Synthetic Form 1120 - U.S. Corporation Income Tax Return Data
Synthetic training data — no real PII, fully coherent identities
Generate synthetic Form 1120 corporate income tax returns across six pages — income, deductions, tax computation, balance sheet, and book-to-tax reconciliation. The most computationally dense form in our catalog, with more arithmetic and conditional bindings than any other.
433
Fields per document
6
Pages
Tax
Category
What this document is
Form 1120 is the U.S. corporation income tax return, filed by every C corporation. It runs six pages: gross receipts through taxable income on the face page, a dividends-and-special-deductions schedule, the tax computation and payments schedule, a compliance questionnaire, the Schedule L balance sheet, and Schedules M-1 and M-2 reconciling book income to taxable income and tracking retained earnings.
Why generate synthetically
Corporate returns are the primary evidence document in commercial lending, M&A diligence, and corporate tax automation, and they are the hardest tax document class to obtain for training — a real 1120 is a company's complete financial position. Synthetic 1120s give you the full six-page structure with hundreds of reconciling currency cells and exact ground truth, with no company's finances involved.
What makes synthetic data useful
The 1120 is the most heavily computed form we ship: gross receipts net of returns produce a balance, cost of goods yields gross profit, the income schedule sums, the deduction stack subtracts, and the tax computation resolves — all through arithmetic bindings evaluated at fill time rather than sampled independently. Balance sheet and reconciliation schedules are generated in proportion to the operating figures, so an extractor trained here learns that corporate returns are systems of equations, not bags of numbers.
Training challenges
Density is the first problem: 302 currency cells over six pages, most of them the same magnitude and format, with meaning carried entirely by an adjacent line label. The second is conditional structure — this form has more conditional bindings than any other in our catalog, so entire regions render or stay blank depending on the corporation's profile, and roughly 43% of fields are legitimately empty on any given return. The third is the reconciliation schedules: M-1 and M-2 relate values that appear on earlier pages, so extraction is only correct if cross-page references resolve. Finally, the compliance questionnaire contributes a large checkbox population whose state changes how downstream schedules should be interpreted.
Generate synthetic Form 1120 - U.S. Corporation Income Tax Return data
Start with 500 free credits. No credit card required.
Generate NowWho uses this data
Corporate tax software and provision tools, commercial-lending and credit-underwriting platforms that size facilities from filed returns, M&A and private-equity diligence document AI, audit and assurance automation at accounting firms, and financial-spreading vendors who turn returns into structured statements. Financial spreading in particular is a direct fit: it is exactly the task of extracting a reconciling set of figures from a return.
Document complexity profile
433 fields across 6 pages: 302 currency amounts, 80 checkbox targets, 40 text, 10 numeric, and 1 date. 565 annotation relations. This is the most computationally dense form in our catalog — 92 arithmetic bindings and 79 conditional bindings, more of each than any other form we publish, driving 244 function calls at a maximum expression depth of 3. Gross profit, total income, total deductions, taxable income, and the tax computation are all derived rather than sampled.
Key stats from our synthetic corpus
Quantitative characteristics of the Form 1120 - U.S. Corporation Income Tax Return documents our generator produces.
| Metric | Value | Detail |
|---|---|---|
| Population coverage | 64% | 64% of identities in our 1,000-identity corpus are attached to an employer entity and therefore 1120-eligible. The remaining 36% have no business relationship at all. |
| Arithmetic bindings | 92 | 92 arithmetic bindings, more than any other form in our catalog and over an order of magnitude beyond the Form 1040. Every one of them is a cross-field constraint an extraction model can be validated against. |
| Conditional bindings | 79 | 79 conditional bindings gate which regions of the return render values. This is the highest conditional density we publish, and it is why nearly half the form is legitimately blank on any single filing. |
| Median total income | $51,756 | Median total income is $51,756, with a p25–p75 range of $30,872 to $84,444 and a maximum of $359,682. Gross profit, cost of goods, and the income components are all generated in proportion, so ratio-based validators have real signal. |
| Currency cells per document | 302 | 302 currency cells spread across six pages, most of the same magnitude and format. Label-to-value association, not character recognition, is the dominant error mode on this form. |
How this document co-occurs with others
Rates at which identities in our corpus that produce a Form 1120 - U.S. Corporation Income Tax Return also produce other documents.
| Correlation | Rate | Detail |
|---|---|---|
| Files quarterly payroll returns | 100% | Every synthetic corporation files Form 941 quarterly. Compensation lines on the 1120 reconcile against aggregated 941 wage figures — the payroll cross-check credit analysts run against a corporate return. |
| Files an annual FUTA return | 100% | Every corporation also files Form 940, extending payroll reconciliation to the annual federal unemployment base. |
| Can file a 7004 extension | 100% | The same entity can produce a Form 7004 business extension, whose form code identifies the return being extended. Pairing the two trains entity-to-extension matching. |
| Issues commercial invoices | 100% | Every corporation issues invoices, so gross receipts on the return can be traced to an invoice corpus — the revenue-verification path used in commercial credit decisioning. |
| Issues W-2s to employees | 100% | Every corporation issues W-2s under the same identity graph, letting you build entity-to-employee document sets from a single generation job. |
| Six-figure principal income | 17% | 17% of synthetic 1120 principals report household income at or above $100K. These are the larger corporations in the corpus and they produce the widest currency magnitudes on the balance sheet. |
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 figures measured over the entity-eligible subset. No real corporate financial 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_f1120_standard_2024`.
Frequently asked questions
- What data format do synthetic Form 1120 documents include?
- Each generated identity produces a rendered PDF plus a structured JSON annotation file with bounding boxes and ground-truth values for all 433 fields across six pages, including the balance sheet and the M-1/M-2 reconciliations. COCO, YOLO, FUNSD, and BIO/NER exports come from the same job.
- Does the arithmetic on a synthetic 1120 actually reconcile?
- Yes. The form carries 92 arithmetic bindings and 244 function calls, the most of any form in our catalog, so gross profit, total income, total deductions, taxable income, and the tax computation all resolve from the same underlying financials. That makes the corpus usable for training cross-field validation, not just field extraction.
- Why are so many fields blank?
- Because most corporations do not have every income and deduction category. About 246 of the form's fields carry a value on a typical filing and the rest render blank, exactly as a real return would. Those blanks are annotated regions with no value, which is what teaches a model that an empty cell is a valid observation.
- How does labeling work across the six pages?
- Every field is annotated with page index, bounding box, type, and its relation to the printed line label. Multi-page linking models get direct supervision on which schedule a value came from, which is what the M-1 and M-2 reconciliations require.
- Can I use this data commercially?
- Yes. All synthetic data is generated from statistical models, contains no real corporate financial data, and is licensed for commercial use including ML model training and benchmarking.