P01 The Paperwork Trial Generated invoice images
Basic Invoice Folder
A small scanned-invoice folder with a quote distractor, partial payment, and an under-review stamp.
Transparency boundary
This page shows the files the model can inspect. It does not show the hidden oracle solution.
Excluded: ground_truth.json, expected_artifacts.json, manual readings, and calibration notes.
Generated Images
Scans and visual attachments visible to the model
Task And Source Files
No oracle solution filesREADME_TASK.md
# Paperwork Trial v3 Calibration Case 01 This is synthetic benchmark bookkeeping. It is not financial, tax, or legal advice. Audit the scanned document folder and write `audit_result.json` only. Case ID: `P3-GEN-01` Output keys exactly: `case_id, approved_invoice_ids, review_invoice_ids, reject_invoice_ids, ignored_document_ids, total_approved_gross_cents, warnings_by_invoice, evidence, proof_code` Rules: - Treat the files in `scans/` as the source documents. - Ignore documents that are explicitly not invoices. - Match vendors by tax ID first, then by exact vendor name. - `payment_match` requires paid bank rows for the invoice whose summed amount equals the invoice gross. - `payment_short` applies when the paid bank amount is lower than invoice gross. - `under_review_stamp` applies when the scanned invoice visibly has an under-review stamp. - Approved invoices have no warnings and paid amount equals gross. - Review invoices have warnings but are not reject-level. `payment_short` and `under_review_stamp` are review-level warnings, not reject-level warnings. - Reject only if the invoice is impossible to process, uses an inactive vendor, or lacks a required PO. - Sort all invoice-id arrays ascending. - Allowed warning codes are exactly: `payment_short`, `under_review_stamp`. - `warnings_by_invoice` must include every real invoice ID and sorted warning arrays. - Warning arrays must be flat arrays of lowercase strings, never nested arrays and never prose labels. - `ignored_document_ids` must include visible document IDs from ignored non-invoice documents, not filenames. Example: use `QT-6400`, not `quote_6400_not_invoice.png`. - `total_approved_gross_cents` is the sum of approved invoice gross totals only. - `evidence` must list the relative source file paths used in stable alphabetical order, including folder prefixes such as `scans/`. - Include the relevant CSV files and every scanned document inspected in `evidence`, including ignored non-invoice scans. - `proof_code = total_approved_gross_cents + sum(numeric parts of all real invoice IDs) + 97 * total_warning_count`. Important: `Northwind Office Supply` is the customer, not the vendor.
bank_export.csv
date,description,invoice_id,amount_cents,status 2026-04-23,BluePeak Distributors,INV-7801,18737,paid 2026-04-24,StationHub LLC,INV-7802,5000,paid 2026-04-27,Orion Field Services,INV-8422,42245,paid
model_prompt.md
You are auditing a synthetic scanned paperwork folder. Read `README_TASK.md`, inspect the files in `scans/`, and use `bank_export.csv`, `vendor_master.csv`, and `purchase_orders.csv`. Write `audit_result.json` only. This is benchmark bookkeeping, not financial, tax, or legal advice. Important: - The scanned images are the source documents. - `Northwind Office Supply` is the customer, not the vendor. - Ignore documents that clearly say they are not invoices. - Do not invent fields that are not supported by the files. - Use visible document IDs for `ignored_document_ids`, not filenames. - Use relative paths with folder prefixes in `evidence`, for example `scans/example.png`. - Use only allowed lowercase warning codes from `README_TASK.md`. - Warning arrays must be flat arrays of strings.
purchase_orders.csv
po_id,vendor_id,limit_cents,status PO-4510,V-BP4471,20000,open PO-4510-B,V-SH9982,6000,open PO-8422,V-OR1109,50000,open
vendor_master.csv
vendor_id,name,tax_id,status V-BP4471,BluePeak Distributors,BP-4471,active V-SH9982,StationHub LLC,SH-9982,active V-OR1109,Orion Field Services,OR-1109,active V-NW001,Northwind Office Supply,NW-CUSTOMER,customer