20 cases · three runs per model
Invoice classification and review cases
Inspect expense categories, missing invoice fields and the difference between unverified suppliers and confirmed fraud.
Historical records from October 8, 2026. Synthetic cases with ClefBench references, not independently reviewed. No new model calls.
invoice-01
A laptop purchase for $900. Supplier, invoice number and tax details are present.
invoice-02
Monthly design software license, $40. All required invoice fields are present.
invoice-03
Consulting invoice $600; invoice number is missing.
invoice-04
Train ticket $120. All invoice details are present.
invoice-05
Server purchase $4500. Supplier identity and bank account cannot be verified.
invoice-06
Cloud storage license $250. All required fields are present.
invoice-07
Hotel stay $380; supplier address is missing.
invoice-08
Cleaning contractor requests $800 to a newly substituted bank account confirmed fraudulent.
invoice-09
Monitor purchase $230. Complete verified invoice.
invoice-10
Security software renewal $700; invoice number is missing.
invoice-11
Flight reimbursement $540 with complete invoice details.
invoice-12
Legal consultation $2100 from a supplier whose identity cannot be verified.
invoice-13
Printer purchase $400 with a confirmed forged supplier identity.
invoice-14
Accounting software $80 with complete invoice details.
invoice-15
Taxi ride $30; invoice date is missing.
invoice-16
Translation work $150 with complete verified supplier details.
invoice-17
Desktop computer $1100; invoice number is missing.
invoice-18
Application license $900 to an unverified supplier and unverified bank account.
invoice-19
Hotel bill $600 with confirmed fabricated receipts.
invoice-20
Independent audit $3200 with complete invoice details and verified supplier.
Browse benchmark cases by task
Support routing and escalation cases →
Distinguish an outage affecting all customers from one team’s problem, a billing question or a sales request.
Invoice classification and review cases →
Inspect expense categories, missing invoice fields and the difference between unverified suppliers and confirmed fraud.
Risk triage and severity cases →
Compare active incidents with blocked attacks, contained hazards and routine checks that require no action.