VynFi/vynfi-group-audit-enterprise-2000

VynFi Group Audit — ACME Enterprise 2000-Entity Archive

A 2,000-entity multinational consolidation generated by the VynFi DataSynth group-audit simulation engine. ACME Inc. is a fictitious US-domiciled holding with subsidiaries across North America, Europe, APAC, and emerging markets, demonstrating every consolidation primitive at enterprise scale: manifest-driven IC matching, IAS 21 translation + CTA, IFRS 10 NCI, IAS 28 equity-method investments — and from v5.10 the consolidated Method-A accounting-network edge list spanning all 2,000 entities and their intercompany flows.

Generated with DataSynth v5.10.0 · GitHub · Companion paper (SSRN) · Generation config.

This is the canonical enterprise-grade reference archive — small enough to download (~3 GB compressed / ~71 GB uncompressed), structured enough to navigate, and heterogeneous enough to exercise every consolidation code path plus a 5.6 M-edge accounting network for graph-ML benchmarking.

What's in the archive

enterprise_2000/
├── manifest.json                                # canonical group manifest (deterministic from config + seed)
├── shard_summary.json                           # per-shard generation summary (4 shards)
├── entities/                                    # 2 000 entity sub-trees (sizes vary by scoping profile)
│   ├── ACME_HQ/                                 # parent (US, USD, "flagship" profile)
│   │   ├── ... (full single-entity output: master_data/, document_flows/, subledger/, audit/, …)
│   │   └── graphs/
│   │       ├── je_network.csv                   # NEW v5.10 — per-entity Method-A edges + ic_pair_id + ic_partner_entity
│   │       └── je_network.parquet               # NEW v5.10 — Zstd-compressed parquet
│   ├── ACME_EUR/                                # 100% EUR sub (DE, "flagship" profile)
│   ├── ACME_UK/                                 # 100% GBP sub ("significant" profile)
│   ├── ACME_JP/                                 # 85% JPY sub ("significant" profile)
│   ├── ACME_NA_SIG0000001 … 50/                 # 50 N. American "significant" subs
│   ├── ACME_EU_SIG0000001 … 25/                 # 25 European "significant" subs
│   ├── ACME_NA0000001 … 700/                    # 700 N. American "material" subs
│   ├── ACME_EU0000001 … 350/                    # 350 European "material" subs
│   ├── ACME_AS0000001 … 200/                    # 200 APAC "material" subs
│   ├── ACME_SMALL0000001 … 471/                 # 471 emerging-market "immaterial" subs
│   └── ACME_JV0000001 … 200/                    # 200 equity-method joint ventures
├── consolidated/                                # group-level outputs
│   ├── consolidated_financial_statements.json   # BS + IS + CF + Statement of Changes in Equity
│   ├── consolidation_schedule.json              # per-account pre/elim/post + per-entity contributions
│   ├── notes_to_consolidated_fs.json            # 8-note disclosure set
│   ├── nci_rollforward.json                     # 1 502 NCI rollforwards (one per fully-consolidated sub <100% owned)
│   ├── cta_rollforward.json                     # CTA per non-presentation-currency entity
│   ├── translation_worksheet.json               # IAS 21 line-by-line worksheet across all entities
│   ├── equity_method_investments.json           # 200 JV investment carrying values
│   ├── je_network.csv                           # NEW v5.10 — 5,604,445 consolidated Method-A edges
│   └── je_network.parquet                       # NEW v5.10 — Zstd-compressed parquet (~210 MB)
└── ic_eliminations/
    └── ic_matching_coverage.json                # diagnostic histogram of matched/unmatched IC pairs

Banking / KYC / AML data is NOT included. The companion banking showcase lives at VynFi/vynfi-aml-100k; this dataset is focused on group-audit specifics.

What changed in v5.10.0

• Per-entity je_network.{csv,parquet} at entities/{code}/graphs/ — the same Method-A 13-column edge list shipped on VynFi/vynfi-journal-entries-1m v5.9.0, plus ic_pair_id + ic_partner_entity columns so the inter-entity flows can be joined into pairs (one edge per side).
• Consolidated consolidated/je_network.{csv,parquet} — every entity's edges concatenated, plus the 368-strong elimination JE set (flagged is_eliminated=true), with entity_code as a partition column. Total 5,604,445 edges (≈ 1.26 GB CSV / 211 MB parquet).
• Apache 2.0 license + graph-ml task category added so the dataset surfaces in HF graph-ML searches.
• The underlying group-audit simulation engine (manifest, shard, aggregate phases, IC matching, NCI, equity-method, IAS 21 translation) is byte-identical to v5.0 — IC matching coverage is 91.59 % (4,359 / 4,759 pairs matched) in both releases.

See the v5.10.0 release notes for the full change list.

Generated under

Engine	VynFi DataSynth datasynth-group v5.10.0
Determinism seed	0xCAFEBABEDEADBEEF
Config	configs/examples/group/enterprise_2000.yaml
Wall-clock	8 min 5 sec for the full pipeline (manifest + 4 shards + aggregate)
Peak RSS	69.4 GiB across the rayon-parallel shard runner
Hardware	Azure Standard_NC40ads_H100_v5 (40 vCPU, 314 GiB RAM) in westeurope
Output	186,369 files / 71 GiB uncompressed
Reproducibility	Bit-for-bit from the pinned config

Standards compliance

The consolidation follows IFRS-equivalent treatment:

• IAS 21 — functional-currency translation with closing/average/historical rates; CTA accumulated to OCI. Non-USD entities use their declared functional currency (EUR/GBP/JPY for the explicit subs, USD pegged for the rest). See consolidated/translation_worksheet.json for the line-by-line worksheet.
• IFRS 10 — fully-consolidated entities (1,800 of them) aggregated at 100 % with NCI separately presented for any sub <100 % owned. See consolidated/nci_rollforward.json (1,502 entries).
• IAS 28 / IFRS 11 — 200 equity-method joint ventures carried as single-line investments with share-of-profit pickup. The IAS 28.38 / ASC 323-10-35-20 carrying-amount-clamped-at-zero rule is applied. See consolidated/equity_method_investments.json.
• IAS 1 — consolidated balance sheet identity (Assets = Liabilities + Equity + NCI). Note: the v5.0 fixture deliberately injects fraud / anomaly entries with mismatched debits and credits, so the literal identity does NOT hold to the cent on this archive — the imbalance IS the ground-truth signal for fraud detection.

IC matching coverage

ACME's IC relationships expand to 4,759 planned pairs under the v5.0 manifest-driven matching strategy. 4,359 (91.59 %) match in this archive. The 400 unmatched pairs are pattern-derived relationships where shard-runner injection didn't produce both sides — usually because one side's entity hit anomaly-injection's "skip this JE" branch. Unchanged from v5.0.

See ic_eliminations/ic_matching_coverage.json for the full histogram.

Quick start (consolidated edge list — graph-ML ready)

from huggingface_hub import hf_hub_download
import pandas as pd

# Pull just the 211 MB consolidated edge parquet — no full download
edges_path = hf_hub_download(
    repo_id="VynFi/vynfi-group-audit-enterprise-2000",
    filename="enterprise_2000/consolidated/je_network.parquet",
    repo_type="dataset",
)
df = pd.read_parquet(edges_path)
print(df.shape)             # (5_604_445, 18)
print(df["entity_code"].nunique())   # 2000+ entity codes
print(df["is_eliminated"].sum())     # 368 elimination edges
print(df["ic_pair_id"].notna().sum())  # ~8K seller+buyer IC edges

Note: the consolidated edge list lives inside the tarball enterprise_2000.tar.zst at the path enterprise_2000/consolidated/je_network.parquet. See the per-entity walkthrough below for partial-download examples.

Quick start (full archive)

from huggingface_hub import snapshot_download
import json, pathlib

# Note: tarball is ~3 GB compressed; uncompressed is ~71 GB
local = pathlib.Path(snapshot_download(
    repo_id="VynFi/vynfi-group-audit-enterprise-2000",
    repo_type="dataset"))

# Extract:
import subprocess
subprocess.run(["tar", "-I", "zstd", "-xf", str(local / "enterprise_2000.tar.zst"), "-C", str(local)])
root = local / "enterprise_2000"

cfs = json.loads(
    (root / "consolidated/consolidated_financial_statements.json")
    .read_text())
print("Group:", cfs["balance_sheet"]["group_id"])
print("Total assets (USD):", cfs["balance_sheet"]["total_assets"])
print("Total L+E+NCI (USD):", cfs["balance_sheet"]["total_liabilities_plus_equity_plus_nci"])
print("NCI separately presented:", cfs["balance_sheet"]["total_nci"])

Quick start (per-entity walkthrough)

import json, tarfile
from huggingface_hub import hf_hub_download

# Download just the tarball, then extract one entity's slice in-memory
tar_path = hf_hub_download(
    repo_id="VynFi/vynfi-group-audit-enterprise-2000",
    filename="enterprise_2000.tar.zst",
    repo_type="dataset",
)
# (Use 'tar -I zstd -xf' on disk for full-archive extraction)

Schema highlights

Per-entity (entities/{code}/) carry the v5.x single-entity output shape unchanged. See the VynFi DataSynth README for the ~20 typed-snapshot subdirectories.

Per-entity graphs/je_network.{csv,parquet} (NEW v5.10) — 15 columns:

edge_id, document_id, posting_date, from_account, to_account,
from_line_id, to_line_id, amount, confidence, predecessor_edge_id,
business_process, is_fraud, is_anomaly,
ic_pair_id, ic_partner_entity

Consolidated consolidated/je_network.{csv,parquet} (NEW v5.10) — 18 columns:

edge_id, document_id, entity_code, posting_date, from_account, to_account,
from_line_id, to_line_id, amount, confidence, predecessor_edge_id,
business_process, is_fraud, is_anomaly,
ic_pair_id, ic_partner_entity, is_eliminated, eliminates_ic_pair_id

Group-level files under consolidated/ and ic_eliminations/ match the v5.0 spec §9 schema. Field-by-field documentation lives in the engine's crates/datasynth-group/src/aggregate/ modules at the v5.10.0 release tag.

What this dataset is good for

• Audit ML benchmarks — large-scale group-audit simulation with known ground truth (every fraud / anomaly / IC pair labelled).
• Graph-ML benchmarks (NEW) — 5.6 M-edge accounting network spanning 2,000 entities with explicit IC pair linkage and elimination-edge labels. See companion model VynFi/je-fraud-gnn for a single-entity baseline.
• Consolidation engine validation — drop-in reference for testing custom IFRS / ASC 810 consolidation logic.
• Education — concrete example of a 2 000-entity multinational consolidation for accounting / audit pedagogy.
• Performance benchmarking — the engine's published 69 GiB peak / 8-minute wall-clock profile on Standard_NC40ads_H100_v5 is reproducible against this exact fixture.

What this dataset is NOT

• Real-world data. Every value is synthetic and deterministically generated. Statistical distributions are approximate models, not samples from any specific company. "ACME" is a fictitious name and has no relationship to any real entity.
• A research-grade fraud benchmark. Fraud labels are injected by construction, not discovered via investigation.
• A regulatory filing. The IFRS treatment is faithful to the published standards but the underlying numbers are fictitious; do not use for any compliance purpose.

License

Apache 2.0. Free for commercial use, modification, distribution, private use; see LICENSE for the full terms.

Citation

@misc{ivertowski2026datasynth,
  author       = {Ivertowski, Michael},
  title        = {{DataSynth}: Reference Knowledge Graphs for Enterprise
                  Audit Analytics through Synthetic Data Generation
                  with Provable Statistical Properties},
  year         = {2026},
  month        = {April},
  howpublished = {SSRN Working Paper},
  url          = {https://ssrn.com/abstract=6538639}
}

Related VynFi datasets

• VynFi/vynfi-journal-entries-1m — single-entity 1 M JE lines + COA + TB + cost / profit centres + Method-A accounting-network edge list
• VynFi/je-fraud-gnn — trained GraphSAGE fraud + GAE anomaly model (companion to the journal-entries dataset)
• VynFi/vynfi-audit-p2p — P2P document-flow corpus
• VynFi/vynfi-supply-chain-ocel — Native OCEL 2.0 event log
• VynFi/vynfi-aml-100k — Banking + AML labels
• VynFi/vynfi-sar-narratives — Banking + AML labels + SAR narratives
• VynFi/vynfi-ocel-manufacturing — Lightweight reconstructed-events prototyping companion

Related VynFi Spaces

• 🔗 VynFi/accounting-network-explorer — Interactive ISO 21378 account-class graph
• 🛡️ VynFi/fraud-gnn-demo — Gradio fraud-GNN inference demo
• 📊 VynFi/process-mining-demo — pm4py process-mining showcase