Does Draxlr drop an earlier filter when the user changes direction?
Draxlr dropped the Texas filter when asked about all states, but reported 20.04% rather than the overall rate of 21.75%. It used an unweighted average of state percentages.
1 of 1 test case failed
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.
Drop the filter, go all segmentsFailEvidence
Input 1 · Texas: “What's our cancellation rate for orders shipped to Texas?”
Input 2 · All states: “What's our cancellation rate across all states?”
Texas answer
The cancellation rate for orders shipped to Texas is 27.78%. That's 5 cancelled orders out of 18 total orders, all-time.
All-states answer
Across all states shown, the average cancellation rate is 20.04%. The highest rate is OH at 29.41%, while NV is lowest at 0.00%. This is all-time, with blank or missing states grouped as Unknown.
Transcribed from Proof 1 · Screen recording at 0:07
Every message the tool sent, in order (2)
- Texas answer
Printed above.
- All-states answer
The final response, printed above.
| ✗ | Expected: Drop the Texas filter and give the overall cancellation rate across all states: 21.75%.Found: It removed the Texas filter. The 15 rows total 67 cancellations among 308 orders (21.75%); its reported 20.04% is the unweighted mean of the 15 state percentages. |
Draxlr dropped the filter, yet its average of state percentages did not supply the required overall rate. The percentage column is clipped in the recording, so the per-state and overall rates were checked from the visible cancelled and total counts.
Accurate state breakdown. Every per-state figure in the all-states table is correct.
Texas lookup coverage. The initial Texas lookup searched for both the full state name and its abbreviation, returning 5 cancelled orders out of 18.
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 | AI Database Agents → | 28 scenarios · 11 tools |
| Capability | Conversational Interaction → | |
| Scenario | The user changes direction mid-conversation → | |
| Tool | Draxlr → |
Global scenario definition → · Global capability definition → · Draxlr 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.