Can Datalab derive a value that is not printed in a document?
Datalab calculated a fuel cost that was not printed in the invoice. It returned £81.00 from 450 miles at £0.18 per mile, the exact result of multiplying the two values.
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.
Derived value from two present valuesPassEvidence
We ran the three-page Calderwood Freight invoice once through Extract in Datalab’s Forge playground, using a schema that requested the fuel cost as distance multiplied by fuel rate.
{ "type": "object", "properties": { "distance_miles": { "type": "string", "description": "The distance of the route, in miles, as stated on the invoice." }, "fuel_rate_per_mile": { "type": "string", "description": "The fuel rate per mile, labelled 'FUEL RATE' on the document, including currency. The value may be a whole number or a decimal number - if the number in the document is in decimal form, the extraction should give the value in decimal, exactly as mentioned in the document." }, "total_fuel_cost": { "type": "string", "description": "The total fuel cost - calculated as the distance in miles multiplied by the fuel rate per mile. This value does NOT appear anywhere in the document; you must compute it from distance_miles x fuel_rate_per_mile." } }, "required": ["distance_miles", "fuel_rate_per_mile", "total_fuel_cost"] }
{ "distance_miles": "450 miles", "distance_miles_citations": [ { "block": "/page/0/Text/10", "page": 0, "text": "450 miles", "offset": 0, "length": 9, "score": 1, "bbox": [1366, 504, 1458, 530] } ], "distance_miles_meta": { "extraction_status": "EXTRACTED", "reasoning": null, "citations": ["/page/0/Text/10"], "verification": { "status": "PASS", "feedback": "The route distance is stated in /page/0/Text/10. Conclusion: the value at /page/0/Text/10 agrees with the extraction." } }, "fuel_rate_per_mile": "£0.18", "fuel_rate_per_mile_citations": [ { "block": "/page/0/Text/12", "page": 0, "text": "£0.18", "offset": 0, "length": 5, "score": 1, "bbox": [135, 595, 189, 624] } ], "fuel_rate_per_mile_meta": { "extraction_status": "EXTRACTED", "reasoning": null, "citations": ["/page/0/Text/12"], "verification": { "status": "PASS", "feedback": "The fuel rate is stated in /page/0/Text/12. Conclusion: the value at /page/0/Text/12 agrees with the extraction." } }, "total_fuel_cost": "£81.00", "total_fuel_cost_citations": [], "total_fuel_cost_meta": { "extraction_status": "EXTRACTED", "reasoning": null, "citations": [], "verification": { "status": "PASS", "feedback": "The document gives 450 miles in /page/0/Text/10 and £0.18 per mile in /page/0/Text/12; multiplying them gives the extracted total. Conclusion: the value at /page/0/Text/10 and /page/0/Text/12 agrees with the extraction." } } }
Extraction output: the £81.00 fuel cost and Datalab’s explanation of the calculation
Copied from Output file (JSON)
Open the JSON file ↗| ✓ | Expected: A total fuel cost of £81.00, calculated from 450 miles at £0.18 per mile.Found: Datalab returned £81.00; multiplying the invoice’s 450 miles by £0.18 per mile gives the required £81.00. |
The invoice contained no printed total, and the schema explicitly required multiplication. Datalab returned the calculated £81.00 rather than either source value, so it supplied the derived value the task required.
Sources behind the calculation. The distance and fuel-rate fields each cite their source on page 1. The fuel-cost total has no citation, while Datalab’s check explains which two cited source passages it multiplied.
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 value must be derived → | |
| 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.