test.jsonl

{"id": "bank_statement_fin_006_p1_classify", "task": "classify", "class": "bank_statement", "doc": "bank_statement_fin_006.pdf", "images": ["images/test/bank_statement/bank_statement_fin_006_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"bank_statement\", \"confidence\": 1.0, \"issuer\": \"Commonwealth Bank\", \"holder\": \"MR R D GEE\", \"page_marker\": \"\"}]}"}]}
{"id": "bank_statement_fin_006_extract", "task": "extract", "class": "bank_statement", "doc": "bank_statement_fin_006.pdf", "images": ["images/test/bank_statement/bank_statement_fin_006_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('bank_statement', 'The document class.')\nfields = (['account_holder_name', 'bank_name', 'statement_period_start', 'statement_period_end', 'ending_balance', 'address', 'bank_account_type', 'bank_account_number', 'transaction_amount', 'transaction_date', 'transaction_type', 'transaction_sender'], 'The exact field names to extract.')\nhints = ({'account_holder_name': 'Name on the account', 'bank_name': 'Name of the bank', 'statement_period_start': 'Start of statement period', 'statement_period_end': 'End of statement period', 'ending_balance': 'Closing balance for the period', 'address': 'Address of the account holder', 'bank_account_type': 'Type of bank account - Checking/Saving', 'bank_account_number': 'Bank account number', 'transaction_amount': 'Transaction amount', 'transaction_date': 'Transaction date', 'transaction_type': 'Transaction type', 'transaction_sender': 'Transaction sender'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"account_holder_name\", \"value\": \"MR R D GEE\"}, {\"field\": \"bank_name\", \"value\": \"Commonwealth Bank\"}, {\"field\": \"statement_period_start\", \"value\": \"2018-07-01\"}, {\"field\": \"statement_period_end\", \"value\": \"2018-12-30\"}, {\"field\": \"ending_balance\", \"value\": \"782.92\"}, {\"field\": \"address\", \"value\": \"14 STODA ST HEATHMONT VIC 3135\"}, {\"field\": \"bank_account_type\", \"value\": \"\"}, {\"field\": \"bank_account_number\", \"value\": \"06 3167 10781391\"}, {\"field\": \"transaction_amount\", \"value\": \"372.0\"}, {\"field\": \"transaction_date\", \"value\": \"2018-07-02\"}, {\"field\": \"transaction_type\", \"value\": \"Transfer to other Bank NetBank\"}, {\"field\": \"transaction_sender\", \"value\": \"Rob Ubank Transfer\"}]}"}]}
{"id": "bank_statement_fin_022_p1_classify", "task": "classify", "class": "bank_statement", "doc": "bank_statement_fin_022.pdf", "images": ["images/test/bank_statement/bank_statement_fin_022_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"bank_statement\", \"confidence\": 1.0, \"issuer\": \"The Royal Bank of Scotland\", \"holder\": \"MR T TESTER\", \"page_marker\": \"\"}]}"}]}
{"id": "bank_statement_fin_022_extract", "task": "extract", "class": "bank_statement", "doc": "bank_statement_fin_022.pdf", "images": ["images/test/bank_statement/bank_statement_fin_022_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('bank_statement', 'The document class.')\nfields = (['account_holder_name', 'bank_name', 'statement_period_start', 'statement_period_end', 'ending_balance', 'address', 'bank_account_type', 'bank_account_number', 'transaction_amount', 'transaction_date', 'transaction_type', 'transaction_sender'], 'The exact field names to extract.')\nhints = ({'account_holder_name': 'Name on the account', 'bank_name': 'Name of the bank', 'statement_period_start': 'Start of statement period', 'statement_period_end': 'End of statement period', 'ending_balance': 'Closing balance for the period', 'address': 'Address of the account holder', 'bank_account_type': 'Type of bank account - Checking/Saving', 'bank_account_number': 'Bank account number', 'transaction_amount': 'Transaction amount', 'transaction_date': 'Transaction date', 'transaction_type': 'Transaction type', 'transaction_sender': 'Transaction sender'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"account_holder_name\", \"value\": \"MR T TESTER\"}, {\"field\": \"bank_name\", \"value\": \"The Royal Bank of Scotland\"}, {\"field\": \"statement_period_start\", \"value\": \"2014-10-22\"}, {\"field\": \"statement_period_end\", \"value\": \"2014-12-21\"}, {\"field\": \"ending_balance\", \"value\": \"300.2\"}, {\"field\": \"address\", \"value\": \"1 TEST STREET, TEST TOWN, TE5 7ER\"}, {\"field\": \"bank_account_type\", \"value\": \"Checking\"}, {\"field\": \"bank_account_number\", \"value\": \"1111111\"}, {\"field\": \"transaction_amount\", \"value\": \"190.4\"}, {\"field\": \"transaction_date\", \"value\": \"2014-10-22\"}, {\"field\": \"transaction_type\", \"value\": \"AUTOMATED PAY IN\"}, {\"field\": \"transaction_sender\", \"value\": \"650274051211-CHB\"}]}"}]}
{"id": "bank_statement_fin_025_p1_classify", "task": "classify", "class": "bank_statement", "doc": "bank_statement_fin_025.pdf", "images": ["images/test/bank_statement/bank_statement_fin_025_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"bank_statement\", \"confidence\": 1.0, \"issuer\": \"HSBC\", \"holder\": \"AHMAD MUSTAFA\", \"page_marker\": \"\"}]}"}]}
{"id": "bank_statement_fin_025_extract", "task": "extract", "class": "bank_statement", "doc": "bank_statement_fin_025.pdf", "images": ["images/test/bank_statement/bank_statement_fin_025_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('bank_statement', 'The document class.')\nfields = (['account_holder_name', 'bank_name', 'statement_period_start', 'statement_period_end', 'ending_balance', 'address', 'bank_account_type', 'bank_account_number', 'transaction_amount', 'transaction_date', 'transaction_type', 'transaction_sender'], 'The exact field names to extract.')\nhints = ({'account_holder_name': 'Name on the account', 'bank_name': 'Name of the bank', 'statement_period_start': 'Start of statement period', 'statement_period_end': 'End of statement period', 'ending_balance': 'Closing balance for the period', 'address': 'Address of the account holder', 'bank_account_type': 'Type of bank account - Checking/Saving', 'bank_account_number': 'Bank account number', 'transaction_amount': 'Transaction amount', 'transaction_date': 'Transaction date', 'transaction_type': 'Transaction type', 'transaction_sender': 'Transaction sender'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"account_holder_name\", \"value\": \"AHMAD MUSTAFA\"}, {\"field\": \"bank_name\", \"value\": \"HSBC\"}, {\"field\": \"statement_period_start\", \"value\": \"\"}, {\"field\": \"statement_period_end\", \"value\": \"\"}, {\"field\": \"ending_balance\", \"value\": \"3100.0\"}, {\"field\": \"address\", \"value\": \"43 Great Sutton Street? London? dsjfh United Kingdom\"}, {\"field\": \"bank_account_type\", \"value\": \"Checking\"}, {\"field\": \"bank_account_number\", \"value\": \"00990274\"}, {\"field\": \"transaction_amount\", \"value\": \"1000.0\"}, {\"field\": \"transaction_date\", \"value\": \"2018-08-11\"}, {\"field\": \"transaction_type\", \"value\": \"Credit\"}, {\"field\": \"transaction_sender\", \"value\": \"IBM Uk\"}]}"}]}
{"id": "bank_statement_fin_032_p1_classify", "task": "classify", "class": "bank_statement", "doc": "bank_statement_fin_032.pdf", "images": ["images/test/bank_statement/bank_statement_fin_032_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"bank_statement\", \"confidence\": 1.0, \"issuer\": \"DBS Bank India Limited\", \"holder\": \"Kulvinder Singh\", \"page_marker\": \"\"}]}"}]}
{"id": "bank_statement_fin_032_extract", "task": "extract", "class": "bank_statement", "doc": "bank_statement_fin_032.pdf", "images": ["images/test/bank_statement/bank_statement_fin_032_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('bank_statement', 'The document class.')\nfields = (['account_holder_name', 'bank_name', 'statement_period_start', 'statement_period_end', 'ending_balance', 'address', 'bank_account_type', 'bank_account_number', 'transaction_amount', 'transaction_date', 'transaction_type', 'transaction_sender'], 'The exact field names to extract.')\nhints = ({'account_holder_name': 'Name on the account', 'bank_name': 'Name of the bank', 'statement_period_start': 'Start of statement period', 'statement_period_end': 'End of statement period', 'ending_balance': 'Closing balance for the period', 'address': 'Address of the account holder', 'bank_account_type': 'Type of bank account - Checking/Saving', 'bank_account_number': 'Bank account number', 'transaction_amount': 'Transaction amount', 'transaction_date': 'Transaction date', 'transaction_type': 'Transaction type', 'transaction_sender': 'Transaction sender'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"account_holder_name\", \"value\": \"Kulvinder Singh\"}, {\"field\": \"bank_name\", \"value\": \"DBS Bank India Limited\"}, {\"field\": \"statement_period_start\", \"value\": \"2025-02-01\"}, {\"field\": \"statement_period_end\", \"value\": \"2025-06-06\"}, {\"field\": \"ending_balance\", \"value\": \"5258.85\"}, {\"field\": \"address\", \"value\": \"Ward No - 08, Bilochiya (10 BLM), GANGANAGAR - 335704, RAJASTHAN, INDIA\"}, {\"field\": \"bank_account_type\", \"value\": \"Saving\"}, {\"field\": \"bank_account_number\", \"value\": \"8811010003049028\"}, {\"field\": \"transaction_amount\", \"value\": \"\"}, {\"field\": \"transaction_date\", \"value\": \"\"}, {\"field\": \"transaction_type\", \"value\": \"\"}, {\"field\": \"transaction_sender\", \"value\": \"\"}]}"}]}
{"id": "bank_statement_fin_037_p1_classify", "task": "classify", "class": "bank_statement", "doc": "bank_statement_fin_037.pdf", "images": ["images/test/bank_statement/bank_statement_fin_037_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"bank_statement\", \"confidence\": 1.0, \"issuer\": \"TD Bank\", \"holder\": \"KE 2 CONTRACT LLC\", \"page_marker\": \"\"}]}"}]}
{"id": "bank_statement_fin_037_extract", "task": "extract", "class": "bank_statement", "doc": "bank_statement_fin_037.pdf", "images": ["images/test/bank_statement/bank_statement_fin_037_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('bank_statement', 'The document class.')\nfields = (['account_holder_name', 'bank_name', 'statement_period_start', 'statement_period_end', 'ending_balance', 'address', 'bank_account_type', 'bank_account_number', 'transaction_amount', 'transaction_date', 'transaction_type', 'transaction_sender'], 'The exact field names to extract.')\nhints = ({'account_holder_name': 'Name on the account', 'bank_name': 'Name of the bank', 'statement_period_start': 'Start of statement period', 'statement_period_end': 'End of statement period', 'ending_balance': 'Closing balance for the period', 'address': 'Address of the account holder', 'bank_account_type': 'Type of bank account - Checking/Saving', 'bank_account_number': 'Bank account number', 'transaction_amount': 'Transaction amount', 'transaction_date': 'Transaction date', 'transaction_type': 'Transaction type', 'transaction_sender': 'Transaction sender'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"account_holder_name\", \"value\": \"KE 2 CONTRACT LLC\"}, {\"field\": \"bank_name\", \"value\": \"TD Bank\"}, {\"field\": \"statement_period_start\", \"value\": \"2018-07-01\"}, {\"field\": \"statement_period_end\", \"value\": \"2018-07-31\"}, {\"field\": \"ending_balance\", \"value\": \"51176.35\"}, {\"field\": \"address\", \"value\": \"3418 NORTHERN BLWD STE 412 LONG ISLAND CITY NY 11101\"}, {\"field\": \"bank_account_type\", \"value\": \"Checking\"}, {\"field\": \"bank_account_number\", \"value\": \"435-4366572\"}, {\"field\": \"transaction_amount\", \"value\": \"3750.0\"}, {\"field\": \"transaction_date\", \"value\": \"2018-07-02\"}, {\"field\": \"transaction_type\", \"value\": \"Deposit\"}, {\"field\": \"transaction_sender\", \"value\": \"CHECK DEPOSIT, 59317 LARSON CONTRACTORS\"}]}"}]}
{"id": "bank_statement_fin_070_p1_classify", "task": "classify", "class": "bank_statement", "doc": "bank_statement_fin_070.pdf", "images": ["images/test/bank_statement/bank_statement_fin_070_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"bank_statement\", \"confidence\": 1.0, \"issuer\": \"BARCLAYS\", \"holder\": \"ATIF ALI\", \"page_marker\": \"\"}]}"}]}
{"id": "bank_statement_fin_070_extract", "task": "extract", "class": "bank_statement", "doc": "bank_statement_fin_070.pdf", "images": ["images/test/bank_statement/bank_statement_fin_070_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('bank_statement', 'The document class.')\nfields = (['account_holder_name', 'bank_name', 'statement_period_start', 'statement_period_end', 'ending_balance', 'address', 'bank_account_type', 'bank_account_number', 'transaction_amount', 'transaction_date', 'transaction_type', 'transaction_sender'], 'The exact field names to extract.')\nhints = ({'account_holder_name': 'Name on the account', 'bank_name': 'Name of the bank', 'statement_period_start': 'Start of statement period', 'statement_period_end': 'End of statement period', 'ending_balance': 'Closing balance for the period', 'address': 'Address of the account holder', 'bank_account_type': 'Type of bank account - Checking/Saving', 'bank_account_number': 'Bank account number', 'transaction_amount': 'Transaction amount', 'transaction_date': 'Transaction date', 'transaction_type': 'Transaction type', 'transaction_sender': 'Transaction sender'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"account_holder_name\", \"value\": \"ATIF ALI\"}, {\"field\": \"bank_name\", \"value\": \"BARCLAYS\"}, {\"field\": \"statement_period_start\", \"value\": \"2024-01-05\"}, {\"field\": \"statement_period_end\", \"value\": \"2024-08-02\"}, {\"field\": \"ending_balance\", \"value\": \"4378.1\"}, {\"field\": \"address\", \"value\": \"32 SCOTTS ROAD, LONDON, E10 6LW\"}, {\"field\": \"bank_account_type\", \"value\": \"Checking\"}, {\"field\": \"bank_account_number\", \"value\": \"83605973\"}, {\"field\": \"transaction_amount\", \"value\": \"394.24\"}, {\"field\": \"transaction_date\", \"value\": \"2024-01-05\"}, {\"field\": \"transaction_type\", \"value\": \"Bank Giro\"}, {\"field\": \"transaction_sender\", \"value\": \"Decora Blinds\"}]}"}]}
{"id": "bank_statement_fin_071_p1_classify", "task": "classify", "class": "bank_statement", "doc": "bank_statement_fin_071.pdf", "images": ["images/test/bank_statement/bank_statement_fin_071_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"bank_statement\", \"confidence\": 1.0, \"issuer\": \"BARCLAYS\", \"holder\": \"Mr Oluwagbemiga Oluwaseun Ashiru\", \"page_marker\": \"\"}]}"}]}
{"id": "bank_statement_fin_071_extract", "task": "extract", "class": "bank_statement", "doc": "bank_statement_fin_071.pdf", "images": ["images/test/bank_statement/bank_statement_fin_071_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('bank_statement', 'The document class.')\nfields = (['account_holder_name', 'bank_name', 'statement_period_start', 'statement_period_end', 'ending_balance', 'address', 'bank_account_type', 'bank_account_number', 'transaction_amount', 'transaction_date', 'transaction_type', 'transaction_sender'], 'The exact field names to extract.')\nhints = ({'account_holder_name': 'Name on the account', 'bank_name': 'Name of the bank', 'statement_period_start': 'Start of statement period', 'statement_period_end': 'End of statement period', 'ending_balance': 'Closing balance for the period', 'address': 'Address of the account holder', 'bank_account_type': 'Type of bank account - Checking/Saving', 'bank_account_number': 'Bank account number', 'transaction_amount': 'Transaction amount', 'transaction_date': 'Transaction date', 'transaction_type': 'Transaction type', 'transaction_sender': 'Transaction sender'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"account_holder_name\", \"value\": \"Mr Oluwagbemiga Oluwaseun Ashiru\"}, {\"field\": \"bank_name\", \"value\": \"BARCLAYS\"}, {\"field\": \"statement_period_start\", \"value\": \"2014-01-09\"}, {\"field\": \"statement_period_end\", \"value\": \"2014-02-07\"}, {\"field\": \"ending_balance\", \"value\": \"117.07\"}, {\"field\": \"address\", \"value\": \"MR O O ASHIRU, VICTORIA CARE HOME, 52 VICTORIA ROAD, KEIGHLEY, BD21 1JB\"}, {\"field\": \"bank_account_type\", \"value\": \"Checking\"}, {\"field\": \"bank_account_number\", \"value\": \"93678296\"}, {\"field\": \"transaction_amount\", \"value\": \"6.0\"}, {\"field\": \"transaction_date\", \"value\": \"2014-01-09\"}, {\"field\": \"transaction_type\", \"value\": \"Debit Card\"}, {\"field\": \"transaction_sender\", \"value\": \"Sainsburys\"}]}"}]}
{"id": "bank_statement_fin_076_p1_classify", "task": "classify", "class": "bank_statement", "doc": "bank_statement_fin_076.pdf", "images": ["images/test/bank_statement/bank_statement_fin_076_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"bank_statement\", \"confidence\": 1.0, \"issuer\": \"BARCLAYS\", \"holder\": \"JOHN DOE LEE\", \"page_marker\": \"\"}]}"}]}
{"id": "bank_statement_fin_076_extract", "task": "extract", "class": "bank_statement", "doc": "bank_statement_fin_076.pdf", "images": ["images/test/bank_statement/bank_statement_fin_076_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('bank_statement', 'The document class.')\nfields = (['account_holder_name', 'bank_name', 'statement_period_start', 'statement_period_end', 'ending_balance', 'address', 'bank_account_type', 'bank_account_number', 'transaction_amount', 'transaction_date', 'transaction_type', 'transaction_sender'], 'The exact field names to extract.')\nhints = ({'account_holder_name': 'Name on the account', 'bank_name': 'Name of the bank', 'statement_period_start': 'Start of statement period', 'statement_period_end': 'End of statement period', 'ending_balance': 'Closing balance for the period', 'address': 'Address of the account holder', 'bank_account_type': 'Type of bank account - Checking/Saving', 'bank_account_number': 'Bank account number', 'transaction_amount': 'Transaction amount', 'transaction_date': 'Transaction date', 'transaction_type': 'Transaction type', 'transaction_sender': 'Transaction sender'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"account_holder_name\", \"value\": \"JOHN DOE LEE\"}, {\"field\": \"bank_name\", \"value\": \"BARCLAYS\"}, {\"field\": \"statement_period_start\", \"value\": \"2022-01-09\"}, {\"field\": \"statement_period_end\", \"value\": \"2022-02-07\"}, {\"field\": \"ending_balance\", \"value\": \"452.66\"}, {\"field\": \"address\", \"value\": \"10 HUTON GARDENS, HARROW, HA3 6RA\"}, {\"field\": \"bank_account_type\", \"value\": \"Checking\"}, {\"field\": \"bank_account_number\", \"value\": \"12345678\"}, {\"field\": \"transaction_amount\", \"value\": \"540.71\"}, {\"field\": \"transaction_date\", \"value\": \"2022-01-12\"}, {\"field\": \"transaction_type\", \"value\": \"Deposit\"}, {\"field\": \"transaction_sender\", \"value\": \"Barclays Sheffield 32 - Ref: 100017\"}]}"}]}
{"id": "bank_statement_fin_079_p1_classify", "task": "classify", "class": "bank_statement", "doc": "bank_statement_fin_079.pdf", "images": ["images/test/bank_statement/bank_statement_fin_079_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"bank_statement\", \"confidence\": 1.0, \"issuer\": \"HSBC UK\", \"holder\": \"Mr John Doe\", \"page_marker\": \"\"}]}"}]}
{"id": "bank_statement_fin_079_extract", "task": "extract", "class": "bank_statement", "doc": "bank_statement_fin_079.pdf", "images": ["images/test/bank_statement/bank_statement_fin_079_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('bank_statement', 'The document class.')\nfields = (['account_holder_name', 'bank_name', 'statement_period_start', 'statement_period_end', 'ending_balance', 'address', 'bank_account_type', 'bank_account_number', 'transaction_amount', 'transaction_date', 'transaction_type', 'transaction_sender'], 'The exact field names to extract.')\nhints = ({'account_holder_name': 'Name on the account', 'bank_name': 'Name of the bank', 'statement_period_start': 'Start of statement period', 'statement_period_end': 'End of statement period', 'ending_balance': 'Closing balance for the period', 'address': 'Address of the account holder', 'bank_account_type': 'Type of bank account - Checking/Saving', 'bank_account_number': 'Bank account number', 'transaction_amount': 'Transaction amount', 'transaction_date': 'Transaction date', 'transaction_type': 'Transaction type', 'transaction_sender': 'Transaction sender'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"account_holder_name\", \"value\": \"Mr John Doe\"}, {\"field\": \"bank_name\", \"value\": \"HSBC UK\"}, {\"field\": \"statement_period_start\", \"value\": \"2020-10-15\"}, {\"field\": \"statement_period_end\", \"value\": \"2020-11-14\"}, {\"field\": \"ending_balance\", \"value\": \"17.03\"}, {\"field\": \"address\", \"value\": \"20 Sherwood St, London W1F 7ED\"}, {\"field\": \"bank_account_type\", \"value\": \"\"}, {\"field\": \"bank_account_number\", \"value\": \"123456979\"}, {\"field\": \"transaction_amount\", \"value\": \"9000.0\"}, {\"field\": \"transaction_date\", \"value\": \"2020-10-15\"}, {\"field\": \"transaction_type\", \"value\": \"BP\"}, {\"field\": \"transaction_sender\", \"value\": \"GBCLONLTD\"}]}"}]}
{"id": "bank_statement_fin_087_p1_classify", "task": "classify", "class": "bank_statement", "doc": "bank_statement_fin_087.pdf", "images": ["images/test/bank_statement/bank_statement_fin_087_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"bank_statement\", \"confidence\": 1.0, \"issuer\": \"BARCLAYS\", \"holder\": \"Mr Ahad Abdul\", \"page_marker\": \"\"}]}"}]}
{"id": "bank_statement_fin_087_extract", "task": "extract", "class": "bank_statement", "doc": "bank_statement_fin_087.pdf", "images": ["images/test/bank_statement/bank_statement_fin_087_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('bank_statement', 'The document class.')\nfields = (['account_holder_name', 'bank_name', 'statement_period_start', 'statement_period_end', 'ending_balance', 'address', 'bank_account_type', 'bank_account_number', 'transaction_amount', 'transaction_date', 'transaction_type', 'transaction_sender'], 'The exact field names to extract.')\nhints = ({'account_holder_name': 'Name on the account', 'bank_name': 'Name of the bank', 'statement_period_start': 'Start of statement period', 'statement_period_end': 'End of statement period', 'ending_balance': 'Closing balance for the period', 'address': 'Address of the account holder', 'bank_account_type': 'Type of bank account - Checking/Saving', 'bank_account_number': 'Bank account number', 'transaction_amount': 'Transaction amount', 'transaction_date': 'Transaction date', 'transaction_type': 'Transaction type', 'transaction_sender': 'Transaction sender'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"account_holder_name\", \"value\": \"Mr Ahad Abdul\"}, {\"field\": \"bank_name\", \"value\": \"BARCLAYS\"}, {\"field\": \"statement_period_start\", \"value\": \"2024-07-02\"}, {\"field\": \"statement_period_end\", \"value\": \"2024-07-18\"}, {\"field\": \"ending_balance\", \"value\": \"37.27\"}, {\"field\": \"address\", \"value\": \"38 UNIT 10 ST PAUL ST LONDON N1 7AB\"}, {\"field\": \"bank_account_type\", \"value\": \"Checking\"}, {\"field\": \"bank_account_number\", \"value\": \"06766862\"}, {\"field\": \"transaction_amount\", \"value\": \"73.8\"}, {\"field\": \"transaction_date\", \"value\": \"2024-07-02\"}, {\"field\": \"transaction_type\", \"value\": \"Card Payment\"}, {\"field\": \"transaction_sender\", \"value\": \"B & Q 1255\"}]}"}]}
{"id": "drivers_license_fin_012_p1_classify", "task": "classify", "class": "drivers_license", "doc": "drivers_license_fin_012.pdf", "images": ["images/test/drivers_license/drivers_license_fin_012_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"drivers_license\", \"confidence\": 1.0, \"issuer\": \"Florida\", \"holder\": \"RACHEL DAVIS\", \"page_marker\": \"\"}]}"}]}
{"id": "drivers_license_fin_012_extract", "task": "extract", "class": "drivers_license", "doc": "drivers_license_fin_012.pdf", "images": ["images/test/drivers_license/drivers_license_fin_012_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('drivers_license', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"RACHEL DAVIS\"}, {\"field\": \"date_of_birth\", \"value\": \"1991-04-06\"}, {\"field\": \"address\", \"value\": \"1234 STREET, SPRING HILL, FL 34606\"}, {\"field\": \"issue_date\", \"value\": \"2018-01-28\"}, {\"field\": \"expiration_date\", \"value\": \"2027-11-06\"}, {\"field\": \"state\", \"value\": \"Florida\"}]}"}]}
{"id": "drivers_license_fin_019_p1_classify", "task": "classify", "class": "drivers_license", "doc": "drivers_license_fin_019.pdf", "images": ["images/test/drivers_license/drivers_license_fin_019_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"drivers_license\", \"confidence\": 1.0, \"issuer\": \"ALASKA\", \"holder\": \"JANE DIANE SAMPLECARD\", \"page_marker\": \"\"}]}"}]}
{"id": "drivers_license_fin_019_extract", "task": "extract", "class": "drivers_license", "doc": "drivers_license_fin_019.pdf", "images": ["images/test/drivers_license/drivers_license_fin_019_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('drivers_license', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"JANE DIANE SAMPLECARD\"}, {\"field\": \"date_of_birth\", \"value\": \"2000-05-25\"}, {\"field\": \"address\", \"value\": \"1234 EXAMPLE ST, EXAMPLE CITY, AK 12345-0000\"}, {\"field\": \"issue_date\", \"value\": \"2024-05-25\"}, {\"field\": \"expiration_date\", \"value\": \"2032-05-25\"}, {\"field\": \"state\", \"value\": \"ALASKA\"}]}"}]}
{"id": "drivers_license_fin_026_p1_classify", "task": "classify", "class": "drivers_license", "doc": "drivers_license_fin_026.pdf", "images": ["images/test/drivers_license/drivers_license_fin_026_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"drivers_license\", \"confidence\": 1.0, \"issuer\": \"MAINE\", \"holder\": \"SUMMERS JULIE W\", \"page_marker\": \"\"}]}"}]}
{"id": "drivers_license_fin_026_extract", "task": "extract", "class": "drivers_license", "doc": "drivers_license_fin_026.pdf", "images": ["images/test/drivers_license/drivers_license_fin_026_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('drivers_license', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"SUMMERS JULIE W\"}, {\"field\": \"date_of_birth\", \"value\": \"1981-02-17\"}, {\"field\": \"address\", \"value\": \"6969 SUMMERSIDE BLVD PORTLAND, ME 04101\"}, {\"field\": \"issue_date\", \"value\": \"2011-08-05\"}, {\"field\": \"expiration_date\", \"value\": \"2019-02-17\"}, {\"field\": \"state\", \"value\": \"MAINE\"}]}"}]}
{"id": "drivers_license_fin_043_p1_classify", "task": "classify", "class": "drivers_license", "doc": "drivers_license_fin_043.pdf", "images": ["images/test/drivers_license/drivers_license_fin_043_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"drivers_license\", \"confidence\": 1.0, \"issuer\": \"South Carolina\", \"holder\": \"SAMPLE, REGULAR LICENSE\", \"page_marker\": \"\"}]}"}]}
{"id": "drivers_license_fin_043_extract", "task": "extract", "class": "drivers_license", "doc": "drivers_license_fin_043.pdf", "images": ["images/test/drivers_license/drivers_license_fin_043_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('drivers_license', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"SAMPLE, REGULAR LICENSE\"}, {\"field\": \"date_of_birth\", \"value\": \"1972-03-14\"}, {\"field\": \"address\", \"value\": \"110 MAPLE STREET, ANYTOWN, SC 292221111\"}, {\"field\": \"issue_date\", \"value\": \"2008-02-27\"}, {\"field\": \"expiration_date\", \"value\": \"2018-03-14\"}, {\"field\": \"state\", \"value\": \"South Carolina\"}]}"}]}
{"id": "drivers_license_fin_058_p1_classify", "task": "classify", "class": "drivers_license", "doc": "drivers_license_fin_058.pdf", "images": ["images/test/drivers_license/drivers_license_fin_058_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"drivers_license\", \"confidence\": 1.0, \"issuer\": \"Texas\", \"holder\": \"DRIVER TEXAS\", \"page_marker\": \"\"}]}"}]}
{"id": "drivers_license_fin_058_extract", "task": "extract", "class": "drivers_license", "doc": "drivers_license_fin_058.pdf", "images": ["images/test/drivers_license/drivers_license_fin_058_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('drivers_license', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"DRIVER TEXAS\"}, {\"field\": \"date_of_birth\", \"value\": \"1995-01-01\"}, {\"field\": \"address\", \"value\": \"2120 OLD MAIN STREET, ANYTOWN, TX 12345-6789\"}, {\"field\": \"issue_date\", \"value\": \"2020-03-09\"}, {\"field\": \"expiration_date\", \"value\": \"2027-01-01\"}, {\"field\": \"state\", \"value\": \"Texas\"}]}"}]}
{"id": "drivers_license_fin_059_p1_classify", "task": "classify", "class": "drivers_license", "doc": "drivers_license_fin_059.pdf", "images": ["images/test/drivers_license/drivers_license_fin_059_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"drivers_license\", \"confidence\": 1.0, \"issuer\": \"Texas\", \"holder\": \"JANE SAMPLE\", \"page_marker\": \"\"}]}"}]}
{"id": "drivers_license_fin_059_extract", "task": "extract", "class": "drivers_license", "doc": "drivers_license_fin_059.pdf", "images": ["images/test/drivers_license/drivers_license_fin_059_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('drivers_license', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"JANE SAMPLE\"}, {\"field\": \"date_of_birth\", \"value\": \"1990-01-01\"}, {\"field\": \"address\", \"value\": \"123 SAMPLE STREET, SAMPLE CITY, TX 12345\"}, {\"field\": \"issue_date\", \"value\": \"2025-01-04\"}, {\"field\": \"expiration_date\", \"value\": \"2028-01-01\"}, {\"field\": \"state\", \"value\": \"Texas\"}]}"}]}
{"id": "drivers_license_fin_061_p1_classify", "task": "classify", "class": "drivers_license", "doc": "drivers_license_fin_061.pdf", "images": ["images/test/drivers_license/drivers_license_fin_061_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"drivers_license\", \"confidence\": 1.0, \"issuer\": \"COLORADO\", \"holder\": \"SAMPLE ADULT DRIVER LICENSE\", \"page_marker\": \"\"}]}"}]}
{"id": "drivers_license_fin_061_extract", "task": "extract", "class": "drivers_license", "doc": "drivers_license_fin_061.pdf", "images": ["images/test/drivers_license/drivers_license_fin_061_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('drivers_license', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"SAMPLE ADULT DRIVER LICENSE\"}, {\"field\": \"date_of_birth\", \"value\": \"1995-04-20\"}, {\"field\": \"address\", \"value\": \"1881 PIERCE ST, LAKEWOOD, CO 80214\"}, {\"field\": \"issue_date\", \"value\": \"2021-12-16\"}, {\"field\": \"expiration_date\", \"value\": \"2024-12-16\"}, {\"field\": \"state\", \"value\": \"COLORADO\"}]}"}]}
{"id": "drivers_license_fin_068_p1_classify", "task": "classify", "class": "drivers_license", "doc": "drivers_license_fin_068.pdf", "images": ["images/test/drivers_license/drivers_license_fin_068_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"drivers_license\", \"confidence\": 1.0, \"issuer\": \"GEORGIA\", \"holder\": \"IMA GEORGIA SAMPLE\", \"page_marker\": \"\"}]}"}]}
{"id": "drivers_license_fin_068_extract", "task": "extract", "class": "drivers_license", "doc": "drivers_license_fin_068.pdf", "images": ["images/test/drivers_license/drivers_license_fin_068_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('drivers_license', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"IMA GEORGIA SAMPLE\"}, {\"field\": \"date_of_birth\", \"value\": \"1984-06-01\"}, {\"field\": \"address\", \"value\": \"2206 EASTVIEW PARKWAY, CONYERS, GA 30013-5755, ROCKDALE\"}, {\"field\": \"issue_date\", \"value\": \"2018-05-25\"}, {\"field\": \"expiration_date\", \"value\": \"2023-05-25\"}, {\"field\": \"state\", \"value\": \"GEORGIA\"}]}"}]}
{"id": "drivers_license_fin_070_p1_classify", "task": "classify", "class": "drivers_license", "doc": "drivers_license_fin_070.pdf", "images": ["images/test/drivers_license/drivers_license_fin_070_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"drivers_license\", \"confidence\": 1.0, \"issuer\": \"South Carolina\", \"holder\": \"SAMPLE CAROLINE\", \"page_marker\": \"\"}]}"}]}
{"id": "drivers_license_fin_070_extract", "task": "extract", "class": "drivers_license", "doc": "drivers_license_fin_070.pdf", "images": ["images/test/drivers_license/drivers_license_fin_070_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('drivers_license', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"SAMPLE CAROLINE\"}, {\"field\": \"date_of_birth\", \"value\": \"2000-01-01\"}, {\"field\": \"address\", \"value\": \"123 MAIN STREET, COLUMBIA, SC 29201\"}, {\"field\": \"issue_date\", \"value\": \"2025-08-27\"}, {\"field\": \"expiration_date\", \"value\": \"2033-08-27\"}, {\"field\": \"state\", \"value\": \"South Carolina\"}]}"}]}
{"id": "monthly_bill_fin_014_p1_classify", "task": "classify", "class": "monthly_bill", "doc": "monthly_bill_fin_014.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_014_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"monthly_bill\", \"confidence\": 1.0, \"issuer\": \"British Gas\", \"holder\": \"\", \"page_marker\": \"\"}]}"}]}
{"id": "monthly_bill_fin_014_extract", "task": "extract", "class": "monthly_bill", "doc": "monthly_bill_fin_014.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_014_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('monthly_bill', 'The document class.')\nfields = (['full_name', 'address', 'issue_date', 'company_name', 'account_number'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name on the bill', 'address': 'Address', 'issue_date': 'Issued date', 'company_name': 'Name of the issuing company', 'account_number': 'Account number'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"\"}, {\"field\": \"address\", \"value\": \"HEAD OFFICE, ANOTHER PLC, 10 EXAMPLE STREET, ANYTOWN, COUNTYSHIRE, AB12 3CD\"}, {\"field\": \"issue_date\", \"value\": \"2014-03-31\"}, {\"field\": \"company_name\", \"value\": \"British Gas\"}, {\"field\": \"account_number\", \"value\": \"600485895\"}]}"}]}
{"id": "monthly_bill_fin_053_p1_classify", "task": "classify", "class": "monthly_bill", "doc": "monthly_bill_fin_053.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_053_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"monthly_bill\", \"confidence\": 1.0, \"issuer\": \"airtel\", \"holder\": \"Mr  G Ajith Kumar\", \"page_marker\": \"\"}]}"}]}
{"id": "monthly_bill_fin_053_extract", "task": "extract", "class": "monthly_bill", "doc": "monthly_bill_fin_053.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_053_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('monthly_bill', 'The document class.')\nfields = (['full_name', 'address', 'issue_date', 'company_name', 'account_number'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name on the bill', 'address': 'Address', 'issue_date': 'Issued date', 'company_name': 'Name of the issuing company', 'account_number': 'Account number'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"Mr  G Ajith Kumar\"}, {\"field\": \"address\", \"value\": \"F3 2nd Aristo Builders 1Street State Bank Colony, Ondikuppam Chennai Tamil Nadu,600019, Chennai,Tamil Nadu, 600019,\"}, {\"field\": \"issue_date\", \"value\": \"2024-08-03\"}, {\"field\": \"company_name\", \"value\": \"airtel\"}, {\"field\": \"account_number\", \"value\": \"\"}]}"}]}
{"id": "monthly_bill_fin_060_p1_classify", "task": "classify", "class": "monthly_bill", "doc": "monthly_bill_fin_060.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_060_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"monthly_bill\", \"confidence\": 1.0, \"issuer\": \"airtel\", \"holder\": \"Mr. Avinash Tatyaram Jambale\", \"page_marker\": \"\"}]}"}]}
{"id": "monthly_bill_fin_060_extract", "task": "extract", "class": "monthly_bill", "doc": "monthly_bill_fin_060.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_060_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('monthly_bill', 'The document class.')\nfields = (['full_name', 'address', 'issue_date', 'company_name', 'account_number'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name on the bill', 'address': 'Address', 'issue_date': 'Issued date', 'company_name': 'Name of the issuing company', 'account_number': 'Account number'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"Mr. Avinash Tatyaram Jambale\"}, {\"field\": \"address\", \"value\": \"Road No-2 Panchwati Society Spine Road, Sharadnagar,Chikhali Pune, Chikhali Bk, Maharashtra, 412114\"}, {\"field\": \"issue_date\", \"value\": \"2024-04-27\"}, {\"field\": \"company_name\", \"value\": \"airtel\"}, {\"field\": \"account_number\", \"value\": \"\"}]}"}]}
{"id": "monthly_bill_fin_070_p1_classify", "task": "classify", "class": "monthly_bill", "doc": "monthly_bill_fin_070.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_070_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"monthly_bill\", \"confidence\": 1.0, \"issuer\": \"AT&T\", \"holder\": \"JOHN CITIZEN\", \"page_marker\": \"\"}]}"}]}
{"id": "monthly_bill_fin_070_extract", "task": "extract", "class": "monthly_bill", "doc": "monthly_bill_fin_070.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_070_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('monthly_bill', 'The document class.')\nfields = (['full_name', 'address', 'issue_date', 'company_name', 'account_number'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name on the bill', 'address': 'Address', 'issue_date': 'Issued date', 'company_name': 'Name of the issuing company', 'account_number': 'Account number'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"JOHN CITIZEN\"}, {\"field\": \"address\", \"value\": \"2208 HORIZON PL, BURNSVILLE, MN 55337-3097\"}, {\"field\": \"issue_date\", \"value\": \"2025-01-21\"}, {\"field\": \"company_name\", \"value\": \"AT&T\"}, {\"field\": \"account_number\", \"value\": \"199067547094\"}]}"}]}
{"id": "monthly_bill_fin_075_p1_classify", "task": "classify", "class": "monthly_bill", "doc": "monthly_bill_fin_075.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_075_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"monthly_bill\", \"confidence\": 1.0, \"issuer\": \"verizon\", \"holder\": \"GURAM KIKNADZE\", \"page_marker\": \"\"}]}"}]}
{"id": "monthly_bill_fin_075_extract", "task": "extract", "class": "monthly_bill", "doc": "monthly_bill_fin_075.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_075_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('monthly_bill', 'The document class.')\nfields = (['full_name', 'address', 'issue_date', 'company_name', 'account_number'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name on the bill', 'address': 'Address', 'issue_date': 'Issued date', 'company_name': 'Name of the issuing company', 'account_number': 'Account number'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"GURAM KIKNADZE\"}, {\"field\": \"address\", \"value\": \"2626 KINGS HY, BROOKLYN, NY 11229-1750\"}, {\"field\": \"issue_date\", \"value\": \"2024-07-27\"}, {\"field\": \"company_name\", \"value\": \"verizon\"}, {\"field\": \"account_number\", \"value\": \"889926639-00001\"}]}"}]}
{"id": "monthly_bill_fin_076_p1_classify", "task": "classify", "class": "monthly_bill", "doc": "monthly_bill_fin_076.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_076_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"monthly_bill\", \"confidence\": 1.0, \"issuer\": \"verizon\", \"holder\": \"FIRST MIDDLE LAST\", \"page_marker\": \"\"}]}"}]}
{"id": "monthly_bill_fin_076_extract", "task": "extract", "class": "monthly_bill", "doc": "monthly_bill_fin_076.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_076_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('monthly_bill', 'The document class.')\nfields = (['full_name', 'address', 'issue_date', 'company_name', 'account_number'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name on the bill', 'address': 'Address', 'issue_date': 'Issued date', 'company_name': 'Name of the issuing company', 'account_number': 'Account number'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"FIRST MIDDLE LAST\"}, {\"field\": \"address\", \"value\": \"123 STREET, CITY, STATE ZIP\"}, {\"field\": \"issue_date\", \"value\": \"2020-10-16\"}, {\"field\": \"company_name\", \"value\": \"verizon\"}, {\"field\": \"account_number\", \"value\": \"3213213213-21321\"}]}"}]}
{"id": "monthly_bill_fin_078_p1_classify", "task": "classify", "class": "monthly_bill", "doc": "monthly_bill_fin_078.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_078_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"monthly_bill\", \"confidence\": 1.0, \"issuer\": \"Verizon Business\", \"holder\": \"TIMOTHY BUCKLEY\", \"page_marker\": \"\"}]}"}]}
{"id": "monthly_bill_fin_078_extract", "task": "extract", "class": "monthly_bill", "doc": "monthly_bill_fin_078.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_078_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('monthly_bill', 'The document class.')\nfields = (['full_name', 'address', 'issue_date', 'company_name', 'account_number'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name on the bill', 'address': 'Address', 'issue_date': 'Issued date', 'company_name': 'Name of the issuing company', 'account_number': 'Account number'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"TIMOTHY BUCKLEY\"}, {\"field\": \"address\", \"value\": \"7569 MIDLAND TRAIL RD, ASHLAND, KY 41102\"}, {\"field\": \"issue_date\", \"value\": \"2024-11-29\"}, {\"field\": \"company_name\", \"value\": \"Verizon Business\"}, {\"field\": \"account_number\", \"value\": \"242752319-00001\"}]}"}]}
{"id": "monthly_bill_fin_089_p1_classify", "task": "classify", "class": "monthly_bill", "doc": "monthly_bill_fin_089.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_089_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"monthly_bill\", \"confidence\": 1.0, \"issuer\": \"Comcast\", \"holder\": \"ROY H ANDREW\", \"page_marker\": \"\"}]}"}]}
{"id": "monthly_bill_fin_089_extract", "task": "extract", "class": "monthly_bill", "doc": "monthly_bill_fin_089.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_089_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('monthly_bill', 'The document class.')\nfields = (['full_name', 'address', 'issue_date', 'company_name', 'account_number'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name on the bill', 'address': 'Address', 'issue_date': 'Issued date', 'company_name': 'Name of the issuing company', 'account_number': 'Account number'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"ROY H ANDREW\"}, {\"field\": \"address\", \"value\": \"1807 W 36TH AVE APT 1, ANCHORAGE, AK 99517\"}, {\"field\": \"issue_date\", \"value\": \"2022-07-26\"}, {\"field\": \"company_name\", \"value\": \"Comcast\"}, {\"field\": \"account_number\", \"value\": \"8778 10 702 0253213\"}]}"}]}
{"id": "monthly_bill_fin_092_p1_classify", "task": "classify", "class": "monthly_bill", "doc": "monthly_bill_fin_092.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_092_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"monthly_bill\", \"confidence\": 1.0, \"issuer\": \"Comcast\", \"holder\": \"JOSHUA LEE WINTERHEIMER\", \"page_marker\": \"\"}]}"}]}
{"id": "monthly_bill_fin_092_extract", "task": "extract", "class": "monthly_bill", "doc": "monthly_bill_fin_092.pdf", "images": ["images/test/monthly_bill/monthly_bill_fin_092_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('monthly_bill', 'The document class.')\nfields = (['full_name', 'address', 'issue_date', 'company_name', 'account_number'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name on the bill', 'address': 'Address', 'issue_date': 'Issued date', 'company_name': 'Name of the issuing company', 'account_number': 'Account number'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"JOSHUA LEE WINTERHEIMER\"}, {\"field\": \"address\", \"value\": \"5800 Tower Rd Apt# 2411, Denver, CO 80249\"}, {\"field\": \"issue_date\", \"value\": \"2019-06-04\"}, {\"field\": \"company_name\", \"value\": \"Comcast\"}, {\"field\": \"account_number\", \"value\": \"01691 202408-02-4\"}]}"}]}
{"id": "paystub_fin_008_p1_classify", "task": "classify", "class": "paystub", "doc": "paystub_fin_008.pdf", "images": ["images/test/paystub/paystub_fin_008_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"paystub\", \"confidence\": 1.0, \"issuer\": \"PaystubHero\", \"holder\": \"Jane Doe\", \"page_marker\": \"\"}]}"}]}
{"id": "paystub_fin_008_extract", "task": "extract", "class": "paystub", "doc": "paystub_fin_008.pdf", "images": ["images/test/paystub/paystub_fin_008_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('paystub', 'The document class.')\nfields = (['pay_date', 'employee_name', 'employer_name', 'gross_pay', 'ytd_gross', 'pay_frequency', 'pay_period_start', 'pay_period_end', 'federal_tax', 'state_tax', 'social_security_tax', 'medicare_tax', 'hire_date', 'direct_deposit_bank_type'], 'The exact field names to extract.')\nhints = ({'pay_date': 'Date the paycheck was issued', 'employee_name': 'Name of Employee', 'employer_name': 'Name of Employer', 'gross_pay': 'Gross pay per period (current period, not YTD)', 'ytd_gross': 'Gross pay on paystub year to date (cumulative, not current period)', 'pay_frequency': 'The frequency if provided: weekly, biweekly, monthly, semi-monthly', 'pay_period_start': 'Pay period start date', 'pay_period_end': 'Pay period end date', 'hire_date': 'Job hire date', 'direct_deposit_bank_type': 'Direct deposit bank type'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"pay_date\", \"value\": \"2025-03-26\"}, {\"field\": \"employee_name\", \"value\": \"Jane Doe\"}, {\"field\": \"employer_name\", \"value\": \"PaystubHero\"}, {\"field\": \"gross_pay\", \"value\": \"1600.0\"}, {\"field\": \"ytd_gross\", \"value\": \"8000.0\"}, {\"field\": \"pay_frequency\", \"value\": \"\"}, {\"field\": \"pay_period_start\", \"value\": \"2025-03-08\"}, {\"field\": \"pay_period_end\", \"value\": \"2025-03-21\"}, {\"field\": \"federal_tax\", \"value\": \"119.62\"}, {\"field\": \"state_tax\", \"value\": \"33.7\"}, {\"field\": \"social_security_tax\", \"value\": \"99.2\"}, {\"field\": \"medicare_tax\", \"value\": \"23.2\"}, {\"field\": \"hire_date\", \"value\": \"\"}, {\"field\": \"direct_deposit_bank_type\", \"value\": \"\"}]}"}]}
{"id": "paystub_fin_023_p1_classify", "task": "classify", "class": "paystub", "doc": "paystub_fin_023.pdf", "images": ["images/test/paystub/paystub_fin_023_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"paystub\", \"confidence\": 1.0, \"issuer\": \"Acme Corporation\", \"holder\": \"John M. Smith\", \"page_marker\": \"\"}]}"}]}
{"id": "paystub_fin_023_extract", "task": "extract", "class": "paystub", "doc": "paystub_fin_023.pdf", "images": ["images/test/paystub/paystub_fin_023_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('paystub', 'The document class.')\nfields = (['pay_date', 'employee_name', 'employer_name', 'gross_pay', 'ytd_gross', 'pay_frequency', 'pay_period_start', 'pay_period_end', 'federal_tax', 'state_tax', 'social_security_tax', 'medicare_tax', 'hire_date', 'direct_deposit_bank_type'], 'The exact field names to extract.')\nhints = ({'pay_date': 'Date the paycheck was issued', 'employee_name': 'Name of Employee', 'employer_name': 'Name of Employer', 'gross_pay': 'Gross pay per period (current period, not YTD)', 'ytd_gross': 'Gross pay on paystub year to date (cumulative, not current period)', 'pay_frequency': 'The frequency if provided: weekly, biweekly, monthly, semi-monthly', 'pay_period_start': 'Pay period start date', 'pay_period_end': 'Pay period end date', 'hire_date': 'Job hire date', 'direct_deposit_bank_type': 'Direct deposit bank type'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"pay_date\", \"value\": \"2024-12-20\"}, {\"field\": \"employee_name\", \"value\": \"John M. Smith\"}, {\"field\": \"employer_name\", \"value\": \"Acme Corporation\"}, {\"field\": \"gross_pay\", \"value\": \"2450.0\"}, {\"field\": \"ytd_gross\", \"value\": \"50450.0\"}, {\"field\": \"pay_frequency\", \"value\": \"biweekly\"}, {\"field\": \"pay_period_start\", \"value\": \"2024-12-01\"}, {\"field\": \"pay_period_end\", \"value\": \"2024-12-15\"}, {\"field\": \"federal_tax\", \"value\": \"294.0\"}, {\"field\": \"state_tax\", \"value\": \"122.5\"}, {\"field\": \"social_security_tax\", \"value\": \"151.9\"}, {\"field\": \"medicare_tax\", \"value\": \"35.53\"}, {\"field\": \"hire_date\", \"value\": \"\"}, {\"field\": \"direct_deposit_bank_type\", \"value\": \"\"}]}"}]}
{"id": "paystub_fin_045_p1_classify", "task": "classify", "class": "paystub", "doc": "paystub_fin_045.pdf", "images": ["images/test/paystub/paystub_fin_045_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"paystub\", \"confidence\": 1.0, \"issuer\": \"Blue Wisp Cleaners\", \"holder\": \"Lucinda Smith\", \"page_marker\": \"\"}]}"}]}
{"id": "paystub_fin_045_extract", "task": "extract", "class": "paystub", "doc": "paystub_fin_045.pdf", "images": ["images/test/paystub/paystub_fin_045_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('paystub', 'The document class.')\nfields = (['pay_date', 'employee_name', 'employer_name', 'gross_pay', 'ytd_gross', 'pay_frequency', 'pay_period_start', 'pay_period_end', 'federal_tax', 'state_tax', 'social_security_tax', 'medicare_tax', 'hire_date', 'direct_deposit_bank_type'], 'The exact field names to extract.')\nhints = ({'pay_date': 'Date the paycheck was issued', 'employee_name': 'Name of Employee', 'employer_name': 'Name of Employer', 'gross_pay': 'Gross pay per period (current period, not YTD)', 'ytd_gross': 'Gross pay on paystub year to date (cumulative, not current period)', 'pay_frequency': 'The frequency if provided: weekly, biweekly, monthly, semi-monthly', 'pay_period_start': 'Pay period start date', 'pay_period_end': 'Pay period end date', 'hire_date': 'Job hire date', 'direct_deposit_bank_type': 'Direct deposit bank type'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"pay_date\", \"value\": \"1997-11-16\"}, {\"field\": \"employee_name\", \"value\": \"Lucinda Smith\"}, {\"field\": \"employer_name\", \"value\": \"Blue Wisp Cleaners\"}, {\"field\": \"gross_pay\", \"value\": \"294.0\"}, {\"field\": \"ytd_gross\", \"value\": \"1095.0\"}, {\"field\": \"pay_frequency\", \"value\": \"\"}, {\"field\": \"pay_period_start\", \"value\": \"1997-11-09\"}, {\"field\": \"pay_period_end\", \"value\": \"1997-11-16\"}, {\"field\": \"federal_tax\", \"value\": \"29.0\"}, {\"field\": \"state_tax\", \"value\": \"10.8\"}, {\"field\": \"social_security_tax\", \"value\": \"18.23\"}, {\"field\": \"medicare_tax\", \"value\": \"4.26\"}, {\"field\": \"hire_date\", \"value\": \"\"}, {\"field\": \"direct_deposit_bank_type\", \"value\": \"\"}]}"}]}
{"id": "paystub_fin_062_p1_classify", "task": "classify", "class": "paystub", "doc": "paystub_fin_062.pdf", "images": ["images/test/paystub/paystub_fin_062_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"paystub\", \"confidence\": 1.0, \"issuer\": \"Sample Company Name\", \"holder\": \"Employee Name\", \"page_marker\": \"\"}]}"}]}
{"id": "paystub_fin_062_extract", "task": "extract", "class": "paystub", "doc": "paystub_fin_062.pdf", "images": ["images/test/paystub/paystub_fin_062_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('paystub', 'The document class.')\nfields = (['pay_date', 'employee_name', 'employer_name', 'gross_pay', 'ytd_gross', 'pay_frequency', 'pay_period_start', 'pay_period_end', 'federal_tax', 'state_tax', 'social_security_tax', 'medicare_tax', 'hire_date', 'direct_deposit_bank_type'], 'The exact field names to extract.')\nhints = ({'pay_date': 'Date the paycheck was issued', 'employee_name': 'Name of Employee', 'employer_name': 'Name of Employer', 'gross_pay': 'Gross pay per period (current period, not YTD)', 'ytd_gross': 'Gross pay on paystub year to date (cumulative, not current period)', 'pay_frequency': 'The frequency if provided: weekly, biweekly, monthly, semi-monthly', 'pay_period_start': 'Pay period start date', 'pay_period_end': 'Pay period end date', 'hire_date': 'Job hire date', 'direct_deposit_bank_type': 'Direct deposit bank type'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"pay_date\", \"value\": \"2018-02-14\"}, {\"field\": \"employee_name\", \"value\": \"Employee Name\"}, {\"field\": \"employer_name\", \"value\": \"Sample Company Name\"}, {\"field\": \"gross_pay\", \"value\": \"800.0\"}, {\"field\": \"ytd_gross\", \"value\": \"5600.0\"}, {\"field\": \"pay_frequency\", \"value\": \"\"}, {\"field\": \"pay_period_start\", \"value\": \"2018-02-07\"}, {\"field\": \"pay_period_end\", \"value\": \"2018-02-13\"}, {\"field\": \"federal_tax\", \"value\": \"108.1\"}, {\"field\": \"state_tax\", \"value\": \"40.0\"}, {\"field\": \"social_security_tax\", \"value\": \"49.6\"}, {\"field\": \"medicare_tax\", \"value\": \"11.6\"}, {\"field\": \"hire_date\", \"value\": \"\"}, {\"field\": \"direct_deposit_bank_type\", \"value\": \"\"}]}"}]}
{"id": "paystub_fin_064_p1_classify", "task": "classify", "class": "paystub", "doc": "paystub_fin_064.pdf", "images": ["images/test/paystub/paystub_fin_064_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"paystub\", \"confidence\": 1.0, \"issuer\": \"Test Company\", \"holder\": \"Test Name\", \"page_marker\": \"\"}]}"}]}
{"id": "paystub_fin_064_extract", "task": "extract", "class": "paystub", "doc": "paystub_fin_064.pdf", "images": ["images/test/paystub/paystub_fin_064_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('paystub', 'The document class.')\nfields = (['pay_date', 'employee_name', 'employer_name', 'gross_pay', 'ytd_gross', 'pay_frequency', 'pay_period_start', 'pay_period_end', 'federal_tax', 'state_tax', 'social_security_tax', 'medicare_tax', 'hire_date', 'direct_deposit_bank_type'], 'The exact field names to extract.')\nhints = ({'pay_date': 'Date the paycheck was issued', 'employee_name': 'Name of Employee', 'employer_name': 'Name of Employer', 'gross_pay': 'Gross pay per period (current period, not YTD)', 'ytd_gross': 'Gross pay on paystub year to date (cumulative, not current period)', 'pay_frequency': 'The frequency if provided: weekly, biweekly, monthly, semi-monthly', 'pay_period_start': 'Pay period start date', 'pay_period_end': 'Pay period end date', 'hire_date': 'Job hire date', 'direct_deposit_bank_type': 'Direct deposit bank type'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"pay_date\", \"value\": \"2021-06-23\"}, {\"field\": \"employee_name\", \"value\": \"Test Name\"}, {\"field\": \"employer_name\", \"value\": \"Test Company\"}, {\"field\": \"gross_pay\", \"value\": \"400.0\"}, {\"field\": \"ytd_gross\", \"value\": \"10000.0\"}, {\"field\": \"pay_frequency\", \"value\": \"\"}, {\"field\": \"pay_period_start\", \"value\": \"2021-06-16\"}, {\"field\": \"pay_period_end\", \"value\": \"2021-06-22\"}, {\"field\": \"federal_tax\", \"value\": \"44.5\"}, {\"field\": \"state_tax\", \"value\": \"20.0\"}, {\"field\": \"social_security_tax\", \"value\": \"24.8\"}, {\"field\": \"medicare_tax\", \"value\": \"5.8\"}, {\"field\": \"hire_date\", \"value\": \"\"}, {\"field\": \"direct_deposit_bank_type\", \"value\": \"\"}]}"}]}
{"id": "paystub_fin_065_p1_classify", "task": "classify", "class": "paystub", "doc": "paystub_fin_065.pdf", "images": ["images/test/paystub/paystub_fin_065_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"paystub\", \"confidence\": 1.0, \"issuer\": \"Test company\", \"holder\": \"emp test\", \"page_marker\": \"\"}]}"}]}
{"id": "paystub_fin_065_extract", "task": "extract", "class": "paystub", "doc": "paystub_fin_065.pdf", "images": ["images/test/paystub/paystub_fin_065_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('paystub', 'The document class.')\nfields = (['pay_date', 'employee_name', 'employer_name', 'gross_pay', 'ytd_gross', 'pay_frequency', 'pay_period_start', 'pay_period_end', 'federal_tax', 'state_tax', 'social_security_tax', 'medicare_tax', 'hire_date', 'direct_deposit_bank_type'], 'The exact field names to extract.')\nhints = ({'pay_date': 'Date the paycheck was issued', 'employee_name': 'Name of Employee', 'employer_name': 'Name of Employer', 'gross_pay': 'Gross pay per period (current period, not YTD)', 'ytd_gross': 'Gross pay on paystub year to date (cumulative, not current period)', 'pay_frequency': 'The frequency if provided: weekly, biweekly, monthly, semi-monthly', 'pay_period_start': 'Pay period start date', 'pay_period_end': 'Pay period end date', 'hire_date': 'Job hire date', 'direct_deposit_bank_type': 'Direct deposit bank type'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"pay_date\", \"value\": \"2024-12-16\"}, {\"field\": \"employee_name\", \"value\": \"emp test\"}, {\"field\": \"employer_name\", \"value\": \"Test company\"}, {\"field\": \"gross_pay\", \"value\": \"400.0\"}, {\"field\": \"ytd_gross\", \"value\": \"20000.0\"}, {\"field\": \"pay_frequency\", \"value\": \"\"}, {\"field\": \"pay_period_start\", \"value\": \"2024-12-03\"}, {\"field\": \"pay_period_end\", \"value\": \"2025-01-20\"}, {\"field\": \"federal_tax\", \"value\": \"15.1\"}, {\"field\": \"state_tax\", \"value\": \"19.23\"}, {\"field\": \"social_security_tax\", \"value\": \"24.8\"}, {\"field\": \"medicare_tax\", \"value\": \"5.8\"}, {\"field\": \"hire_date\", \"value\": \"\"}, {\"field\": \"direct_deposit_bank_type\", \"value\": \"\"}]}"}]}
{"id": "paystub_fin_073_p1_classify", "task": "classify", "class": "paystub", "doc": "paystub_fin_073.pdf", "images": ["images/test/paystub/paystub_fin_073_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"paystub\", \"confidence\": 1.0, \"issuer\": \"Sample Company Name\", \"holder\": \"James Robert\", \"page_marker\": \"\"}]}"}]}
{"id": "paystub_fin_073_extract", "task": "extract", "class": "paystub", "doc": "paystub_fin_073.pdf", "images": ["images/test/paystub/paystub_fin_073_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('paystub', 'The document class.')\nfields = (['pay_date', 'employee_name', 'employer_name', 'gross_pay', 'ytd_gross', 'pay_frequency', 'pay_period_start', 'pay_period_end', 'federal_tax', 'state_tax', 'social_security_tax', 'medicare_tax', 'hire_date', 'direct_deposit_bank_type'], 'The exact field names to extract.')\nhints = ({'pay_date': 'Date the paycheck was issued', 'employee_name': 'Name of Employee', 'employer_name': 'Name of Employer', 'gross_pay': 'Gross pay per period (current period, not YTD)', 'ytd_gross': 'Gross pay on paystub year to date (cumulative, not current period)', 'pay_frequency': 'The frequency if provided: weekly, biweekly, monthly, semi-monthly', 'pay_period_start': 'Pay period start date', 'pay_period_end': 'Pay period end date', 'hire_date': 'Job hire date', 'direct_deposit_bank_type': 'Direct deposit bank type'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"pay_date\", \"value\": \"2014-01-31\"}, {\"field\": \"employee_name\", \"value\": \"James Robert\"}, {\"field\": \"employer_name\", \"value\": \"Sample Company Name\"}, {\"field\": \"gross_pay\", \"value\": \"1000.0\"}, {\"field\": \"ytd_gross\", \"value\": \"5000.0\"}, {\"field\": \"pay_frequency\", \"value\": \"\"}, {\"field\": \"pay_period_start\", \"value\": \"2014-01-23\"}, {\"field\": \"pay_period_end\", \"value\": \"2014-01-29\"}, {\"field\": \"federal_tax\", \"value\": \"159.5\"}, {\"field\": \"state_tax\", \"value\": \"44.26\"}, {\"field\": \"social_security_tax\", \"value\": \"62.0\"}, {\"field\": \"medicare_tax\", \"value\": \"14.5\"}, {\"field\": \"hire_date\", \"value\": \"\"}, {\"field\": \"direct_deposit_bank_type\", \"value\": \"\"}]}"}]}
{"id": "paystub_fin_076_p1_classify", "task": "classify", "class": "paystub", "doc": "paystub_fin_076.pdf", "images": ["images/test/paystub/paystub_fin_076_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"paystub\", \"confidence\": 1.0, \"issuer\": \"Guardian National Bank\", \"holder\": \"Julie Jones\", \"page_marker\": \"\"}]}"}]}
{"id": "paystub_fin_076_extract", "task": "extract", "class": "paystub", "doc": "paystub_fin_076.pdf", "images": ["images/test/paystub/paystub_fin_076_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('paystub', 'The document class.')\nfields = (['pay_date', 'employee_name', 'employer_name', 'gross_pay', 'ytd_gross', 'pay_frequency', 'pay_period_start', 'pay_period_end', 'federal_tax', 'state_tax', 'social_security_tax', 'medicare_tax', 'hire_date', 'direct_deposit_bank_type'], 'The exact field names to extract.')\nhints = ({'pay_date': 'Date the paycheck was issued', 'employee_name': 'Name of Employee', 'employer_name': 'Name of Employer', 'gross_pay': 'Gross pay per period (current period, not YTD)', 'ytd_gross': 'Gross pay on paystub year to date (cumulative, not current period)', 'pay_frequency': 'The frequency if provided: weekly, biweekly, monthly, semi-monthly', 'pay_period_start': 'Pay period start date', 'pay_period_end': 'Pay period end date', 'hire_date': 'Job hire date', 'direct_deposit_bank_type': 'Direct deposit bank type'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"pay_date\", \"value\": \"2012-12-13\"}, {\"field\": \"employee_name\", \"value\": \"Julie Jones\"}, {\"field\": \"employer_name\", \"value\": \"Guardian National Bank\"}, {\"field\": \"gross_pay\", \"value\": \"1083.0\"}, {\"field\": \"ytd_gross\", \"value\": \"\"}, {\"field\": \"pay_frequency\", \"value\": \"\"}, {\"field\": \"pay_period_start\", \"value\": \"\"}, {\"field\": \"pay_period_end\", \"value\": \"\"}, {\"field\": \"federal_tax\", \"value\": \"122.05\"}, {\"field\": \"state_tax\", \"value\": \"42.27\"}, {\"field\": \"social_security_tax\", \"value\": \"67.15\"}, {\"field\": \"medicare_tax\", \"value\": \"15.7\"}, {\"field\": \"hire_date\", \"value\": \"\"}, {\"field\": \"direct_deposit_bank_type\", \"value\": \"Guardian National Bank\"}]}"}]}
{"id": "paystub_fin_079_p1_classify", "task": "classify", "class": "paystub", "doc": "paystub_fin_079.pdf", "images": ["images/test/paystub/paystub_fin_079_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"paystub\", \"confidence\": 1.0, \"issuer\": \"Your Employer\", \"holder\": \"John R. Doe\", \"page_marker\": \"\"}]}"}]}
{"id": "paystub_fin_079_extract", "task": "extract", "class": "paystub", "doc": "paystub_fin_079.pdf", "images": ["images/test/paystub/paystub_fin_079_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('paystub', 'The document class.')\nfields = (['pay_date', 'employee_name', 'employer_name', 'gross_pay', 'ytd_gross', 'pay_frequency', 'pay_period_start', 'pay_period_end', 'federal_tax', 'state_tax', 'social_security_tax', 'medicare_tax', 'hire_date', 'direct_deposit_bank_type'], 'The exact field names to extract.')\nhints = ({'pay_date': 'Date the paycheck was issued', 'employee_name': 'Name of Employee', 'employer_name': 'Name of Employer', 'gross_pay': 'Gross pay per period (current period, not YTD)', 'ytd_gross': 'Gross pay on paystub year to date (cumulative, not current period)', 'pay_frequency': 'The frequency if provided: weekly, biweekly, monthly, semi-monthly', 'pay_period_start': 'Pay period start date', 'pay_period_end': 'Pay period end date', 'hire_date': 'Job hire date', 'direct_deposit_bank_type': 'Direct deposit bank type'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"pay_date\", \"value\": \"2006-06-19\"}, {\"field\": \"employee_name\", \"value\": \"John R. Doe\"}, {\"field\": \"employer_name\", \"value\": \"Your Employer\"}, {\"field\": \"gross_pay\", \"value\": \"450.0\"}, {\"field\": \"ytd_gross\", \"value\": \"900.0\"}, {\"field\": \"pay_frequency\", \"value\": \"biweekly\"}, {\"field\": \"pay_period_start\", \"value\": \"2006-06-02\"}, {\"field\": \"pay_period_end\", \"value\": \"2006-06-16\"}, {\"field\": \"federal_tax\", \"value\": \"\"}, {\"field\": \"state_tax\", \"value\": \"\"}, {\"field\": \"social_security_tax\", \"value\": \"25.92\"}, {\"field\": \"medicare_tax\", \"value\": \"6.08\"}, {\"field\": \"hire_date\", \"value\": \"\"}, {\"field\": \"direct_deposit_bank_type\", \"value\": \"\"}]}"}]}
{"id": "state_id_fin_006_p1_classify", "task": "classify", "class": "state_id", "doc": "state_id_fin_006.pdf", "images": ["images/test/state_id/state_id_fin_006_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"state_id\", \"confidence\": 1.0, \"issuer\": \"Massachusetts\", \"holder\": \"BRENDAN KELLY\", \"page_marker\": \"\"}]}"}]}
{"id": "state_id_fin_006_extract", "task": "extract", "class": "state_id", "doc": "state_id_fin_006.pdf", "images": ["images/test/state_id/state_id_fin_006_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('state_id', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"BRENDAN KELLY\"}, {\"field\": \"date_of_birth\", \"value\": \"2016-06-15\"}, {\"field\": \"address\", \"value\": \"123 YOUR STREET, BOSTON, MA 02101\"}, {\"field\": \"issue_date\", \"value\": \"2022-12-25\"}, {\"field\": \"expiration_date\", \"value\": \"2036-06-15\"}, {\"field\": \"state\", \"value\": \"Massachusetts\"}]}"}]}
{"id": "state_id_fin_026_p1_classify", "task": "classify", "class": "state_id", "doc": "state_id_fin_026.pdf", "images": ["images/test/state_id/state_id_fin_026_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"state_id\", \"confidence\": 1.0, \"issuer\": \"GEORGIA\", \"holder\": \"IMA GEORGIA SAMPLE\", \"page_marker\": \"\"}]}"}]}
{"id": "state_id_fin_026_extract", "task": "extract", "class": "state_id", "doc": "state_id_fin_026.pdf", "images": ["images/test/state_id/state_id_fin_026_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('state_id', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"IMA GEORGIA SAMPLE\"}, {\"field\": \"date_of_birth\", \"value\": \"1993-02-22\"}, {\"field\": \"address\", \"value\": \"123 MAIN STREET, ANYTOWN, GA 39999, ANYCOUNTY\"}, {\"field\": \"issue_date\", \"value\": \"2018-02-22\"}, {\"field\": \"expiration_date\", \"value\": \"2028-02-22\"}, {\"field\": \"state\", \"value\": \"GEORGIA\"}]}"}]}
{"id": "state_id_fin_035_p1_classify", "task": "classify", "class": "state_id", "doc": "state_id_fin_035.pdf", "images": ["images/test/state_id/state_id_fin_035_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"state_id\", \"confidence\": 1.0, \"issuer\": \"Arizona\", \"holder\": \"SAMPLE CITIZEN\", \"page_marker\": \"\"}]}"}]}
{"id": "state_id_fin_035_extract", "task": "extract", "class": "state_id", "doc": "state_id_fin_035.pdf", "images": ["images/test/state_id/state_id_fin_035_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('state_id', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"SAMPLE CITIZEN\"}, {\"field\": \"date_of_birth\", \"value\": \"1978-02-14\"}, {\"field\": \"address\", \"value\": \"1234 EXAMPLE ST, PHOENIX, AZ 85007\"}, {\"field\": \"issue_date\", \"value\": \"2023-02-01\"}, {\"field\": \"expiration_date\", \"value\": \"2028-02-01\"}, {\"field\": \"state\", \"value\": \"Arizona\"}]}"}]}
{"id": "state_id_fin_036_p1_classify", "task": "classify", "class": "state_id", "doc": "state_id_fin_036.pdf", "images": ["images/test/state_id/state_id_fin_036_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"state_id\", \"confidence\": 1.0, \"issuer\": \"GEORGIA\", \"holder\": \"IMA GEORGIA SAMPLE\", \"page_marker\": \"\"}]}"}]}
{"id": "state_id_fin_036_extract", "task": "extract", "class": "state_id", "doc": "state_id_fin_036.pdf", "images": ["images/test/state_id/state_id_fin_036_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('state_id', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"IMA GEORGIA SAMPLE\"}, {\"field\": \"date_of_birth\", \"value\": \"1993-02-22\"}, {\"field\": \"address\", \"value\": \"123 MAIN STREET, ANYTOWN, GA 39999, ANYCOUNTY\"}, {\"field\": \"issue_date\", \"value\": \"2018-02-22\"}, {\"field\": \"expiration_date\", \"value\": \"2028-02-22\"}, {\"field\": \"state\", \"value\": \"GEORGIA\"}]}"}]}
{"id": "state_id_fin_043_p1_classify", "task": "classify", "class": "state_id", "doc": "state_id_fin_043.pdf", "images": ["images/test/state_id/state_id_fin_043_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"state_id\", \"confidence\": 1.0, \"issuer\": \"Virginia\", \"holder\": \"ZACHARY MAYA ROSE\", \"page_marker\": \"\"}]}"}]}
{"id": "state_id_fin_043_extract", "task": "extract", "class": "state_id", "doc": "state_id_fin_043.pdf", "images": ["images/test/state_id/state_id_fin_043_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('state_id', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"ZACHARY MAYA ROSE\"}, {\"field\": \"date_of_birth\", \"value\": \"2006-04-10\"}, {\"field\": \"address\", \"value\": \"2348 YORK CROSSING DR, HAYES, VA 23072\"}, {\"field\": \"issue_date\", \"value\": \"2023-05-14\"}, {\"field\": \"expiration_date\", \"value\": \"2031-04-10\"}, {\"field\": \"state\", \"value\": \"Virginia\"}]}"}]}
{"id": "state_id_fin_054_p1_classify", "task": "classify", "class": "state_id", "doc": "state_id_fin_054.pdf", "images": ["images/test/state_id/state_id_fin_054_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"state_id\", \"confidence\": 1.0, \"issuer\": \"Hawai'i\", \"holder\": \"SAMPLE LANI KALEA\", \"page_marker\": \"\"}]}"}]}
{"id": "state_id_fin_054_extract", "task": "extract", "class": "state_id", "doc": "state_id_fin_054.pdf", "images": ["images/test/state_id/state_id_fin_054_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('state_id', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"SAMPLE LANI KALEA\"}, {\"field\": \"date_of_birth\", \"value\": \"1974-10-15\"}, {\"field\": \"address\", \"value\": \"123 ANYSTREET ANYCITY, HI 12345\"}, {\"field\": \"issue_date\", \"value\": \"2026-02-15\"}, {\"field\": \"expiration_date\", \"value\": \"2031-12-21\"}, {\"field\": \"state\", \"value\": \"Hawai'i\"}]}"}]}
{"id": "state_id_fin_062_p1_classify", "task": "classify", "class": "state_id", "doc": "state_id_fin_062.pdf", "images": ["images/test/state_id/state_id_fin_062_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"state_id\", \"confidence\": 1.0, \"issuer\": \"Virginia\", \"holder\": \"DAN OLIVIA, JUNE\", \"page_marker\": \"\"}]}"}]}
{"id": "state_id_fin_062_extract", "task": "extract", "class": "state_id", "doc": "state_id_fin_062.pdf", "images": ["images/test/state_id/state_id_fin_062_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('state_id', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"DAN OLIVIA, JUNE\"}, {\"field\": \"date_of_birth\", \"value\": \"1989-09-03\"}, {\"field\": \"address\", \"value\": \"874 GARRISONVILLE ROAD, STAFFORD, VA 22554\"}, {\"field\": \"issue_date\", \"value\": \"2009-06-14\"}, {\"field\": \"expiration_date\", \"value\": \"2011-06-14\"}, {\"field\": \"state\", \"value\": \"Virginia\"}]}"}]}
{"id": "state_id_fin_072_p1_classify", "task": "classify", "class": "state_id", "doc": "state_id_fin_072.pdf", "images": ["images/test/state_id/state_id_fin_072_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"state_id\", \"confidence\": 1.0, \"issuer\": \"North Carolina\", \"holder\": \"DOE JEFFERY JAMES\", \"page_marker\": \"\"}]}"}]}
{"id": "state_id_fin_072_extract", "task": "extract", "class": "state_id", "doc": "state_id_fin_072.pdf", "images": ["images/test/state_id/state_id_fin_072_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('state_id', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"DOE JEFFERY JAMES\"}, {\"field\": \"date_of_birth\", \"value\": \"1975-06-17\"}, {\"field\": \"address\", \"value\": \"1100 NEW BERN AVE, RALEIGH, NC 27697-0002\"}, {\"field\": \"issue_date\", \"value\": \"2017-02-13\"}, {\"field\": \"expiration_date\", \"value\": \"2025-06-17\"}, {\"field\": \"state\", \"value\": \"North Carolina\"}]}"}]}
{"id": "state_id_fin_081_p1_classify", "task": "classify", "class": "state_id", "doc": "state_id_fin_081.pdf", "images": ["images/test/state_id/state_id_fin_081_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"state_id\", \"confidence\": 1.0, \"issuer\": \"California\", \"holder\": \"IVAN ALBERTO CASTANEDA\", \"page_marker\": \"\"}]}"}]}
{"id": "state_id_fin_081_extract", "task": "extract", "class": "state_id", "doc": "state_id_fin_081.pdf", "images": ["images/test/state_id/state_id_fin_081_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('state_id', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"IVAN ALBERTO CASTANEDA\"}, {\"field\": \"date_of_birth\", \"value\": \"1954-07-16\"}, {\"field\": \"address\", \"value\": \"254 HOLYOKE ST, SAN FRANCISCO, CA 94134\"}, {\"field\": \"issue_date\", \"value\": \"2019-04-17\"}, {\"field\": \"expiration_date\", \"value\": \"2026-07-16\"}, {\"field\": \"state\", \"value\": \"California\"}]}"}]}
{"id": "state_id_fin_097_p1_classify", "task": "classify", "class": "state_id", "doc": "state_id_fin_097.pdf", "images": ["images/test/state_id/state_id_fin_097_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Classify the document image into exactly one of the given class names (unknown if none fit), and read its identity fields (issuer, holder, page marker) as printed on the page.\n\nclasses = (['bank_statement', 'drivers_license', 'monthly_bill', 'paystub', 'state_id', 'unknown'], 'Allowed class names to choose from.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"cls\": \"state_id\", \"confidence\": 1.0, \"issuer\": \"Missouri\", \"holder\": \"SAM NONDRIVER SAMPLE\", \"page_marker\": \"\"}]}"}]}
{"id": "state_id_fin_097_extract", "task": "extract", "class": "state_id", "doc": "state_id_fin_097.pdf", "images": ["images/test/state_id/state_id_fin_097_p1.png"], "messages": [{"role": "system", "content": "    This is a task you must complete by returning only the output.\n    Do not include explanations, code, or extra text\u2014only the result.\n    "}, {"role": "user", "content": [{"type": "text", "text": "Read the financial document image and extract the requested fields from what is actually printed on it. Return exactly one entry per requested field, using an empty string when a field is absent. Values are short verbatim strings from the document - never sentences or repeated text. Dates as ISO 8601 (YYYY-MM-DD). Amounts as plain numbers without currency symbols or thousands separators.\n\ncls = ('state_id', 'The document class.')\nfields = (['full_name', 'date_of_birth', 'address', 'issue_date', 'expiration_date', 'state'], 'The exact field names to extract.')\nhints = ({'full_name': 'Name', 'date_of_birth': 'Date of Birth', 'address': 'Address', 'issue_date': 'Issued Date', 'expiration_date': 'Valid Until Date', 'state': 'Issuing state'}, 'Per-field guidance mapping a field name to a short description.')"}, {"type": "image"}, {"type": "text", "text": "Schema requirements:\n- schema_object_wrapper (array)"}]}, {"role": "assistant", "content": "{\"schema_object_wrapper\": [{\"field\": \"full_name\", \"value\": \"SAM NONDRIVER SAMPLE\"}, {\"field\": \"date_of_birth\", \"value\": \"2002-03-26\"}, {\"field\": \"address\", \"value\": \"123 SAMPLE ST, ANYWHERE, MO 65105\"}, {\"field\": \"issue_date\", \"value\": \"2020-03-26\"}, {\"field\": \"expiration_date\", \"value\": \"2023-03-26\"}, {\"field\": \"state\", \"value\": \"Missouri\"}]}"}]}