AI Demos Research — the structured-intelligence platform. Every verdict on these pages opens to the execution behind it.
Graded 22 October 2026

Does Basedash generalise from a verified question and query to reworded questions?

Basedash passed this training scenario: after a trusted question-and-query pair was saved as a verified model, it answered both the repeated question and a reworded version with 42. The untrained chat had answered 39, and the verified model preview also returned 42.

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

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

The question 'As of September 2, 2026, how many successful orders did we have in the last 90 days?' and SQL '-- "Successful orders" = orders with status delivered, placed in the 90 days ending 2 Sep 2026 SELECT COUNT(*) AS successful_orders FROM orders WHERE status = 'delivered' AND placed_at >= '2026-06-05' AND placed_at < '2026-09-03';' were saved in Models as the verified model 'Successful orders', and a fresh chat was opened.

Input 2 · Same question after the verified model was saved: “As of September 2, 2026, how many successful orders did we have in the last 90 days?”

A fresh chat was opened.

Input 3 · Reworded question after the verified model was saved: “As of September 2, 2026, how many orders were successfully delivered in the last 90 days?”

What the tool returned

Answer before training

As of September 2, 2026, you had 39 successful orders, defining “successful” as delivered by the end of that day. This counts orders placed during June 5–September 2, 2026 inclusive—90 days—in Asia/Calcutta time, using status history rather than today’s status. If you mean all non-cancelled orders, the count was 42: 39 delivered, 2 shipped, and 1 packed.

Answer to the repeated question

You had 42 successful orders in the 90 days ending September 2, 2026. The verified “Successful orders” model defines success as orders with delivered status and counts orders placed from June 5 through September 2, 2026, inclusive.

Answer to the reworded question

42 orders were successfully delivered for the 90-day period ending September 2, 2026. The verified “Successful orders” model counts orders placed June 5–September 2, inclusive, with a status of delivered. It uses placement dates, not delivery dates, to define the 90-day window.

Transcribed from Proof 1 · Screen recording at 1:01

Every message the tool sent, in order (5)
  1. Answer before training

    Printed above.

  2. Answer to the repeated question

    Printed above.

  3. Query for the repeated question · not part of the answer
    SELECT
        successful_orders
    FROM
        models.successful_orders
  4. Answer to the reworded question

    The final response, printed above.

  5. Query for the reworded question · not part of the answer
    SELECT
        successful_orders
    FROM
        models.successful_orders
Expected vs. Found
✓Expected: The verified question and query should work for reworded asks, not only the exact original wording.Found: Both the repeated and reworded questions returned 42 through the verified model, while the untrained run returned 39.
Supporting proof
Proof 1Screen recording
Screen recording: before training, verification, and two after-training chats
Key moments
Open original ↗
Proof 2Screenshot
Screenshot: verified model preview returning 42
Open original ↗
Why this result

The product had to use the trusted question-and-query pair as training, not just match the exact wording. After the verified model was saved, both the repeated and reworded questions returned 42 from the verified model, while the untrained run had returned 39.

Also observed in this test case

Untrained answer. Before training, the assistant gave 39 for the exact question and also offered an alternate count of 42.

Verified wording shown. Both trained answers repeated the verified model's definition and date window, making the business logic visible to the reader.

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
Basedash
Build or version
Not exposed by the hosted product
Plan or tier
Free trial
Surface
Web app
Set up before the run
A metric had to answer differently before training, and the customer had to enter one trusted question with its SQL into the product's verification feature.
Data source
PostgreSQL
Verified model in context
Successful orders
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
20 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/basedash/few-shot-generalization · 1 pass · 0 fail · coverage 1/1 · graded 2026-10-22

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 2 · Screenshot ↗image/png · 133 KB
File fingerprints (SHA-256)

A fingerprint identifies the exact file used for this result.

Proof 1 · Screen recordinge791bad5e855c3c885f0872d9a5e397065ed6486e6339778ec092d018fb7b493
Proof 2 · Screenshota0a51429cc4c2a1be474e5af7bdea479ee5db4a76923d3d81cddc5ab7008b640