DeepRetrieval: Hacking Real Search Engines and Retrievers with Large Language Models via Reinforcement Learning
Paper
•
2503.00223
•
Published
•
1
Search is not available for this dataset
qid
string | query
string | target
sequence | score
sequence | data_source
string | prompt
list | ability
string | reward_model
dict | extra_info
dict |
|---|---|---|---|---|---|---|---|---|
152297
|
Chinese people are people only associated with Russia.
|
[
"Chinese_people"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nChinese people are people only associated with Russia.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Chinese_people"
]
},
"style": "rule"
}
|
{
"index": 40658,
"split": "train"
}
|
97075
|
Sense and Sensibility originally printed 750 copies in its 1813 release.
|
[
"Sense_and_Sensibility"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nSense and Sensibility originally printed 750 copies in its 1813 release.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Sense_and_Sensibility"
]
},
"style": "rule"
}
|
{
"index": 61728,
"split": "train"
}
|
95538
|
Samuel L. Jackson was in Die Hard with a Vengeance, the third film in the Die Hard film series.
|
[
"Die_Hard_with_a_Vengeance",
"Samuel_L._Jackson"
] |
[
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nSamuel L. Jackson was in Die Hard with a Vengeance, the third film in the Die Hard film series.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1
],
"target": [
"Die_Hard_with_a_Vengeance",
"Samuel_L._Jackson"
]
},
"style": "rule"
}
|
{
"index": 106898,
"split": "train"
}
|
111814
|
Chokher Bali is entirely in English.
|
[
"Chokher_Bali_(film)"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nChokher Bali is entirely in English.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Chokher_Bali_(film)"
]
},
"style": "rule"
}
|
{
"index": 24859,
"split": "train"
}
|
109218
|
Tina Turner launched a major book deal.
|
[
"Tina_Turner"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nTina Turner launched a major book deal.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Tina_Turner"
]
},
"style": "rule"
}
|
{
"index": 31498,
"split": "train"
}
|
34930
|
Adrien Broner was born in 1989.
|
[
"Adrien_Broner"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nAdrien Broner was born in 1989.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Adrien_Broner"
]
},
"style": "rule"
}
|
{
"index": 96178,
"split": "train"
}
|
111119
|
Newspapers include criticism and reviews of the arts.
|
[
"Newspaper"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nNewspapers include criticism and reviews of the arts.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Newspaper"
]
},
"style": "rule"
}
|
{
"index": 19529,
"split": "train"
}
|
101944
|
James Marsden gained prominence with his portrayal of Peter Parker.
|
[
"James_Marsden"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nJames Marsden gained prominence with his portrayal of Peter Parker.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"James_Marsden"
]
},
"style": "rule"
}
|
{
"index": 74265,
"split": "train"
}
|
163785
|
Wishful drinking is based on a one-woman performance.
|
[
"Wishful_Drinking"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nWishful drinking is based on a one-woman performance.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Wishful_Drinking"
]
},
"style": "rule"
}
|
{
"index": 77451,
"split": "train"
}
|
189492
|
Karl Malone was teammates with John Stockton.
|
[
"Karl_Malone"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nKarl Malone was teammates with John Stockton.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Karl_Malone"
]
},
"style": "rule"
}
|
{
"index": 3316,
"split": "train"
}
|
47323
|
Aliens (film) has a character named Bishop.
|
[
"Bishop_(Aliens)",
"Aliens_(film)"
] |
[
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nAliens (film) has a character named Bishop.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1
],
"target": [
"Bishop_(Aliens)",
"Aliens_(film)"
]
},
"style": "rule"
}
|
{
"index": 19874,
"split": "train"
}
|
114949
|
One of the original three Nicktoons was The Ren & Stimpy Show.
|
[
"The_Ren_&_Stimpy_Show"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nOne of the original three Nicktoons was The Ren & Stimpy Show.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"The_Ren_&_Stimpy_Show"
]
},
"style": "rule"
}
|
{
"index": 96588,
"split": "train"
}
|
24824
|
Julius Caesar's office was also called dictator in perpetuum and he died.
|
[
"Dictator_perpetuo",
"Julius_Caesar"
] |
[
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nJulius Caesar's office was also called dictator in perpetuum and he died.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1
],
"target": [
"Dictator_perpetuo",
"Julius_Caesar"
]
},
"style": "rule"
}
|
{
"index": 84027,
"split": "train"
}
|
83029
|
Game of thrones has received 38 Primetime Emmy Awards.
|
[
"Game_of_Thrones"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nGame of thrones has received 38 Primetime Emmy Awards.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Game_of_Thrones"
]
},
"style": "rule"
}
|
{
"index": 78667,
"split": "train"
}
|
9538
|
Suits (TV series) features actress Gina Torres.
|
[
"Suits_(TV_series)",
"Gina_Torres"
] |
[
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nSuits (TV series) features actress Gina Torres.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1
],
"target": [
"Suits_(TV_series)",
"Gina_Torres"
]
},
"style": "rule"
}
|
{
"index": 75314,
"split": "train"
}
|
53928
|
Gemma Arterton refuses to act.
|
[
"Gemma_Arterton"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nGemma Arterton refuses to act.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Gemma_Arterton"
]
},
"style": "rule"
}
|
{
"index": 32850,
"split": "train"
}
|
21818
|
The creators of Spider-Man gave the character the ability to fly.
|
[
"Spider-Man"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nThe creators of Spider-Man gave the character the ability to fly.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Spider-Man"
]
},
"style": "rule"
}
|
{
"index": 106950,
"split": "train"
}
|
43180
|
Adolf Hitler aimed to eliminate only pets.
|
[
"Adolf_Hitler"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nAdolf Hitler aimed to eliminate only pets.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Adolf_Hitler"
]
},
"style": "rule"
}
|
{
"index": 70082,
"split": "train"
}
|
211960
|
Jean-Martin Charcot was born in 1825.
|
[
"Jean-Martin_Charcot"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nJean-Martin Charcot was born in 1825.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Jean-Martin_Charcot"
]
},
"style": "rule"
}
|
{
"index": 23558,
"split": "train"
}
|
130033
|
Ink Master had eight seasons.
|
[
"Ink_Master"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nInk Master had eight seasons.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Ink_Master"
]
},
"style": "rule"
}
|
{
"index": 89024,
"split": "train"
}
|
90624
|
Take Me High was released on Warner Home Video as a VHS in 1988.
|
[
"Take_Me_High"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nTake Me High was released on Warner Home Video as a VHS in 1988.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Take_Me_High"
]
},
"style": "rule"
}
|
{
"index": 63004,
"split": "train"
}
|
165984
|
Bad and Boujee failed to include Lil Uzi Vert.
|
[
"Bad_and_Boujee"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nBad and Boujee failed to include Lil Uzi Vert.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Bad_and_Boujee"
]
},
"style": "rule"
}
|
{
"index": 102270,
"split": "train"
}
|
214378
|
Chandni is an American film.
|
[
"Chandni"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nChandni is an American film.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Chandni"
]
},
"style": "rule"
}
|
{
"index": 75335,
"split": "train"
}
|
187867
|
Devdas won five National Film Awards.
|
[
"Devdas_(2002_Hindi_film)"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nDevdas won five National Film Awards.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Devdas_(2002_Hindi_film)"
]
},
"style": "rule"
}
|
{
"index": 31707,
"split": "train"
}
|
25803
|
Andrew Wood and band members from Green River performed together.
|
[
"Andrew_Wood_(singer)"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nAndrew Wood and band members from Green River performed together.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Andrew_Wood_(singer)"
]
},
"style": "rule"
}
|
{
"index": 4037,
"split": "train"
}
|
37142
|
The Czech Republic has inhabitants.
|
[
"Czech_Republic"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nThe Czech Republic has inhabitants.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Czech_Republic"
]
},
"style": "rule"
}
|
{
"index": 10238,
"split": "train"
}
|
157994
|
Serbia is highly ranked on the Human Development Index.
|
[
"Serbia"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nSerbia is highly ranked on the Human Development Index.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Serbia"
]
},
"style": "rule"
}
|
{
"index": 99805,
"split": "train"
}
|
143669
|
Avatar was released in theaters.
|
[
"Avatar_(2009_film)"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nAvatar was released in theaters.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Avatar_(2009_film)"
]
},
"style": "rule"
}
|
{
"index": 102558,
"split": "train"
}
|
206058
|
Burnaby achieved City status in 1992, one hundred years after its incorporation.
|
[
"Burnaby"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nBurnaby achieved City status in 1992, one hundred years after its incorporation.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Burnaby"
]
},
"style": "rule"
}
|
{
"index": 1478,
"split": "train"
}
|
40614
|
Ministates include San Marino in Europe.
|
[
"San_Marino",
"Ministate"
] |
[
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nMinistates include San Marino in Europe.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1
],
"target": [
"San_Marino",
"Ministate"
]
},
"style": "rule"
}
|
{
"index": 80692,
"split": "train"
}
|
182889
|
Dog Day Afternoon stars at least one actor.
|
[
"Dog_Day_Afternoon",
"John_Cazale"
] |
[
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nDog Day Afternoon stars at least one actor.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1
],
"target": [
"Dog_Day_Afternoon",
"John_Cazale"
]
},
"style": "rule"
}
|
{
"index": 69828,
"split": "train"
}
|
202888
|
The Woman in Red is a 1954 film.
|
[
"The_Woman_in_Red_(1984_film)"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nThe Woman in Red is a 1954 film.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"The_Woman_in_Red_(1984_film)"
]
},
"style": "rule"
}
|
{
"index": 101504,
"split": "train"
}
|
216027
|
Iris Murdoch was illiterate.
|
[
"Iris_Murdoch"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nIris Murdoch was illiterate.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Iris_Murdoch"
]
},
"style": "rule"
}
|
{
"index": 54576,
"split": "train"
}
|
22990
|
Iron Man 3 is the first movie in a series.
|
[
"Iron_Man_(2008_film)",
"Iron_Man_3"
] |
[
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nIron Man 3 is the first movie in a series.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1
],
"target": [
"Iron_Man_(2008_film)",
"Iron_Man_3"
]
},
"style": "rule"
}
|
{
"index": 89746,
"split": "train"
}
|
193818
|
Forbes rated Dr. Dre as the second-richest figure in American hip-hop one year.
|
[
"Dr._Dre"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nForbes rated Dr. Dre as the second-richest figure in American hip-hop one year.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Dr._Dre"
]
},
"style": "rule"
}
|
{
"index": 2099,
"split": "train"
}
|
7200
|
Ringo Starr experienced two life-threatening illnesses during childhood.
|
[
"Ringo_Starr"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nRingo Starr experienced two life-threatening illnesses during childhood.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Ringo_Starr"
]
},
"style": "rule"
}
|
{
"index": 16752,
"split": "train"
}
|
193300
|
In 1942, Richard Nixon's wife moved to Washington.
|
[
"Richard_Nixon"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nIn 1942, Richard Nixon's wife moved to Washington.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Richard_Nixon"
]
},
"style": "rule"
}
|
{
"index": 9712,
"split": "train"
}
|
205234
|
Audrey Horne was introduced in Twin Peaks and she was in a David Lynch production.
|
[
"Audrey_Horne",
"Pilot_(Twin_Peaks)"
] |
[
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nAudrey Horne was introduced in Twin Peaks and she was in a David Lynch production.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1
],
"target": [
"Audrey_Horne",
"Pilot_(Twin_Peaks)"
]
},
"style": "rule"
}
|
{
"index": 38687,
"split": "train"
}
|
212483
|
Romelu Lukaku's club is Everton.
|
[
"Romelu_Lukaku"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nRomelu Lukaku's club is Everton.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Romelu_Lukaku"
]
},
"style": "rule"
}
|
{
"index": 108679,
"split": "train"
}
|
139323
|
Uzbekistan's economy relies on cotton, gold, uranium, and natural gas production.
|
[
"Uzbekistan"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nUzbekistan's economy relies on cotton, gold, uranium, and natural gas production.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Uzbekistan"
]
},
"style": "rule"
}
|
{
"index": 40668,
"split": "train"
}
|
145320
|
IBM invented the personal computer.
|
[
"IBM",
"Personal_computer"
] |
[
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nIBM invented the personal computer.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1
],
"target": [
"IBM",
"Personal_computer"
]
},
"style": "rule"
}
|
{
"index": 71003,
"split": "train"
}
|
6727
|
Anorexia nervosa is often referred to complicatedly as anorexia.
|
[
"Anorexia_nervosa"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nAnorexia nervosa is often referred to complicatedly as anorexia.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Anorexia_nervosa"
]
},
"style": "rule"
}
|
{
"index": 91141,
"split": "train"
}
|
169704
|
Brandon Lee lived throughout the entire filming of The Crow.
|
[
"Brandon_Lee"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nBrandon Lee lived throughout the entire filming of The Crow.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Brandon_Lee"
]
},
"style": "rule"
}
|
{
"index": 16400,
"split": "train"
}
|
72681
|
21st Century Fox owned the channel Sons of Anarchy was on.
|
[
"FX_(TV_channel)",
"Sons_of_Anarchy"
] |
[
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\n21st Century Fox owned the channel Sons of Anarchy was on.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1
],
"target": [
"FX_(TV_channel)",
"Sons_of_Anarchy"
]
},
"style": "rule"
}
|
{
"index": 68952,
"split": "train"
}
|
68631
|
Red Velvet is a solo act.
|
[
"Red_Velvet_(band)",
"Joy_(entertainer)",
"Ice_Cream_Cake_(EP)"
] |
[
1,
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nRed Velvet is a solo act.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1,
1
],
"target": [
"Red_Velvet_(band)",
"Joy_(entertainer)",
"Ice_Cream_Cake_(EP)"
]
},
"style": "rule"
}
|
{
"index": 67156,
"split": "train"
}
|
187132
|
Heartland was a drama.
|
[
"Heartland_(Canadian_TV_series)"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nHeartland was a drama.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Heartland_(Canadian_TV_series)"
]
},
"style": "rule"
}
|
{
"index": 40243,
"split": "train"
}
|
60303
|
Stockard Channing was in movies.
|
[
"Stockard_Channing",
"The_Matthew_Shepard_Story",
"Jack_(2004_film)"
] |
[
1,
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nStockard Channing was in movies.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1,
1
],
"target": [
"Stockard_Channing",
"The_Matthew_Shepard_Story",
"Jack_(2004_film)"
]
},
"style": "rule"
}
|
{
"index": 25541,
"split": "train"
}
|
208698
|
Tim Rice wrote Joseph and the Amazing Technicolor Dreamcoat.
|
[
"Tim_Rice"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nTim Rice wrote Joseph and the Amazing Technicolor Dreamcoat.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Tim_Rice"
]
},
"style": "rule"
}
|
{
"index": 9337,
"split": "train"
}
|
176994
|
Marilyn Monroe appeared in an American romantic comedy work.
|
[
"The_Seven_Year_Itch",
"Marilyn_Monroe"
] |
[
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nMarilyn Monroe appeared in an American romantic comedy work.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1
],
"target": [
"The_Seven_Year_Itch",
"Marilyn_Monroe"
]
},
"style": "rule"
}
|
{
"index": 50481,
"split": "train"
}
|
210024
|
Afghanistan is the source of the Saffarid Kingdom.
|
[
"Afghanistan"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nAfghanistan is the source of the Saffarid Kingdom.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Afghanistan"
]
},
"style": "rule"
}
|
{
"index": 36396,
"split": "train"
}
|
52198
|
Broadchurch does not have any seasons.
|
[
"Broadchurch"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nBroadchurch does not have any seasons.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Broadchurch"
]
},
"style": "rule"
}
|
{
"index": 107354,
"split": "train"
}
|
115470
|
19 Kids and Counting was an America reality television series.
|
[
"19_Kids_and_Counting"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\n19 Kids and Counting was an America reality television series.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"19_Kids_and_Counting"
]
},
"style": "rule"
}
|
{
"index": 65643,
"split": "train"
}
|
118410
|
U2 did not have a reputation as a politically and socially conscious group.
|
[
"U2"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nU2 did not have a reputation as a politically and socially conscious group.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"U2"
]
},
"style": "rule"
}
|
{
"index": 2676,
"split": "train"
}
|
7061
|
Josh Homme has been involved with numerous projects.
|
[
"Josh_Homme"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nJosh Homme has been involved with numerous projects.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Josh_Homme"
]
},
"style": "rule"
}
|
{
"index": 98715,
"split": "train"
}
|
171760
|
Rob Cavallo refuses to work with Eric Clapton.
|
[
"Rob_Cavallo"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nRob Cavallo refuses to work with Eric Clapton.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Rob_Cavallo"
]
},
"style": "rule"
}
|
{
"index": 87651,
"split": "train"
}
|
171221
|
Mr. Nobody was directed by a man.
|
[
"Jaco_Van_Dormael",
"Mr._Nobody_(film)"
] |
[
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nMr. Nobody was directed by a man.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1
],
"target": [
"Jaco_Van_Dormael",
"Mr._Nobody_(film)"
]
},
"style": "rule"
}
|
{
"index": 57405,
"split": "train"
}
|
211513
|
Thomas DeSimone was Irish.
|
[
"Thomas_DeSimone"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nThomas DeSimone was Irish.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Thomas_DeSimone"
]
},
"style": "rule"
}
|
{
"index": 2395,
"split": "train"
}
|
81749
|
Dracula is a written work.
|
[
"Dracula"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nDracula is a written work.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Dracula"
]
},
"style": "rule"
}
|
{
"index": 40355,
"split": "train"
}
|
19062
|
South America contains at least five countries.
|
[
"South_America"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nSouth America contains at least five countries.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"South_America"
]
},
"style": "rule"
}
|
{
"index": 42126,
"split": "train"
}
|
173972
|
Susan Sarandon debuted in An Evening with Richard Nixon in 1972.
|
[
"Susan_Sarandon"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nSusan Sarandon debuted in An Evening with Richard Nixon in 1972.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Susan_Sarandon"
]
},
"style": "rule"
}
|
{
"index": 57532,
"split": "train"
}
|
115833
|
CeeLo Green is on the billboard charts multiple times.
|
[
"CeeLo_Green"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nCeeLo Green is on the billboard charts multiple times.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"CeeLo_Green"
]
},
"style": "rule"
}
|
{
"index": 52347,
"split": "train"
}
|
21351
|
Pompeii had an amphitheater.
|
[
"Pompeii"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nPompeii had an amphitheater.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Pompeii"
]
},
"style": "rule"
}
|
{
"index": 12786,
"split": "train"
}
|
206622
|
Titanium is on David Guetta's 2011 album.
|
[
"Titanium_(song)",
"Nothing_but_the_Beat"
] |
[
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nTitanium is on David Guetta's 2011 album.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1
],
"target": [
"Titanium_(song)",
"Nothing_but_the_Beat"
]
},
"style": "rule"
}
|
{
"index": 92290,
"split": "train"
}
|
100595
|
Chicago White Sox have home games at Guaranteed Rate Field.
|
[
"Chicago_White_Sox"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nChicago White Sox have home games at Guaranteed Rate Field.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Chicago_White_Sox"
]
},
"style": "rule"
}
|
{
"index": 102821,
"split": "train"
}
|
16021
|
Socrates was Athenian.
|
[
"Socrates"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nSocrates was Athenian.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Socrates"
]
},
"style": "rule"
}
|
{
"index": 32865,
"split": "train"
}
|
19012
|
Broadchurch is a television crime drama.
|
[
"Broadchurch"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nBroadchurch is a television crime drama.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Broadchurch"
]
},
"style": "rule"
}
|
{
"index": 72083,
"split": "train"
}
|
223458
|
The iPhone 5C is a phone.
|
[
"IPhone_5C"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nThe iPhone 5C is a phone.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"IPhone_5C"
]
},
"style": "rule"
}
|
{
"index": 99201,
"split": "train"
}
|
160678
|
Precious premiered at a film festival in Texas.
|
[
"Cannes_Film_Festival",
"2009_Sundance_Film_Festival",
"Precious_(film)"
] |
[
1,
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nPrecious premiered at a film festival in Texas.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1,
1
],
"target": [
"Cannes_Film_Festival",
"2009_Sundance_Film_Festival",
"Precious_(film)"
]
},
"style": "rule"
}
|
{
"index": 48643,
"split": "train"
}
|
43430
|
The Berlin Wall divided Munich from 1961 to 1989.
|
[
"Berlin_Wall"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nThe Berlin Wall divided Munich from 1961 to 1989.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Berlin_Wall"
]
},
"style": "rule"
}
|
{
"index": 49290,
"split": "train"
}
|
48943
|
The Second Punic War lasted from 218 BC to 201 BC.
|
[
"Second_Punic_War"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nThe Second Punic War lasted from 218 BC to 201 BC.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Second_Punic_War"
]
},
"style": "rule"
}
|
{
"index": 12474,
"split": "train"
}
|
79723
|
Thor: The Dark World is a 2013 superhero movie.
|
[
"Thor:_The_Dark_World"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nThor: The Dark World is a 2013 superhero movie.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Thor:_The_Dark_World"
]
},
"style": "rule"
}
|
{
"index": 37137,
"split": "train"
}
|
33209
|
A fantasy film usually involves exotic fantasy worlds.
|
[
"Fantasy_film"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nA fantasy film usually involves exotic fantasy worlds.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Fantasy_film"
]
},
"style": "rule"
}
|
{
"index": 108712,
"split": "train"
}
|
142287
|
Joe Manganiello has worked with an actor.
|
[
"Channing_Tatum",
"Joe_Manganiello",
"Magic_Mike",
"How_I_Met_Your_Mother",
"Sam_Raimi"
] |
[
1,
1,
1,
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nJoe Manganiello has worked with an actor.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1,
1,
1,
1
],
"target": [
"Channing_Tatum",
"Joe_Manganiello",
"Magic_Mike",
"How_I_Met_Your_Mother",
"Sam_Raimi"
]
},
"style": "rule"
}
|
{
"index": 4641,
"split": "train"
}
|
87855
|
Keanu Reeves is often cast as one who is annihilating the world.
|
[
"Keanu_Reeves"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nKeanu Reeves is often cast as one who is annihilating the world.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Keanu_Reeves"
]
},
"style": "rule"
}
|
{
"index": 53878,
"split": "train"
}
|
198649
|
Jordan Knight sings and writes songs.
|
[
"Jordan_Knight"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nJordan Knight sings and writes songs.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Jordan_Knight"
]
},
"style": "rule"
}
|
{
"index": 91090,
"split": "train"
}
|
71346
|
Neymar plays forward.
|
[
"Neymar"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nNeymar plays forward.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Neymar"
]
},
"style": "rule"
}
|
{
"index": 62054,
"split": "train"
}
|
12659
|
Usain Bolt has won the 200 m title.
|
[
"Usain_Bolt"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nUsain Bolt has won the 200 m title.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Usain_Bolt"
]
},
"style": "rule"
}
|
{
"index": 90391,
"split": "train"
}
|
113450
|
Steve Irwin was Bolivian.
|
[
"Steve_Irwin"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nSteve Irwin was Bolivian.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Steve_Irwin"
]
},
"style": "rule"
}
|
{
"index": 5610,
"split": "train"
}
|
88064
|
Futurama is an animated science fiction series.
|
[
"Futurama"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nFuturama is an animated science fiction series.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Futurama"
]
},
"style": "rule"
}
|
{
"index": 48437,
"split": "train"
}
|
138659
|
Clint Eastwood was born in the first half of the 20th century.
|
[
"Clint_Eastwood"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nClint Eastwood was born in the first half of the 20th century.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Clint_Eastwood"
]
},
"style": "rule"
}
|
{
"index": 31419,
"split": "train"
}
|
49698
|
Nicole Kidman is an actress in the motion picture industry.
|
[
"Nicole_Kidman"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nNicole Kidman is an actress in the motion picture industry.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Nicole_Kidman"
]
},
"style": "rule"
}
|
{
"index": 79748,
"split": "train"
}
|
39367
|
Megan Fox played a regular role on the Hope & Faith television sitcom.
|
[
"Megan_Fox"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nMegan Fox played a regular role on the Hope & Faith television sitcom.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Megan_Fox"
]
},
"style": "rule"
}
|
{
"index": 12166,
"split": "train"
}
|
97526
|
Pope John Paul II is only the name of a pope.
|
[
"Pope_John_Paul_II_(miniseries)",
"Pope_John_Paul_II"
] |
[
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nPope John Paul II is only the name of a pope.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1
],
"target": [
"Pope_John_Paul_II_(miniseries)",
"Pope_John_Paul_II"
]
},
"style": "rule"
}
|
{
"index": 51154,
"split": "train"
}
|
69997
|
A Game of Thrones was nominated for the 1995 Nebula Award.
|
[
"A_Game_of_Thrones"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nA Game of Thrones was nominated for the 1995 Nebula Award.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"A_Game_of_Thrones"
]
},
"style": "rule"
}
|
{
"index": 109155,
"split": "train"
}
|
38458
|
Debrecen is the second largest city in Hungary.
|
[
"Debrecen",
"Hungary"
] |
[
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nDebrecen is the second largest city in Hungary.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1
],
"target": [
"Debrecen",
"Hungary"
]
},
"style": "rule"
}
|
{
"index": 83769,
"split": "train"
}
|
28410
|
Jeanine Pirro was not involved in cases about crimes against the elderly.
|
[
"Jeanine_Pirro"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nJeanine Pirro was not involved in cases about crimes against the elderly.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Jeanine_Pirro"
]
},
"style": "rule"
}
|
{
"index": 19869,
"split": "train"
}
|
91793
|
Quincy Jones came to prominence in the Constitutional Convention.
|
[
"Quincy_Jones"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nQuincy Jones came to prominence in the Constitutional Convention.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Quincy_Jones"
]
},
"style": "rule"
}
|
{
"index": 74729,
"split": "train"
}
|
226608
|
Valerian and the City of a Thousand Planets features Rihanna.
|
[
"Valerian_and_the_City_of_a_Thousand_Planets"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nValerian and the City of a Thousand Planets features Rihanna.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Valerian_and_the_City_of_a_Thousand_Planets"
]
},
"style": "rule"
}
|
{
"index": 45923,
"split": "train"
}
|
18115
|
Katy Perry is a singer.
|
[
"Katy_Perry"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nKaty Perry is a singer.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Katy_Perry"
]
},
"style": "rule"
}
|
{
"index": 39307,
"split": "train"
}
|
167066
|
Parker has served as the primary writer and director for South Park episodes.
|
[
"South_Park"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nParker has served as the primary writer and director for South Park episodes.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"South_Park"
]
},
"style": "rule"
}
|
{
"index": 19722,
"split": "train"
}
|
50071
|
Brittany Murphy worked on Don't Say a Word with Michael Douglas.
|
[
"Brittany_Murphy"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nBrittany Murphy worked on Don't Say a Word with Michael Douglas.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Brittany_Murphy"
]
},
"style": "rule"
}
|
{
"index": 91253,
"split": "train"
}
|
152582
|
Rashida Jones plays the lead role in the comedy series Angie Tribeca.
|
[
"Rashida_Jones"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nRashida Jones plays the lead role in the comedy series Angie Tribeca.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Rashida_Jones"
]
},
"style": "rule"
}
|
{
"index": 106387,
"split": "train"
}
|
171267
|
Cougar Town was an American television sitcom that ran for 102 episodes.
|
[
"Cougar_Town"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nCougar Town was an American television sitcom that ran for 102 episodes.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Cougar_Town"
]
},
"style": "rule"
}
|
{
"index": 19831,
"split": "train"
}
|
223134
|
Carrie Fisher had an emergency.
|
[
"Carrie_Fisher"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nCarrie Fisher had an emergency.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Carrie_Fisher"
]
},
"style": "rule"
}
|
{
"index": 57911,
"split": "train"
}
|
121137
|
Jennifer Hudson became famous in 2004.
|
[
"Jennifer_Hudson"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nJennifer Hudson became famous in 2004.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Jennifer_Hudson"
]
},
"style": "rule"
}
|
{
"index": 65677,
"split": "train"
}
|
14452
|
Rosamund Pike never worked with Laurence Dunmore.
|
[
"The_Libertine_(2004_film)",
"Rosamund_Pike"
] |
[
1,
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nRosamund Pike never worked with Laurence Dunmore.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1,
1
],
"target": [
"The_Libertine_(2004_film)",
"Rosamund_Pike"
]
},
"style": "rule"
}
|
{
"index": 85589,
"split": "train"
}
|
102117
|
Leukemia usually begins in the bone marrow and produces abnormal white blood cells.
|
[
"Leukemia"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nLeukemia usually begins in the bone marrow and produces abnormal white blood cells.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Leukemia"
]
},
"style": "rule"
}
|
{
"index": 71452,
"split": "train"
}
|
221700
|
Tom Felton worked for a cast member of The Borrowers.
|
[
"Tom_Felton"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nTom Felton worked for a cast member of The Borrowers.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Tom_Felton"
]
},
"style": "rule"
}
|
{
"index": 28600,
"split": "train"
}
|
26564
|
Missing is a French film.
|
[
"Missing_(film)"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nMissing is a French film.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Missing_(film)"
]
},
"style": "rule"
}
|
{
"index": 1934,
"split": "train"
}
|
209551
|
Sleep can be disrupted by dyssomnia, insomnia, hypersomnia, narcolepsy, sleep apnea, parasomnia, sleepwalking, REM behavior disorder, bruxism, and circadian rhythm sleep disorders.
|
[
"Sleep"
] |
[
1
] |
fever_dense_train
|
[
{
"content": "<|im_start|>system\nYou are a helpful assistant. You first think about the reasoning process in the mind and then provide the user with the answer.<|im_end|>\n<|im_start|>user\n\nYou are an expert in generating queries for dense retrieval. Given a claim, your task is to retain the original query while expanding it with additional semantically relevant information, to retrieve documents that support or refute the claim. If no useful expansion is needed, return the original query as is.\nShow your work in <think> </think> tags. Your final response must be in JSON format within <answer> </answer> tags. For example,\n<answer>\n{\n \"query\": \"....\"\n} \n</answer>. \n\nHere's the claim:\nSleep can be disrupted by dyssomnia, insomnia, hypersomnia, narcolepsy, sleep apnea, parasomnia, sleepwalking, REM behavior disorder, bruxism, and circadian rhythm sleep disorders.\nAssistant: Let me think step by step. \n<think>\n",
"role": "user"
}
] |
literature_mining
|
{
"ground_truth": {
"score": [
1
],
"target": [
"Sleep"
]
},
"style": "rule"
}
|
{
"index": 40596,
"split": "train"
}
|
End of preview. Expand
in Data Studio
DeepRetrieval
Overview
DeepRetrieval is a novel approach that uses reinforcement learning (RL) to train Large Language Models (LLMs) for query generation without requiring supervised data. Instead of relying on expensive human-annotated or distilled reference queries, DeepRetrieval enables LLMs to learn through direct trial and error, using retrieval metrics as rewards.
Key Features
- No Supervision Required: Eliminates the need for expensive human-annotated or distilled reference queries
- RL-Based Framework: Uses reinforcement learning to optimize query generation directly for retrieval performance
- State-of-the-Art Performance: Achieves remarkable results across diverse retrieval tasks
Please view our GitHub page for instructions.
@article{jiang2025deepretrievalhackingrealsearch,
title={DeepRetrieval: Hacking Real Search Engines and Retrievers with Large Language Models via Reinforcement Learning},
author={Pengcheng Jiang and Jiacheng Lin and Lang Cao and Runchu Tian and SeongKu Kang and Zifeng Wang and Jimeng Sun and Jiawei Han},
year={2025},
journal = {arXiv preprint arXiv: 2503.00223},
url={https://arxiv.org/abs/2503.00223}
}
- Downloads last month
- 67
Size of downloaded dataset files:
1.14 GB
Size of the auto-converted Parquet files:
1.14 GB
Number of rows:
6,462,470