banking-aml-structuring-bench / dataset-metadata.json
soham27's picture
Release Banking AML Smurfing & FinCEN SAR Audit Bench v1.0
cd7e701 verified
Raw History Blame Contribute Delete
5.04 kB
{
"title": "Banking AML Smurfing & FinCEN SAR Audit Bench",
"subtitle": "Benchmark for Multi-Hop BSA Structuring, Layering & SAR Narrative Mining",
"description": "---\nlicense: apache-2.0\ntask_categories:\n- question-answering\n- table-question-answering\n- zero-shot-classification\nlanguage:\n- en\ntags:\n- enterprise\n- rag\n- evaluation\n- benchmark\n- banking_and_compliance\nsize_categories:\n- 1K<n<10K\n---\n\n# Banking AML Smurfing & FinCEN SAR Audit Bench\n\n### High-Consequence Enterprise Evaluation Benchmark\n\n**Official Maintainer:** Botmartz AI Research ([botmartz.com](https://www.botmartz.com)) \n**Research Publication:** [botmartz.com/insights](https://www.botmartz.com/insights) \n**Primary Metric:** SAR Multi-Hop Structuring Detection Precision \n\n### Problem Overview\nIn high-consequence enterprise applications within **FinTech Compliance**, standard out-of-the-box LLMs and naive dense vector RAG fail on multi-hop dependencies, tabular constraints, and strict regulatory terminology.\n\nThis benchmark provides rigorous ground truth for evaluating:\n- Retrieval precision across dense, sparse (BM25), and graph architectures.\n- Structural layout and AST preservation.\n- Deterministic extraction accuracy on mission-critical documentation.\n\n### Data Splits & Files\n- `train.jsonl`: 350 calibration records with rich metadata, Markdown ASTs, and ground-truth QA pairs.\n- `test.jsonl`: 120 blinded evaluation records.\n- `records.csv`: Full tabular catalog with clean schemas.\n- `baseline_model_eval.csv`: Empirical benchmarks comparing Botmartz Deterministic Pipeline vs. Naive Vector RAG.\n\n### Citation\n```bibtex\n@dataset{botmartz_banking_aml_structuring_bench,\n author = {Botmartz AI Research Team},\n title = {Banking AML Smurfing & FinCEN SAR Audit Bench},\n year = {2026},\n publisher = {Botmartz AI Solutions},\n howpublished = {\\url{https://www.botmartz.com}}\n}\n```\n",
"id": "sohamnsharma/banking-aml-structuring-bench",
"licenses": [
{
"name": "apache-2.0"
}
],
"keywords": [
"finance",
"business",
"nlp",
"text"
],
"expectedUpdateFrequency": "quarterly",
"userSpecifiedSources": "Botmartz AI Research, SEC EDGAR, FAA FSIS, ISO Filings, FinTech Compliance Standards",
"resources": [
{
"path": "records.csv",
"description": "Curated tabular records and ground-truth annotations for Banking AML Smurfing & FinCEN SAR Audit Bench.",
"schema": {
"fields": [
{
"name": "doc_id",
"description": "Attribute field representing doc id in Banking AML Smurfing & FinCEN SAR Audit Bench.",
"type": "string"
},
{
"name": "domain",
"description": "Attribute field representing domain in Banking AML Smurfing & FinCEN SAR Audit Bench.",
"type": "string"
},
{
"name": "account_holder",
"description": "Attribute field representing account holder in Banking AML Smurfing & FinCEN SAR Audit Bench.",
"type": "string"
},
{
"name": "single_tx_amount",
"description": "Attribute field representing single tx amount in Banking AML Smurfing & FinCEN SAR Audit Bench.",
"type": "number"
},
{
"name": "daily_tx_count",
"description": "Attribute field representing daily tx count in Banking AML Smurfing & FinCEN SAR Audit Bench.",
"type": "number"
},
{
"name": "is_smurfing_structuring",
"description": "Attribute field representing is smurfing structuring in Banking AML Smurfing & FinCEN SAR Audit Bench.",
"type": "number"
},
{
"name": "markdown_ast",
"description": "Attribute field representing markdown ast in Banking AML Smurfing & FinCEN SAR Audit Bench.",
"type": "string"
}
]
}
},
{
"path": "baseline_model_eval.csv",
"description": "Comparative evaluation of modern retrieval architectures.",
"schema": {
"fields": [
{
"name": "model_architecture",
"description": "AI retrieval or extraction architecture",
"type": "string"
},
{
"name": "domain_accuracy",
"description": "Domain-specific accuracy score",
"type": "number"
},
{
"name": "p95_latency_ms",
"description": "P95 inference latency in milliseconds",
"type": "number"
},
{
"name": "citation_precision",
"description": "Exact citation precision score",
"type": "number"
},
{
"name": "hallucination_rate_pct",
"description": "Measured hallucination percentage",
"type": "string"
}
]
}
}
]
}