AI Demos Research — the structured-intelligence platform. Every verdict on these pages opens to the execution behind it.
Graded 26 October 2026

Does Draxlr apply a user's corrected metric definition to later questions?

Draxlr retained the user’s corrected metric definition for both a repeated question and a differently worded one. Both revenue answers were $23,194.52, the same as the reference total.

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.

Corrected metric definition survives a rephrasingPassEvidence
What we sent

The revenue definition was saved in Train AI: “Our revenue definition excludes cancelled orders and refunded orders.”

Input 1 · Repeated revenue question: “As of September 2, 2026, what was our total revenue over the last 90 days?”

A new chat was opened.

Input 2 · Rephrased revenue question: “As of September 2, 2026, how much did we earn in total over the last 90 days?”

What the tool returned

Answer to the repeated question

As of September 2, 2026, total revenue over the last 90 days was $23,194.52. This excludes cancelled and refunded orders.

Answer to the rephrased question

You earned $23,194.52 in total revenue over the last 90 days as of September 2, 2026. This includes orders through September 2 and excludes cancelled or refunded orders.

Transcribed from Proof 1 · Screen recording at 0:20

Everything the tool returned, in order (4)
  1. Answer to the repeated question

    Printed above.

  2. SQL behind the repeated-question answer · not part of the answer
    SELECT SUM(CAST(total_amount AS NUMERIC)) AS total_revenue
    FROM public.orders
    WHERE placed_at >= (DATE '2026-09-02' - INTERVAL '90 days')
      AND placed_at < (DATE '2026-09-02' + INTERVAL '1 day')
      AND status::text NOT IN ('cancelled', 'refunded')
      AND refunded_at IS NULL
  3. Answer to the rephrased question

    The final response, printed above.

  4. SQL behind the rephrased-question answer · not part of the answer
    SELECT COALESCE(SUM(total_amount), 0) AS total_revenue
    FROM public.orders
    WHERE placed_at >= (DATE '2026-09-02' - INTERVAL '90 days')
      AND placed_at < (DATE '2026-09-02' + INTERVAL '1 day')
      AND status::text NOT IN ('cancelled', 'refunded')
      AND refunded_at IS NULL
Expected vs. Found
✓Expected: The repeated and reworded questions must use the corrected definition and return the reference total of $23,194.52.Found: Both answers excluded cancelled and refunded orders and returned $23,194.52; the reference total was $23,194.52 across 41 orders.
Supporting proof
Proof 1Screen recording
Screen recording: revenue training, both question versions, answers and generated queries
Key moments
Open original ↗
Proof 2Screenshot
Screenshot: the untrained chat asking which revenue definition to use
Open original ↗
Why this result

Both fresh chats used the taught exclusions without another reminder and returned the $23,194.52 reference total. The reference total is reported without the independent database query being shown; the recording shows Draxlr’s own queries and results.

Also observed in this test case

Longer calculation window. Both revenue queries covered 91 days rather than 90. No orders fell on the extra day, so the total was unaffected; the reference calculation used the same window.

Tested by Ayush Ghosh · evidence dated 26 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
Draxlr
Build or version
Not exposed by the hosted product
Plan or tier
Premium
Model
gpt-5.5-1
Surface
Web app
Set up before the run
After an untrained revenue question prompted clarification, Train AI received the rule excluding cancelled and refunded orders. The September 2, 2026 question about the last 90 days was repeated and reworded in separate new chats.
Train AI context
Our revenue definition excludes cancelled orders and refunded orders. · no other context saved
Tested
26 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
19 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/draxlr/definition-correction · 1 pass · 0 fail · coverage 1/1 · graded 2026-10-26

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 · 31 KB
File fingerprints (SHA-256)

A fingerprint identifies the exact file used for this result.

Proof 1 · Screen recordingc048feb77d40573fc681e08740b9f9759fff6f996b1b0c6b01f099a48f7326d9
Proof 2 · Screenshotd5bbb518438075d6aac3eb44a62593c34ed017b085f8514ed7b22b0333476a69