AI Demos Research — the structured-intelligence platform. Every verdict on these pages opens to the execution behind it.
Graded 12 November 2026

Does AskYourDatabase use trusted query examples for reworded questions?

AskYourDatabase used a supplied question-and-query example beyond the original wording. The repeated and reworded questions each returned 42 orders, equal to the trusted query’s result.

1 of 1 test case passed

Every test case in this scenario has a result.

Pass rate100%1 of 1 gradable
Coverage1 of 1test case with a result
1PassDid everything it was expected to do.
0FailNo test case failed.
0Not gradableEvery result here could be graded.
0UntestedEvery test case in this scenario has a result.
The pass rate is a summary. The evidence is the test case below: what we sent, what we checked, what the tool returned, and the proof.

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.

Verified question/query pair generalisesPassEvidence
What we sent

Earlier plain-text entries were removed from Documentation, and a new chat was opened.

Input 1 · Before training: “As of September 2, 2026, how many successful orders did we have in the last 90 days?”

The trusted question-and-query pair was saved in SQL Examples, and a new chat was opened.

Input 2 · After training: repeated question: “As of September 2, 2026, how many successful orders did we have in the last 90 days?”

A new chat was opened.

Input 3 · After training: reworded question: “As of September 2, 2026, how many orders were successfully delivered in the last 90 days?”

What the tool returned

Answer before training

42 successful orders were placed in the 90 days leading up to September 2, 2026 (June 4 – September 1, 2026), excluding cancelled orders.

Answer to the repeated question

You had 42 successful orders in the 90 days ending September 2, 2026 (June 5 – September 2, 2026). This counts orders with a delivered status placed within that window.

Answer to the reworded question

42 orders were successfully delivered in the 90 days ending September 2, 2026 (i.e., placed between June 5 and September 2, 2026, with status "delivered").

Transcribed from Proof 3 · Screenshot

Every message the tool sent, in order (6)
  1. SQL before training · not part of the answer
    SELECT COUNT(*) AS successful_orders FROM orders WHERE status <> 'cancelled' AND placed_at >= '2026-06-04' AND placed_at < '2026-09-02';
  2. Answer before training

    Printed above.

  3. SQL for the repeated question · not part of the answer
    SELECT COUNT(*) AS successful_orders FROM orders WHERE status = 'delivered' AND placed_at >= '2026-06-05' AND placed_at < '2026-09-03';
  4. Answer to the repeated question

    Printed above.

  5. Displayed SQL for the reworded question · not part of the answer
    SELECT COUNT(*) AS successful_orders FROM orders WHERE status = 'delivered' AND placed_at >= '2026-06-05' AND placed_at < '2026-09-03';
  6. Answer to the reworded question

    The final response, printed above.

Expected vs. Found
✓Expected: The repeated question and an unseen rephrasing must both return the trusted query’s result: 42 orders.Found: The repeated and reworded questions both counted delivered orders and returned 42; the trusted reference query returned 42.
Supporting proof
Proof 1Screenshot
Chat screenshot: the untrained question returns 42 using a different query
Open original ↗
Proof 2Screenshot
Chat screenshot: the repeated question uses the trusted query and returns 42 orders
Open original ↗
Proof 3Screenshot
Chat screenshot: the reworded question uses the trusted query and returns 42 orders
Open original ↗
Proof 4Screenshot
Training screen: the question and trusted query entered in SQL Examples
Open original ↗
Why this result

Both trained questions used the trusted query, including unseen wording: more than exact-question recall. All three answers were 42, so the change is in the executed query’s definition and window, not the count. The reworded chat displays one SQL block, without a separate access-controlled version.

Also observed in this test case

As-of date included. The untrained successful-orders query left out September 2, 2026 and included June 4. Both trained queries include September 2 and start on June 5.

Generated and access-controlled queries. In the before-training and repeated-question chats, the generated query and the access-controlled query are displayed separately and have the same meaning within each chat.

Answers explain their scope. The repeated-question and reworded-question answers both explain what counts as a successful order and which dates are included.

Tested by Ayush Ghosh · evidence dated 22 October 2026

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.

Software that produced the output
AskYourDatabase
Build or version
Not exposed by the hosted product
Surface
Web app
Set up before the run
A trusted question-and-query pair for successful orders had to be supplied through SQL Examples, for a metric previously answered with a different query: non-cancelled rather than delivered orders, despite the same count.
Training question
As of September 2, 2026, how many successful orders did we have in the last 90 days?
Trusted query
Count delivered orders placed from June 5 through September 2, 2026.
Chat scope
new chat
Documentation
Earlier entries removed before the questions.
Tested
22 October 2026 · Ayush Ghosh

How this scenario is graded

How we decide Pass, Fail and Not gradable. The same rules apply to every tool tested on this scenario.

How results are decided

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.

The rules
  • 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.

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.

from the publication record
16 November 2026First publishedAI Database Agents v1

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.

This matches what I see

You run the same kind of test against your own setup and get the same behaviour.

Agree →
This does not match

Yours behaves differently. Tell us what you got, with a screenshot if you have one.

Disagree →
Point out an issue

Something here is wrong — a reference value, a transcription, a grade.

Report an issue →
Request a retest

On a newer build, a larger dataset, or your own setup.

Request a retest →
We have fixed this

Tell us what changed and we schedule a rerun of the failing test case. The old result stays as history.

Vendor notice →
Filed against this evidenceNothing yet. Challenges, counter-evidence and fix notices appear here with their outcome, and stay on the page after they are resolved.
Cite this result
aidemos.com/benchmarks/ai-database-agents/results/askyourdatabase/few-shot-generalization · 1 pass · 0 fail · coverage 1/1 · graded 2026-11-12

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.

Proof 1 · Screenshot ↗image/png · 66 KB
Proof 2 · Screenshot ↗image/png · 45 KB
Proof 3 · Screenshot ↗image/png · 42 KB
Proof 4 · Screenshot ↗image/png · 88 KB
File fingerprints (SHA-256)

A fingerprint identifies the exact file used for this result.

Proof 1 · Screenshot8c4a86c6408ebd4b9f03c6dff7528e5b98d20d469c97194908c264b8c54e9620
Proof 2 · Screenshote0225641839706b254a8d17ad087607d9784ce8ff406f35ea2df6d18fdfbbd30
Proof 3 · Screenshot7a31addcd7550deee9183028908a51d572d86a5ea2fd69b8a27db61613ef331e
Proof 4 · Screenshot905189dfbc775580ebc81dc8107fad24e11c1818e3751ce7ec5865e83346558f