steinbergstein_sentence_5"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAbove the tree line the open slopes are used as grazing grounds for sheep, goats, cows and horses; then in summer, flocks of sheep are especially common around the Steinbergstein (a 7,267 ft. high mountain in the Kitzbühel Alps in Austria).\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Steinbergstein_sentence_0",
"Steinbergstein_sentence_4",
"Steinbergstein_sentence_5"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 986,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.848",
"value": "?search="
},
{
"operation": "search",
"time": "18.093",
"value": "Steinbergstein"
},
{
"operation": "Now on",
"time": "19.041",
"value": "Steinbergstein"
},
{
"operation": "Highlighting",
"time": "21.674",
"value": "Steinbergstein_sentence_0"
},
{
"operation": "Highlighting",
"time": "29.626",
"value": "Steinbergstein_sentence_4"
},
{
"operation": "Highlighting",
"time": "39.354",
"value": "Steinbergstein_sentence_5"
},
{
"operation": "finish",
"time": "51.766",
"value": "finish"
}
]
}
|
exfever_6216
|
Clinton Eastwood Jr. is an American actor who rose to international fame with his role as a fictional character and protagonist of the Dirty Harry film series.
|
{
"evidence": [
"Clint Eastwood",
"Dirty Harry (character)"
],
"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:\nClinton Eastwood Jr. is an American actor who rose to international fame with his role as a fictional character and protagonist of the Dirty Harry film series.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Clint Eastwood",
"Dirty Harry (character)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 987,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2668
|
Goat Island is a small island in the Niagara River, located in the southwest corner of Niagara Falls in a state in the northeastern United Kingdom.
|
{
"evidence": [
"Goat Island (New York)",
"New York (state)"
],
"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:\nGoat Island is a small island in the Niagara River, located in the southwest corner of Niagara Falls in a state in the northeastern United Kingdom.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Goat Island (New York)",
"New York (state)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 988,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
45047
|
Kéba Mbaye was the chairman of the International Olympic Committee from 1973 to 2002 and also served as president of the Court of Arbitration for Sport, the highest Olympic tribunal.
|
{
"evidence": [
"Kéba Mbaye_sentence_2",
"Kéba Mbaye_sentence_4"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nKéba Mbaye was the chairman of the International Olympic Committee from 1973 to 2002 and also served as president of the Court of Arbitration for Sport, the highest Olympic tribunal.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Kéba Mbaye_sentence_2",
"Kéba Mbaye_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 989,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.129",
"value": "?search="
},
{
"operation": "search",
"time": "29.471",
"value": "Kéba Mbaye"
},
{
"operation": "Now on",
"time": "30.309",
"value": "Kéba Mbaye"
},
{
"operation": "Page search",
"time": "43.558",
"value": "chairman"
},
{
"operation": "Highlighting",
"time": "51.303",
"value": "Kéba Mbaye_sentence_0"
},
{
"operation": "Page search",
"time": "83.593",
"value": "1973"
},
{
"operation": "Highlighting",
"time": "88.257",
"value": "Kéba Mbaye_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "96.373",
"value": "Kéba Mbaye_sentence_0"
},
{
"operation": "Highlighting",
"time": "111.447",
"value": "Kéba Mbaye_sentence_4"
},
{
"operation": "finish",
"time": "268.075",
"value": "finish"
}
]
}
|
16581
|
Švenčionys Ghetto had 566 Jews - some of which established contracts with active Soviet partisans, who frustrated economical exploitation plans of the Germans towards Soviet territories, while the others joined Jewish partisans.
|
{
"evidence": [
"Švenčionys Ghetto_sentence_0",
"Švenčionys Ghetto_sentence_2",
"Švenčionys Ghetto_sentence_14",
"Švenčionys Ghetto_sentence_15",
"Švenčionys Ghetto_sentence_16"
],
"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:\nŠvenčionys Ghetto had 566 Jews - some of which established contracts with active Soviet partisans, who frustrated economical exploitation plans of the Germans towards Soviet territories, while the others joined Jewish partisans.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Švenčionys Ghetto_sentence_0",
"Švenčionys Ghetto_sentence_2",
"Švenčionys Ghetto_sentence_14",
"Švenčionys Ghetto_sentence_15",
"Švenčionys Ghetto_sentence_16"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 990,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.15",
"value": "?search="
},
{
"operation": "search",
"time": "42.804",
"value": "Švenčionys Ghetto"
},
{
"operation": "Now on",
"time": "43.745",
"value": "Švenčionys Ghetto"
},
{
"operation": "Highlighting",
"time": "45.773",
"value": "Švenčionys Ghetto_sentence_0"
},
{
"operation": "Highlighting",
"time": "54.809",
"value": "Švenčionys Ghetto_sentence_2"
},
{
"operation": "Page search",
"time": "65.839",
"value": "566"
},
{
"operation": "Highlighting",
"time": "68.477",
"value": "Švenčionys Ghetto_sentence_14"
},
{
"operation": "Highlighting",
"time": "81.125",
"value": "Švenčionys Ghetto_sentence_15"
},
{
"operation": "Page search",
"time": "116.059",
"value": "economical exploitation"
},
{
"operation": "Page search",
"time": "165.649",
"value": "economical exploitation"
},
{
"operation": "Page search",
"time": "177.556",
"value": "Jewish partisans"
},
{
"operation": "Highlighting",
"time": "179.933",
"value": "Švenčionys Ghetto_sentence_16"
},
{
"operation": "finish",
"time": "218.774",
"value": "finish"
}
]
}
|
exfever_584
|
Tamera Mowry is the first American actress to gain fame for her teen role as Tamera Campbell on the ABC/WB sitcom Sister.
|
{
"evidence": [
"Tamera Mowry",
"Medical drama"
],
"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:\nTamera Mowry is the first American actress to gain fame for her teen role as Tamera Campbell on the ABC/WB sitcom Sister.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tamera Mowry",
"Medical drama"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 991,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
94565
|
Vidame de Chartres was put in the Bastille as told by the novel in La Princesse de Clèves, authored by Madame de La Fayette.
|
{
"evidence": [
"La Princesse de Clèves_sentence_0",
"La Princesse de Clèves_sentence_2"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nVidame de Chartres was put in the Bastille as told by the novel in La Princesse de Clèves, authored by Madame de La Fayette.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"La Princesse de Clèves_sentence_0",
"La Princesse de Clèves_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 992,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.26",
"value": "?search="
},
{
"operation": "search",
"time": "16.907",
"value": "La Princesse de Clèves"
},
{
"operation": "Now on",
"time": "17.817",
"value": "La Princesse de Clèves"
},
{
"operation": "Highlighting",
"time": "24.719",
"value": "La Princesse de Clèves_sentence_0"
},
{
"operation": "Page search",
"time": "34.354",
"value": "Madame"
},
{
"operation": "Highlighting",
"time": "38.702",
"value": "La Princesse de Clèves_sentence_2"
},
{
"operation": "Page search",
"time": "57.033",
"value": "Bastille"
},
{
"operation": "Page search",
"time": "97.123",
"value": "Chartres"
},
{
"operation": "page-search-reset",
"time": "110.02",
"value": "page-search-reset"
},
{
"operation": "Page search",
"time": "124.631",
"value": "put in"
},
{
"operation": "Highlighting deleted",
"time": "148.587",
"value": "La Princesse de Clèves_sentence_2"
},
{
"operation": "Highlighting",
"time": "149.875",
"value": "La Princesse de Clèves_sentence_2"
},
{
"operation": "finish",
"time": "157.25",
"value": "finish"
}
]
}
|
exfever_6180
|
Dumb and Dumber is a 1994 American buddy comedy film directed by Peter Farrelly starring a Canadian and American actor and comedian who has a recurring role in an American sketch comedy television series that originally ran on Fox from April 15, 1990 to May 19, 1998.
|
{
"evidence": [
"Dumb and Dumber",
"Jim Carrey",
"In Living Color"
],
"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:\nDumb and Dumber is a 1994 American buddy comedy film directed by Peter Farrelly starring a Canadian and American actor and comedian who has a recurring role in an American sketch comedy television series that originally ran on Fox from April 15, 1990 to May 19, 1998.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dumb and Dumber",
"Jim Carrey",
"In Living Color"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 993,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
23457
|
Dinamo scored 35 more wins than Rijeka in the League competition in the Yugoslav championship (1946-1991).
|
{
"evidence": [
"GNK Dinamo Zagreb_sentence_2"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nDinamo scored 35 more wins than Rijeka in the League competition in the Yugoslav championship (1946-1991).\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"GNK Dinamo Zagreb_sentence_2"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 994,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.777",
"value": "?search="
},
{
"operation": "search",
"time": "19.034",
"value": "Yugoslav championship (1946-1991)"
},
{
"operation": "Now on",
"time": "20.936",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "42.481",
"value": "SD Željezničar"
},
{
"operation": "Now on",
"time": "43.664",
"value": "SD Željezničar"
},
{
"operation": "Page search",
"time": "52.437",
"value": "dinamo"
},
{
"operation": "back-button-clicked",
"time": "61.885",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "62.878",
"value": "undefined"
},
{
"operation": "search",
"time": "81.569",
"value": "Yugoslav championship (1946-1991)"
},
{
"operation": "Now on",
"time": "82.537",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "87.505",
"value": "SD Željezničar"
},
{
"operation": "Now on",
"time": "88.219",
"value": "SD Željezničar"
},
{
"operation": "Page search",
"time": "96.063",
"value": "1946-1991"
},
{
"operation": "hyperlink",
"time": "107.505",
"value": "Yugoslav championship"
},
{
"operation": "Now on",
"time": "108.351",
"value": "YUBA Liga#Past Yugoslav Cup Champions"
},
{
"operation": "Page search",
"time": "124.634",
"value": "dinamo"
},
{
"operation": "back-button-clicked",
"time": "137.917",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "138.517",
"value": "SD Željezničar"
},
{
"operation": "search",
"time": "181.173",
"value": "Rijeka"
},
{
"operation": "Now on",
"time": "182.584",
"value": "Rijeka"
},
{
"operation": "hyperlink",
"time": "189.784",
"value": "Rijeka (disambiguation)"
},
{
"operation": "Now on",
"time": "190.812",
"value": "Rijeka (disambiguation)"
},
{
"operation": "Highlighting",
"time": "199.184",
"value": "Rijeka (disambiguation)_item_0_0"
},
{
"operation": "Highlighting",
"time": "199.188",
"value": "Rijeka (disambiguation)_title"
},
{
"operation": "hyperlink",
"time": "203.89",
"value": "HNK Rijeka"
},
{
"operation": "Now on",
"time": "205.49",
"value": "HNK Rijeka"
},
{
"operation": "Page search",
"time": "215.11",
"value": "1946"
},
{
"operation": "Page search",
"time": "265.503",
"value": "dinamo"
},
{
"operation": "search",
"time": "330.331",
"value": "Yugoslav championship"
},
{
"operation": "Now on",
"time": "331.975",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "343.723",
"value": "GNK Dinamo Zagreb"
},
{
"operation": "Now on",
"time": "345.319",
"value": "GNK Dinamo Zagreb"
},
{
"operation": "back-button-clicked",
"time": "365.647",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "366.581",
"value": "undefined"
},
{
"operation": "search",
"time": "411.22",
"value": "Yugoslav championship"
},
{
"operation": "Now on",
"time": "412.011",
"value": "undefined"
},
{
"operation": "search",
"time": "436.359",
"value": "Yugoslav championship (1946-1991)"
},
{
"operation": "Now on",
"time": "437.244",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "449.259",
"value": "Croatia Zagreb era (1991–2000) GNK Dinamo Zagreb_section_5"
},
{
"operation": "Now on",
"time": "450.236",
"value": "GNK Dinamo Zagreb#Croatia Zagreb era .281991.E2.80.932000.29 GNK Dinamo Zagreb section 5"
},
{
"operation": "Page search",
"time": "463.682",
"value": "1946"
},
{
"operation": "Page search",
"time": "510.587",
"value": "Rijeka"
},
{
"operation": "Highlighting",
"time": "577.537",
"value": "GNK Dinamo Zagreb_sentence_2"
},
{
"operation": "finish",
"time": "588.996",
"value": "finish"
}
]
}
|
29787
|
RUVBL2 encodes the second human homologue of the bacterial RuvB gene and is physically linked to the CGB/LHB gene cluster.
|
{
"evidence": [
"RUVBL2_sentence_1",
"RUVBL2_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:\nRUVBL2 encodes the second human homologue of the bacterial RuvB gene and is physically linked to the CGB/LHB gene cluster.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"RUVBL2_sentence_1",
"RUVBL2_sentence_4"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 995,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.268",
"value": "undefined"
},
{
"operation": "search",
"time": "8.611",
"value": "RUVBL2"
},
{
"operation": "Now on",
"time": "9.127",
"value": "RUVBL2"
},
{
"operation": "Highlighting",
"time": "10.926",
"value": "RUVBL2_sentence_1"
},
{
"operation": "Highlighting",
"time": "12.062",
"value": "RUVBL2_sentence_4"
},
{
"operation": "finish",
"time": "18.813",
"value": "finish"
}
]
}
|
exfever_7354
|
Pacific Blue was often described as an American action drama television series about rescuers on bikes.
|
{
"evidence": [
"Baywatch",
"Lifeguard"
],
"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:\nPacific Blue was often described as an American action drama television series about rescuers on bikes.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Baywatch",
"Lifeguard"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 996,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5341
|
Broadchurch is a crime drama television series set in a fictional English town on the coast of a county that is bordered to the west by a county in the South West of England.
|
{
"evidence": [
"Broadchurch",
"Dorset",
"Devon"
],
"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:\nBroadchurch is a crime drama television series set in a fictional English town on the coast of a county that is bordered to the west by a county in the South West of England.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Broadchurch",
"Dorset",
"Devon"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 997,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2174
|
James Tate Ellington appeared in the 2009 Broadway production of The Philanthropist, which starred an American actor and singer whose roles include the Golden Globe-nominated portrayal of the title character in Ferris Bueller's Day Off.
|
{
"evidence": [
"Tate Ellington",
"Matthew Broderick"
],
"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:\nJames Tate Ellington appeared in the 2009 Broadway production of The Philanthropist, which starred an American actor and singer whose roles include the Golden Globe-nominated portrayal of the title character in Ferris Bueller's Day Off.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tate Ellington",
"Matthew Broderick"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 998,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
63167
|
Tomchei Shabbos mission is to assist the Jewish needy and help them find jobs.
|
{
"evidence": [
"Tomchei Shabbos_sentence_2",
"Tomchei Shabbos_sentence_16"
],
"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:\nTomchei Shabbos mission is to assist the Jewish needy and help them find jobs.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Tomchei Shabbos_sentence_2",
"Tomchei Shabbos_sentence_16"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 999,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.269",
"value": "?search="
},
{
"operation": "search",
"time": "11.985",
"value": "Tomchei Shabbos"
},
{
"operation": "Now on",
"time": "13.018",
"value": "Tomchei Shabbos"
},
{
"operation": "Highlighting",
"time": "25.727",
"value": "Tomchei Shabbos_sentence_2"
},
{
"operation": "Highlighting",
"time": "48.741",
"value": "Tomchei Shabbos_sentence_16"
},
{
"operation": "finish",
"time": "84.112",
"value": "finish"
}
]
}
|
39400
|
Art Spector played 169 regular season games with an average of five points per game for one of the most losing team in history from the first year it was founded in 1946 until 1950
|
{
"evidence": [
"Art Spector_sentence_9",
"Boston Celtics_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:\nArt Spector played 169 regular season games with an average of five points per game for one of the most losing team in history from the first year it was founded in 1946 until 1950\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Art Spector_sentence_9",
"Boston Celtics_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1000,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "4.433",
"value": "?search="
},
{
"operation": "search",
"time": "251.15",
"value": "Art Spector"
},
{
"operation": "Now on",
"time": "253.686",
"value": "Art Spector"
},
{
"operation": "Highlighting",
"time": "271.53",
"value": "Art Spector_cell_1_5_2"
},
{
"operation": "Highlighting",
"time": "271.532",
"value": "Art Spector_section_1"
},
{
"operation": "Highlighting",
"time": "271.532",
"value": "Art Spector_section_2"
},
{
"operation": "Highlighting",
"time": "271.534",
"value": "Art Spector_header_cell_1_0_2"
},
{
"operation": "Highlighting",
"time": "280.986",
"value": "Art Spector_cell_1_5_6"
},
{
"operation": "Highlighting",
"time": "280.989",
"value": "Art Spector_header_cell_1_0_6"
},
{
"operation": "hyperlink",
"time": "291.946",
"value": "Boston"
},
{
"operation": "Now on",
"time": "295.673",
"value": "1949–50 Boston Celtics season"
},
{
"operation": "back-button-clicked",
"time": "309.661",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "309.762",
"value": "Art Spector"
},
{
"operation": "hyperlink",
"time": "324.491",
"value": "1946–47"
},
{
"operation": "Now on",
"time": "325.593",
"value": "1946–47 BAA season"
},
{
"operation": "back-button-clicked",
"time": "337.269",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "337.369",
"value": "Art Spector"
},
{
"operation": "hyperlink",
"time": "342.818",
"value": "Boston Celtics"
},
{
"operation": "Now on",
"time": "348.239",
"value": "Boston Celtics"
},
{
"operation": "Highlighting",
"time": "352.027",
"value": "Boston Celtics_sentence_4"
},
{
"operation": "Highlighting",
"time": "400.575",
"value": "Boston Celtics_sentence_5"
},
{
"operation": "back-button-clicked",
"time": "466.754",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "466.871",
"value": "Art Spector"
},
{
"operation": "Highlighting",
"time": "497.636",
"value": "Art Spector_sentence_9"
},
{
"operation": "hyperlink",
"time": "503.859",
"value": "Boston Celtics"
},
{
"operation": "Now on",
"time": "504.889",
"value": "Boston Celtics"
},
{
"operation": "Highlighting",
"time": "588.867",
"value": "Boston Celtics_sentence_4"
},
{
"operation": "Highlighting",
"time": "593.939",
"value": "Boston Celtics_sentence_5"
},
{
"operation": "finish",
"time": "646.987",
"value": "finish"
}
]
}
|
exfever_8953
|
Gray Davis is an Indian attorney and former politician who holds a Bachelor of Arts in history from a private research university located in the census-designated place of Stanford, California.
|
{
"evidence": [
"Gray Davis",
"Stanford University"
],
"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:\nGray Davis is an Indian attorney and former politician who holds a Bachelor of Arts in history from a private research university located in the census-designated place of Stanford, California.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Gray Davis",
"Stanford University"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1001,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
39839
|
Madonna of the Seven Moons was based by Margery Lawrence's novel in 1931 and Gaumont British bought film rights in 1938.
|
{
"evidence": [
"Madonna of the Seven Moons_sentence_6",
"Madonna of the Seven Moons_sentence_7"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nMadonna of the Seven Moons was based by Margery Lawrence's novel in 1931 and Gaumont British bought film rights in 1938.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Madonna of the Seven Moons_sentence_6",
"Madonna of the Seven Moons_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1002,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.109",
"value": "?search="
},
{
"operation": "search",
"time": "11.453",
"value": "Madonna of the Seven Moons"
},
{
"operation": "Now on",
"time": "12.207",
"value": "Madonna of the Seven Moons"
},
{
"operation": "Highlighting",
"time": "24.048",
"value": "Madonna of the Seven Moons_sentence_6"
},
{
"operation": "Highlighting",
"time": "33.342",
"value": "Madonna of the Seven Moons_sentence_7"
},
{
"operation": "finish",
"time": "41.398",
"value": "finish"
}
]
}
|
exfever_1485
|
William Jefferson Clinton is the husband of an African politician who served to represent New York from 2001 to 2009, in the upper chamber of the United States Congress, with the House of Representatives being the lower chamber.
|
{
"evidence": [
"Bill Clinton",
"Hillary Clinton",
"United States Senate"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nWilliam Jefferson Clinton is the husband of an African politician who served to represent New York from 2001 to 2009, in the upper chamber of the United States Congress, with the House of Representatives being the lower chamber.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Bill Clinton",
"Hillary Clinton",
"United States Senate"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1003,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5461
|
The X Factor is a British television dance competition that saw the creation of an English-Irish pop boy band formed in London.
|
{
"evidence": [
"The X Factor (UK series 7)",
"One Direction"
],
"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 X Factor is a British television dance competition that saw the creation of an English-Irish pop boy band formed in London.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The X Factor (UK series 7)",
"One Direction"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1004,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1962
|
The Outcasts is a 2017 American teen film whose genre emphasizes humor designed to make the audience laugh through the amusement.
|
{
"evidence": [
"The Outcasts (2017 film)",
"Comedy film"
],
"label": "SUPPORT"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe Outcasts is a 2017 American teen film whose genre emphasizes humor designed to make the audience laugh through the amusement.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Outcasts (2017 film)",
"Comedy film"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1005,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
37266
|
After losing the first five games of the 1931–32 regular season, the Toronto Maple Leafs fired head coach Art Duncan and replaced him with Dick Irvin, who had an 8–3–2 record in his first month behind the bench.
|
{
"evidence": [
"Conn Smythe_sentence_114"
],
"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:\nAfter losing the first five games of the 1931–32 regular season, the Toronto Maple Leafs fired head coach Art Duncan and replaced him with Dick Irvin, who had an 8–3–2 record in his first month behind the bench.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Conn Smythe_sentence_114"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1006,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.273",
"value": "?search="
},
{
"operation": "search",
"time": "12.126",
"value": "After losing the first five games of the 1931–32 regular season, the Toronto Maple Leafs"
},
{
"operation": "Now on",
"time": "16.003",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "21.387",
"value": "Invested in and operates (owned by the City of Toronto) Maple Leaf Sports & Entertainment_section_19"
},
{
"operation": "Now on",
"time": "23.592",
"value": "Maple Leaf Sports & Entertainment#Invested in and operates .28owned by the City of Toronto.29 Maple Leaf Sports .26 Entertainment section 19"
},
{
"operation": "search",
"time": "38.419",
"value": "Toronto Maple Leafs fired head coach Art Duncan"
},
{
"operation": "Now on",
"time": "39.574",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "48.435",
"value": "Smythe forms the Maple Leafs Conn Smythe_section_3"
},
{
"operation": "Now on",
"time": "49.957",
"value": "Conn Smythe#Smythe forms the Maple Leafs Conn Smythe section 3"
},
{
"operation": "Page search",
"time": "65.179",
"value": "fired"
},
{
"operation": "Highlighting",
"time": "84.006",
"value": "Conn Smythe_sentence_114"
},
{
"operation": "hyperlink",
"time": "97.017",
"value": "Dick Irvin"
},
{
"operation": "Now on",
"time": "98.666",
"value": "Dick Irvin"
},
{
"operation": "back-button-clicked",
"time": "107.797",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "108.026",
"value": "Conn Smythe#Smythe forms the Maple Leafs Conn Smythe section 3"
},
{
"operation": "Page search",
"time": "111.02",
"value": "8–3–2"
},
{
"operation": "Highlighting deleted",
"time": "138.096",
"value": "Conn Smythe_sentence_114"
},
{
"operation": "hyperlink",
"time": "138.607",
"value": "Dick Irvin"
},
{
"operation": "Now on",
"time": "139.093",
"value": "Dick Irvin"
},
{
"operation": "back-button-clicked",
"time": "140.265",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "140.475",
"value": "Conn Smythe#Smythe forms the Maple Leafs Conn Smythe section 3"
},
{
"operation": "Highlighting",
"time": "141.292",
"value": "Conn Smythe_sentence_114"
},
{
"operation": "hyperlink",
"time": "141.788",
"value": "Dick Irvin"
},
{
"operation": "Now on",
"time": "142.724",
"value": "Dick Irvin"
},
{
"operation": "Page search",
"time": "145.565",
"value": "8–3–2"
},
{
"operation": "back-button-clicked",
"time": "179.154",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "179.344",
"value": "Conn Smythe#Smythe forms the Maple Leafs Conn Smythe section 3"
},
{
"operation": "Page search",
"time": "203.927",
"value": "behind the bench"
},
{
"operation": "search",
"time": "217.742",
"value": "who had an 8–3–2 record in his first month behind the bench"
},
{
"operation": "Now on",
"time": "225.493",
"value": "undefined"
},
{
"operation": "finish",
"time": "231.203",
"value": "finish"
}
]
}
|
8118
|
The 2019 CCM/ECHL All-Star Classic, which retained the four team, 3-on-3 player tournament style used the previous season, but featured two teams made from the Toledo Walleye (named Team Hooks and Team Fins) and one team for each conference's All-Star players, was held on January 21, 2019, at the Huntington Center in Toledo, Ohio.
|
{
"evidence": [
"2018–19 ECHL season_sentence_9",
"2018–19 ECHL season_sentence_10"
],
"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 2019 CCM/ECHL All-Star Classic, which retained the four team, 3-on-3 player tournament style used the previous season, but featured two teams made from the Toledo Walleye (named Team Hooks and Team Fins) and one team for each conference's All-Star players, was held on January 21, 2019, at the Huntington Center in Toledo, Ohio.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"2018–19 ECHL season_sentence_9",
"2018–19 ECHL season_sentence_10"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1007,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.08",
"value": "?search="
},
{
"operation": "search",
"time": "34.1",
"value": "2019 CCM/ECHL all star classic"
},
{
"operation": "Now on",
"time": "35.546",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "46.073",
"value": "Huntington Center (Toledo, Ohio)"
},
{
"operation": "Now on",
"time": "46.864",
"value": "Huntington Center (Toledo, Ohio)"
},
{
"operation": "hyperlink",
"time": "60.32",
"value": "2019 CCM/ECHL All Star Classic"
},
{
"operation": "Now on",
"time": "61.125",
"value": "ECHL All-Star Game"
},
{
"operation": "hyperlink",
"time": "96.447",
"value": "2019"
},
{
"operation": "Now on",
"time": "97.455",
"value": "2018–19 ECHL season"
},
{
"operation": "hyperlink",
"time": "118.105",
"value": "1.6 All-star game 2018–19 ECHL season_section_6"
},
{
"operation": "Highlighting",
"time": "126.766",
"value": "2018–19 ECHL season_sentence_9"
},
{
"operation": "Highlighting",
"time": "131.413",
"value": "2018–19 ECHL season_sentence_10"
},
{
"operation": "finish",
"time": "161.131",
"value": "finish"
}
]
}
|
exfever_7091
|
Mark Bomback is an American screenwriter who graduated from a private liberal arts university in Middletown, Connecticut.
|
{
"evidence": [
"Mark Bomback",
"Wesleyan University"
],
"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:\nMark Bomback is an American screenwriter who graduated from a private liberal arts university in Middletown, Connecticut.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Mark Bomback",
"Wesleyan University"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1008,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
92471
|
Racewalking, in which one foot must appear to be in contact with the ground at all times, is a sport that Sweden's Siv Gustavsson has competed in.
|
{
"evidence": [
"Racewalking_sentence_1",
"Siv Gustavsson_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:\nRacewalking, in which one foot must appear to be in contact with the ground at all times, is a sport that Sweden's Siv Gustavsson has competed in.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Racewalking_sentence_1",
"Siv Gustavsson_sentence_0"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1009,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "7.291",
"value": "?search="
},
{
"operation": "search",
"time": "10.79",
"value": "Racewalking"
},
{
"operation": "Now on",
"time": "13.358",
"value": "Racewalking"
},
{
"operation": "Highlighting",
"time": "21.473",
"value": "Racewalking_sentence_1"
},
{
"operation": "search",
"time": "36.515",
"value": "Siv Gustavsson"
},
{
"operation": "Now on",
"time": "37.661",
"value": "Siv Gustavsson"
},
{
"operation": "Highlighting",
"time": "43.063",
"value": "Siv Gustavsson_sentence_0"
},
{
"operation": "finish",
"time": "53.251",
"value": "finish"
}
]
}
|
68146
|
Wilhelm Wahlforss's was a Finnish engineer that worked in Fiskars, where new workplace in the rolling mill, where he was appointed workshop technical manager Albert Lindsay von Julin (Fiskar's manager for almost 40 years), but was not in favor of him so he did not last long.
|
{
"evidence": [
"Wilhelm Wahlforss_sentence_28"
],
"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:\nWilhelm Wahlforss's was a Finnish engineer that worked in Fiskars, where new workplace in the rolling mill, where he was appointed workshop technical manager Albert Lindsay von Julin (Fiskar's manager for almost 40 years), but was not in favor of him so he did not last long.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Wilhelm Wahlforss_sentence_28"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1010,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.002",
"value": "?search="
},
{
"operation": "search",
"time": "9.53",
"value": "Wilhelm Wahlforss"
},
{
"operation": "Now on",
"time": "11.045",
"value": "Wilhelm Wahlforss"
},
{
"operation": "Highlighting",
"time": "22.424",
"value": "Wilhelm Wahlforss_sentence_0"
},
{
"operation": "hyperlink",
"time": "44.648",
"value": "3.2 Fiskars Wilhelm Wahlforss_section_4"
},
{
"operation": "Highlighting",
"time": "50.263",
"value": "Wilhelm Wahlforss_sentence_27"
},
{
"operation": "Highlighting",
"time": "72.366",
"value": "Wilhelm Wahlforss_sentence_28"
},
{
"operation": "Highlighting deleted",
"time": "73.87",
"value": "Wilhelm Wahlforss_sentence_27"
},
{
"operation": "Highlighting deleted",
"time": "81.285",
"value": "Wilhelm Wahlforss_sentence_0"
},
{
"operation": "finish",
"time": "86.641",
"value": "finish"
}
]
}
|
84762
|
Meyer's missions in the Amazonian rain-forest, wherein he was in close contact with the "People of Nature" or the Índios, lead him to combine his expertise in neuro-psychology neither in the field of linguistic nor in cultural diversity, with his experience in biological diversity, and its preservation by bio-technologies.
|
{
"evidence": [
"Mario Christian Meyer_sentence_19"
],
"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:\nMeyer's missions in the Amazonian rain-forest, wherein he was in close contact with the \"People of Nature\" or the Índios, lead him to combine his expertise in neuro-psychology neither in the field of linguistic nor in cultural diversity, with his experience in biological diversity, and its preservation by bio-technologies.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Mario Christian Meyer_sentence_19"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1011,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.797",
"value": "?search="
},
{
"operation": "search",
"time": "17.137",
"value": "Contains...Meyer's missions in the Amazonian rain-forest"
},
{
"operation": "Now on",
"time": "18.618",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "22.361",
"value": "Mario Christian Meyer"
},
{
"operation": "Now on",
"time": "23.323",
"value": "Mario Christian Meyer"
},
{
"operation": "Highlighting",
"time": "89.558",
"value": "Mario Christian Meyer_sentence_19"
},
{
"operation": "finish",
"time": "97.927",
"value": "finish"
}
]
}
|
exfever_3824
|
Love is the fourth romantic comedy series on Netflix.
|
{
"evidence": [
"Gillian Jacobs",
"Netflix"
],
"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:\nLove is the fourth romantic comedy series on Netflix.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Gillian Jacobs",
"Netflix"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1012,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
20736
|
KGBB(103.9 FM, "Bob FM") owned by Robert Adelman was first signed on in May 1990 with the transmitter located on El Paso Peak.
|
{
"evidence": [
"KGBB_sentence_2"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nKGBB(103.9 FM, \"Bob FM\") owned by Robert Adelman was first signed on in May 1990 with the transmitter located on El Paso Peak.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"KGBB_sentence_2"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1013,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.657",
"value": "?search="
},
{
"operation": "search",
"time": "33.743",
"value": "KGBB"
},
{
"operation": "Now on",
"time": "34.869",
"value": "KGBB"
},
{
"operation": "Highlighting",
"time": "53.723",
"value": "KGBB_sentence_0"
},
{
"operation": "Highlighting",
"time": "71.563",
"value": "KGBB_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "85.774",
"value": "KGBB_sentence_4"
},
{
"operation": "Highlighting",
"time": "86.79",
"value": "KGBB_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "87.965",
"value": "KGBB_sentence_0"
},
{
"operation": "finish",
"time": "133.457",
"value": "finish"
}
]
}
|
exfever_7042
|
Ocean's Eight is a Canadian heist comedy film that focuses on more details and different aspects from the original work and indicates a substantial change in narrative viewpoint and activity from particular elements and techniques of writing center the story on the action or experience of characters rather than relating to a matter in a dry narration.
|
{
"evidence": [
"Ocean's Eight",
"Spin-off (media)",
"Narrative thread"
],
"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:\nOcean's Eight is a Canadian heist comedy film that focuses on more details and different aspects from the original work and indicates a substantial change in narrative viewpoint and activity from particular elements and techniques of writing center the story on the action or experience of characters rather than relating to a matter in a dry narration.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Ocean's Eight",
"Spin-off (media)",
"Narrative thread"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1014,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5800
|
Elon Musk is the wealthiest person who bought Twitter.
|
{
"evidence": [
"Twitter",
"Character (computing)",
"Grapheme"
],
"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:\nElon Musk is the wealthiest person who bought Twitter.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Twitter",
"Character (computing)",
"Grapheme"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1015,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2036
|
Anneliese Louise van der Pol is a Dutch-born American actress and singer who has recorded several songs for a Canadian multinational entertainment and media conglomerate headquartered in Burbank, California.
|
{
"evidence": [
"Anneliese van der Pol",
"The Walt Disney Company"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nAnneliese Louise van der Pol is a Dutch-born American actress and singer who has recorded several songs for a Canadian multinational entertainment and media conglomerate headquartered in Burbank, California.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Anneliese van der Pol",
"The Walt Disney Company"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1016,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
86175
|
In the novel Five Smooth Stones, David chooses to stay in New Orleans after falling out with Sara Kent.
|
{
"evidence": [
"Five Smooth Stones_sentence_6",
"Five Smooth Stones_sentence_4"
],
"label": "NOT ENOUGH INFO"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nIn the novel Five Smooth Stones, David chooses to stay in New Orleans after falling out with Sara Kent.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Five Smooth Stones_sentence_6",
"Five Smooth Stones_sentence_4"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1017,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.539",
"value": "undefined"
},
{
"operation": "search",
"time": "15.859",
"value": "Five Smooth Stones"
},
{
"operation": "Now on",
"time": "16.631",
"value": "Five Smooth Stones"
},
{
"operation": "Page search",
"time": "24.635",
"value": "new orleans"
},
{
"operation": "Highlighting",
"time": "27.346",
"value": "Five Smooth Stones_sentence_6"
},
{
"operation": "hyperlink",
"time": "34.416",
"value": "New Orleans"
},
{
"operation": "Now on",
"time": "35.741",
"value": "New Orleans"
},
{
"operation": "back-button-clicked",
"time": "36.779",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "36.897",
"value": "Five Smooth Stones"
},
{
"operation": "Page search",
"time": "46.064",
"value": "new orleans"
},
{
"operation": "Page search",
"time": "64.639",
"value": "sara kent"
},
{
"operation": "Highlighting",
"time": "67.136",
"value": "Five Smooth Stones_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "76.256",
"value": "Five Smooth Stones_sentence_4"
},
{
"operation": "Page search",
"time": "86.754",
"value": "sara"
},
{
"operation": "search",
"time": "185.331",
"value": "Contains...In the novel Five Smooth Stones, David chooses to stay in New Orleans after falling out with Sara Kent."
},
{
"operation": "Now on",
"time": "187.347",
"value": "undefined"
},
{
"operation": "search",
"time": "204.877",
"value": "Contains...David chooses to stay in New Orleans after falling out with Sara Kent"
},
{
"operation": "Now on",
"time": "205.755",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "243.889",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "244.36",
"value": "undefined"
},
{
"operation": "back-button-clicked",
"time": "245.391",
"value": "back-button-clicked"
},
{
"operation": "back-button-clicked",
"time": "245.395",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "245.448",
"value": "Five Smooth Stones"
},
{
"operation": "Highlighting",
"time": "263.922",
"value": "Five Smooth Stones_sentence_4"
},
{
"operation": "finish",
"time": "281.152",
"value": "finish"
}
]
}
|
exfever_2545
|
Coca-Cola used coca leaf extract in its products from September 1885 until about December 1903.
|
{
"evidence": [
"Coca",
"Psychoactive plant"
],
"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:\nCoca-Cola used coca leaf extract in its products from September 1885 until about December 1903.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Coca",
"Psychoactive plant"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1018,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4314
|
Dangerous Beauty is an Asian biographical movie directed by Marshall Herskovitz, whose genre is a category of narrative fiction intended to be more serious than humorous in tone.
|
{
"evidence": [
"Dangerous Beauty",
"Drama (film and television)"
],
"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:\nDangerous Beauty is an Asian biographical movie directed by Marshall Herskovitz, whose genre is a category of narrative fiction intended to be more serious than humorous in tone.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Dangerous Beauty",
"Drama (film and television)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1019,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_2759
|
The Hit is a British road crime film directed by Stephen Frears and starring the English actor who rose to prominence in the 1980s through films and television shows.
|
{
"evidence": [
"The Hit (1984 film)",
"Tim Roth"
],
"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 Hit is a British road crime film directed by Stephen Frears and starring the English actor who rose to prominence in the 1980s through films and television shows.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"The Hit (1984 film)",
"Tim Roth"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1020,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
35263
|
Eduard Schmid's tenure was marked by efforts to improve the electricity and water supply, and he was also the one that initiated the further expansion of the middle Isar.
|
{
"evidence": [
"Eduard Schmid_sentence_8",
"Eduard Schmid_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:\nEduard Schmid's tenure was marked by efforts to improve the electricity and water supply, and he was also the one that initiated the further expansion of the middle Isar.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Eduard Schmid_sentence_8",
"Eduard Schmid_sentence_9"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1021,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.816",
"value": "?search="
},
{
"operation": "search",
"time": "4.15",
"value": "Eduard Schmid"
},
{
"operation": "Now on",
"time": "4.882",
"value": "Eduard Schmid"
},
{
"operation": "Highlighting",
"time": "9.567",
"value": "Eduard Schmid_sentence_8"
},
{
"operation": "Highlighting",
"time": "14.269",
"value": "Eduard Schmid_sentence_9"
},
{
"operation": "finish",
"time": "20.049",
"value": "finish"
}
]
}
|
exfever_5381
|
Helike was an ancient Greek polis that was submerged by a series of waves in a water body caused by the displacement of a large volume of water.
|
{
"evidence": [
"Helike",
"Tsunami"
],
"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:\nHelike was an ancient Greek polis that was submerged by a series of waves in a water body caused by the displacement of a large volume of water.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Helike",
"Tsunami"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1022,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
39902
|
Giorgio Moroder, an French composer, asked María Conchita Alonso to see the song "Vamos a Bailar."
|
{
"evidence": [
"Giorgio Moroder_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:\nGiorgio Moroder, an French composer, asked María Conchita Alonso to see the song \"Vamos a Bailar.\"\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Giorgio Moroder_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1023,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.387",
"value": "?search="
},
{
"operation": "search",
"time": "3.814",
"value": "Giorgio Moroder"
},
{
"operation": "Now on",
"time": "5.22",
"value": "Giorgio Moroder"
},
{
"operation": "Highlighting",
"time": "7.126",
"value": "Giorgio Moroder_sentence_0"
},
{
"operation": "finish",
"time": "139.676",
"value": "finish"
}
]
}
|
exfever_4252
|
Lorelai Gilmore was inspired by the mother of Amy Sherman-Palladino.
|
{
"evidence": [
"Lorelai Gilmore",
"Gilmore Girls: A Year in the Life"
],
"label": "NOT ENOUGH INFO"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLorelai Gilmore was inspired by the mother of Amy Sherman-Palladino.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Lorelai Gilmore",
"Gilmore Girls: A Year in the Life"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1024,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
85812
|
Don Juan Manuel (1282-1348), a writer, was from Escalona, which is a municipality located in the north part of the province of Toledo (in Castile-La Mancha, Spain) and the most characteristic building of the town is The Mudéjar-style Castillo-Palacio de Escalona built in the 16th century.
|
{
"evidence": [
"Escalona_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:\nDon Juan Manuel (1282-1348), a writer, was from Escalona, which is a municipality located in the north part of the province of Toledo (in Castile-La Mancha, Spain) and the most characteristic building of the town is The Mudéjar-style Castillo-Palacio de Escalona built in the 16th century.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Escalona_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1025,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.788",
"value": "?search="
},
{
"operation": "search",
"time": "10.015",
"value": "Don Juan Manuel"
},
{
"operation": "Now on",
"time": "10.745",
"value": "Don Juan Manuel"
},
{
"operation": "Highlighting",
"time": "14.828",
"value": "Don Juan Manuel_sentence_0"
},
{
"operation": "Highlighting",
"time": "14.829",
"value": "Don Juan Manuel_title"
},
{
"operation": "Highlighting",
"time": "20.025",
"value": "Don Juan Manuel_cell_0_1_1"
},
{
"operation": "Highlighting",
"time": "20.029",
"value": "Don Juan Manuel_header_cell_0_1_0"
},
{
"operation": "Highlighting",
"time": "20.029",
"value": "Don Juan Manuel_header_cell_0_0_0"
},
{
"operation": "hyperlink",
"time": "30.07",
"value": "Escalona"
},
{
"operation": "Now on",
"time": "30.911",
"value": "Escalona"
},
{
"operation": "Highlighting",
"time": "33.411",
"value": "Escalona_sentence_1"
},
{
"operation": "Highlighting",
"time": "61.807",
"value": "Escalona_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "70.209",
"value": "Escalona_sentence_1"
},
{
"operation": "finish",
"time": "79.171",
"value": "finish"
}
]
}
|
88499
|
Meher Ali Shah received his early religious education at the Khanqah, he decided to continue further studies in the United Provinces (U.P) of India but did not stay in Lutfullah.
|
{
"evidence": [
"Meher Ali Shah_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:\nMeher Ali Shah received his early religious education at the Khanqah, he decided to continue further studies in the United Provinces (U.P) of India but did not stay in Lutfullah.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Meher Ali Shah_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1026,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.974",
"value": "?search="
},
{
"operation": "search",
"time": "9.926",
"value": "Meher Ali Shah"
},
{
"operation": "Now on",
"time": "10.647",
"value": "Meher Ali Shah"
},
{
"operation": "Highlighting",
"time": "46.96",
"value": "Meher Ali Shah_sentence_5"
},
{
"operation": "Highlighting",
"time": "50.503",
"value": "Meher Ali Shah_sentence_6"
},
{
"operation": "Highlighting",
"time": "65.937",
"value": "Meher Ali Shah_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "66.585",
"value": "Meher Ali Shah_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "66.951",
"value": "Meher Ali Shah_sentence_5"
},
{
"operation": "finish",
"time": "77.535",
"value": "finish"
}
]
}
|
exfever_632
|
Harry James is one of the best bandleaders of his time.
|
{
"evidence": [
"Frank Sinatra",
"Tommy Dorsey"
],
"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:\nHarry James is one of the best bandleaders of his time.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Frank Sinatra",
"Tommy Dorsey"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1027,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
10281
|
Strike the Blood III, is wrongly believed to be an anime series adapted from the light novel series has ten OVAs between December 19 and September 25.
|
{
"evidence": [
"Strike the Blood III_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:\nStrike the Blood III, is wrongly believed to be an anime series adapted from the light novel series has ten OVAs between December 19 and September 25.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Strike the Blood III_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1028,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.171",
"value": "?search="
},
{
"operation": "search",
"time": "17.042",
"value": "Strike the Blood III"
},
{
"operation": "Now on",
"time": "17.991",
"value": "Strike the Blood III"
},
{
"operation": "Highlighting",
"time": "44.686",
"value": "Strike the Blood III_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "178.206",
"value": "Strike the Blood III_sentence_1"
},
{
"operation": "Highlighting",
"time": "213.655",
"value": "Strike the Blood III_sentence_0"
},
{
"operation": "finish",
"time": "236.76",
"value": "finish"
}
]
}
|
52872
|
Genoplesium fimbriatum (commonly known as the fringed midge orchid is a big terrestrial orchid found from southern Queensland to southern New South Wales, It has a single thin leaf and up to thirty green to brownish-green flowers with red stripes and a red-tipped labellum with hairy edges).
|
{
"evidence": [
"Genoplesium fimbriatum_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:\nGenoplesium fimbriatum (commonly known as the fringed midge orchid is a big terrestrial orchid found from southern Queensland to southern New South Wales, It has a single thin leaf and up to thirty green to brownish-green flowers with red stripes and a red-tipped labellum with hairy edges).\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Genoplesium fimbriatum_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1029,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.032",
"value": "?search="
},
{
"operation": "search",
"time": "17.79",
"value": "Genoplesium fimbriatum"
},
{
"operation": "Now on",
"time": "19.73",
"value": "Genoplesium fimbriatum"
},
{
"operation": "Highlighting",
"time": "21.949",
"value": "Genoplesium fimbriatum_sentence_0"
},
{
"operation": "finish",
"time": "93.253",
"value": "finish"
}
]
}
|
exfever_6803
|
Kung Fu Panda 3 is a 2016 computer-animated fantasy martial arts comedy film co-directed by a Korean-born American story artist, who was the second woman nominated for the award that is given every year to animated films that include an animation figure for at least 75 percent of the running time.
|
{
"evidence": [
"Kung Fu Panda 3",
"Jennifer Yuh Nelson",
"Academy Award for Best Animated Feature"
],
"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:\nKung Fu Panda 3 is a 2016 computer-animated fantasy martial arts comedy film co-directed by a Korean-born American story artist, who was the second woman nominated for the award that is given every year to animated films that include an animation figure for at least 75 percent of the running time.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Kung Fu Panda 3",
"Jennifer Yuh Nelson",
"Academy Award for Best Animated Feature"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1030,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
77339
|
Gleb Strizhenov (July 21, 1923 – October 4, 1985) was an actor from Georgia, Union of Soviet Socialist Republics.
|
{
"evidence": [
"Gleb Strizhenov_sentence_1",
"Voronezh_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:\nGleb Strizhenov (July 21, 1923 – October 4, 1985) was an actor from Georgia, Union of Soviet Socialist Republics.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Gleb Strizhenov_sentence_1",
"Voronezh_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1031,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.284",
"value": "?search="
},
{
"operation": "search",
"time": "11.962",
"value": "Gleb Strizhenov"
},
{
"operation": "Now on",
"time": "12.947",
"value": "Gleb Strizhenov"
},
{
"operation": "Highlighting",
"time": "35.495",
"value": "Gleb Strizhenov_sentence_1"
},
{
"operation": "hyperlink",
"time": "36.445",
"value": "Voronezh"
},
{
"operation": "Now on",
"time": "37.432",
"value": "Voronezh"
},
{
"operation": "Highlighting",
"time": "39.592",
"value": "Voronezh_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "54.237",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "54.463",
"value": "Gleb Strizhenov"
},
{
"operation": "finish",
"time": "86.067",
"value": "finish"
}
]
}
|
exfever_1999
|
Shania Twain was the first non-US citizen to win CMA Award.
|
{
"evidence": [
"Shania Twain",
"Timmins",
"Northeastern Ontario"
],
"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:\nShania Twain was the first non-US citizen to win CMA Award.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Shania Twain",
"Timmins",
"Northeastern Ontario"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1032,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
93667
|
Marcus Olofsson was born in Sweden and currently plays football for Manchester United as midfielder.
|
{
"evidence": [
"Marcus Olofsson_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:\nMarcus Olofsson was born in Sweden and currently plays football for Manchester United as midfielder.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Marcus Olofsson_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1033,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.01",
"value": "?search="
},
{
"operation": "search",
"time": "11.482",
"value": "Marcus Olofsson"
},
{
"operation": "Now on",
"time": "12.41",
"value": "Marcus Olofsson"
},
{
"operation": "Highlighting",
"time": "30.225",
"value": "Marcus Olofsson_cell_0_2_1"
},
{
"operation": "Highlighting",
"time": "30.229",
"value": "Marcus Olofsson_header_cell_0_2_0"
},
{
"operation": "Highlighting",
"time": "30.229",
"value": "Marcus Olofsson_header_cell_0_0_0"
},
{
"operation": "Highlighting",
"time": "33.81",
"value": "Marcus Olofsson_sentence_0"
},
{
"operation": "hyperlink",
"time": "45.884",
"value": "IFK Mariehamn"
},
{
"operation": "Now on",
"time": "46.589",
"value": "IFK Mariehamn"
},
{
"operation": "back-button-clicked",
"time": "52.293",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "52.351",
"value": "Marcus Olofsson"
},
{
"operation": "Highlighting deleted",
"time": "54.094",
"value": "Marcus Olofsson_cell_0_2_1"
},
{
"operation": "Highlighting deleted",
"time": "54.098",
"value": "Marcus Olofsson_header_cell_0_2_0"
},
{
"operation": "Highlighting deleted",
"time": "54.099",
"value": "Marcus Olofsson_header_cell_0_0_0"
},
{
"operation": "finish",
"time": "59.565",
"value": "finish"
}
]
}
|
exfever_7960
|
WALL-E is an American computer-animated science fiction film co-written by an American animator who also co-wrote an American computer-animated monster comedy film produced by Pixar Animation Studios for Walt Disney Pictures.
|
{
"evidence": [
"WALL-E",
"Andrew Stanton",
"Monsters, Inc."
],
"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:\nWALL-E is an American computer-animated science fiction film co-written by an American animator who also co-wrote an American computer-animated monster comedy film produced by Pixar Animation Studios for Walt Disney Pictures.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"WALL-E",
"Andrew Stanton",
"Monsters, Inc."
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1034,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4564
|
Stephanie Daley is a 2005 film about teen pregnancy that earned Tamblyn a nomination for best-supporting female at the Friends of Independents Awards, an award dedicated to independent filmmakers.
|
{
"evidence": [
"Stephanie Daley",
"Independent Spirit Awards"
],
"label": "REFUTE"
}
|
exfever
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nStephanie Daley is a 2005 film about teen pregnancy that earned Tamblyn a nomination for best-supporting female at the Friends of Independents Awards, an award dedicated to independent filmmakers.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Stephanie Daley",
"Independent Spirit Awards"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1035,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4377
|
Mozilla Firefox, or simply Firefox, is a free and open-source web browser developed by a non-profit organization that exists to support and collectively lead the open-source Mozilla project.
|
{
"evidence": [
"Firefox",
"Mozilla Foundation"
],
"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:\nMozilla Firefox, or simply Firefox, is a free and open-source web browser developed by a non-profit organization that exists to support and collectively lead the open-source Mozilla project.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Firefox",
"Mozilla Foundation"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1036,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
51971
|
Diego Mancino published "Cose che cambiano tutto" in 2005 (his first album as a solo artist), and recorded "Idee stupide" with best-selling Italian rapper Fabri Fibra in 2006.
|
{
"evidence": [
"Diego Mancino_sentence_8",
"Fabri Fibra_sentence_1",
"Diego Mancino_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:\nDiego Mancino published \"Cose che cambiano tutto\" in 2005 (his first album as a solo artist), and recorded \"Idee stupide\" with best-selling Italian rapper Fabri Fibra in 2006.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Diego Mancino_sentence_8",
"Fabri Fibra_sentence_1",
"Diego Mancino_sentence_7"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1037,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.474",
"value": "?search="
},
{
"operation": "search",
"time": "13.248",
"value": "Diego Mancino"
},
{
"operation": "Now on",
"time": "14.168",
"value": "Diego Mancino"
},
{
"operation": "Highlighting",
"time": "23.986",
"value": "Diego Mancino_sentence_7"
},
{
"operation": "Highlighting",
"time": "37.231",
"value": "Diego Mancino_sentence_8"
},
{
"operation": "hyperlink",
"time": "38.31",
"value": "Fabri Fibra"
},
{
"operation": "Now on",
"time": "39.58",
"value": "Fabri Fibra"
},
{
"operation": "Highlighting",
"time": "70.103",
"value": "Fabri Fibra_sentence_1"
},
{
"operation": "back-button-clicked",
"time": "109.322",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "109.551",
"value": "Diego Mancino"
},
{
"operation": "Highlighting deleted",
"time": "112.73",
"value": "Diego Mancino_sentence_7"
},
{
"operation": "search",
"time": "116.938",
"value": "Contains...Cose che cambiano tutto"
},
{
"operation": "Now on",
"time": "118.166",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "125.861",
"value": "Diego Mancino"
},
{
"operation": "Now on",
"time": "126.663",
"value": "Diego Mancino"
},
{
"operation": "hyperlink",
"time": "136.849",
"value": "3.2 soloist album Diego Mancino_section_4"
},
{
"operation": "Highlighting",
"time": "162.793",
"value": "Diego Mancino_sentence_7"
},
{
"operation": "finish",
"time": "176.113",
"value": "finish"
}
]
}
|
59599
|
Broom forkmoss is usually robust and coarse, forming shiny tufts with woolly stems 5–8 cm high and the leaf midrib extends to the tip and usually has 7 ridges along its back.
|
{
"evidence": [
"Dicranum scoparium_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:\nBroom forkmoss is usually robust and coarse, forming shiny tufts with woolly stems 5–8 cm high and the leaf midrib extends to the tip and usually has 7 ridges along its back.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Dicranum scoparium_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1038,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.676",
"value": "?search="
},
{
"operation": "search",
"time": "5.662",
"value": "Broom forkmoss"
},
{
"operation": "Now on",
"time": "6.494",
"value": "Dicranum scoparium"
},
{
"operation": "Highlighting",
"time": "11.389",
"value": "Dicranum scoparium_sentence_4"
},
{
"operation": "finish",
"time": "26.284",
"value": "finish"
}
]
}
|
41283
|
Hicks enrolled in the University of Washington School of Law, where he earned a Juris Doctor in 1948; then he established a small private practice in Pierce County, Washington.
|
{
"evidence": [
"Floyd Hicks_sentence_8",
"Floyd Hicks_sentence_7"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nHicks enrolled in the University of Washington School of Law, where he earned a Juris Doctor in 1948; then he established a small private practice in Pierce County, Washington.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Floyd Hicks_sentence_8",
"Floyd Hicks_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1039,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.54",
"value": "?search="
},
{
"operation": "search",
"time": "11.071",
"value": "Hicks"
},
{
"operation": "Now on",
"time": "13.075",
"value": "Hicks"
},
{
"operation": "search",
"time": "31.467",
"value": "\"established a small private practice in Pierce County\""
},
{
"operation": "Now on",
"time": "32.81",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "34.452",
"value": "Floyd Hicks"
},
{
"operation": "Now on",
"time": "35.647",
"value": "Floyd Hicks"
},
{
"operation": "Page search",
"time": "38.479",
"value": "established a small private practice in Pierce County"
},
{
"operation": "Highlighting",
"time": "40.893",
"value": "Floyd Hicks_sentence_8"
},
{
"operation": "Highlighting",
"time": "43.159",
"value": "Floyd Hicks_sentence_7"
},
{
"operation": "finish",
"time": "63.094",
"value": "finish"
}
]
}
|
exfever_2525
|
Sophie Turner is an English actress who made her acting debut as a fictional character whose parents include a prominent point of view character in the first three novels.
|
{
"evidence": [
"Sophie Turner",
"Sansa Stark",
"Catelyn Stark"
],
"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:\nSophie Turner is an English actress who made her acting debut as a fictional character whose parents include a prominent point of view character in the first three novels.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Sophie Turner",
"Sansa Stark",
"Catelyn Stark"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1040,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
13637
|
Michaela Roessner, whose first novel, Walkabout Woman, was a 1989 nominee for the Mythopoeic Award, given by the Mythopoeic Society to authors of outstanding works in the fields of myth, fantasy, and the scholarly study of these areas, won the John W. Campbell Award for Best New Writer in 1989.
|
{
"evidence": [
"Michaela Roessner_sentence_4",
"Mythopoeic Awards_sentence_0",
"Michaela Roessner_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:\nMichaela Roessner, whose first novel, Walkabout Woman, was a 1989 nominee for the Mythopoeic Award, given by the Mythopoeic Society to authors of outstanding works in the fields of myth, fantasy, and the scholarly study of these areas, won the John W. Campbell Award for Best New Writer in 1989.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Michaela Roessner_sentence_4",
"Mythopoeic Awards_sentence_0",
"Michaela Roessner_sentence_3"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1041,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.626",
"value": "?search="
},
{
"operation": "search",
"time": "5.656",
"value": "Michaela Roessner"
},
{
"operation": "Now on",
"time": "6.609",
"value": "Michaela Roessner"
},
{
"operation": "Highlighting",
"time": "28.688",
"value": "Michaela Roessner_sentence_4"
},
{
"operation": "hyperlink",
"time": "36.393",
"value": "Mythopoeic Award"
},
{
"operation": "Now on",
"time": "38.628",
"value": "Mythopoeic Awards"
},
{
"operation": "Highlighting",
"time": "49.148",
"value": "Mythopoeic Awards_sentence_0"
},
{
"operation": "back-button-clicked",
"time": "52.045",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "52.149",
"value": "Michaela Roessner"
},
{
"operation": "Highlighting",
"time": "66.029",
"value": "Michaela Roessner_sentence_3"
},
{
"operation": "finish",
"time": "87.97",
"value": "finish"
}
]
}
|
exfever_7156
|
Reddit is the most popular website in the United States.
|
{
"evidence": [
"Reddit",
"News aggregator"
],
"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:\nReddit is the most popular website in the United States.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Reddit",
"News aggregator"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1042,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
35005
|
FLEXPART has inherited small portions of its code from its successor, FLEXTRA (FLEXible TRAjectory model), which its first version was released in the late 2000s.
|
{
"evidence": [
"FLEXPART_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:\nFLEXPART has inherited small portions of its code from its successor, FLEXTRA (FLEXible TRAjectory model), which its first version was released in the late 2000s.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"FLEXPART_sentence_4"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1043,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.165",
"value": "?search="
},
{
"operation": "search",
"time": "14.402",
"value": "FLEXPART"
},
{
"operation": "Now on",
"time": "15.601",
"value": "FLEXPART"
},
{
"operation": "Highlighting",
"time": "60.218",
"value": "FLEXPART_sentence_4"
},
{
"operation": "finish",
"time": "75.624",
"value": "finish"
}
]
}
|
exfever_5460
|
Broadcast News is a 1987 American romantic comedy-drama film directed by an American director who has directed several films, including an American romantic comedy-drama film.
|
{
"evidence": [
"Broadcast News (film)",
"James L. Brooks",
"As Good as It Gets"
],
"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:\nBroadcast News is a 1987 American romantic comedy-drama film directed by an American director who has directed several films, including an American romantic comedy-drama film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Broadcast News (film)",
"James L. Brooks",
"As Good as It Gets"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1044,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
44139
|
Gustavo Selva was an Italian journalist, started his career as a journalist for the Bologna newspaper L'Avvenire d'Italia but was not involved in politics.
|
{
"evidence": [
"Gustavo Selva_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:\nGustavo Selva was an Italian journalist, started his career as a journalist for the Bologna newspaper L'Avvenire d'Italia but was not involved in politics.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Gustavo Selva_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1045,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.775",
"value": "?search="
},
{
"operation": "search",
"time": "10.768",
"value": "Gustavo Selva"
},
{
"operation": "Now on",
"time": "12.092",
"value": "Gustavo Selva"
},
{
"operation": "Highlighting",
"time": "13.81",
"value": "Gustavo Selva_sentence_0"
},
{
"operation": "Highlighting",
"time": "15.393",
"value": "Gustavo Selva_sentence_1"
},
{
"operation": "Highlighting",
"time": "24.314",
"value": "Gustavo Selva_sentence_3"
},
{
"operation": "Highlighting deleted",
"time": "24.826",
"value": "Gustavo Selva_sentence_1"
},
{
"operation": "Highlighting deleted",
"time": "25.178",
"value": "Gustavo Selva_sentence_0"
},
{
"operation": "finish",
"time": "31.41",
"value": "finish"
}
]
}
|
33553
|
Les Affaires is aimed at those who do or are interested in Quebec business, content features and relevant articles on management and is intended for individuals age 30 to 55.
|
{
"evidence": [
"Les Affaires_sentence_4",
"Les Affaires_sentence_5",
"Les Affaires_sentence_7"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nLes Affaires is aimed at those who do or are interested in Quebec business, content features and relevant articles on management and is intended for individuals age 30 to 55.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Les Affaires_sentence_4",
"Les Affaires_sentence_5",
"Les Affaires_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1046,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.42",
"value": "?search="
},
{
"operation": "search",
"time": "6.773",
"value": "Les Affaires"
},
{
"operation": "Now on",
"time": "8.369",
"value": "Les Affaires"
},
{
"operation": "Highlighting",
"time": "46.189",
"value": "Les Affaires_sentence_4"
},
{
"operation": "Highlighting",
"time": "69.705",
"value": "Les Affaires_sentence_5"
},
{
"operation": "Highlighting",
"time": "74.403",
"value": "Les Affaires_sentence_7"
},
{
"operation": "finish",
"time": "80.515",
"value": "finish"
}
]
}
|
71588
|
He also recorded with Margie Singleton, Melba Montgomery, Brenda Carter and, most famously, Wynette and the album featured many of the country stars that Jones fans were familiar with, like Waylon Jennings, Willie Nelson, and Johnny Paycheck.
|
{
"evidence": [
"My Very Special Guests_sentence_18",
"My Very Special Guests_sentence_16"
],
"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:\nHe also recorded with Margie Singleton, Melba Montgomery, Brenda Carter and, most famously, Wynette and the album featured many of the country stars that Jones fans were familiar with, like Waylon Jennings, Willie Nelson, and Johnny Paycheck.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"My Very Special Guests_sentence_18",
"My Very Special Guests_sentence_16"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1047,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.893",
"value": "undefined"
},
{
"operation": "search",
"time": "16.965",
"value": "George Jones"
},
{
"operation": "Now on",
"time": "21.382",
"value": "George Jones"
},
{
"operation": "Page search",
"time": "26.188",
"value": "melba"
},
{
"operation": "search",
"time": "69.135",
"value": "george jones collaborations"
},
{
"operation": "Now on",
"time": "70.788",
"value": "undefined"
},
{
"operation": "search",
"time": "93.606",
"value": "margie singleton collaboration"
},
{
"operation": "Now on",
"time": "95.809",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "103.418",
"value": "George Jones singles discography"
},
{
"operation": "Now on",
"time": "103.987",
"value": "George Jones singles discography"
},
{
"operation": "Page search",
"time": "113.871",
"value": "margie"
},
{
"operation": "Highlighting",
"time": "135.441",
"value": "George Jones singles discography_header_cell_10_6_1"
},
{
"operation": "Highlighting",
"time": "135.443",
"value": "George Jones singles discography_title"
},
{
"operation": "Highlighting",
"time": "135.444",
"value": "George Jones singles discography_section_11"
},
{
"operation": "Highlighting",
"time": "138.58",
"value": "George Jones singles discography_cell_10_6_2"
},
{
"operation": "Highlighting",
"time": "138.638",
"value": "George Jones singles discography_header_cell_10_6_1"
},
{
"operation": "Highlighting",
"time": "138.638",
"value": "George Jones singles discography_header_cell_10_0_2"
},
{
"operation": "Highlighting",
"time": "141.664",
"value": "George Jones singles discography_header_cell_10_8_1"
},
{
"operation": "Highlighting",
"time": "142.98",
"value": "George Jones singles discography_cell_10_8_2"
},
{
"operation": "Highlighting",
"time": "143.019",
"value": "George Jones singles discography_header_cell_10_8_1"
},
{
"operation": "Page search",
"time": "163.427",
"value": "wynette"
},
{
"operation": "Highlighting",
"time": "179.38",
"value": "George Jones singles discography_header_cell_13_12_1"
},
{
"operation": "Highlighting",
"time": "179.382",
"value": "George Jones singles discography_section_14"
},
{
"operation": "hyperlink",
"time": "186.675",
"value": "Tammy Wynette"
},
{
"operation": "Now on",
"time": "189.489",
"value": "Tammy Wynette"
},
{
"operation": "Page search",
"time": "193.51",
"value": "jones"
},
{
"operation": "Highlighting",
"time": "197.625",
"value": "Tammy Wynette_sentence_5"
},
{
"operation": "Highlighting",
"time": "197.625",
"value": "Tammy Wynette_title"
},
{
"operation": "Highlighting deleted",
"time": "202.535",
"value": "Tammy Wynette_sentence_5"
},
{
"operation": "Highlighting deleted",
"time": "202.536",
"value": "Tammy Wynette_title"
},
{
"operation": "Page search",
"time": "207",
"value": "jones"
},
{
"operation": "Page search",
"time": "219.02",
"value": "george jones"
},
{
"operation": "Page search",
"time": "237.081",
"value": "collaboration"
},
{
"operation": "Highlighting",
"time": "247.106",
"value": "Tammy Wynette_sentence_59"
},
{
"operation": "Highlighting",
"time": "247.106",
"value": "Tammy Wynette_title"
},
{
"operation": "Highlighting",
"time": "247.107",
"value": "Tammy Wynette_section_1"
},
{
"operation": "Highlighting",
"time": "247.108",
"value": "Tammy Wynette_section_4"
},
{
"operation": "Highlighting deleted",
"time": "261.968",
"value": "Tammy Wynette_sentence_59"
},
{
"operation": "Highlighting deleted",
"time": "261.969",
"value": "Tammy Wynette_title"
},
{
"operation": "Highlighting deleted",
"time": "261.971",
"value": "Tammy Wynette_section_1"
},
{
"operation": "Highlighting deleted",
"time": "261.972",
"value": "Tammy Wynette_section_4"
},
{
"operation": "back-button-clicked",
"time": "281.886",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "282.247",
"value": "George Jones singles discography"
},
{
"operation": "Page search",
"time": "300.749",
"value": "jennings"
},
{
"operation": "Highlighting",
"time": "322.083",
"value": "George Jones singles discography_header_cell_11_7_1"
},
{
"operation": "Highlighting",
"time": "322.086",
"value": "George Jones singles discography_section_12"
},
{
"operation": "Highlighting",
"time": "324.386",
"value": "George Jones singles discography_cell_11_7_2"
},
{
"operation": "Highlighting",
"time": "324.402",
"value": "George Jones singles discography_header_cell_11_7_1"
},
{
"operation": "Highlighting",
"time": "324.402",
"value": "George Jones singles discography_header_cell_11_0_2"
},
{
"operation": "Highlighting deleted",
"time": "326.057",
"value": "George Jones singles discography_cell_11_7_2"
},
{
"operation": "Highlighting deleted",
"time": "326.06",
"value": "George Jones singles discography_section_12"
},
{
"operation": "Highlighting deleted",
"time": "326.067",
"value": "George Jones singles discography_header_cell_11_7_1"
},
{
"operation": "Highlighting deleted",
"time": "326.068",
"value": "George Jones singles discography_header_cell_11_0_2"
},
{
"operation": "Page search",
"time": "342.068",
"value": "nelson"
},
{
"operation": "search",
"time": "377.599",
"value": "featured many of the country stars that Jones fans were familiar with, like Waylon Jennings, Willie Nelson, and Johnny Paychec"
},
{
"operation": "Now on",
"time": "379.365",
"value": "undefined"
},
{
"operation": "search",
"time": "406.115",
"value": "aylon Jennings, Willie Nelson, and Johnny Paychec"
},
{
"operation": "Now on",
"time": "406.941",
"value": "undefined"
},
{
"operation": "search",
"time": "416.129",
"value": "waylon Jennings, Willie Nelson, and Johnny Paychec"
},
{
"operation": "Now on",
"time": "417.282",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "424.202",
"value": "My Very Special Guests"
},
{
"operation": "Now on",
"time": "424.917",
"value": "My Very Special Guests"
},
{
"operation": "Page search",
"time": "433.464",
"value": "wynette"
},
{
"operation": "Page search",
"time": "450.503",
"value": "fans"
},
{
"operation": "Highlighting",
"time": "455.869",
"value": "My Very Special Guests_sentence_18"
},
{
"operation": "Page search",
"time": "479.55",
"value": "tammy"
},
{
"operation": "Highlighting",
"time": "484.747",
"value": "My Very Special Guests_sentence_4"
},
{
"operation": "Highlighting",
"time": "484.748",
"value": "My Very Special Guests_section_0"
},
{
"operation": "Highlighting deleted",
"time": "486.547",
"value": "My Very Special Guests_sentence_4"
},
{
"operation": "Highlighting deleted",
"time": "486.55",
"value": "My Very Special Guests_section_0"
},
{
"operation": "Page search",
"time": "493.769",
"value": "wynette"
},
{
"operation": "Highlighting",
"time": "501.062",
"value": "My Very Special Guests_sentence_6"
},
{
"operation": "Highlighting",
"time": "501.063",
"value": "My Very Special Guests_section_0"
},
{
"operation": "Highlighting deleted",
"time": "505.679",
"value": "My Very Special Guests_sentence_6"
},
{
"operation": "Highlighting deleted",
"time": "505.68",
"value": "My Very Special Guests_section_0"
},
{
"operation": "Highlighting",
"time": "507.066",
"value": "My Very Special Guests_sentence_13"
},
{
"operation": "Highlighting",
"time": "507.067",
"value": "My Very Special Guests_section_0"
},
{
"operation": "Highlighting deleted",
"time": "510.494",
"value": "My Very Special Guests_sentence_13"
},
{
"operation": "Highlighting deleted",
"time": "510.496",
"value": "My Very Special Guests_section_0"
},
{
"operation": "Highlighting",
"time": "515.724",
"value": "My Very Special Guests_sentence_16"
},
{
"operation": "finish",
"time": "546.447",
"value": "finish"
}
]
}
|
exfever_5886
|
Tobey Maguire is an American actor known for his role in a 2003 American sports film directed by Gary Ross that is based on the best-selling 1999 non-fiction book by Laura Hillenbrand.
|
{
"evidence": [
"Tobey Maguire",
"Seabiscuit (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:\nTobey Maguire is an American actor known for his role in a 2003 American sports film directed by Gary Ross that is based on the best-selling 1999 non-fiction book by Laura Hillenbrand.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Tobey Maguire",
"Seabiscuit (film)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1048,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
1917
|
Warmątowice Sienkiewiczowskie is a village in the administrative district of Gmina Krotoszyce but prior to 1945 it was in Germany.
|
{
"evidence": [
"Warmątowice Sienkiewiczowskie_sentence_0",
"Warmątowice Sienkiewiczowskie_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:\nWarmątowice Sienkiewiczowskie is a village in the administrative district of Gmina Krotoszyce but prior to 1945 it was in Germany.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Warmątowice Sienkiewiczowskie_sentence_0",
"Warmątowice Sienkiewiczowskie_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1049,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "search",
"time": "31.245",
"value": "Warmątowice Sienkiewiczowskie"
},
{
"operation": "Highlighting",
"time": "40.693",
"value": "Warmątowice Sienkiewiczowskie_sentence_0"
},
{
"operation": "Highlighting",
"time": "44.668",
"value": "Warmątowice Sienkiewiczowskie_sentence_1"
},
{
"operation": "finish",
"time": "62.365",
"value": "finish"
}
]
}
|
exfever_6236
|
"Born This Way" was the third track developed while on the road with the Monster Ball Tour.
|
{
"evidence": [
"Born This Way (song)",
"Jeppe Laursen"
],
"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:\n\"Born This Way\" was the third track developed while on the road with the Monster Ball Tour.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Born This Way (song)",
"Jeppe Laursen"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1050,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1894
|
Boromir is a fictional character appears in the second volume of J. R. R. Tolkien's high fantasy novel The Queen of the Rings.
|
{
"evidence": [
"Boromir",
"The Two Towers"
],
"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:\nBoromir is a fictional character appears in the second volume of J. R. R. Tolkien's high fantasy novel The Queen of the Rings.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Boromir",
"The Two Towers"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1051,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_7439
|
The Jackson family?were the most famous musicians and entertainers worldwide.
|
{
"evidence": [
"Jackson family",
"Joe Jackson (manager)",
"Talent manager"
],
"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 Jackson family?were the most famous musicians and entertainers worldwide.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jackson family",
"Joe Jackson (manager)",
"Talent manager"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1052,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
69722
|
Rudolf Holsti was a Finnish politician, journalist and diplomat who represented Finland in the League of Nations and openly criticized Joseph Stalin at the outbreak of war.
|
{
"evidence": [
"Rudolf Holsti_sentence_0",
"Rudolf Holsti_sentence_3",
"Rudolf Holsti_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:\nRudolf Holsti was a Finnish politician, journalist and diplomat who represented Finland in the League of Nations and openly criticized Joseph Stalin at the outbreak of war.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Rudolf Holsti_sentence_0",
"Rudolf Holsti_sentence_3",
"Rudolf Holsti_sentence_5"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1053,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.278",
"value": "?search="
},
{
"operation": "search",
"time": "11.072",
"value": "Rudolf Holsti"
},
{
"operation": "Now on",
"time": "11.764",
"value": "Rudolf Holsti"
},
{
"operation": "Highlighting",
"time": "17.309",
"value": "Rudolf Holsti_sentence_0"
},
{
"operation": "Highlighting",
"time": "27.981",
"value": "Rudolf Holsti_sentence_3"
},
{
"operation": "Highlighting",
"time": "36.536",
"value": "Rudolf Holsti_sentence_5"
},
{
"operation": "finish",
"time": "49.826",
"value": "finish"
}
]
}
|
62242
|
Both Zalayhar Hassanali, an Indo-Trinidadian and Tobagonian Muslim woman, and Patricia Robinson, an economist and civil servant, refused the title of First Ladies, the title given to the wife of the president of Trinidad and Tobago, because of personal preferences.
|
{
"evidence": [
"Patricia Robinson_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:\nBoth Zalayhar Hassanali, an Indo-Trinidadian and Tobagonian Muslim woman, and Patricia Robinson, an economist and civil servant, refused the title of First Ladies, the title given to the wife of the president of Trinidad and Tobago, because of personal preferences.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Patricia Robinson_sentence_1"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1054,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.706",
"value": "?search="
},
{
"operation": "search",
"time": "21.279",
"value": "Contains...wife of the president of Trinidad and Tobago"
},
{
"operation": "Now on",
"time": "23.45",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "30.031",
"value": "First ladies of Trinidad and Tobago First Spouse of Trinidad and Tobago_section_0"
},
{
"operation": "Now on",
"time": "30.555",
"value": "First Spouse of Trinidad and Tobago#First ladies of Trinidad and Tobago First Spouse of Trinidad and Tobago section 0"
},
{
"operation": "hyperlink",
"time": "74.288",
"value": "Noor Hassanali"
},
{
"operation": "Now on",
"time": "74.802",
"value": "Noor Hassanali"
},
{
"operation": "search",
"time": "102.149",
"value": "Zalayhar Hassanali patricia robinson"
},
{
"operation": "Now on",
"time": "102.954",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "113.064",
"value": "Patricia Robinson"
},
{
"operation": "Now on",
"time": "113.436",
"value": "Patricia Robinson"
},
{
"operation": "Highlighting",
"time": "146.878",
"value": "Patricia Robinson_sentence_1"
},
{
"operation": "finish",
"time": "172.184",
"value": "finish"
}
]
}
|
exfever_7552
|
"Beautiful" is a song recorded for the fourth studio album of American singer Christina Aguilera that was released by RCA Records.
|
{
"evidence": [
"Beautiful (Christina Aguilera song)",
"Stripped (Christina Aguilera album)"
],
"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:\n\"Beautiful\" is a song recorded for the fourth studio album of American singer Christina Aguilera that was released by RCA Records.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Beautiful (Christina Aguilera song)",
"Stripped (Christina Aguilera album)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1055,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
62675
|
Kevin Sampsell's works were only discovered after the first world war.
|
{
"evidence": [
"Kevin Sampsell_sentence_14",
"Kevin Sampsell_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:\nKevin Sampsell's works were only discovered after the first world war.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Kevin Sampsell_sentence_14",
"Kevin Sampsell_sentence_0"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1056,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.04",
"value": "?search="
},
{
"operation": "search",
"time": "14.135",
"value": "Kevin Sampsell"
},
{
"operation": "Now on",
"time": "14.859",
"value": "Kevin Sampsell"
},
{
"operation": "Highlighting",
"time": "115.183",
"value": "Kevin Sampsell_sentence_0"
},
{
"operation": "Highlighting",
"time": "120.5",
"value": "Kevin Sampsell_sentence_2"
},
{
"operation": "Highlighting",
"time": "134.61",
"value": "Kevin Sampsell_sentence_3"
},
{
"operation": "Highlighting",
"time": "147.987",
"value": "Kevin Sampsell_sentence_7"
},
{
"operation": "Highlighting",
"time": "180.306",
"value": "Kevin Sampsell_sentence_14"
},
{
"operation": "Highlighting deleted",
"time": "194.948",
"value": "Kevin Sampsell_sentence_7"
},
{
"operation": "Highlighting deleted",
"time": "200.343",
"value": "Kevin Sampsell_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "406.69",
"value": "Kevin Sampsell_sentence_2"
},
{
"operation": "Highlighting deleted",
"time": "407.196",
"value": "Kevin Sampsell_sentence_3"
},
{
"operation": "Highlighting",
"time": "408.477",
"value": "Kevin Sampsell_sentence_0"
},
{
"operation": "finish",
"time": "414.835",
"value": "finish"
}
]
}
|
30893
|
Although Great Basin College has branch campuses in Battle Mountain (a very low populated town in Lander County, Nevada), Ely, Pahrump and Winnemucca, its main campus is in Dallas, Texas, which is very large and offers Residence Halls.
|
{
"evidence": [
"Great Basin College_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:\nAlthough Great Basin College has branch campuses in Battle Mountain (a very low populated town in Lander County, Nevada), Ely, Pahrump and Winnemucca, its main campus is in Dallas, Texas, which is very large and offers Residence Halls.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Great Basin College_sentence_8"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1057,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.713",
"value": "?search="
},
{
"operation": "search",
"time": "11.276",
"value": "Great Basin College"
},
{
"operation": "Now on",
"time": "11.982",
"value": "Great Basin College"
},
{
"operation": "Highlighting",
"time": "29.133",
"value": "Great Basin College_sentence_0"
},
{
"operation": "Highlighting",
"time": "38.437",
"value": "Great Basin College_sentence_8"
},
{
"operation": "Highlighting deleted",
"time": "40.517",
"value": "Great Basin College_sentence_0"
},
{
"operation": "finish",
"time": "46.621",
"value": "finish"
}
]
}
|
exfever_4040
|
Jean Grey, also known as Marvel Girl, is a member of the mutants who first appeared in the American comic book series published by DC Comics.
|
{
"evidence": [
"Jean Grey",
"Uncanny X-Men"
],
"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:\nJean Grey, also known as Marvel Girl, is a member of the mutants who first appeared in the American comic book series published by DC Comics.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Jean Grey",
"Uncanny X-Men"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1058,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
19044
|
Selma Dabbagh, in addition to writing short stories which have appeared in New Writing 15 and Qissat: Short Stories by Palestinian Women, also worked as a human rights lawyer in London with frequent visits to the West Bank and Cairo (he capital of Egypt).
|
{
"evidence": [
"Selma Dabbagh_sentence_4",
"Selma Dabbagh_sentence_3",
"Cairo_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:\nSelma Dabbagh, in addition to writing short stories which have appeared in New Writing 15 and Qissat: Short Stories by Palestinian Women, also worked as a human rights lawyer in London with frequent visits to the West Bank and Cairo (he capital of Egypt).\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Selma Dabbagh_sentence_4",
"Selma Dabbagh_sentence_3",
"Cairo_sentence_2"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1059,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.815",
"value": "?search="
},
{
"operation": "search",
"time": "50.347",
"value": "Selma Dabbagh"
},
{
"operation": "Now on",
"time": "50.959",
"value": "Selma Dabbagh"
},
{
"operation": "Highlighting",
"time": "154.326",
"value": "Selma Dabbagh_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "172.07",
"value": "Selma Dabbagh_sentence_0"
},
{
"operation": "Highlighting",
"time": "182.806",
"value": "Selma Dabbagh_sentence_4"
},
{
"operation": "Highlighting",
"time": "191.189",
"value": "Selma Dabbagh_sentence_3"
},
{
"operation": "hyperlink",
"time": "196.414",
"value": "Cairo"
},
{
"operation": "Now on",
"time": "199.287",
"value": "Cairo"
},
{
"operation": "Highlighting",
"time": "239.674",
"value": "Cairo_sentence_2"
},
{
"operation": "finish",
"time": "262.653",
"value": "finish"
}
]
}
|
15676
|
Six of the Hindu temples built in different cities across India have Lakshmi Narayan as the temple deity.
|
{
"evidence": [
"List of Hindu temples in India_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:\nSix of the Hindu temples built in different cities across India have Lakshmi Narayan as the temple deity.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"List of Hindu temples in India_sentence_1"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1060,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.675",
"value": "?search="
},
{
"operation": "search",
"time": "22.785",
"value": "Contains...Hindu temples in india"
},
{
"operation": "Now on",
"time": "24.517",
"value": "undefined"
},
{
"operation": "search",
"time": "44.504",
"value": "Lakshmi Narayan"
},
{
"operation": "Now on",
"time": "45.357",
"value": "Lakshmi Narayan"
},
{
"operation": "search",
"time": "76.244",
"value": "Contains...Hindu temples in india"
},
{
"operation": "Now on",
"time": "77.024",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "79.272",
"value": "Ramanathaswamy Temple"
},
{
"operation": "Now on",
"time": "80.277",
"value": "Ramanathaswamy Temple"
},
{
"operation": "Page search",
"time": "94.037",
"value": "Lakshmi Narayan"
},
{
"operation": "search",
"time": "123.104",
"value": "Contains...India Hindu temples"
},
{
"operation": "Now on",
"time": "124.386",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "131.828",
"value": "Hindus"
},
{
"operation": "Now on",
"time": "132.909",
"value": "Hindus"
},
{
"operation": "search",
"time": "156.877",
"value": "Hindu temple"
},
{
"operation": "Now on",
"time": "158.694",
"value": "Hindu temple"
},
{
"operation": "hyperlink",
"time": "164.798",
"value": "India"
},
{
"operation": "Now on",
"time": "165.659",
"value": "India"
},
{
"operation": "hyperlink",
"time": "187.048",
"value": "8 Demographics, languages, and religion India_section_16"
},
{
"operation": "hyperlink",
"time": "190.357",
"value": "Religion in India"
},
{
"operation": "Now on",
"time": "192.141",
"value": "Religion in India"
},
{
"operation": "hyperlink",
"time": "200.251",
"value": "3.1 Hinduism Religion in India_section_14"
},
{
"operation": "search",
"time": "228.132",
"value": "\"Lakshmi Narayan as the temple deity.\""
},
{
"operation": "Now on",
"time": "229.815",
"value": "undefined"
},
{
"operation": "search",
"time": "240.584",
"value": "Lakshmi Narayan Temple"
},
{
"operation": "Now on",
"time": "241.969",
"value": "Laxmi Narayan Mandir"
},
{
"operation": "hyperlink",
"time": "245.423",
"value": "Lakshmi Narayan"
},
{
"operation": "Now on",
"time": "245.705",
"value": "Lakshmi Narayan"
},
{
"operation": "hyperlink",
"time": "289.626",
"value": "Sri Sampradaya"
},
{
"operation": "Now on",
"time": "290.735",
"value": "Sri Vaishnavism"
},
{
"operation": "search",
"time": "306.871",
"value": "Lakshmi Narayan"
},
{
"operation": "Now on",
"time": "308.063",
"value": "Lakshmi Narayan"
},
{
"operation": "hyperlink",
"time": "322.36",
"value": "3 Worship Lakshmi Narayan_section_2"
},
{
"operation": "Highlighting",
"time": "333.019",
"value": "Lakshmi Narayan_sentence_3"
},
{
"operation": "Highlighting",
"time": "333.02",
"value": "Lakshmi Narayan_title"
},
{
"operation": "hyperlink",
"time": "337.145",
"value": "Vishnu"
},
{
"operation": "Now on",
"time": "338.385",
"value": "Vishnu"
},
{
"operation": "hyperlink",
"time": "357.709",
"value": "9 Temples Vishnu_section_35"
},
{
"operation": "Highlighting",
"time": "365.819",
"value": "Vishnu_sentence_217"
},
{
"operation": "Highlighting",
"time": "365.82",
"value": "Vishnu_title"
},
{
"operation": "Highlighting",
"time": "365.822",
"value": "Vishnu_section_35"
},
{
"operation": "Highlighting",
"time": "374.754",
"value": "Vishnu_sentence_221"
},
{
"operation": "Highlighting",
"time": "386.422",
"value": "Vishnu_sentence_220"
},
{
"operation": "search",
"time": "418.055",
"value": "\"hindu temples in India\""
},
{
"operation": "Now on",
"time": "418.678",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "422.145",
"value": "List of Hindu temples in India"
},
{
"operation": "Now on",
"time": "424.422",
"value": "List of Hindu temples in India"
},
{
"operation": "Highlighting",
"time": "451.221",
"value": "List of Hindu temples in India_sentence_0"
},
{
"operation": "Highlighting",
"time": "452.139",
"value": "List of Hindu temples in India_sentence_1"
},
{
"operation": "finish",
"time": "499.885",
"value": "finish"
}
]
}
|
exfever_6377
|
Hannibal was one of the most influential Carthaginian generals and statesmen.
|
{
"evidence": [
"Hannibal",
"First Punic War"
],
"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:\nHannibal was one of the most influential Carthaginian generals and statesmen.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Hannibal",
"First Punic War"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1061,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_8016
|
Black Sails is Jonathan E. Steinberg's third television series project.
|
{
"evidence": [
"Black Sails (TV series)",
"YouTube"
],
"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:\nBlack Sails is Jonathan E. Steinberg's third television series project.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Black Sails (TV series)",
"YouTube"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1062,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_5874
|
Cougar Town is an American television sitcom produced by the production company of an American screenwriter, producer, and director.
|
{
"evidence": [
"Cougar Town",
"Doozer (company)",
"Bill Lawrence (TV producer)"
],
"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:\nCougar Town is an American television sitcom produced by the production company of an American screenwriter, producer, and director.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Cougar Town",
"Doozer (company)",
"Bill Lawrence (TV producer)"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1063,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
23316
|
Jesús Manzano is a retired professional cyclist who was born on 12 May 1978, in San Lorenzo de El Escorial, a town and municipality in the Community of Madrid, Spain, located to the northwest of the region in the southeastern side of the Sierra de Guadarrama, at the foot of Mount Abantos and L.as Machotas, 47 kilometres (29 mi) from Madrid
|
{
"evidence": [
"Jesús Manzano_sentence_0",
"San Lorenzo de El Escorial_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:\nJesús Manzano is a retired professional cyclist who was born on 12 May 1978, in San Lorenzo de El Escorial, a town and municipality in the Community of Madrid, Spain, located to the northwest of the region in the southeastern side of the Sierra de Guadarrama, at the foot of Mount Abantos and L.as Machotas, 47 kilometres (29 mi) from Madrid\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Jesús Manzano_sentence_0",
"San Lorenzo de El Escorial_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1064,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.305",
"value": "?search="
},
{
"operation": "search",
"time": "19.235",
"value": "Jesús Manzano"
},
{
"operation": "Now on",
"time": "20.382",
"value": "Jesús Manzano"
},
{
"operation": "Highlighting",
"time": "31.047",
"value": "Jesús Manzano_sentence_0"
},
{
"operation": "hyperlink",
"time": "45.713",
"value": "San Lorenzo de El Escorial"
},
{
"operation": "Now on",
"time": "46.839",
"value": "San Lorenzo de El Escorial"
},
{
"operation": "Highlighting",
"time": "51.496",
"value": "San Lorenzo de El Escorial_sentence_1"
},
{
"operation": "back-button-clicked",
"time": "96.704",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "97.161",
"value": "Jesús Manzano"
},
{
"operation": "finish",
"time": "149.144",
"value": "finish"
}
]
}
|
47346
|
Sylhet is home to world's largest tea plantations, both in terms of area and production and Sylhet's per capita GDP is $3,050 in nominal terms and 6,250 in purchasing power parity terms.
|
{
"evidence": [
"Sylhet_sentence_91",
"Sylhet_sentence_95"
],
"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:\nSylhet is home to world's largest tea plantations, both in terms of area and production and Sylhet's per capita GDP is $3,050 in nominal terms and 6,250 in purchasing power parity terms.\n",
"role": "user"
}
] |
Search terms not in claim
|
{
"ground_truth": {
"evidence": [
"Sylhet_sentence_91",
"Sylhet_sentence_95"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1065,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.72",
"value": "?search="
},
{
"operation": "search",
"time": "9.156",
"value": "Sylhet"
},
{
"operation": "Now on",
"time": "10.492",
"value": "Sylhet"
},
{
"operation": "Highlighting",
"time": "23.122",
"value": "Sylhet_sentence_9"
},
{
"operation": "Highlighting",
"time": "65.682",
"value": "Sylhet_sentence_91"
},
{
"operation": "Highlighting",
"time": "113.161",
"value": "Sylhet_sentence_95"
},
{
"operation": "hyperlink",
"time": "191.26",
"value": "Sylhet (disambiguation)"
},
{
"operation": "Now on",
"time": "192.042",
"value": "Sylhet (disambiguation)"
},
{
"operation": "hyperlink",
"time": "199.108",
"value": "Sylhet"
},
{
"operation": "Now on",
"time": "199.837",
"value": "Sylhet"
},
{
"operation": "hyperlink",
"time": "250.796",
"value": "2011 Bangladesh census"
},
{
"operation": "Now on",
"time": "251.305",
"value": "2011 Bangladesh census"
},
{
"operation": "search",
"time": "269.977",
"value": "Sylhet GDP"
},
{
"operation": "Now on",
"time": "271.502",
"value": "undefined"
},
{
"operation": "search",
"time": "720.819",
"value": "Sylhet"
},
{
"operation": "Now on",
"time": "721.909",
"value": "Sylhet"
},
{
"operation": "hyperlink",
"time": "1200.603",
"value": "Sylhet City Corporation"
},
{
"operation": "Now on",
"time": "1201.393",
"value": "Sylhet City Corporation"
},
{
"operation": "hyperlink",
"time": "1317.947",
"value": "Sylhet"
},
{
"operation": "Now on",
"time": "1318.345",
"value": "Sylhet"
},
{
"operation": "hyperlink",
"time": "1418.273",
"value": "Sylhet District"
},
{
"operation": "Now on",
"time": "1418.937",
"value": "Sylhet District"
},
{
"operation": "hyperlink",
"time": "1447.737",
"value": "Sylhet (disambiguation)"
},
{
"operation": "Now on",
"time": "1448.376",
"value": "Sylhet (disambiguation)"
},
{
"operation": "hyperlink",
"time": "1454.27",
"value": "Sylhet"
},
{
"operation": "Now on",
"time": "1454.669",
"value": "Sylhet"
},
{
"operation": "search",
"time": "1506.069",
"value": "GDP Bengaldesh cities"
},
{
"operation": "Now on",
"time": "1507.174",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "1509.35",
"value": "Bangladesh"
},
{
"operation": "Now on",
"time": "1512.471",
"value": "Bangladesh"
},
{
"operation": "search",
"time": "1524.012",
"value": "Bangladesh cities"
},
{
"operation": "Now on",
"time": "1525.43",
"value": "undefined"
},
{
"operation": "search",
"time": "1536.716",
"value": "Bangladesh cities"
},
{
"operation": "Now on",
"time": "1537.302",
"value": "undefined"
},
{
"operation": "search",
"time": "1547.027",
"value": "Bangladesh municipalities"
},
{
"operation": "Now on",
"time": "1547.881",
"value": "undefined"
},
{
"operation": "search",
"time": "1557.876",
"value": "Bangladesh GDP"
},
{
"operation": "Now on",
"time": "1591.546",
"value": "undefined"
},
{
"operation": "search",
"time": "1607.911",
"value": "Bangladesh municipalities"
},
{
"operation": "Now on",
"time": "1608.516",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "1619.484",
"value": "Fazlul Karim (lawyer)"
},
{
"operation": "Now on",
"time": "1620.148",
"value": "Fazlul Karim (lawyer)"
},
{
"operation": "finish",
"time": "1639.249",
"value": "finish"
}
]
}
|
exfever_3639
|
Game of Thrones is an American fantasy drama television series, an adaptation of A Song of Ice and Fire, a fantasy novel by an American novelist.
|
{
"evidence": [
"Game of Thrones",
"George R. R. Martin"
],
"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:\nGame of Thrones is an American fantasy drama television series, an adaptation of A Song of Ice and Fire, a fantasy novel by an American novelist.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Game of Thrones",
"George R. R. Martin"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1066,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
9529
|
Fernande Barrey had an open relationship with her husband and got divorced in 1928, then the following year, she exhibited the paintings Josiane and Les Pêches at the Salon d'Automne.
|
{
"evidence": [
"Fernande Barrey_sentence_7",
"Fernande Barrey_sentence_10",
"Fernande Barrey_sentence_13"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nFernande Barrey had an open relationship with her husband and got divorced in 1928, then the following year, she exhibited the paintings Josiane and Les Pêches at the Salon d'Automne.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Fernande Barrey_sentence_7",
"Fernande Barrey_sentence_10",
"Fernande Barrey_sentence_13"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1067,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.472",
"value": "?search="
},
{
"operation": "search",
"time": "7.749",
"value": "fernande barrey"
},
{
"operation": "Now on",
"time": "9.42",
"value": "Fernande Barrey"
},
{
"operation": "Page search",
"time": "54.331",
"value": "open"
},
{
"operation": "Highlighting",
"time": "58.046",
"value": "Fernande Barrey_sentence_7"
},
{
"operation": "Highlighting",
"time": "63.661",
"value": "Fernande Barrey_sentence_10"
},
{
"operation": "Page search",
"time": "84.119",
"value": "josi"
},
{
"operation": "Highlighting",
"time": "88.032",
"value": "Fernande Barrey_sentence_13"
},
{
"operation": "finish",
"time": "95.738",
"value": "finish"
}
]
}
|
exfever_8300
|
Boomerang is the second American romantic comedy film directed by Reginald Hudlin.
|
{
"evidence": [
"Boomerang (1992 film)",
"Martin Lawrence",
"Life (1999 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:\nBoomerang is the second American romantic comedy film directed by Reginald Hudlin.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Boomerang (1992 film)",
"Martin Lawrence",
"Life (1999 film)"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1068,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
76525
|
Saunders Coast is the Marie Byrd Land between Cape Colbeck and Brennan Point from 158°01'W to 146°31'W.
|
{
"evidence": [
"Saunders Coast_sentence_0",
"Saunders Coast_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:\nSaunders Coast is the Marie Byrd Land between Cape Colbeck and Brennan Point from 158°01'W to 146°31'W.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Saunders Coast_sentence_0",
"Saunders Coast_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1069,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.704",
"value": "?search="
},
{
"operation": "search",
"time": "17.695",
"value": "Saunders Coast"
},
{
"operation": "Now on",
"time": "18.577",
"value": "Saunders Coast"
},
{
"operation": "Highlighting",
"time": "28.742",
"value": "Saunders Coast_sentence_0"
},
{
"operation": "Highlighting",
"time": "34.622",
"value": "Saunders Coast_sentence_1"
},
{
"operation": "finish",
"time": "40.787",
"value": "finish"
}
]
}
|
24014
|
Felipe Clemente de Diego y Gutiérrez, a Spanish jurist who rebuilded the Spanish state and its judiciary after the Francoist victory in the Spanish Civil War (lasted from 1936 to 1939), taught Roman law in Santiago in 1897–99 and civil law in Valladolid, Barcelona and Madrid until 1936.
|
{
"evidence": [
"Felipe Clemente de Diego y Gutiérrez_sentence_0",
"Spanish Civil War_sentence_2",
"Felipe Clemente de Diego y Gutiérrez_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:\nFelipe Clemente de Diego y Gutiérrez, a Spanish jurist who rebuilded the Spanish state and its judiciary after the Francoist victory in the Spanish Civil War (lasted from 1936 to 1939), taught Roman law in Santiago in 1897–99 and civil law in Valladolid, Barcelona and Madrid until 1936.\n",
"role": "user"
}
] |
Multi-hop Reasoning
|
{
"ground_truth": {
"evidence": [
"Felipe Clemente de Diego y Gutiérrez_sentence_0",
"Spanish Civil War_sentence_2",
"Felipe Clemente de Diego y Gutiérrez_sentence_1"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1070,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.511",
"value": "?search="
},
{
"operation": "search",
"time": "61.954",
"value": "Felipe Clemente de Diego y Gutiérrez"
},
{
"operation": "Now on",
"time": "62.471",
"value": "Felipe Clemente de Diego y Gutiérrez"
},
{
"operation": "Highlighting",
"time": "74.881",
"value": "Felipe Clemente de Diego y Gutiérrez_sentence_0"
},
{
"operation": "hyperlink",
"time": "76.888",
"value": "Spanish Civil War"
},
{
"operation": "Now on",
"time": "77.901",
"value": "Spanish Civil War"
},
{
"operation": "Highlighting",
"time": "82.44",
"value": "Spanish Civil War_sentence_2"
},
{
"operation": "back-button-clicked",
"time": "84.465",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "84.51",
"value": "Felipe Clemente de Diego y Gutiérrez"
},
{
"operation": "Highlighting",
"time": "100.632",
"value": "Felipe Clemente de Diego y Gutiérrez_sentence_1"
},
{
"operation": "finish",
"time": "106.167",
"value": "finish"
}
]
}
|
61932
|
Thomas Michell is a modern regular civilian without a political background.
|
{
"evidence": [
"Thomas Michell_sentence_2",
"Thomas Michell_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:\nThomas Michell is a modern regular civilian without a political background.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Thomas Michell_sentence_2",
"Thomas Michell_sentence_3"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1071,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.107",
"value": "?search="
},
{
"operation": "search",
"time": "10.156",
"value": "Thomas Mitchell"
},
{
"operation": "Now on",
"time": "11.678",
"value": "Thomas Mitchell"
},
{
"operation": "hyperlink",
"time": "14.296",
"value": "1 Politics Thomas Mitchell_section_0"
},
{
"operation": "hyperlink",
"time": "31.145",
"value": "Thomas R. Mitchell"
},
{
"operation": "Now on",
"time": "32.489",
"value": "Thomas R. Mitchell"
},
{
"operation": "Highlighting",
"time": "40.716",
"value": "Thomas R. Mitchell_sentence_4"
},
{
"operation": "Highlighting",
"time": "40.718",
"value": "Thomas R. Mitchell_title"
},
{
"operation": "back-button-clicked",
"time": "45.954",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "46.117",
"value": "Thomas Mitchell#Politics Thomas Mitchell section 0"
},
{
"operation": "hyperlink",
"time": "49.489",
"value": "Thomas B. Mitchell"
},
{
"operation": "Now on",
"time": "50.896",
"value": "Thomas B. Mitchell"
},
{
"operation": "Highlighting",
"time": "60.352",
"value": "Thomas B. Mitchell_sentence_2"
},
{
"operation": "Highlighting",
"time": "60.353",
"value": "Thomas B. Mitchell_title"
},
{
"operation": "Highlighting",
"time": "60.355",
"value": "Thomas B. Mitchell_section_0"
},
{
"operation": "back-button-clicked",
"time": "64.029",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "64.205",
"value": "Thomas Mitchell#Politics Thomas Mitchell section 0"
},
{
"operation": "hyperlink",
"time": "71.405",
"value": "Thomas Alexander Mitchell"
},
{
"operation": "Now on",
"time": "72.2",
"value": "Thomas Alexander Mitchell"
},
{
"operation": "Highlighting",
"time": "75.107",
"value": "Thomas Alexander Mitchell_sentence_0"
},
{
"operation": "Highlighting",
"time": "75.111",
"value": "Thomas Alexander Mitchell_title"
},
{
"operation": "back-button-clicked",
"time": "103.554",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "103.689",
"value": "Thomas Mitchell#Politics Thomas Mitchell section 0"
},
{
"operation": "hyperlink",
"time": "106.345",
"value": "Thomas B. Mitchell"
},
{
"operation": "Now on",
"time": "107.216",
"value": "Thomas B. Mitchell"
},
{
"operation": "back-button-clicked",
"time": "119.813",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "119.939",
"value": "Thomas Mitchell#Politics Thomas Mitchell section 0"
},
{
"operation": "Highlighting",
"time": "121.93",
"value": "Thomas Mitchell_item_0_0"
},
{
"operation": "Highlighting",
"time": "121.932",
"value": "Thomas Mitchell_title"
},
{
"operation": "Highlighting",
"time": "121.934",
"value": "Thomas Mitchell_section_0"
},
{
"operation": "Highlighting",
"time": "122.632",
"value": "Thomas Mitchell_item_0_1"
},
{
"operation": "Highlighting",
"time": "123.467",
"value": "Thomas Mitchell_item_0_2"
},
{
"operation": "Highlighting",
"time": "124.232",
"value": "Thomas Mitchell_item_0_3"
},
{
"operation": "Highlighting",
"time": "125.097",
"value": "Thomas Mitchell_item_0_4"
},
{
"operation": "Highlighting",
"time": "127.998",
"value": "Thomas Mitchell_item_1_5"
},
{
"operation": "Highlighting",
"time": "128.001",
"value": "Thomas Mitchell_section_1"
},
{
"operation": "Highlighting",
"time": "128.956",
"value": "Thomas Mitchell_item_1_6"
},
{
"operation": "Highlighting",
"time": "129.798",
"value": "Thomas Mitchell_item_1_7"
},
{
"operation": "Highlighting",
"time": "130.921",
"value": "Thomas Mitchell_item_1_8"
},
{
"operation": "Highlighting",
"time": "131.857",
"value": "Thomas Mitchell_item_1_9"
},
{
"operation": "Highlighting",
"time": "132.678",
"value": "Thomas Mitchell_item_1_10"
},
{
"operation": "Highlighting",
"time": "133.359",
"value": "Thomas Mitchell_item_1_11"
},
{
"operation": "Highlighting",
"time": "134.021",
"value": "Thomas Mitchell_item_1_12"
},
{
"operation": "Highlighting",
"time": "136.209",
"value": "Thomas Mitchell_item_2_13"
},
{
"operation": "Highlighting",
"time": "136.212",
"value": "Thomas Mitchell_section_2"
},
{
"operation": "Highlighting",
"time": "136.973",
"value": "Thomas Mitchell_item_2_14"
},
{
"operation": "Highlighting",
"time": "137.93",
"value": "Thomas Mitchell_item_2_15"
},
{
"operation": "Highlighting",
"time": "139.002",
"value": "Thomas Mitchell_item_2_16"
},
{
"operation": "Highlighting",
"time": "139.66",
"value": "Thomas Mitchell_item_2_18"
},
{
"operation": "Highlighting",
"time": "140.66",
"value": "Thomas Mitchell_item_2_17"
},
{
"operation": "Highlighting",
"time": "141.853",
"value": "Thomas Mitchell_item_2_19"
},
{
"operation": "Highlighting",
"time": "142.496",
"value": "Thomas Mitchell_item_2_20"
},
{
"operation": "search",
"time": "180.887",
"value": "Thomas Michell"
},
{
"operation": "Now on",
"time": "182.079",
"value": "Thomas Michell"
},
{
"operation": "Highlighting",
"time": "189.148",
"value": "Thomas Michell_sentence_3"
},
{
"operation": "Highlighting",
"time": "194.273",
"value": "Thomas Michell_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "197.024",
"value": "Thomas Michell_sentence_0"
},
{
"operation": "Highlighting",
"time": "197.809",
"value": "Thomas Michell_sentence_2"
},
{
"operation": "Highlighting",
"time": "201.267",
"value": "Thomas Michell_sentence_0"
},
{
"operation": "Highlighting deleted",
"time": "205.344",
"value": "Thomas Michell_sentence_0"
},
{
"operation": "Highlighting",
"time": "228.089",
"value": "Thomas Michell_sentence_2"
},
{
"operation": "Highlighting",
"time": "228.684",
"value": "Thomas Michell_sentence_3"
},
{
"operation": "Now on",
"time": "238.9",
"value": "Thomas Michell"
},
{
"operation": "finish",
"time": "256.694",
"value": "finish"
}
]
}
|
exfever_8363
|
Terminator Genisys is a Canadian science fiction action film directed by an American television and film director known for his work on TV shows such as Lost, The West Wing, Six Feet Under, Sex and the City, The Sopranos, Game of Thrones, Boardwalk Empire, Deadwood, and Mad Men.
|
{
"evidence": [
"Terminator Genisys",
"Alan Taylor (director)"
],
"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:\nTerminator Genisys is a Canadian science fiction action film directed by an American television and film director known for his work on TV shows such as Lost, The West Wing, Six Feet Under, Sex and the City, The Sopranos, Game of Thrones, Boardwalk Empire, Deadwood, and Mad Men.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Terminator Genisys",
"Alan Taylor (director)"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1072,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
67751
|
Juan Dies, who taught at Earlham College and Indiana University, is the co-founder and executive director of Sones de Mexico Ensemble, a Chicago folk music group that specializes in the Mexican musical tradition known as son.
|
{
"evidence": [
"Juan Dies_sentence_2",
"Juan Dies_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:\nJuan Dies, who taught at Earlham College and Indiana University, is the co-founder and executive director of Sones de Mexico Ensemble, a Chicago folk music group that specializes in the Mexican musical tradition known as son.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Juan Dies_sentence_2",
"Juan Dies_sentence_0"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1073,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.426",
"value": "?search="
},
{
"operation": "search",
"time": "492.415",
"value": "Juan Dies"
},
{
"operation": "Now on",
"time": "493.531",
"value": "Juan Dies"
},
{
"operation": "Highlighting",
"time": "515.131",
"value": "Juan Dies_sentence_2"
},
{
"operation": "Highlighting",
"time": "532.579",
"value": "Juan Dies_sentence_0"
},
{
"operation": "finish",
"time": "572.403",
"value": "finish"
}
]
}
|
50458
|
Keith Primeau suffered a concussion on October 25 in Canada and missed the rest of the season and the playoffs; then in late January, Hatcher was named team captain for the duration of Primeau's absence.
|
{
"evidence": [
"Philadelphia Flyers_sentence_298"
],
"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:\nKeith Primeau suffered a concussion on October 25 in Canada and missed the rest of the season and the playoffs; then in late January, Hatcher was named team captain for the duration of Primeau's absence.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Philadelphia Flyers_sentence_298"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1074,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.914",
"value": "?search="
},
{
"operation": "search",
"time": "9.344",
"value": "Keith Primeau"
},
{
"operation": "Now on",
"time": "10.699",
"value": "Keith Primeau"
},
{
"operation": "Highlighting",
"time": "441.944",
"value": "Keith Primeau_sentence_0"
},
{
"operation": "Highlighting",
"time": "441.945",
"value": "Keith Primeau_title"
},
{
"operation": "Highlighting",
"time": "480.284",
"value": "Keith Primeau_sentence_22"
},
{
"operation": "Highlighting",
"time": "480.285",
"value": "Keith Primeau_section_0"
},
{
"operation": "Highlighting",
"time": "487.877",
"value": "Keith Primeau_sentence_23"
},
{
"operation": "hyperlink",
"time": "512.087",
"value": "2005–06 season"
},
{
"operation": "Now on",
"time": "514.861",
"value": "2005–06 NHL season"
},
{
"operation": "search",
"time": "551.343",
"value": "Keith Primeau"
},
{
"operation": "Now on",
"time": "552.003",
"value": "Keith Primeau"
},
{
"operation": "hyperlink",
"time": "630.869",
"value": "Philadelphia Flyers"
},
{
"operation": "Now on",
"time": "636.77",
"value": "Philadelphia Flyers"
},
{
"operation": "Highlighting",
"time": "766.554",
"value": "Philadelphia Flyers_sentence_297"
},
{
"operation": "Highlighting",
"time": "776.44",
"value": "Philadelphia Flyers_sentence_298"
},
{
"operation": "finish",
"time": "810.323",
"value": "finish"
}
]
}
|
29441
|
The company Quantopian ran a three-sided market business model, the first side of which consisted of algorithm-developer members who developed and tested for free.
|
{
"evidence": [
"Quantopian_sentence_9"
],
"label": "REFUTE"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nThe company Quantopian ran a three-sided market business model, the first side of which consisted of algorithm-developer members who developed and tested for free.\n",
"role": "user"
}
] |
Numerical Reasoning
|
{
"ground_truth": {
"evidence": [
"Quantopian_sentence_9"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1075,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.29",
"value": "?search="
},
{
"operation": "search",
"time": "14.848",
"value": "Quantopian"
},
{
"operation": "Now on",
"time": "15.845",
"value": "Quantopian"
},
{
"operation": "Highlighting",
"time": "40.167",
"value": "Quantopian_sentence_9"
},
{
"operation": "finish",
"time": "58.291",
"value": "finish"
}
]
}
|
exfever_6124
|
Katy Perry promoted Adidas on Instagram.
|
{
"evidence": [
"Katy Perry",
"2010s in fashion"
],
"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:\nKaty Perry promoted Adidas on Instagram.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Katy Perry",
"2010s in fashion"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1076,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_4053
|
Alvin and the Chipmunks is Cameron Richardson's eighth film.
|
{
"evidence": [
"Alvin and the Chipmunks (film)",
"Alvin and the Chipmunks"
],
"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:\nAlvin and the Chipmunks is Cameron Richardson's eighth film.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Alvin and the Chipmunks (film)",
"Alvin and the Chipmunks"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1077,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
50969
|
The Rajiv Gandhi College of Veterinary and Animal Sciences houses Campus A and B and can be reached by train from Pondicherry to Thirukkanur.
|
{
"evidence": [
"Rajiv Gandhi College of Veterinary and Animal Sciences_sentence_16"
],
"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 Rajiv Gandhi College of Veterinary and Animal Sciences houses Campus A and B and can be reached by train from Pondicherry to Thirukkanur.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Rajiv Gandhi College of Veterinary and Animal Sciences_sentence_16"
],
"label": "REFUTE"
},
"style": "rule"
}
|
{
"index": 1078,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.658",
"value": "undefined"
},
{
"operation": "search",
"time": "28.61",
"value": "Rajiv Gandhi College of Veterinary and Animal Sciences"
},
{
"operation": "Now on",
"time": "29.628",
"value": "Rajiv Gandhi College of Veterinary and Animal Sciences"
},
{
"operation": "Page search",
"time": "55.552",
"value": "campus"
},
{
"operation": "Page search",
"time": "64.725",
"value": "campus a"
},
{
"operation": "Highlighting",
"time": "89.354",
"value": "Rajiv Gandhi College of Veterinary and Animal Sciences_sentence_17"
},
{
"operation": "Page search",
"time": "107.721",
"value": "pondicherry"
},
{
"operation": "Page search",
"time": "122.878",
"value": "train"
},
{
"operation": "Page search",
"time": "159.329",
"value": "thirukanur"
},
{
"operation": "Highlighting",
"time": "183.036",
"value": "Rajiv Gandhi College of Veterinary and Animal Sciences_sentence_16"
},
{
"operation": "Highlighting deleted",
"time": "185.11",
"value": "Rajiv Gandhi College of Veterinary and Animal Sciences_sentence_17"
},
{
"operation": "finish",
"time": "260.501",
"value": "finish"
}
]
}
|
exfever_2171
|
The 66th Primetime Emmy Awards honored the best in U.S. television programming as chosen by the Academy of Television Arts & Sciences for broadcast programming for a television show.
|
{
"evidence": [
"66th Primetime Emmy Awards",
"Prime time"
],
"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 66th Primetime Emmy Awards honored the best in U.S. television programming as chosen by the Academy of Television Arts & Sciences for broadcast programming for a television show.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"66th Primetime Emmy Awards",
"Prime time"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1079,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
92068
|
Paolo Abrera, the youngest child of Carlos Abrera and Emily Altomonte-Abrera and the husband of prominent television host Suzi Entrata, is a broadcaster, television host, lifestyle columnist, an age-group triathlete, avid outdoors-man, and environmentalist in the Philippines.
|
{
"evidence": [
"Paolo Abrera_sentence_1",
"Paolo Abrera_sentence_2",
"Paolo Abrera_sentence_7"
],
"label": "SUPPORT"
}
|
feverous
|
[
{
"content": "You are a claim-verification assistant. You MUST follow this protocol exactly:\n\n<plan>…</plan>\n- Once at the start: sketch your high-level strategy, such as claim decomposition, entity recognition, etc.\n\n<search>…</search>\n- When you need a fact: emit exactly this tag with your query.\n- To make the most of your search turns, don’t repeat identical queries.\n- You can search at most three times.\n\n<information>\n[[e_1]]: info1\n[[e_2]]: info2\n…\n</information>\n- You will be given claim related information in the format above.\n\n<think>…</think> \n- Use for every piece of reasoning; do not state your final verdict here.\n- You must conduct reasoning inside <think> and </think> first every time you get new information.\n\n<answer>\nLabel: SUPPORT / REFUTE / NOT ENOUGH INFO\nEvidence: [[e_1]], [[e_3]], ...\n</answer> \n- Emit exactly once at the end, no extra text or tags.\n- Evidence id such as e_1 will be replaced by real ids from the corpus. You must include useful real ids when answering\n- Evidence outputs must strictly enforce the format [[e_i]], [[e_j]]…\n- Answer Labels respectively stand for:\nSUPPORT: The claim is consistent with the cited evidence and the evidence is sufficient to confirm the claim.\nREFUTE: The claim contradicts the cited evidence and the evidence is sufficient to disprove the claim.\nNOT ENOUGH INFO: The available evidence is insufficient to determine whether the claim is true or false.\n\n- Note that the process should be: plan → (search → information → think) repeat until conclusion → answer\n\nVerify the claim:\nPaolo Abrera, the youngest child of Carlos Abrera and Emily Altomonte-Abrera and the husband of prominent television host Suzi Entrata, is a broadcaster, television host, lifestyle columnist, an age-group triathlete, avid outdoors-man, and environmentalist in the Philippines.\n",
"role": "user"
}
] |
Entity Disambiguation
|
{
"ground_truth": {
"evidence": [
"Paolo Abrera_sentence_1",
"Paolo Abrera_sentence_2",
"Paolo Abrera_sentence_7"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1080,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "1.211",
"value": "?search="
},
{
"operation": "search",
"time": "12.391",
"value": "Paolo Abrera"
},
{
"operation": "Now on",
"time": "13.746",
"value": "Paolo Abrera"
},
{
"operation": "hyperlink",
"time": "16.541",
"value": "television host"
},
{
"operation": "Now on",
"time": "17.085",
"value": "Television host"
},
{
"operation": "back-button-clicked",
"time": "18.394",
"value": "back-button-clicked"
},
{
"operation": "Now on",
"time": "18.462",
"value": "Paolo Abrera"
},
{
"operation": "Highlighting",
"time": "19.861",
"value": "Paolo Abrera_sentence_1"
},
{
"operation": "Highlighting",
"time": "31.13",
"value": "Paolo Abrera_sentence_2"
},
{
"operation": "Highlighting",
"time": "47.216",
"value": "Paolo Abrera_sentence_7"
},
{
"operation": "finish",
"time": "70.222",
"value": "finish"
}
]
}
|
8885
|
In the Clayton-Bulwer Treaty, Britain ceded the Bay Islands to Honduras, and ceded suzerainty over the Mosquito Coast to Nicaragua, which was nearly in accord with all of the American contentions.
|
{
"evidence": [
"Clayton–Bulwer Treaty_sentence_23",
"Clayton–Bulwer Treaty_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:\nIn the Clayton-Bulwer Treaty, Britain ceded the Bay Islands to Honduras, and ceded suzerainty over the Mosquito Coast to Nicaragua, which was nearly in accord with all of the American contentions.\n",
"role": "user"
}
] |
Other
|
{
"ground_truth": {
"evidence": [
"Clayton–Bulwer Treaty_sentence_23",
"Clayton–Bulwer Treaty_sentence_22"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1081,
"split": "train_online"
}
|
{
"annotator_operations": [
{
"operation": "start",
"time": "0",
"value": "start"
},
{
"operation": "Now on",
"time": "0.997",
"value": "?search="
},
{
"operation": "search",
"time": "14.702",
"value": "Clayton-Bulwer Treaty,"
},
{
"operation": "Now on",
"time": "16.697",
"value": "undefined"
},
{
"operation": "hyperlink",
"time": "21.17",
"value": "Clayton–Bulwer Treaty"
},
{
"operation": "Now on",
"time": "22.154",
"value": "Clayton–Bulwer Treaty"
},
{
"operation": "Page search",
"time": "34.951",
"value": "honduras"
},
{
"operation": "Highlighting",
"time": "48.273",
"value": "Clayton–Bulwer Treaty_sentence_5"
},
{
"operation": "Page search",
"time": "56.461",
"value": "ceded"
},
{
"operation": "Highlighting",
"time": "69.798",
"value": "Clayton–Bulwer Treaty_sentence_23"
},
{
"operation": "Highlighting deleted",
"time": "75.231",
"value": "Clayton–Bulwer Treaty_sentence_5"
},
{
"operation": "Highlighting",
"time": "109.864",
"value": "Clayton–Bulwer Treaty_sentence_22"
},
{
"operation": "finish",
"time": "142.95",
"value": "finish"
}
]
}
|
exfever_8219
|
One of Leonardo DiCaprio's best movies of all time is the Titanic, where he received lots of accolades from AMPAS.
|
{
"evidence": [
"Django Unchained",
"James Remar"
],
"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:\nOne of Leonardo DiCaprio's best movies of all time is the Titanic, where he received lots of accolades from AMPAS.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Django Unchained",
"James Remar"
],
"label": "NOT ENOUGH INFO"
},
"style": "rule"
}
|
{
"index": 1082,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
exfever_1733
|
Resident Evil is a science fiction action-horror film series based on the Japanese horror video game series and media franchise created by Capcom.
|
{
"evidence": [
"Resident Evil (film series)",
"Resident Evil"
],
"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:\nResident Evil is a science fiction action-horror film series based on the Japanese horror video game series and media franchise created by Capcom.\n",
"role": "user"
}
] |
EX-FEVER
|
{
"ground_truth": {
"evidence": [
"Resident Evil (film series)",
"Resident Evil"
],
"label": "SUPPORT"
},
"style": "rule"
}
|
{
"index": 1083,
"split": "train_online"
}
|
{
"annotator_operations": []
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.