Synthetic Form 1065 - U.S. Return of Partnership Income Data
Synthetic training data — no real PII, fully coherent identities
Generate synthetic Form 1065 partnership returns across six pages of income, deductions, Schedule B yes/no questions, balance sheet, and book-to-tax reconciliation. Every generated partnership carries a realistic partner count, asset base, and receipts figure.
373
Fields per document
6
Pages
Tax
Category
What this document is
Form 1065 is the U.S. return of partnership income, filed by partnerships and multi-member LLCs. It is a six-page document: an income-and-deductions face page, a long Schedule B of yes/no compliance questions, Schedule K summarizing partners' distributive shares, Schedule L balance sheet, and Schedules M-1 and M-2 reconciling book income to tax income and tracking partner capital accounts. Each partner then receives a Schedule K-1 derived from it.
Why generate synthetically
Partnership returns are the document at the center of private-fund administration, small-business lending, and K-1 processing, and there is essentially no public corpus for them. They are also structurally punishing: hundreds of currency cells, a checkbox-dominated compliance section, and a balance sheet that has to balance. Synthetic 1065s give you the whole six-page structure with ground truth, and with no real partnership's financials exposed.
What makes synthetic data useful
Each return is generated for an identity whose employer is a partnership, so the filing entity is coherent rather than invented per-field. Gross receipts flow through returns and allowances to a balance, cost of goods produces gross profit, and the deduction stack — salaries, guaranteed payments to partners, rent, repairs — is scaled to the partnership's revenue rather than sampled independently. Partner count drives the number of K-1s reported, so the entity's size is consistent across the schedules that reference it.
Training challenges
Schedule B is a wall of yes/no checkbox pairs, and checkboxes make up more than a quarter of the form's fields — an unusual balance for a tax return, and one that punishes models tuned for value extraction over state classification. Roughly half the fields on a 1065 are legitimately blank on any given filing, because most partnerships do not have farm income, oil and gas depletion, or foreign transactions; a model has to treat those empty regions as expected rather than as recall failures. The six pages are visually heterogeneous — a face page, a questionnaire, a summary schedule, a balance sheet, and two reconciliations — so a single page classifier trained on 'tax return' as one class will not separate them. Line labels repeat across schedules with different meanings, so context beyond the label string is required to disambiguate.
Generate synthetic Form 1065 - U.S. Return of Partnership Income data
Start with 500 free credits. No credit card required.
Generate NowWho uses this data
Partnership and K-1 tax software, private-fund and family-office administrators, business-lending underwriters who read partnership returns to size credit, commercial real estate and syndication platforms, and accounting-firm IDP pipelines. K-1 processing season is the single largest concentrated document-extraction workload in the alternative-investment industry, and it starts with the 1065 the K-1s are cut from.
Document complexity profile
373 fields across 6 pages: 185 currency amounts, 98 checkbox targets, 72 text, and 18 numeric counters. 374 annotation relations. The binding graph is deliberately shallow — 1 arithmetic binding and 13 function calls, at a maximum expression depth of 1 — because a partnership return's difficulty is volumetric rather than computational: the challenge is 185 same-typed currency cells spread over six visually distinct schedules, not deep expression trees.
Key stats from our synthetic corpus
Quantitative characteristics of the Form 1065 - U.S. Return of Partnership Income documents our generator produces.
| Metric | Value | Detail |
|---|---|---|
| Population coverage | 9% | 9% of identities in our 1,000-identity corpus work for a partnership and are therefore 1065-eligible. Requesting the form directly constrains the population to partnership entities, so a dedicated job yields a full corpus. |
| Median Schedule K-1 count | 4 | The median synthetic partnership issues 4 K-1s, p25–p75 of 2 to 6, maximum 15. Partner count is the variable that determines downstream K-1 processing volume, so it drives how a fund administrator sizes a document pipeline. |
| Median total assets | $601,250 | Median Schedule L total assets is $601,250, with a p25–p75 range of $367,500 to $905,000. The balance sheet is generated in proportion to the partnership's operations rather than sampled independently. |
| Median gross receipts | $240,500 | Median gross receipts are $240,500, against median salaries and wages of $67,340 and guaranteed payments to partners of $16,835. Those ratios hold across the corpus, so expense-to-revenue validators have a real signal to learn. |
| Checkbox share of fields | 26% | 98 of the form's fields are checkboxes, over a quarter of the total — almost all of them Schedule B yes/no compliance answers. Extractors that treat checkbox state as a secondary concern lose most of the 1065's compliance signal. |
How this document co-occurs with others
Rates at which identities in our corpus that produce a Form 1065 - U.S. Return of Partnership Income also produce other documents.
| Correlation | Rate | Detail |
|---|---|---|
| Amended counterpart (1065-X) | 100% | Every 1065 partnership can also produce a Form 1065-X administrative adjustment request. Generating the pair gives original-and-amended partnership filings, the setup restatement-detection models need. |
| Files quarterly payroll returns | 100% | Every synthetic partnership is an employer and files Form 941 quarterly. Salaries and wages on the 1065 reconcile against aggregated 941 wage lines — a cross-form check underwriters use to sanity-test reported payroll. |
| Files an annual FUTA return | 100% | Every partnership also files Form 940, extending the payroll reconciliation to the annual federal unemployment base. |
| Issues commercial invoices | 100% | Every synthetic partnership issues invoices, so gross receipts on the 1065 can be traced to a corresponding invoice corpus — the revenue-verification path used in small-business credit decisioning. |
| Partners file personal 1040s | 100% | Every partnership identity also produces a personal Form 1040, letting you build the entity-to-owner document trail that flows from the partnership return through the K-1 to an individual filing. |
| Married principals | 51% | 51% of synthetic 1065 principals are married and 31% are in dual-income households — the family-partnership pattern that entity-resolution logic has to keep distinct from unrelated co-partners. |
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 1065 figures measured over the partnership-eligible subset. No real partnership or partner 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_f1065_standard_2023`.
Frequently asked questions
- What data format do synthetic Form 1065 documents include?
- Each generated identity produces a rendered PDF plus a structured JSON annotation file with bounding boxes and ground-truth values for all 373 fields across six pages, covering the income face page, Schedule B, Schedule K, the balance sheet, and the M-1/M-2 reconciliations. COCO, YOLO, FUNSD, and BIO/NER exports come from the same job.
- How many partners do generated partnerships have?
- The median synthetic partnership reports 4 Schedule K-1s, with an interquartile range of 2 to 6 and a maximum of 15. That distribution matches the small-partnership segment that fund administrators and business lenders actually process.
- Are all the fields populated on every return?
- No, and that is deliberate. Fewer than half the form's fields carry a value on a typical filing, because most partnerships have no farm income, no depletion, and no foreign transactions. Those blanks are annotated regions with no value, which is what lets a model learn the difference between an empty field and a missed one.
- How does labeling work across six pages?
- Every field is annotated with its page index, bounding box, type, and label relation, so multi-page linking models get supervision on which schedule a value came from — not just where it sat on a sheet.
- Can I use this data commercially?
- Yes. All synthetic data is generated from statistical models, contains no real business or partner data, and is licensed for commercial use including ML model training and benchmarking.