Can Reducto point to where a returned value comes from?
Reducto gave a precise source location for the extracted value. It returned the £6,103.00 subtotal with a citation to the amount itself on page 2.
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.
Pointer to where the value sitsPassEvidence
{
"instructions": {
"schema": {
"type": "object",
"properties": {
"subtotal": { "type": "string", "description": "The subtotal amount of the invoice, before tax" }
},
"required": ["subtotal"]
}
},
"settings": {
"citations": { "enabled": true }
}
}Copied from Output file (JSON), result › subtotal › value
£6,103.00
Copied from Output file (JSON), result › subtotal › citations 1 › bbox
{ "left": 0.8680672268907563, "top": 0.25771971496437057, "width": 0.05042016806722689, "height": 0.009501187648456057, "page": 2, "original_page": 2 }
Extraction response: the subtotal value and its citation box on page 2
Open the whole JSON file ↗| ✓ | Expected: A pointer to the returned value’s actual location in the document, at the precision Reducto offers.Found: It returned £6,103.00 and a box within the invoice’s £6,103.00 subtotal on page 2. |
- result › subtotal › citations 1
- { "type": "Text", "content": "£6,103.00", "image_url": null, "chart_data": null, "confidence": "high", "granular_confidence": { "extract_confidence": 0.995, "parse_confidence": 1 }, "extra": null, "parentBlock": { "type": "Table", "bbox": { "left": 0.074, "top": 0.118, "width": 0.851, "height": 0.20600000000000002, "page": 2, "original_page": 2 }, "content": "<table><tr><th>Sl.</th><th>Item Description</th><th>Qty</th><th>Unit Price (£)</th><th>Amount (£)</th></tr><tr><td>7</td><td>Barcode Scanner - Wireless</td><td>10</td><td>58.00</td><td>580.00</td></tr><tr><td>8</td><td>Receipt Paper Roll - … [685 characters in the file]", "image_url": null, "chart_data": null, "confidence": "high", "granular_confidence": { "extract_confidence": null, "parse_confidence": 1 }, "extra": null } }
- result › subtotal › citations 1 › parentBlock › content
- <table><tr><th>Sl.</th><th>Item Description</th><th>Qty</th><th>Unit Price (£)</th><th>Amount (£)</th></tr><tr><td>7</td><td>Barcode Scanner - Wireless</td><td>10</td><td>58.00</td><td>580.00</td></tr><tr><td>8</td><td>Receipt Paper Roll - Case of 50</td><td>15</td><td>22.00</td><td>330.00</td></tr><tr><td>9</td><td>Mannequin - Adjustable, Full Body</td><td>6</td><td>95.00</td><td>570.00</td></tr><tr><td>10</td><td>Display Signage Kit - Window Set</td><td>5</td><td>76.00</td><td>380.00</td></tr><tr><td colspan="4">Subtotal</td><td>£6,103.00</td></tr><tr><td colspan="4">VAT (8%)</td><td>£488.24</td></tr><tr><td colspan="4"><b>TOTAL</b></td><td><b>£6,591.24</b></td></tr></table>
The invoice had two pages and one occurrence of the amount. Reducto’s citation located the amount itself, clear of the VAT row, meeting the requirement for a usable source pointer.
Citation confidence. For the subtotal citation, Reducto reported high confidence, with extraction confidence of 0.995 and parsing confidence of 1.
Surrounding table. The citation also names the totals table it sits in (subtotal, VAT and total rows), but its box covers only the subtotal amount.
Visible source highlight. Studio highlighted the subtotal amount in green on the rendered invoice.
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 | Source Grounding → | |
| Scenario | Trace a value to its location → | |
| Tool | Reducto → |
Global scenario definition → · Global capability definition → · Reducto 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.