Can FutureSmart Database Agent ask for clarification when a follow-up is ambiguous?
FutureSmart Database Agent did not ask for clarification when the follow-up was ambiguous. Turn 2 named the two rankings it had chosen and repeated the same comparison instead of asking which was meant.
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.
"Compare them" with two candidate setsFailEvidence
Input 1: “Compare our top 3 customers by order count against our top 3 customers by total spend.”
Input 2: “Compare them.”
Response 1 is a progress message the tool sent for Input 1; it is not part of the answer.
Response 2 answers Input 1.
Response 3 is the query the tool generated for Input 1; it is not part of the answer.
Response 4 is a progress message the tool sent for Input 2; it is not part of the answer.
Response 5, the final response, answers Input 2.
Response 6 is the query the tool generated for Input 2; it is not part of the answer.
Comparison of Top Customers We looked at the two rankings returned by the query: Top 3 customers by order count Top 3 customers by total spend Below is a combined view that shows each customer's order count and total spend where the data is available. Key Insights Wei Zhang leads both rankings, with the highest number of orders (10) and the highest spend ($7,866.02). The other two customers in the order-count list (Ingrid Sorensen and Priya Raghavan) do not appear in the spend list, indicating they placed many orders but with lower average values. Conversely, Deshawn Carter and Hannah Kowalski rank high on spend but are not in the top-order-count group, suggesting fewer but higher-value purchases. Overall, the data shows that while order volume and monetary value often overlap (as with Wei Zhang), they can also highlight different customer behaviours.
Transcribed from Proof 2 · Screenshot
Every message the tool sent, in order (6)
- Progress message for input 1
Generated SQL WITH top_by_orders AS (SELECT c.id AS customer_id, c.name, … | Read tables top_by_orders, top_by_spend | Returned 6 rows
- Response to input 1 (message 2 of 3)
Comparison of the Top 3 Customers Top 3 by Order Count (most orders placed) Top 3 by Total Spend (highest amount spent) Key Insight: Wei Zhang is the only customer who appears in both top-3 lists, leading in both order count (10 orders) and total spend ($7,866.02). All other customers are unique to each ranking. Summary Shared top performer: Wei Zhang (dominates both metrics). Order-focused leaders: Ingrid Sorensen and Priya Raghavan have high order counts but their spend data is not in the top-3 spend list. Spend-focused leaders: Deshawn Carter and Hannah Kowalski have the highest spend amounts but did not rank in the top-3 for order count. This comparison highlights that while a high number of orders often correlates with higher spend, the two metrics can also identify different customer segments.
- Generated query for input 1 · not part of the answer
WITH top_by_orders AS (SELECT c.id AS customer_id, c.name, COUNT(o.id) AS order_count, NULL::numeric AS total_spend FROM customers c JOIN orders o ON c.id = o.customer_id GROUP BY c.id, c.name ORDER BY order_count DESC LIMIT 3), top_by_spend AS (SELECT c.id AS customer_id, c.name, NULL::bigint AS order_count, SUM(o.total_amount) AS total_spend FROM customers c JOIN orders o ON c.id = o.customer_id GROUP BY c.id, c.name, o.currency ORDER BY total_spend DESC LIMIT 3) SELECT * FROM top_by_orders UNION SELECT * FROM top_by_spend
- Progress message for input 2
Generated SQL (SELECT c.id AS customer_id, c.name AS customer_name, COUNT… | Returned 6 rows
- Final response
Comparison of Top Customers We looked at the two rankings returned by the query: Top 3 customers by order count Top 3 customers by total spend Below is a combined view that shows each customer's order count and total spend where the data is available. Key Insights Wei Zhang leads both rankings, with the highest number of orders (10) and the highest spend ($7,866.02). The other two customers in the order-count list (Ingrid Sorensen and Priya Raghavan) do not appear in the spend list, indicating they placed many orders but with lower average values. Conversely, Deshawn Carter and Hannah Kowalski rank high on spend but are not in the top-order-count group, suggesting fewer but higher-value purchases. Overall, the data shows that while order volume and monetary value often overlap (as with Wei Zhang), they can also highlight different customer behaviours.
- Generated query for input 2 · not part of the answer
(SELECT c.id AS customer_id, c.name AS customer_name, COUNT(o.id) AS order_count, NULL::numeric AS total_spend, 'top_by_orders' AS ranking_type FROM customers c JOIN orders o ON c.id = o.customer_id GROUP BY c.id, c.name ORDER BY order_count DESC LIMIT 3) UNION ALL (SELECT c.id AS customer_id, c.name AS customer_name, NULL::bigint AS order_count, SUM(o.total_amount) AS total_spend, 'top_by_spend' AS ranking_type FROM customers c JOIN orders o ON c.id = o.customer_id GROUP BY c.id, c.name ORDER BY total_spend DESC LIMIT 3)
| ✗ | Expected: Asks which set is meant instead of silently picking one.Found: It named the two rankings it chose and repeated the comparison without asking for clarification. |
Because the first turn produced two different sets, the follow-up was ambiguous and needed clarification. The second turn instead named the two rankings it had chosen and repeated the same comparison without asking which was meant, so this is a fail.
Chosen reading. The second turn named the two rankings it had chosen, but it still did not ask which one was meant.
The first capture shows only five of the six result rows.
The second capture shows the full turn-2 response, so the lack of clarification is visible there.
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 follow-up is ambiguous → | |
| Tool | FutureSmart Database Agent → |
Global scenario definition → · Global capability definition → · FutureSmart Database Agent 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.