LEDGER
Collection
A Long-Context Benchmark of Corporate Annual Reports for Grounded Financial Retrieval and Extraction • 3 items • Updated • 6
Datasets:
ticker stringlengths 2 6 | exchange stringclasses 4
values | company_name stringlengths 8 43 | industry stringclasses 6
values | year int64 2.02k 2.02k | accounts_payable float64 102k 5.88B ⌀ | accounts_receivable float64 113 10.5B ⌀ | capex float64 -55,684,000 3.9B ⌀ | cash_and_equivalents float64 -985,000 3.04B ⌀ | cash_incl_restricted float64 0 3.04B ⌀ | cost_of_revenue float64 -1,490,000 28.7B ⌀ | depreciation_amortization float64 -53,083,000 3.84B ⌀ | dividends_paid float64 -100,000,000 399B ⌀ | eps_basic float64 -148.08 41.1 ⌀ | eps_diluted float64 -148.08 41.1 ⌀ | financing_cash_flow float64 -37,230,000,000 27B | gross_profit float64 -903,928,000 9.33B ⌀ | income_tax_expense float64 -1,461,000,000 3.04B ⌀ | interest_expense float64 -7,097,000 2.15B ⌀ | inventory float64 -8,995,103,000 14.3B ⌀ | investing_cash_flow float64 -27,936,000,000 36.5B ⌀ | long_term_debt_current float64 0 86.8B ⌀ | long_term_debt_noncurrent float64 0 16B ⌀ | long_term_debt_total float64 0 21.8B ⌀ | net_income float64 -10,518,000,000 3.67B ⌀ | operating_cash_flow float64 -257,824,000 5.71B | operating_income float64 -6,279,000,000 5.57B ⌀ | rd_expense float64 0 787M ⌀ | revenue float64 -877,954,000 32.9B | sga_expense float64 347k 6.33B ⌀ | shares_outstanding float64 0 1.42B ⌀ | short_term_borrowings float64 0 96.3B ⌀ | stockholders_equity float64 -1,639,000,000 36.8B ⌀ | stockholders_equity_incl_nci float64 -717,000,000 37.3B ⌀ | total_assets float64 1.84M 113B | total_liabilities float64 800k 102B | mmd_text stringlengths 19.4k 531k |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
AAP | NYSE | Advance Auto Parts, Inc. | Consumer Cyclical / Auto Parts | 2,017 | 2,894,582,000 | 606,357,000 | 189,758,000 | 546,937,000 | 546,937,000 | 5,288,735,000 | 249,260,000 | 17,854,000 | 6.44 | 6.42 | -14,872,000 | 4,085,049,000 | 44,754,000 | null | 4,168,492,000 | -178,639,000 | 350,000 | 1,044,327,000 | null | 475,505,000 | 600,805,000 | 570,212,000 | null | 9,373,784,000 | 3,514,837,000 | 73,936,000 | null | 3,415,196,000 | null | 8,482,301,000 | 5,067,105,000 | "\n## 2017\nANNUAL\nREPORT\n\n\n\nADVANCE AUTO PARTS, INC.\n\n<--- Page Split ---(...TRUNCATED) |
AAP | NYSE | Advance Auto Parts, Inc. | Consumer Cyclical / Auto Parts | 2,018 | 3,172,790,000 | 624,972,000 | 193,715,000 | 896,527,000 | 896,527,000 | 5,361,141,000 | 238,184,000 | 17,819,000 | 5.75 | 5.73 | -263,915,000 | 4,219,413,000 | 131,417,000 | null | 4,362,547,000 | -191,827,000 | 210,000 | 1,045,720,000 | null | 423,847,000 | 811,028,000 | 604,275,000 | null | 9,580,554,000 | 3,615,138,000 | 72,460,000 | null | 3,550,813,000 | null | 9,040,648,000 | 5,489,835,000 | "\n## THINK AHEAD.\nTHINK ADVANCE.\n2018 ANNUAL REPORT\n\nAdvance Auto Parts!\n\n<--- Page Split ---(...TRUNCATED) |
AAP | NYSE | Advance Auto Parts, Inc. | Consumer Cyclical / Auto Parts | 2,019 | 3,421,987,000 | 689,469,000 | 270,129,000 | 418,665,000 | 418,665,000 | 5,454,257,000 | 238,371,000 | 17,185,000 | 6.87 | 6.84 | -882,153,000 | 4,254,746,000 | 150,850,000 | null | 4,432,168,000 | -462,939,000 | 0 | 747,320,000 | null | 486,896,000 | 866,909,000 | 677,180,000 | null | 9,709,003,000 | 3,577,566,000 | 69,232,000 | null | 3,549,081,000 | null | 11,248,525,000 | 7,699,444,000 | "\n\n## 2019\nANNUAL REPORT\n\n<--- Page Split --->\n\n\n“At(...TRUNCATED) |
AAP | NYSE | Advance Auto Parts, Inc. | Consumer Cyclical / Auto Parts | 2,020 | 3,640,639,000 | 749,999,000 | 267,576,000 | 834,992,000 | 826,980,000 | 5,624,707,000 | 250,081,000 | 56,347,000 | 7.17 | 7.14 | -285,997,000 | 4,481,614,000 | 157,994,000 | null | 4,538,199,000 | -266,897,000 | null | 1,032,984,000 | null | 493,021,000 | 969,688,000 | 749,907,000 | null | 10,106,321,000 | 3,731,707,000 | 66,361,000 | null | 3,536,961,000 | null | 11,839,636,000 | 8,280,124,000 | "\n## ANNUAL\nREPORT\n\nAdvance Auto Parts!\n\n<--- Page Split --->\n\n\n4,976\nS(...TRUNCATED) |
AAP | NYSE | Advance Auto Parts, Inc. | Consumer Cyclical / Auto Parts | 2,021 | 3,966,574,000 | 754,114,000 | 289,639,000 | 588,050,000 | 588,050,000 | 6,074,039,000 | 259,933,000 | 160,925,000 | 9.32 | 9.25 | -1,064,112,000 | 4,923,950,000 | 185,878,000 | null | 4,682,635,000 | -287,314,000 | 0 | 1,034,320,000 | null | 596,615,000 | 1,107,022,000 | 822,365,000 | null | 10,997,989,000 | 4,101,585,000 | 62,009,000 | null | 3,081,792,000 | null | 12,175,777,000 | 9,093,985,000 | "\n## 2021\nANNUAL REPORT\n\n<--- Page Split --->\n\nNEW YORK\nSTOCK\nEXCHANGE\nDEBUT\n\n |
AAP | NYSE | Advance Auto Parts, Inc. | Consumer Cyclical / Auto Parts | 2,022 | 4,178,907,000 | 684,048,000 | 398,757,000 | 270,805,000 | 220,134,000 | 4,916,004,000 | 248,327,000 | 336,230,000 | 7.7 | 7.65 | -620,704,000 | 4,232,870,000 | 99,657,000 | null | 4,896,269,000 | -424,448,000 | 185,000,000 | 1,188,283,000 | null | 464,402,000 | 736,571,000 | 524,618,000 | null | 9,148,874,000 | 3,708,252,000 | 59,264,000 | null | 2,599,000,000 | null | 11,986,447,000 | 9,387,255,000 | "\n## 2022\nANNUAL REPORT\n\n## Advance Auto Parts\n\n<--- Page Split --->\n\nADVANCEIVERSARY\n\n |
ABDP.L | LSE | AB Dynamics plc | Consumer Cyclical / Auto Parts | 2,022 | 4,956,000 | 10,092,000 | -3,809,000 | 30,141,000 | null | 36,085,000 | 9,136,000 | -1,131,000 | 0.173 | 0.171 | -2,045,000 | 47,141,000 | 1,039,000 | 0 | 13,651,000 | -8,923,000 | null | null | null | 4,741,000 | 17,908,000 | 6,154,000 | 1,670,000 | 83,226,000 | 40,987,000 | 22,626,466 | null | 112,382,000 | null | 141,723,000 | 29,341,000 | "\n## Delivering\nstrategic growth\n\nAB Dynamics plc Annual Report 2022\n\n<--- Page Split --->\n\n(...TRUNCATED) |
ADNT | NYSE | Adient plc | Consumer Cyclical / Auto Parts | 2,022 | 2,478,000,000 | 1,852,000,000 | 227,000,000 | 947,000,000 | 947,000,000 | 13,314,000,000 | 298,000,000 | 106,000,000 | -1.27 | -1.27 | -1,273,000,000 | 807,000,000 | 94,000,000 | 192,000,000 | 953,000,000 | 484,000,000 | 14,000,000 | null | 2,564,000,000 | -120,000,000 | 274,000,000 | 209,000,000 | 322,000,000 | 14,121,000,000 | 598,000,000 | 94,858,156 | 3,000,000 | 2,073,000,000 | 2,375,000,000 | 9,158,000,000 | 6,738,000,000 | "\n# UNITED STATES\nSECURITIES AND EXCHANGE COMMISSION\nWashington, D.C. 20549\n\nFORM 10-K\n\n# ☑(...TRUNCATED) |
AGNC | NASDAQ | AGNC Investment Corp. | Real Estate / REIT - Mortgage | 2,018 | 518,000,000 | 489,000,000 | 0 | 921,000,000 | 1,520,000,000 | 50,000,000 | 25,000,000 | 974,000,000 | 0.21 | 0.21 | 26,980,000,000 | 1,870,000,000 | 24,839,000 | 1,173,000,000 | null | -27,936,000,000 | 72,892,000,000 | null | 21,790,000,000 | 129,000,000 | 1,113,000,000 | 1,302,000,000 | 0.07 | 1,920,000,000 | 55,000,000 | 536,300,000 | 75,717,000,000 | 9,906,000,000 | null | 109,241,000,000 | 99,335,000,000 | "\nProviding Private Capital to the U.S. Housing Market Through\n\n# INVESTMENT EXCELLENCE\n\n2018 A(...TRUNCATED) |
AGNC | NASDAQ | AGNC Investment Corp. | Real Estate / REIT - Mortgage | 2,019 | 424,000,000 | 10,181,000,000 | 0 | 831,000,000 | 1,282,000,000 | 2,149,000,000 | 601,000,000 | 1,139,000,000 | 1.16 | 1.16 | 12,800,000,000 | 771,000,000 | -1,461,000,000 | 2,149,000,000 | 1,282,000,000 | -14,218,000,000 | 86,760,000,000 | null | 228,000,000 | 688,000,000 | 1,180,000,000 | 688,000,000 | 0.64 | 771,000,000 | 36,000,000 | 540,900,000 | 96,303,000,000 | 11,041,000,000 | null | 113,082,000,000 | 102,041,000,000 | "\nPROVIDING PRIVATE CAPITAL TO\nTHE U.S. HOUSING MARKET\nTHROUGH INVESTMENT EXCELLENCE\n\n## 2019\n(...TRUNCATED) |
End of preview. Expand in Data Studio
the LEDGER Long-Context Multi-KPI extraction datasets and benchmarks.
OCR'd annual reports with ground-truth KPI values for financial information extraction benchmarking.
Dataset Description
This dataset pairs OCR-extracted annual report text (from DeepSeek OCR) with structured KPI ground-truth values. It is designed for evaluating LLM-based financial information extraction, retrieval, and needle-in-a-haystack tasks.
Configs
| Config | Reports | Companies | KPI rows | Years | Purpose |
|---|---|---|---|---|---|
no_eval |
4,505 | 725 | 104,529 | 2009–2024 | Training / development |
eval |
494 | 111 | 13,519 | 2017–2022 | Benchmark evaluation |
Schema
Each row in the parquet files contains:
| Column | Type | Description |
|---|---|---|
ticker |
string | Stock ticker symbol |
exchange |
string | Stock exchange (NYSE, NASDAQ, LSE, AMEX, ASX, OTC) |
company_name |
string | Company long name |
industry |
string | Industry classification |
year |
int | Fiscal year |
revenue |
float64 | Total revenue |
net_income |
float64 | Net income |
total_assets |
float64 | Total assets |
total_liabilities |
float64 | Total liabilities |
| ... | float64 | 31 KPI columns total (see below) |
mmd_text |
string | Full OCR text of the annual report (Markdown with page splits) |
KPI columns (31): accounts_payable, accounts_receivable, capex, cash_and_equivalents, cash_incl_restricted, cost_of_revenue, depreciation_amortization, dividends_paid, eps_basic, eps_diluted, financing_cash_flow, gross_profit, income_tax_expense, interest_expense, inventory, investing_cash_flow, long_term_debt_current, long_term_debt_noncurrent, long_term_debt_total, net_income, operating_cash_flow, operating_income, rd_expense, revenue, sga_expense, shares_outstanding, short_term_borrowings, stockholders_equity, stockholders_equity_incl_nci, total_assets, total_liabilities.
KPI values are in millions (as-reported, no FX conversion). NaN means the KPI was not available for that report/year.
Additional Files
no_eval/mmd/andeval/mmd/: Raw.mmdfiles (same text as themmd_textcolumn, for direct file access).eval/images/: Page-level JPEG images for eval reports (not included for no_eval to save space).
OCR Format
The .mmd files use Markdown with page boundaries marked by <--- Page Split --->. Images are referenced as .
Usage
from datasets import load_dataset
# Load training set
ds = load_dataset("artefactory/ledger-long-context-multi-kpi", "no_eval")
# Load eval set
ds_eval = load_dataset("artefactory/ledger-long-context-multi-kpi", "eval")
# Example: filter to reports with revenue data
ds_filtered = ds["train"].filter(lambda x: x["revenue"] is not None)
Data Sources
- OCR text: DeepSeek OCR applied to annual report PDFs from SEC EDGAR, LSE, ASX, and other exchanges.
- KPI values: SEC EDGAR (XBRL companyfacts) for US listings; yfinance for non-US; Alpha Vantage for gap-fill.
Citation
If you use this dataset, please cite LEDGER (citation key comming soon).
- Downloads last month
- 2,137