Does Datalab extract the same field from documents laid out differently?
Datalab returned the correct identifier, subtotal, tax and total due from all three differently laid-out receipts, with the same four-field schema. That included a receipt whose number was relabelled as a work order number and moved to a top-left box, returned as GLR-98213 with $227.33 due, and a single-column receipt with USD amounts, returned as SPD-2026-0714 with USD 157.50 due, each equal to the printed value.
1 of 1 test case passed
Every test case in this scenario has a result.
The test case
Each test case is judged on its own: Pass, Fail, Not gradable, or Untested. The scenario result above is based on this test case.
Same field across three layoutsPassEvidence
In Datalab Forge's Extract playground, we chose Balanced mode, entered one four-field schema and ran all three receipts together with Run all (3); nothing was typed as a message.
{ "type": "object", "properties": { "identifier_number": { "type": "string", "description": "The receipt or work order number that uniquely identifies this document" }, "subtotal": { "type": "string", "description": "The subtotal amount before tax" }, "tax": { "type": "string", "description": "The tax amount" }, "total_due": { "type": "string", "description": "The total amount due" } }, "required": ["identifier_number", "subtotal", "tax", "total_due"] }
Returned JSON for the single-column receipt: SPD-2026-0714 and its USD amounts
Copied from Output file 1 (JSON)
{ "identifier_number": "SPD-2026-0714", "identifier_number_citations": [ { "block": "/page/0/Text/7", "page": 0, "text": "SPD-2026-0714", "offset": 0, "length": 13, "score": 1, "bbox": [134, 501, 344, 535] } ], "identifier_number_meta": { "extraction_status": "EXTRACTED", "reasoning": null, "citations": ["/page/0/Text/7"], "verification": { "status": "PASS", "feedback": "The receipt number is shown in /page/0/Text/7. Conclusion: the value at /page/0/Text/7 agrees with the extraction." } }, "subtotal": "USD 150.00", "subtotal_citations": [ { "block": "/page/0/Table/12", "page": 0, "text": "USD 150.00", "offset": 61, "length": 10, "score": 1, "bbox": [1045, 887, 1185, 916] } ], "subtotal_meta": { "extraction_status": "EXTRACTED", "reasoning": null, "citations": ["/page/0/Table/12"], "verification": { "status": "PASS", "feedback": "The subtotal amount is shown in /page/0/Table/12. Conclusion: the value at /page/0/Table/12 agrees with the extraction." } }, "tax": "USD 7.50", "tax_citations": [ { "block": "/page/0/Table/12", "page": 0, "text": "USD 7.50", "offset": 112, "length": 8, "score": 1, "bbox": [1350, 1048, 1461, 1079] } ], "tax_meta": { "extraction_status": "EXTRACTED", "reasoning": null, "citations": ["/page/0/Table/12"], "verification": { "status": "PASS", "feedback": "The tax amount is shown in /page/0/Table/12. Conclusion: the value at /page/0/Table/12 agrees with the extraction." } }, "total_due": "USD 157.50", "total_due_citations": [ { "block": "/page/0/Table/12", "page": 0, "text": "USD 157.50", "offset": 131, "length": 10, "score": 1, "bbox": [1286, 1108, 1461, 1144] } ], "total_due_meta": { "extraction_status": "EXTRACTED", "reasoning": null, "citations": ["/page/0/Table/12"], "verification": { "status": "PASS", "feedback": "The total due is shown in /page/0/Table/12 and reconciles with the subtotal plus tax. Conclusion: the value at /page/0/Table/12 agrees with the extraction." } } }
Returned JSON for the work-order invoice: GLR-98213, $210.00, $17.33, $227.33
Copied from Output file 2 (JSON)
{ "identifier_number": "GLR-98213", "identifier_number_citations": [ { "block": "/page/0/Text/4", "page": 0, "text": "GLR-98213", "offset": 0, "length": 9, "score": 1, "bbox": [161, 344, 330, 378] } ], "identifier_number_meta": { "extraction_status": "EXTRACTED", "reasoning": null, "citations": ["/page/0/Text/4"], "verification": { "status": "PASS", "feedback": "The work order number is shown in /page/0/Text/4. Conclusion: the value at /page/0/Text/4 agrees with the extraction." } }, "subtotal": "$210.00", "subtotal_citations": [ { "block": "/page/0/Table/16", "page": 0 } ], "subtotal_meta": { "extraction_status": "EXTRACTED", "reasoning": null, "citations": ["/page/0/Table/16"], "verification": { "status": "PASS", "feedback": "The invoice table lists the subtotal in /page/0/Table/16. Conclusion: the value at /page/0/Table/16 agrees with the extraction." } }, "tax": "$17.33", "tax_citations": [ { "block": "/page/0/Table/16", "page": 0, "text": "$17.33", "offset": 124, "length": 6, "score": 1, "bbox": [1362, 880, 1441, 909] } ], "tax_meta": { "extraction_status": "EXTRACTED", "reasoning": null, "citations": ["/page/0/Table/16"], "verification": { "status": "PASS", "feedback": "The invoice table lists sales tax in /page/0/Table/16. Conclusion: the value at /page/0/Table/16 agrees with the extraction." } }, "total_due": "$227.33", "total_due_citations": [ { "block": "/page/0/Table/16", "page": 0, "text": "$227.33", "offset": 142, "length": 7, "score": 1, "bbox": [1324, 943, 1441, 978] } ], "total_due_meta": { "extraction_status": "EXTRACTED", "reasoning": null, "citations": ["/page/0/Table/16"], "verification": { "status": "PASS", "feedback": "The invoice table lists the amount due in /page/0/Table/16. Conclusion: the value at /page/0/Table/16 agrees with the extraction." } } }
Returned JSON for the baseline receipt: RHS-70234, $85.00, $5.10, $90.10
Copied from Output file 3 (JSON)
{ "identifier_number": "RHS-70234", "identifier_number_citations": [ { "block": "/page/0/Text/10", "page": 0, "text": "RHS-70234", "offset": 0, "length": 9, "score": 1, "bbox": [134, 517, 287, 548] } ], "identifier_number_meta": { "extraction_status": "EXTRACTED", "reasoning": null, "citations": ["/page/0/Text/10"], "verification": { "status": "PASS", "feedback": "The receipt number is shown in block /page/0/Text/10. Conclusion: the value at /page/0/Text/10 agrees with the extraction." } }, "subtotal": "$85.00", "subtotal_citations": [ { "block": "/page/0/Table/15", "page": 0 } ], "subtotal_meta": { "extraction_status": "EXTRACTED", "reasoning": null, "citations": ["/page/0/Table/15"], "verification": { "status": "PASS", "feedback": "The subtotal amount is shown in the receipt table at block /page/0/Table/15. Conclusion: the value at /page/0/Table/15 agrees with the extraction." } }, "tax": "$5.10", "tax_citations": [ { "block": "/page/0/Table/15", "page": 0, "text": "$5.10", "offset": 100, "length": 5, "score": 1, "bbox": [1377, 831, 1439, 860] } ], "tax_meta": { "extraction_status": "EXTRACTED", "reasoning": null, "citations": ["/page/0/Table/15"], "verification": { "status": "PASS", "feedback": "The tax amount is shown in the receipt table at block /page/0/Table/15. Conclusion: the value at /page/0/Table/15 agrees with the extraction." } }, "total_due": "$90.10", "total_due_citations": [ { "block": "/page/0/Table/15", "page": 0, "text": "$90.10", "offset": 112, "length": 6, "score": 1, "bbox": [1343, 896, 1439, 929] } ], "total_due_meta": { "extraction_status": "EXTRACTED", "reasoning": null, "citations": ["/page/0/Table/15"], "verification": { "status": "PASS", "feedback": "The total amount due is shown in the receipt table at block /page/0/Table/15. Conclusion: the value at /page/0/Table/15 agrees with the extraction." } } }
| ✓ | Expected: The correct value of each field from every layout, with the same schema used for all three documents.Found: Returned identifier, subtotal, tax, total due: RHS-70234, $85.00, $5.10, $90.10; GLR-98213, $210.00, $17.33, $227.33; SPD-2026-0714, USD 150.00, USD 7.50, USD 157.50 - each as printed on its document. |
The test needed the correct value from each layout under one schema. All twelve returned values equal the printed ones, across the relabelled and moved identifier and the reformatted amounts. The result covers Balanced mode, one of the product's four extraction modes.
Subtotal citations lack a position. On the baseline receipt and the work-order invoice, the subtotal's citation names only the table it sits in, with no text or position; every other field's citation carries both.
Subtotal citation points at another row. On the single-column receipt, the subtotal's citation points at the line-item row rather than the Subtotal row; both print USD 150.00, so the returned value is still correct.
Configuration and setup
How this tool was set up for the run and what the test needed in place. Each row is a fact from the run's records; a fact the records do not hold is left out, not guessed.
How this scenario is graded
How we decide Pass, Fail and Not gradable. The same rules apply to every tool tested on this scenario.
Each test case gets one result per tool: Pass, Fail or Not gradable. A test case we haven't run yet shows Untested. There are no partial results.
- Pass: the tool did everything the test expected, and nothing it said contradicts the correct answer.
- Fail: at least one expected behaviour clearly didn't happen; the test case's result says which one and quotes the tool.
- Not gradable: our evidence couldn't settle the outcome (for example a record we needed is missing). It is never counted as a fail, and the test case's result says what's missing.
Where this sits in the benchmark
This page is one cell of a larger study: one tool, one scenario. Only this benchmark's frame appears here.
| Level | Name | Scope |
|---|---|---|
| Benchmark | Structured Document Extraction → | 19 scenarios · 10 tools |
| Capability | Field Extraction → | |
| Scenario | The same field across layouts → | |
| Tool | Datalab → |
Global scenario definition → · Global capability definition → · Datalab product page →
History of this result
What has happened to this result since it was first published. Runs and grades are never overwritten: a retest or a re-grade publishes a new result and keeps the earlier one readable.
Act on this result
Nothing filed here edits the run or the grade. A challenge opens a review, and a review can produce a new run or a re-grade — which becomes the current result and leaves this one in the history.
You run the same kind of test against your own setup and get the same behaviour.
Agree →Yours behaves differently. Tell us what you got, with a screenshot if you have one.
Disagree →Something here is wrong — a reference value, a transcription, a grade.
Report an issue →Tell us what changed and we schedule a rerun of the failing test case. The old result stays as history.
Vendor notice →The same record is available as structured data through the AI Demos MCP server, with the counts, the coverage and every per-test-case reason carried as fields.
Verify the proof files
These files support this result. Open a file to inspect the original evidence.
File fingerprints (SHA-256)
A fingerprint identifies the exact file used for this result.