Dataset Viewer
Auto-converted to Parquet Duplicate
uid
stringlengths
32
32
paragraphs
listlengths
21
173
tables
listlengths
3
7
table_description
stringlengths
537
63.1k
question
stringlengths
38
269
answer
stringlengths
1
136
program
stringlengths
0
382
text_evidence
listlengths
0
10
table_evidence
listlengths
0
19
17da9c24de4c4a43accd023a2c40b2c7
[ "notes payable.", "Considerable judgment is necessary to estimate the fair value of financial instruments.", "The estimates of fair value presented herein are not necessarily indicative of the amounts that could be realized upon disposition of the financial instruments.", "A summary of the carrying amount and...
[ "<table><tr><td></td><td colspan=\"2\"> December 31, 2007</td><td colspan=\"2\"> December 31, 2006</td></tr><tr><td></td><td> Carrying Value</td><td> Fair Value</td><td> Carrying Value</td><td> Fair Value</td></tr><tr><td></td><td colspan=\"4\"> (In thousands)</td></tr><tr><td>Mortgages and notes payable</td><td>$5...
{"0-3-1": "Table 0 shows Mortgages and notes payable of December 31, 2007 Carrying Value (In thousands) is $584,795 .", "0-3-2": "Table 0 shows Mortgages and notes payable of December 31, 2007 Fair Value (In thousands) is $603,200 .", "0-3-3": "Table 0 shows Mortgages and notes payable of December 31, 2006 Carrying Val...
What was the average of theIntegrated Systems in the years where Technical Services is positive? (in million)
9805
add(10881,8729), divide(#0,const_2)
[ 94, 96 ]
[ "4-7-1", "4-7-2", "4-7-3", "4-7-4", "4-7-5", "4-7-6" ]
49ca72c5d69c43d9bca9d73117f85fc5
[ "Table V Nonperforming Loans, Leases and Foreclosed Properties (1)", "## Table 0 ##", "(1) Balances do not include PCI loans even though the customer may be contractually past due.", "Loans accounted for as PCI loans were written down to fair value upon acquisition and accrete interest income over the remaini...
[ "<table><tr><td></td><td colspan=\"5\">December 31</td></tr><tr><td>(Dollars in millions)</td><td>2011</td><td>2010</td><td>2009</td><td>2008</td><td>2007</td></tr><tr><td>Consumer</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>Residential mortgage</td><td>$15,970</td><td>$17,691</td><td>$16,596</td>...
{"0-3-1": "Table 0 shows Residential mortgage of December 31 2011 is $15,970 .", "0-3-2": "Table 0 shows Residential mortgage of December 31 2010 is $17,691 .", "0-3-3": "Table 0 shows Residential mortgage of December 31 2009 is $16,596 .", "0-3-4": "Table 0 shows Residential mortgage of December 31 2008 is $7,057 .", ...
What's the greatest value of Consumer in 2011? (in million)
21164
[]
[ "1-3-1" ]
bef2f619629c42648ec542893bc07e42
[ "approximately $41.6 million was held outside the U. S. in our foreign subsidiaries.", "We have not repatriated nor do we currently have plans to repatriate these funds, but if we were to repatriate cash held outside of the U. S. for domestic cash operations, we would be required to accrue and pay U. S. income...
[ "<table><tr><td></td><td colspan=\"2\">December 31,</td></tr><tr><td></td><td>2015</td><td>2014</td></tr><tr><td></td><td colspan=\"2\">(in thousands)</td></tr><tr><td>Cash and cash equivalents</td><td>$687,326</td><td>$240,031</td></tr></table>", "<table><tr><td></td><td colspan=\"3\">Year Ended December 31,</td...
{"0-3-1": "Table 0 shows Cash and cash equivalents of December 31, 2015 (in thousands) is $687,326 .", "0-3-2": "Table 0 shows Cash and cash equivalents of December 31, 2014 (in thousands) is $240,031 .", "1-4-1": "Table 1 shows Cash provided by (used in) investing activities of Year Ended December 31, 2015 (in thousan...
What's the increasing rate of Total assets in 2018?
0.30983
add(1956147,2108298), add(#0,3081983), add(#1,37743), add(#2,619822), add(#3,2143389), divide(3081983,#4)
[]
[ "3-4-1", "3-5-1", "3-6-1", "3-7-1", "3-8-1", "3-9-1" ]
28a5d63bb7914cb39ff67363f4c26f6a
[ "The table below presents estimated capital costs that are based on current technology to comply with the federal Clean Air Interstate Rule and related state implementation plans through 2018 as well as federal ambient air quality standards including ozone and fine particulates, and the federal Clean Air Visibility...
[ "<table><tr><td></td><td> 2009</td><td colspan=\"2\"> 2010 - 2013</td><td colspan=\"2\"> 2014 - 2018</td><td colspan=\"2\"> Total</td></tr><tr><td>UE<sup>(a)</sup></td><td>$100</td><td>$525-</td><td>$655</td><td>$1,530-</td><td>$1,885</td><td>$2,155-</td><td>$2,640</td></tr><tr><td>Genco</td><td>230</td><td>875-</t...
{"0-1-1": "Table 0 shows UE of 2009 is $100 .", "0-1-2": "Table 0 shows UE of 2010 - 2013 is $525- .", "0-1-3": "Table 0 shows UE of 2010 - 2013.1 is $655 .", "0-1-4": "Table 0 shows UE of 2014 - 2018 is $1,530- .", "0-1-5": "Table 0 shows UE of 2014 - 2018.1 is $1,885 .", "0-1-6": "Table 0 shows UE of Total is $2,155-...
As As the chart 3 shows,in the year Ended September 30 where Total Revenues is the lowest, what's the Net Revenues? (in thousand)
472821
[ 76 ]
[ "3-13-1", "3-13-3", "3-13-5", "3-15-5" ]
b6dc26a6d6484015a0340ff0ef5f5964
[ "ADOBE SYSTEMS INCORPORATED NOTES TO CONSOLIDATED FINANCIAL STATEMENTS (Continued) 99 NOTE 15.", "COMMITMENTS AND CONTINGENCIES Lease Commitments We lease certain of our facilities and some of our equipment under non-cancellable operating lease arrangements that expire at various dates through 2028.", "We also ...
[ "<table><tr><td></td><td>2014</td><td>2013</td><td>2012</td></tr><tr><td>Rent expense</td><td>$111,149</td><td>$118,976</td><td>$105,809</td></tr><tr><td>Less: sublease income</td><td>1,412</td><td>3,057</td><td>2,330</td></tr><tr><td>Net rent expense</td><td>$109,737</td><td>$115,919</td><td>$103,479</td></tr></ta...
{"0-1-1": "Table 0 shows Rent expense of 2014 is $111,149 .", "0-1-2": "Table 0 shows Rent expense of 2013 is $118,976 .", "0-1-3": "Table 0 shows Rent expense of 2012 is $105,809 .", "0-2-1": "Table 0 shows Less: sublease income of 2014 is 1412 .", "0-2-2": "Table 0 shows Less: sublease income of 2013 is 3057 .", "0-2...
How long does Available for sale for Average balance keep growing?
1
subtract(2006,2005)
[]
[ "2-6-1", "2-6-4" ]
e5881d1f03ad43e1812b1c182962b841
["1.","Organization and Significant Accounting Policies (continued) Commodity Futures Contracts In a(...TRUNCATED)
["<table><tr><td>December 31 (dollars in millions)</td><td colspan=\"2\">2010</td><td colspan=\"2\">(...TRUNCATED)
"{\"0-2-1\": \"Table 0 shows Euro of 2010 Buy is $- .\", \"0-2-2\": \"Table 0 shows Euro of 2010 Sel(...TRUNCATED)
What is the sum of Electric sales to affiliates in 2013 and Total in 2010 for Buy ?
663.2
add(652,11.2)
[ 17, 40 ]
[ "0-5-1", "1-3-1" ]
82e423667c66428dad339f864a02a819
["Corporate & Other The following table presents the results of operations of our Corporate & Other (...TRUNCATED)
["<table><tr><td> </td><td colspan=\"2\">Years Ended December 31,</td><td></td><td></td></tr><tr><td(...TRUNCATED)
"{\"0-4-1\": \"Table 0 shows Management and financial advice fees of Years Ended December 31, 2008 ((...TRUNCATED)
"Without Management and financial advice fees and Other revenues, how much of revenue is there in to(...TRUNCATED)
22
[ 0 ]
[ "0-5-2" ]
3a4fadb978fa4a968f297478e6780181
["THE GOLDMAN SACHS GROUP, INC. AND SUBSIDIARIES Management’s Discussion and Analysis Scenario An(...TRUNCATED)
["<table><tr><td></td><td>As of December</td></tr><tr><td><i>$ in millions</i></td><td>2016</td><td>(...TRUNCATED)
"{\"0-2-1\": \"Table 0 shows Global Core Liquid Assets (GCLA) of As of December 2016 is $226,066 .\"(...TRUNCATED)
"What's the sum of all ReimbursementAgreements that are greater than 0 in Partner Type? (in thousand(...TRUNCATED)
16826
[ 61 ]
[ "3-8-3" ]
69c0918e3a9e4ffcb7e62111e3fc6455
["The following table presents consolidated revenues by our business models and by segment, as a per(...TRUNCATED)
["<table><tr><td></td><td colspan=\"3\">For the Years Ended</td></tr><tr><td></td><td>2018</td><td>2(...TRUNCATED)
"{\"1-2-1\": \"Table 1 shows Software development costs of For the Years Ended 2018 is $747,128 .\",(...TRUNCATED)
what is the ratio of the 2012 contribution to the anticipated employee contributions in 2013
2.4
divide(3.6,1.5)
[ 83, 84, 85 ]
[]
90dfab7d162a4a2f9b48ca37b5d8d44f
["Shareowner Return Performance Graph The following Performance Graph and related information shall (...TRUNCATED)
["<table><tr><td></td><td>12/31/01</td><td>12/31/02</td><td>12/31/03</td><td>12/31/04</td><td>12/31/(...TRUNCATED)
"{\"0-1-1\": \"Table 0 shows United Parcel Service, Inc. of 12/31/01 is $100.00 .\", \"0-1-2\": \"Ta(...TRUNCATED)
"What will Dollars for Balance at end of year for Deferred Compensation Obligations be like in 2007 (...TRUNCATED)
134.217
subtract(147,161), divide(#0,161), add(const_1,#1), multiply(147,#2)
[ 26 ]
[ "2-7-2", "2-7-4" ]
End of preview. Expand in Data Studio

MultiHiertt

A repackaging of the MultiHiertt dataset (Zhao et al., ACL 2022) for numerical reasoning over documents containing multiple hierarchical financial tables.

MultiHiertt is built on FinTabNet (CDLA-Permissive-1.0), extracting 4,791 multi-page documents from S&P 500 annual reports (1999-2019). Each document contains 2-6 hierarchical HTML tables and surrounding text. Questions require reasoning across multiple tables and/or text passages, with answers derived via arithmetic programs or direct span selection.

Why this version

The existing HuggingFace repository (yilunzhao/MultiHiertt) bundles model checkpoints alongside the dataset files, totalling 7.66 GB, and the dataset viewer is broken due to an Arrow conversion error on the HTML table fields. This version packages only the annotation data (172 MB) in a clean parquet format with a documented schema.

Splits

Split Examples
train 7,830
validation 1,044

The reference-free leaderboard test set (1,566 examples, question only) is intentionally excluded.

Schema

Field Type Description
uid string Unique example ID (MD5 hash)
paragraphs list[string] Sentences from the document; table positions marked with ## Table N ## placeholders
tables list[string] Raw HTML string for each table in the document
table_description string JSON-serialized dict mapping "{table}-{row}-{col}" keys to natural-language cell descriptions (e.g. "Table 0 shows Revenue is $446.")
question string The financial question
answer string Numeric answer (e.g. "9805") or text span; coerced to string
program string Flat DSL reasoning program (e.g. add(10881,8729), divide(#0,const_2)); empty string for span-selection questions (~19% of train)
text_evidence list[int] Zero-indexed paragraph indices used as gold evidence
table_evidence list[string] Gold cell references in "{table}-{row}-{col}" format, matching table_description keys

table_description format

Keys follow the pattern "{table_idx}-{row}-{col}" (zero-indexed table, one-indexed row and column). Values are auto-generated natural-language descriptions of the cell in context, produced by the authors' preprocessing script:

import json
td = json.loads(row["table_description"])
# {"0-2-1": "Table 0 shows Revenue of 2007 is $446.", "0-2-2": ...}

Program format

Programs use the same DSL as FinQA: flat token sequences with #N back-references to prior step results and const_* predefined constants. Unlike FinQA, MultiHiertt does not include a steps decomposition or program_re nested form.

Usage

from datasets import load_dataset
import json

ds = load_dataset("rootsautomation/MultiHiertt")
ex = ds["train"][0]

print(ex["question"])
print(ex["answer"])
print(ex["program"])       # empty string if span-selection

# Reconstruct document with tables in position
for para in ex["paragraphs"]:
    if para.startswith("## Table"):
        idx = int(para.split()[2])
        print(f"[TABLE {idx}]: {ex['tables'][idx][:80]}...")
    else:
        print(para)

# Access cell-level evidence
td = json.loads(ex["table_description"])
for cell_ref in ex["table_evidence"]:
    print(f"  {cell_ref}: {td.get(cell_ref, '(no description)')}")

License

QA annotations and preprocessing code: MIT.

Underlying table data is sourced from FinTabNet, released under CDLA-Permissive-1.0. The original annual reports are publicly available SEC filings from S&P 500 companies.

Citation

@inproceedings{zhao-etal-2022-multihiertt,
    title     = "{M}ulti{H}iertt: Numerical Reasoning over Multi Hierarchical Tabular and Textual Data",
    author    = "Zhao, Yilun and Li, Yunxiang and Li, Chenying and Zhang, Rui",
    booktitle = "Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month     = may,
    year      = "2022",
    address   = "Dublin, Ireland",
    publisher = "Association for Computational Linguistics",
    url       = "https://aclanthology.org/2022.acl-long.454",
    pages     = "6588--6600",
}
Downloads last month
7