query
stringclasses 100
values | image
imagewidth (px) 1.2k
3.6k
| image_filename
stringlengths 104
106
| answer
stringclasses 100
values | page
stringclasses 993
values | model
stringclasses 1
value | prompt
stringclasses 1
value | source
stringclasses 1
value |
|---|---|---|---|---|---|---|---|
What types of road improvements are planned for Greenmount Road in Belleville, Illinois? | data/scrapped_pdfs_split/pages_extracted/energy_test/04a016c8-2e61-4e1f-ae28-07b1b6e02cbc.pdf/page_841.jpg | ['Additional lanes', 'Standard overlay/ADA improvements', 'Bridge replacement', 'Land acquisition', 'Utility adjustment'] | 841 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
How did the average miles per shipment for single modes change from 1997 to 2007? | data/scrapped_pdfs_split/pages_extracted/energy_test/ac6c6d88-1148-440f-a28a-29dab2703d76.pdf/page_153.jpg | ['For truck, it increased from 144 to 206 miles, for rail it increased from 485 to 599 miles.'] | 153 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the total greenhouse gas reduction in metric tons projected for 2050 across all measures listed? | data/scrapped_pdfs_split/pages_extracted/energy_test/9707cd2d-3e02-4a48-a5b5-b31040517606.pdf/page_37.jpg | ['80,829'] | 37 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the importance of accurate estimation of available PV power according to the text? | data/scrapped_pdfs_split/pages_extracted/energy_test/771e78a5-d823-4dfc-8dcc-abff6c1819b3.pdf/page_88.jpg | ['All around the world, system operators and utilities are continually adapting their grid codes, interconnection requirements, operational practices, and market mechanisms to make the integration of shares of fast-growing variable renewable generation both reliable and economic'] | 88 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are some potential vulnerabilities of underground nuclear power plants? | data/scrapped_pdfs_split/pages_extracted/energy_test/3068874d-45a3-4f21-8ec2-b8de94e780b8.pdf/page_85.jpg | ['vulnerability to extreme seismic events', 'possibility of releases from a reactor meltdown'] | 85 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are the two different closed-system mechanically reversible processes described in the example? | data/scrapped_pdfs_split/pages_extracted/energy_test/d2aed42e-f707-4790-8fc1-a82a34564599.pdf/page_62.jpg | ['(a) Cooling at constant pressure followed by heating at constant volume.', '(b) Heating at constant volume followed by cooling at constant pressure.'] | 62 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What information is being requested for the controlled correspondence office? | data/scrapped_pdfs_split/pages_extracted/energy_test/cb517aa5-4764-4807-a557-c4c544ea6793.pdf/page_200.jpg | ['copy of our Functional Directory in hard copy', 'assignments of controlled correspondence', 'Information on grants, inventions, solicitations, actions, fuel cells, combustion, turbines, biomass, National Energy Policy'] | 200 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
Which county is the reconstruction project on Pierce Lane located in? | data/scrapped_pdfs_split/pages_extracted/energy_test/04a016c8-2e61-4e1f-ae28-07b1b6e02cbc.pdf/page_871.jpg | ['MADISON'] | 871 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
How is the Authority addressing concerns related to the San Francisco to San Jose environmental document? | data/scrapped_pdfs_split/pages_extracted/energy_test/3a80a33b-462a-47e3-876a-c4b0cc2540cf.pdf/page_102.jpg | ['worked consistently with the City of Brisbane', 'from the inception to its completion to address concerns', 'considers the City of Brisbane to be a critical partner'] | 102 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is different about PG&E's pipeline constraints and operating conditions compared to Southern California Gas Company? | data/scrapped_pdfs_split/pages_extracted/energy_test/1fb1a5b4-dc92-4b2c-a492-e645d641bd2b.pdf/page_174.jpg | ["PG&E's physical pipeline constraints and operating conditions are different than SCG"] | 174 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the expected revenue for the power production company in the given quarter? | data/scrapped_pdfs_split/pages_extracted/energy_test/1264fc19-f79b-42f6-8686-e6c8e112a447.pdf/page_14.jpg | ['$518,880'] | 14 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
At what temperature and pressure conditions is the reaction vessel maintained? | data/scrapped_pdfs_split/pages_extracted/energy_test/d2aed42e-f707-4790-8fc1-a82a34564599.pdf/page_571.jpg | ['200°C and 34.5 bar'] | 571 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What caused the price of wholesale electricity to significantly increase in 2017 and 2018 at SoCal Citygate? | data/scrapped_pdfs_split/pages_extracted/energy_test/1fb1a5b4-dc92-4b2c-a492-e645d641bd2b.pdf/page_144.jpg | ['the higher than average gas prices'] | 144 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the purpose of a configuration footprint in the Staged Project Delivery Process? | data/scrapped_pdfs_split/pages_extracted/energy_test/3a80a33b-462a-47e3-876a-c4b0cc2540cf.pdf/page_25.jpg | ['Map the initial right of way necessary to construct a project', 'Identify utility relocation requirements', 'Develop third-party agreements', 'Initiate requests for environmental permits'] | 25 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
Who is the Governmental Affairs Agent listed on Schedule F - Travel/Lodging? | data/scrapped_pdfs_split/pages_extracted/energy_test/d674bc51-708b-4305-a4f4-4ead6f6ed842.pdf/page_810.jpg | ['David B. Applebaum'] | 810 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What types of highway improvements are listed for US 150 in Champaign County? | data/scrapped_pdfs_split/pages_extracted/energy_test/04a016c8-2e61-4e1f-ae28-07b1b6e02cbc.pdf/page_559.jpg | ['DESIGNED OVERLAY', 'ADA IMPROVEMENTS', 'BRIDGE DECK OVERLAY', 'BRIDGE REPAIR', 'STANDARD OVERLAY', 'WIDENING EXISTING PAVEMENT', 'INTERSECTION IMPROVEMENT', 'TRAF SIGNAL MODERNIZATION', 'UTILITY ADJUSTMENT'] | 559 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
Which car model from Aston Martin is categorized as a Subcompact Car? | data/scrapped_pdfs_split/pages_extracted/energy_test/ac6c6d88-1148-440f-a28a-29dab2703d76.pdf/page_116.jpg | ['Rapide'] | 116 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are the nodes and links in a power grid graph representation? | data/scrapped_pdfs_split/pages_extracted/energy_test/9787a714-a800-477b-9c6d-8816da69a390.pdf/page_11.jpg | ['Nodes: Buses', 'Links: Transmission Lines'] | 11 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
How is the heat transferred in a mechanically reversible, constant-pressure process related to the constant-pressure heat capacity? | data/scrapped_pdfs_split/pages_extracted/energy_test/d2aed42e-f707-4790-8fc1-a82a34564599.pdf/page_62.jpg | ['Q = ΔH = ∫(T2/T1) Cp dT (const P)'] | 62 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are some factors affecting public opinion on nuclear energy in the United States? | data/scrapped_pdfs_split/pages_extracted/energy_test/3068874d-45a3-4f21-8ec2-b8de94e780b8.pdf/page_28.jpg | ['Fukushima raised concerns', 'environmental consequences', 'strength of U.S. government support', 'enthusiasm of the Obama Administration'] | 28 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
Why are the strategies found within this plan meant to address transportation equity? | data/scrapped_pdfs_split/pages_extracted/energy_test/60cb7046-0660-4ff0-aaee-14ecc9f0537a.pdf/page_7.jpg | ['to address transportation equity by making EV charging infrastructure more readily available in diverse locations, affordable, and by educating community members on EV and EVSE technologies, both in English and Spanish'] | 7 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What could cause the 'Transformer Temp unreadable' fault? | data/scrapped_pdfs_split/pages_extracted/energy_test/798b1fe8-a6a7-4275-856e-f7ad6b2a066e.pdf/page_110.jpg | ['Temperature sensor is damaged'] | 110 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What types of highway improvements are listed for District 9 in Illinois? | data/scrapped_pdfs_split/pages_extracted/energy_test/04a016c8-2e61-4e1f-ae28-07b1b6e02cbc.pdf/page_934.jpg | ['CULVERT REPLACEMENT', 'MISCELLANEOUS IMPROVEMENTS', 'STANDARD OVERLAY', 'NEW SHOULDERS', 'REHABILITATION - PAVEMENTS', 'LAND ACQUISITION'] | 934 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
How can energy storage play a role in enabling cost-effective integration of PV generation? | data/scrapped_pdfs_split/pages_extracted/energy_test/771e78a5-d823-4dfc-8dcc-abff6c1819b3.pdf/page_14.jpg | ['enhance system flexibility and reliability', 'reduce needs for spinning reserves by conventional power plants', 'balance generation and load to manage variability of increased renewable generation'] | 14 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the procedure to populate the master data to the operational store? | data/scrapped_pdfs_split/pages_extracted/energy_test/259ed437-1e56-472b-9a2e-620dc124aea3.pdf/page_168.jpg | ['Log into the Jupyter node', 'Move the data in the tmp store to the ana store with the command: /home/<utility_id>/etl/py_MoveTmpDataToAnaStore.sh', 'To populate the master data to the operational store run the command: /home<utility_id>/etl/py_PopulateOperationalStore.sh'] | 168 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What percentage of EIB Group's 2022 gross mobility emissions came from business travel flights? | data/scrapped_pdfs_split/pages_extracted/energy_test/4e0978ef-c1c9-4887-a451-31f9ad2b5002.pdf/page_17.jpg | ['75.8%'] | 17 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is one of the reasons mentioned for the United States to be concerned about the security of its energy supply according to the passage? | data/scrapped_pdfs_split/pages_extracted/energy_test/2298ed89-9562-4a70-a9d5-2535e602c191.pdf/page_15.jpg | ['growth in demand, as a result of an increasing U.S. population, along with increased electrification of our society'] | 15 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What criteria does NERC's standard BAL-003-1 establish? | data/scrapped_pdfs_split/pages_extracted/energy_test/771e78a5-d823-4dfc-8dcc-abff6c1819b3.pdf/page_88.jpg | ["NERC's standard BAL-003-1, Frequency Response and Frequency Bias Setting, establishes target contingency protection criteria for each North American interconnection and individual balancing authorities within interconnections"] | 88 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What criteria should a heat exchanger satisfy for thermal energy storage in solar and low energy buildings? | data/scrapped_pdfs_split/pages_extracted/energy_test/c4f65ebe-4407-4d6b-ad83-a213ad4ece05.pdf/page_24.jpg | ['Rating (max power in and out)', 'Hydraulic head losses (should be in accordance with the pump capabilities)', 'Cost (it is always an optimisation parameter)'] | 24 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are the major greenhouse gases mentioned in the table? | data/scrapped_pdfs_split/pages_extracted/energy_test/5a89ed07-8780-4ad6-8f4b-a43daac513b7.pdf/page_187.jpg | ['CO2', 'CH4', 'N2O', 'HFCs', 'PFCs', 'SF6', 'NF3'] | 187 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are the different ways AC Support behaves depending on the equipment installed with the Conext SW? | data/scrapped_pdfs_split/pages_extracted/energy_test/798b1fe8-a6a7-4275-856e-f7ad6b2a066e.pdf/page_28.jpg | ['SOC - Xanbus-enabled Conext Battery Monitor is installed', 'Enhanced - Xanbus-enabled MPPT solar charge controller is installed', 'Regular - neither Xanbus-enabled battery monitor nor MPPT solar charge controller is installed'] | 28 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What percentage of workers had a commute time of 15-29 minutes in 1990? | data/scrapped_pdfs_split/pages_extracted/energy_test/ac6c6d88-1148-440f-a28a-29dab2703d76.pdf/page_196.jpg | ['51.6%'] | 196 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are some examples of microgrid controller research and development? | data/scrapped_pdfs_split/pages_extracted/energy_test/cb7262a4-1e99-42c6-8213-75fecbf7f98a.pdf/page_40.jpg | ['CERTS microgrid controller using plug-and-play and peer-to-peer concepts', 'CSEISMIC microgrid controller with advanced functions and electric power system interaction', 'VOLTTRON open-source software platform for load and energy resource coordination within buildings'] | 40 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
Which country is Leszek Twardy's organization located in? | data/scrapped_pdfs_split/pages_extracted/energy_test/64043de5-2d0a-4526-92bd-8ee909e65f14.pdf/page_97.jpg | ['Poland'] | 97 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are some characteristics of investors suitable for the Aggressive Growth asset mix? | data/scrapped_pdfs_split/pages_extracted/energy_test/602386fb-f4d8-4607-be0e-bebda4b49699.pdf/page_5.jpg | ['Seek aggressive growth', 'Can tolerate wide fluctuations in market values, especially over the short term'] | 5 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What was the average annual percent change in ton-miles for all modes of transportation between 1997 and 2007? | data/scrapped_pdfs_split/pages_extracted/energy_test/ac6c6d88-1148-440f-a28a-29dab2703d76.pdf/page_153.jpg | ['2.3%'] | 153 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What factors contribute to the projected annual change in U.S. natural gas plant liquids production? | data/scrapped_pdfs_split/pages_extracted/energy_test/1da66e0e-0228-4095-b05a-ab0acc86ca16.pdf/page_21.jpg | ['The components shown are net change, natural gas plant liquids gross refinery output, and net butane/isobutane/isobutylene'] | 21 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the cooling system used in the roof of the building? | data/scrapped_pdfs_split/pages_extracted/energy_test/3c1052da-262e-4231-b532-13a7798ae3a8.pdf/page_185.jpg | ["Eight evaporative coolers in the roof -4' cube (18,000-20,000 CFM)"] | 185 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are the key features of the PV-BESS plant described in the text? | data/scrapped_pdfs_split/pages_extracted/energy_test/771e78a5-d823-4dfc-8dcc-abff6c1819b3.pdf/page_55.jpg | ['combines AC-coupled PV array and BESS', 'allows testing multihour PV energy shifting and scheduling', 'provides dispatchable operation without PV curtailment', 'can be coupled with economic dispatch and optimization controller', 'allows testing with various market rules and metrics'] | 55 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the significance of the μ = 0 region in the inversion curves? | data/scrapped_pdfs_split/pages_extracted/energy_test/d2aed42e-f707-4790-8fc1-a82a34564599.pdf/page_296.jpg | ['When (∂Z/∂T)p is zero, as for the ideal-gas state, then μ = 0', 'In the μ = 0 region, no temperature change accompanies throttling'] | 296 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What was the average commute time for workers in 2011? | data/scrapped_pdfs_split/pages_extracted/energy_test/ac6c6d88-1148-440f-a28a-29dab2703d76.pdf/page_196.jpg | ['25.3 minutes'] | 196 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the address of Ramba Consulting Group? | data/scrapped_pdfs_split/pages_extracted/energy_test/d674bc51-708b-4305-a4f4-4ead6f6ed842.pdf/page_545.jpg | ['120 S Monroe St Tallahassee FL 32301'] | 545 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the estimated total savings for a PV system in Durham under the net metering (flat rate) billing option over the system's useful life of 25 years? | data/scrapped_pdfs_split/pages_extracted/energy_test/0861a746-f62d-440a-8701-53ff0b789482.pdf/page_17.jpg | ['$10,614'] | 17 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the fee range for Everglades Foundation, The? | data/scrapped_pdfs_split/pages_extracted/energy_test/d674bc51-708b-4305-a4f4-4ead6f6ed842.pdf/page_254.jpg | ['$30,000.00 - $39,999.00'] | 254 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the estimated cost for the intersection reconstruction project at Main St in Spaulding, Sangamon County? | data/scrapped_pdfs_split/pages_extracted/energy_test/04a016c8-2e61-4e1f-ae28-07b1b6e02cbc.pdf/page_631.jpg | ['$3,300,000'] | 631 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What measures have been taken to improve security at nuclear power plants since 9/11? | data/scrapped_pdfs_split/pages_extracted/energy_test/3068874d-45a3-4f21-8ec2-b8de94e780b8.pdf/page_85.jpg | ['The Design Basis Threat has been increased', 'force on force exercises by the NRC are now being done every three years instead of every eight years'] | 85 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What type of nuclear reactor was involved in the Chernobyl accident? | data/scrapped_pdfs_split/pages_extracted/energy_test/5d10e46d-8edc-4262-8c17-0975016325ff.pdf/page_25.jpg | ['RBMK-1000', 'water-cooled', 'graphite-moderated reactor'] | 25 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the formula given for the Green's function describing the evolution of an instantaneous point injection of tracer? | data/scrapped_pdfs_split/pages_extracted/energy_test/6ccb1888-e525-45b1-9802-10d473b8f622.pdf/page_116.jpg | ['G(x, t) = 1/√(4πDmt) exp(−((x − vt)^2)/(4Dmt))'] | 116 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
In which county are the Greenmount Road projects located? | data/scrapped_pdfs_split/pages_extracted/energy_test/04a016c8-2e61-4e1f-ae28-07b1b6e02cbc.pdf/page_841.jpg | ['St. Clair'] | 841 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What types of communication expenses are listed on Schedule E? | data/scrapped_pdfs_split/pages_extracted/energy_test/d674bc51-708b-4305-a4f4-4ead6f6ed842.pdf/page_810.jpg | ['Printed Materials', 'Postage', 'Film, Slides, Video, Audio', 'TV - Network', 'TV - Cable', 'Radio', 'Other Broadcast Medium', 'Internet', 'Telephone, Facsimile', 'Pro Rata Overhead Costs of Specific Events Over $100', 'Other'] | 810 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the average OPEC spare production capacity for the years 2014-2023? | data/scrapped_pdfs_split/pages_extracted/energy_test/1da66e0e-0228-4095-b05a-ab0acc86ca16.pdf/page_5.jpg | ['2.8 million barrels per day'] | 5 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the estimated cost for the bridge joint replacement and culvert repair project on Route 55 BUS at Ditch 1.2 Mi S & Fancy Creek in Sherman? | data/scrapped_pdfs_split/pages_extracted/energy_test/04a016c8-2e61-4e1f-ae28-07b1b6e02cbc.pdf/page_598.jpg | ['$250,000'] | 598 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the recommended approach for implementing shared streets with cars? | data/scrapped_pdfs_split/pages_extracted/energy_test/9707cd2d-3e02-4a48-a5b5-b31040517606.pdf/page_199.jpg | ['shared streets with cars should be a last stage development', 'after physical protection for bikes', 'after reaching critical volume of cyclists and pedestrians'] | 199 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What initiatives are being undertaken to address urban freight movement challenges in Orlando? | data/scrapped_pdfs_split/pages_extracted/energy_test/685e37cf-4fe0-427b-999e-2f49eed68230.pdf/page_3.jpg | ['State, county, and local governments are working together to develop a regional approach to transportation issues', 'Several large road projects are underway, such as SR 408 improvements and I-4 expansion', 'The development of SunRail, a commuter rail transit system linking many Central Florida communities'] | 3 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
How does the United Kingdom's electricity grid differ from the United States according to the passage? | data/scrapped_pdfs_split/pages_extracted/energy_test/fb50e359-6a89-4def-9b53-e0ba00ea814a.pdf/page_69.jpg | ['The United Kingdom has only one grid, and one owner', 'The United States has eight or ten grids, eight or ten large regional machines that have scores or hundreds of owners'] | 69 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
Why does the ground appear warmer than the panels in the thermal image? | data/scrapped_pdfs_split/pages_extracted/energy_test/771e78a5-d823-4dfc-8dcc-abff6c1819b3.pdf/page_29.jpg | ['image shows blackbody radiation temperature, not absolute thermodynamic temperature', 'it is conceivable that the ground is radiating more than the panels'] | 29 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the altitude operating limit for these models? | data/scrapped_pdfs_split/pages_extracted/energy_test/798b1fe8-a6a7-4275-856e-f7ad6b2a066e.pdf/page_117.jpg | ['6,562 ft. (2,000 m)'] | 117 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the address of Florida State University Foundation? | data/scrapped_pdfs_split/pages_extracted/energy_test/d674bc51-708b-4305-a4f4-4ead6f6ed842.pdf/page_567.jpg | ['2010 Levy Ave Building B, Suite 300 Tallahassee FL 32310'] | 567 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is mentioned about the need for a carbon pricing framework? | data/scrapped_pdfs_split/pages_extracted/energy_test/3068874d-45a3-4f21-8ec2-b8de94e780b8.pdf/page_28.jpg | ['create incentives for utilities to build more nuclear power projects', "a carbon 'tax' would need to be higher than $30/ton of carbon dioxide"] | 28 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What steps can be taken if a biomass project appears viable after evaluation? | data/scrapped_pdfs_split/pages_extracted/energy_test/ec15b1e9-786c-4d19-ab27-b2046293b72f.pdf/page_250.jpg | ['Assess whether the project will help achieve energy-related goals', 'Enter information already collected into Community Goals Summary Worksheets to identify which goals the project is likely to achieve', 'Learn more about how to move the project forward by going to the Next Steps section'] | 250 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What happens to the voltage of a battery undergoing constant current discharge over time? | data/scrapped_pdfs_split/pages_extracted/energy_test/d21725c4-40e5-4ac5-81a9-8e094d8e1691.pdf/page_14.jpg | ['The voltage experiences an initial drop', 'the voltage gradually decreases', 'the voltage declines rapidly toward the end of discharge'] | 14 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What parking strategies are suggested for new residential development? | data/scrapped_pdfs_split/pages_extracted/energy_test/9707cd2d-3e02-4a48-a5b5-b31040517606.pdf/page_46.jpg | ['Adopt development code standards that do not require a minimum number of general-purpose parking spaces and set a low maximum number of general-purpose passenger vehicle parking spaces for new multifamily development'] | 46 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the projected peak electricity demand in California for the year 2030? | data/scrapped_pdfs_split/pages_extracted/energy_test/1fb1a5b4-dc92-4b2c-a492-e645d641bd2b.pdf/page_97.jpg | ['~60 GW'] | 97 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
Why is the $5/Dth noncompliance charge for operational flow orders not considered productive or beneficial? | data/scrapped_pdfs_split/pages_extracted/energy_test/1fb1a5b4-dc92-4b2c-a492-e645d641bd2b.pdf/page_174.jpg | ['The gas system must be operated within safe maximum and minimum pressures', 'PG&E Gas Operations has zero control over the supplies nominated into the system', 'therefore, there must be sufficient economic signals for supplies to balance demands'] | 174 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What types of accounts or products allow investors to defer paying taxes? | data/scrapped_pdfs_split/pages_extracted/energy_test/602386fb-f4d8-4607-be0e-bebda4b49699.pdf/page_16.jpg | ['401(k) and 403(b) plans', 'IRAs', 'health savings accounts (HSAs)', 'other tax-deferred products such as deferred annuities'] | 16 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are some companies and their locations listed on the page? | data/scrapped_pdfs_split/pages_extracted/energy_test/d674bc51-708b-4305-a4f4-4ead6f6ed842.pdf/page_214.jpg | ['United States Sugar Corporation, 111 Ponce De Leon Ave, Clewiston FL 33440', 'Walmart, 702 SW 8th St, Bentonville AR 72716', 'Wal-Mart Stores, Inc, 702 SW 8th St. MW 215, Bentonville AR 72716-6209', 'Water & Soil Solutions, LLC, 16112 E Duran Rd, Loxahatchee FL 33470-7890', 'Watershed Technologies, LLC, 3208 Westchester Dr, Cocoa FL 32926'] | 214 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the projected regional energy mix for electricity demand scenarios in 2010? | data/scrapped_pdfs_split/pages_extracted/energy_test/64043de5-2d0a-4526-92bd-8ee909e65f14.pdf/page_70.jpg | ['hydro', 'nuclear', 'coal', 'gas'] | 70 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What were some of the unique design features of the Chernobyl reactor that contributed to the accident? | data/scrapped_pdfs_split/pages_extracted/energy_test/5d10e46d-8edc-4262-8c17-0975016325ff.pdf/page_25.jpg | ['graphite moderator', 'water coolant', 'susceptible to runaway chain reaction'] | 25 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What method was used to estimate the accuracy of the baseline method? | data/scrapped_pdfs_split/pages_extracted/energy_test/fecee5f3-1998-4860-92dc-d89fa4d97992.pdf/page_92.jpg | ['ISO-NE baseline'] | 92 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the mission of the Bioenergy Technologies Office (BETO)? | data/scrapped_pdfs_split/pages_extracted/energy_test/54f79286-ae61-4973-924b-e3911f5ae5bc.pdf/page_71.jpg | ['To develop industrially relevant, transformative, and revolutionary bioenergy technologies to enable sustainable, domestically produced biofuels, bioproducts, and biopower for a prosperous nation'] | 71 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are the specific topics mentioned related to the national energy policy options? | data/scrapped_pdfs_split/pages_extracted/energy_test/cb517aa5-4764-4807-a557-c4c544ea6793.pdf/page_237.jpg | ['volume purchasing', 'weatherization'] | 237 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the purpose of decreasing the ash content in sludge from 26% to 15% for the future renovations? | data/scrapped_pdfs_split/pages_extracted/energy_test/ec07ed0b-2624-418b-bec8-4e85ee63da46.pdf/page_138.jpg | ['enable a reduced sludge ash content over time'] | 138 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are the strategies and recommendations for early notification and coordination in Texas? | data/scrapped_pdfs_split/pages_extracted/energy_test/88e0e72a-8ca0-4828-ad89-e4241bfa8096.pdf/page_153.jpg | ['Implement and maintain the geodatabase of energy developments in Texas', 'Implement interagency cooperation agreements with other agencies', 'Improve communication and coordination with energy developers', 'Implement additional proactive mechanisms to learn about energy developments', 'Work with TxDPS to improve traffic safety and protect the transportation infrastructure'] | 153 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are the potential impacts of high geothermal fluid production rates compared to recharge rates? | data/scrapped_pdfs_split/pages_extracted/energy_test/0cf773fd-1aed-4d57-ab1a-c5c20c8cf14c.pdf/page_277.jpg | ['formation may experience consolidation', 'lowering of the surface elevation', 'surface subsidence'] | 277 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are some challenges faced in the integration of solar photovoltaic (PV) power plants into the electric power grid? | data/scrapped_pdfs_split/pages_extracted/energy_test/771e78a5-d823-4dfc-8dcc-abff6c1819b3.pdf/page_14.jpg | ['variability across timescales', 'forecast uncertainty of solar energy resource', 'impacts on distribution and transmission systems', 'limiting PV installations or assigning PV integration costs', 'curtailment due to overgeneration and other constraints'] | 14 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
Which hour of the day had the highest overall electricity generation in 2019? | data/scrapped_pdfs_split/pages_extracted/energy_test/1fb1a5b4-dc92-4b2c-a492-e645d641bd2b.pdf/page_127.jpg | ['Hour 20'] | 127 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
How is China's energy intensity expected to change between 2012 and 2040 for the chemical industry? | data/scrapped_pdfs_split/pages_extracted/energy_test/3bcded3d-c22b-4428-a2d5-1b5802e8fa3f.pdf/page_11.jpg | ['not expected to be significant', 'overall intensity improvement at only 15%'] | 11 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the forecasted annual change for U.S. hydrocarbon gas liquids product supplied (consumption) in 2025? | data/scrapped_pdfs_split/pages_extracted/energy_test/1da66e0e-0228-4095-b05a-ab0acc86ca16.pdf/page_22.jpg | ['0.06'] | 22 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What fuel type had the highest average hourly generation in 2019 according to the chart? | data/scrapped_pdfs_split/pages_extracted/energy_test/1fb1a5b4-dc92-4b2c-a492-e645d641bd2b.pdf/page_127.jpg | ['Natural Gas'] | 127 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the governing equation for the motion of tracers? | data/scrapped_pdfs_split/pages_extracted/energy_test/6ccb1888-e525-45b1-9802-10d473b8f622.pdf/page_116.jpg | ['∂n/∂t = D∇^2n + ∇ · (ũn) = 0'] | 116 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
How can the 'Dead Battery Detected' error be resolved? | data/scrapped_pdfs_split/pages_extracted/energy_test/798b1fe8-a6a7-4275-856e-f7ad6b2a066e.pdf/page_110.jpg | ['Both Line 1 and Line 2 must qualify to charge the battery', 'Recharge the battery with an external charging device', 'If the error persists, then the battery may be defective', 'Replace the battery'] | 110 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What was the cause of the Chernobyl nuclear accident? | data/scrapped_pdfs_split/pages_extracted/energy_test/5d10e46d-8edc-4262-8c17-0975016325ff.pdf/page_25.jpg | ['unauthorized experimentation with the reactor'] | 25 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the trend in natural gas usage for electricity generation according to the scenarios? | data/scrapped_pdfs_split/pages_extracted/energy_test/64043de5-2d0a-4526-92bd-8ee909e65f14.pdf/page_70.jpg | ['natural gas will enjoy the major share in all electricity consumption options', 'capturing the growth in fuel mix requirements'] | 70 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are some challenges mentioned regarding the large-scale implementation of electric vehicles? | data/scrapped_pdfs_split/pages_extracted/energy_test/8f6b20fc-7024-4ddb-af56-8bc057b9b945.pdf/page_23.jpg | ['issues relating to charging infrastructure', 'potential logistical, economic and physical impacts of massive multiple vehicle charging on utility supply'] | 23 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the purpose of the integrated risk management framework mentioned in the passage? | data/scrapped_pdfs_split/pages_extracted/energy_test/5b202385-dd1f-456d-a7dd-401fe78f2856.pdf/page_72.jpg | ['Risk management involves the systematic identification, risk significance evaluation, and control of potential losses that may arise from uncertain problem events, such as fires, explosions, mechanical/electrical failures or natural disasters', 'The practice of risk management permits contractors, employees and managers to anticipate such losses and to evaluate their potential impacts, so they can be managed in an effective manner'] | 72 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
Which region has the highest load growth rate from 2020 to 2026? | data/scrapped_pdfs_split/pages_extracted/energy_test/1fb1a5b4-dc92-4b2c-a492-e645d641bd2b.pdf/page_9.jpg | ['Southwest', '1.5% per year'] | 9 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the lobbyist income reported for Amelia B McKie during the period from January 1 to May 31, 2018? | data/scrapped_pdfs_split/pages_extracted/energy_test/d674bc51-708b-4305-a4f4-4ead6f6ed842.pdf/page_882.jpg | ['$2,500.00'] | 882 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are some key goals and requirements related to renewable energy integration in California? | data/scrapped_pdfs_split/pages_extracted/energy_test/771e78a5-d823-4dfc-8dcc-abff6c1819b3.pdf/page_14.jpg | ['California state legislature passed Senate Bill 350 requiring utilities to produce 50% of electricity sales from renewable sources by 2030', "Senate Bill 100 increased the RPS to 60% by 2030 and required all the state's electricity to come from carbon-free resources by 2045", 'California operators will need to find additional ways to balance generation and load to manage the variability of increased renewable generation'] | 14 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are the different regional and subregional plans mentioned for the San Diego region? | data/scrapped_pdfs_split/pages_extracted/energy_test/90f92e15-60eb-4fb1-8cb0-daaa3a7758e7.pdf/page_53.jpg | ['San Diego County Multi-jurisdictional Hazard Mitigation Plan (MHMP) and Local Hazard Mitigation Plans', 'San Diego Forward: The Regional Plan (RTP/SCS)', 'Regional and Subregional Water Quality Plans: Water Quality Control Plan for the San Diego Basin (Basin Plan), Watershed Management Area Water Quality Improvement Plans (WQIPs)', 'Urban Water Management Plan (UWMP)', 'Regional and Subregional Habitat Conservation Programs: San Diego Multiple Species Conservation Program (MSCP), MSCP Subregional Planning Area Programs', 'San Diego County Climate Change and Public Health Plan', 'San Diego Port Master Plan'] | 53 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
How did the air travel emissions from the EIB Group change in 2022 compared to 2021? | data/scrapped_pdfs_split/pages_extracted/energy_test/4e0978ef-c1c9-4887-a451-31f9ad2b5002.pdf/page_17.jpg | ['increased by 767%'] | 17 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
How did the percentage of workers with a commute time of less than 15 minutes change from 1990 to 2011? | data/scrapped_pdfs_split/pages_extracted/energy_test/ac6c6d88-1148-440f-a28a-29dab2703d76.pdf/page_196.jpg | ['In 1990, 15.9% of workers commuted less than 15 minutes, while in 2011, 28.3% enjoyed the short commute.'] | 196 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the average annual growth rate of regional load across the Western Interconnection from 2020 to 2026? | data/scrapped_pdfs_split/pages_extracted/energy_test/1fb1a5b4-dc92-4b2c-a492-e645d641bd2b.pdf/page_9.jpg | ['0.6% per year'] | 9 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What does the activity coefficient represent for a species in solution? | data/scrapped_pdfs_split/pages_extracted/energy_test/d2aed42e-f707-4790-8fc1-a82a34564599.pdf/page_480.jpg | ['the ratio of its actual fugacity to the value given by the Lewis/Randall rule at the same T, P, and composition'] | 480 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the projected annual growth rate of electricity demand in the country according to the latest Philippine Energy Plan Update? | data/scrapped_pdfs_split/pages_extracted/energy_test/64043de5-2d0a-4526-92bd-8ee909e65f14.pdf/page_60.jpg | ['8.6 per cent from 43,780 GW-h in 1999 to 99,714 GW-h in 2009'] | 60 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What industry does Servico cater to? | data/scrapped_pdfs_split/pages_extracted/energy_test/6dfedf6e-553c-483c-bb23-fd89909c8c9b.pdf/page_30.jpg | ['Restaurant & Hospitality'] | 30 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the spark spread used for in power generation? | data/scrapped_pdfs_split/pages_extracted/energy_test/1264fc19-f79b-42f6-8686-e6c8e112a447.pdf/page_32.jpg | ['a proxy for the heat rate (a measure of efficiency) of a specific generating unit or station'] | 32 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What are some recent advances in HVDC system controllers? | data/scrapped_pdfs_split/pages_extracted/energy_test/cb7262a4-1e99-42c6-8213-75fecbf7f98a.pdf/page_40.jpg | ['providing frequency response and voltage support', 'plug-and-play functionality to lower installation costs', 'smart functionality to provide grid support', 'intelligent controllers to optimize battery charging/discharging'] | 40 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What is the name of the institute where Andrei Vladescu works? | data/scrapped_pdfs_split/pages_extracted/energy_test/64043de5-2d0a-4526-92bd-8ee909e65f14.pdf/page_97.jpg | ['Romanian Electricity Company - CONEL'] | 97 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
How is the drilling cost index calculated? | data/scrapped_pdfs_split/pages_extracted/energy_test/0cf773fd-1aed-4d57-ab1a-c5c20c8cf14c.pdf/page_202.jpg | ['The graph indicates that the MITDD (Minerals Management Service Technical Data) drilling cost index is calculated and adjusted using the GDP deflator index, with 1977 set as the base year 100.'] | 202 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |
|
What does the Authority do after the design work progresses in keeping with the Staged Project Delivery process? | data/scrapped_pdfs_split/pages_extracted/energy_test/3a80a33b-462a-47e3-876a-c4b0cc2540cf.pdf/page_25.jpg | ['Research and explore alternative construction delivery methods prior to the final design phase'] | 25 | sonnet |
You are an assistant specialized in Multimodal RAG tasks.
The task is the following: given an image from a pdf page, you will have to
generate questions that can be asked by a user to retrieve information from
a large documentary corpus.
The question should be relevant to the page, and should not be too specific
or too general. The question should be about the subject of the page, and
the answer need to be found in the page.
Remember that the question is asked by a user to get some information from a
large documentary corpus that contains multimodal data. Generate a question
that could be asked by a user without knowing the existence and the content
of the corpus.
Generate as well the answer to the question, which should be found in the
page. And the format of the answer should be a list of words answering the
question.
Generate at most THREE pairs of questions and answers per page in a
dictionary with the following format, answer ONLY this dictionary
NOTHING ELSE:
{
"questions": [
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
{
"question": "XXXXXX",
"answer": ["YYYYYY"]
},
]
}
where XXXXXX is the question and ['YYYYYY'] is the corresponding list of answers
that could be as long as needed.
Note: If there are no questions to ask about the page, return an empty list.
Focus on making relevant questions concerning the page.
Here is the page:
| pdf |