Synthetic Form 1120-SF - U.S. Income Tax Return for Settlement Funds under Section 468B Data
Synthetic training data — no real PII, fully coherent identities
Generate synthetic Form 1120-SF returns for section 468B settlement funds — investment income, administrative expenses, and distributions, with every amount split into separate dollar and cent columns. Training data for settlement administration and legal-finance document pipelines.
107
Fields per document
2
Pages
Tax
Category
What this document is
Form 1120-SF is the income tax return of a designated or qualified settlement fund under section 468B — the escrow-like entity that holds money between a legal settlement and its distribution to claimants. It reports what the fund earned while holding the money (interest, dividends, capital gains, partnership income), what it spent administering itself, and the resulting tax. Like the 1120-FSC, it prints every amount in separate dollar and cent columns.
Why generate synthetically
Settlement administration is a document-driven industry — class-action administrators, mass-tort trusts, and litigation-finance firms all process fund filings — and its documents are among the least represented in any public corpus. Synthetic 1120-SFs give that industry a labeled training set, and they add a second, smaller split-column form to any corpus built around the dollars/cents extraction problem.
What makes synthetic data useful
Each return is generated for a coherent entity with investment income components scaled together, so gross income is a true sum of its parts rather than an independent draw. Amounts are split into dollar and cent components at generation time, meaning ground truth records exactly which digits belong in which column — the property you need to train or evaluate split-column extraction properly.
Training challenges
The dollars/cents split doubles the column count on every monetary line, and because the cent column is numeric-typed while the dollar column is currency-typed, a pipeline that infers type from the column header will mis-cast half the values. A quarter of the form's fields are cent columns whose values are uniformly distributed across the two-digit range, offering no statistical shortcut for separating them from dollar amounts. Most income categories are legitimately zero on a typical fund — dividends, capital gains, and partnership income all have zero medians — because most settlement funds simply hold cash and earn interest. Nineteen checkboxes sit interleaved with the numeric grid across two pages.
Generate synthetic Form 1120-SF - U.S. Income Tax Return for Settlement Funds under Section 468B data
Start with 500 free credits. No credit card required.
Generate NowWho uses this data
Class-action and mass-tort settlement administrators, litigation-finance and legal-funding platforms, trust and escrow accounting providers, and law-firm back-office automation. Settlement fund reporting is a recurring, compliance-driven document workload with real deadlines, and it is currently handled almost entirely by manual review.
Document complexity profile
107 fields across 2 pages: 48 currency amounts paired with 25 numeric cent columns, plus 19 checkbox targets and 15 text fields. 131 annotation relations — above the field count, because paired dollar and cent columns share one printed label. There are no conditional or arithmetic bindings; the generator supplies a consistent income and expense set, and the extraction difficulty is the split-column geometry and the prevalence of legitimate zeros.
Key stats from our synthetic corpus
Quantitative characteristics of the Form 1120-SF - U.S. Income Tax Return for Settlement Funds under Section 468B documents our generator produces.
| Metric | Value | Detail |
|---|---|---|
| Population coverage | 64% | 64% of identities in our 1,000-identity corpus are attached to an entity and therefore 1120-SF eligible. |
| Median taxable interest | $16,181 | Median taxable interest is $16,181, p25–p75 of $8,325 to $27,122. Interest is the dominant income line because a settlement fund's job is to hold money, not to invest it aggressively. |
| Median gross income | $29,343 | Median gross income is $29,343 with a maximum of $626,322. The wide range reflects funds holding anything from a small local settlement to a large multi-claimant one. |
| Cent columns | 25 | 25 of the form's fields are cent columns, numeric-typed and uniformly distributed across the two-digit range. There is no distributional shortcut for telling them apart from dollar amounts — a model has to learn the column geometry. |
| Income lines at zero | 4 of 6 | Median dividends, capital gains, partnership income, and other income are all zero. Two thirds of the income categories are empty on a typical fund return. |
How this document co-occurs with others
Rates at which identities in our corpus that produce a Form 1120-SF - U.S. Income Tax Return for Settlement Funds under Section 468B also produce other documents.
| Correlation | Rate | Detail |
|---|---|---|
| Other split-column filer (1120-FSC) | 100% | Form 1120-FSC uses the same dollars/cents split at roughly six times the field count. Generating both gives a split-column corpus that spans two very different form scales. |
| Standard corporate return | 100% | The same entity can produce a standard Form 1120, contrasting a single-column corporate layout against the split-column fund return. |
| Can file a 7004 extension | 100% | The same entity can produce a Form 7004 extension whose form code identifies the return being extended — settlement funds routinely extend while distributions are still in progress. |
| Fund W-9 on file | 100% | Every entity can co-generate a W-9 carrying the fund name and TIN printed on the return — the identity document a settlement administrator collects before any distribution. |
| Files quarterly payroll returns | 100% | Every entity in the corpus is an employer filing Form 941 quarterly, so fund tax filings and administrative payroll documents generate as one coherent set. |
All stats above are corpus-derived: they were computed on a local synthetic corpus of 1,000 generated identities produced by SymageDocs' World Simulation Engine, with 1120-SF figures measured over the entity-eligible subset. No real settlement, claimant, or fund data was used. `make corpus-stats` regenerates the shipped corpus_stats.json for the site's analytics form set, which does not include this edition; reproduce these figures with `python -m symagedocs.corpus_analysis --corpus-size 1000 --form-id irs_f1120sf_standard_2024`.
Frequently asked questions
- What data format do synthetic Form 1120-SF documents include?
- Each generated identity produces a rendered PDF plus a structured JSON annotation file with bounding boxes and ground-truth values for all 107 fields across both pages, with dollar and cent columns annotated as separate fields. COCO, YOLO, FUNSD, and BIO/NER exports come from the same job.
- How are the dollar and cent columns annotated?
- As two separate fields under one printed label, each with its own bounding box, type, and label relation. The dollar half is currency-typed and the cent half is numeric-typed, which is deliberate: it exposes any pipeline that assumes a column's type from its header rather than its content.
- Why are so many income lines zero?
- Because a settlement fund typically holds cash in an interest-bearing account and nothing more. Median taxable interest in our corpus is $16,181 while median dividends, capital gains, and partnership income are all zero. That concentration is realistic and it is what a fund-return extractor must expect.
- How does this compare to the Form 1120-FSC corpus?
- Same split-column problem at very different scale: the 1120-SF is a two-page form with a compact numeric grid, while the 1120-FSC runs six pages and several hundred more fields. Training on both gives a split-column model size invariance instead of one memorized layout.
- Can I use this data commercially?
- Yes. All synthetic data is generated from statistical models, contains no real settlement, claimant, or fund data, and is licensed for commercial use including ML model training and benchmarking.