Receipt #2

2 / 2 rules pass Model: auto-accepted
CORD-v2 test receipt · read by Qwen3.5-4B (QLoRA) with best-of-4 and the accounting checker

Receipt photo

Photo of receipt #2 from the CORD-v2 test set

How the checker chose

The model wrote one normal reading and four sampled ones. The checker keeps the reading that passes the most accounting rules; the model's confidence only breaks ties.

  • Greedy50% rules
  • Sample 150% rules
  • Sample 250% rules
  • Sample 350% rules
  • Sample 4100% rulespicked

Extracted fields

ItemQtyUnit priceAmountRemove

Accounting checks

would be accepted
  • Items add up to the subtotalexpected 28,000 · found 28,000
  • Cash − change = totalexpected 28,000 · found 28,000

The rules recheck as you edit, using a TypeScript port of the Python checker that agrees with it on all 277 rule results of the test set. Amounts within 1% are treated as equal, to allow for rounding.