The BaseResearcharXiv
Text-to-SQL Models Make Choices Without Saying So
According to the abstract, Yichen Xu, Shipei Lin and Chenhao Ma introduce SpecBridge, which tests whether text-to-SQL models silently settle ambiguous requests, such as counting rows versus distinct entities. On 204 hand-validated tests, four models reported both readings for open requests in only 15 to 58% of cases where they could.
Why it matters
The same thing, for different readers.
What it means for you
Researchers found AI database assistants often pick one meaning of an ambiguous question without telling the analyst, so a plausible number may answer a different question.
For researchers
Xu, Lin and Ma suggest the failure lies in reporting rather than computing, and that benchmarks with one reference answer can mark right answers wrong.
How the finding was reached
In our words, from the source. Weigh the finding by how it was made.
- What was done
- SpecBridge: three otherwise identical requests per test (one open, one per interpretation), scored by running the answers on real databases; 204 hand-validated tests on four language models.
- What it found
- Models usually compute each interpretation when asked, but for the open request return both in only 15 to 58% of tests where they can; on BIRD, over two in five answers marked wrong are right under the other interpretation.
- What it does not show
- Read from the abstract excerpt only (cut off); four models and 204 tests.
- Status
- A preprint: not yet peer reviewed
The record
- Questions
- Are AI companies putting stronger safeguards into practice? · How is AI changing childhood, health and daily life?
- People
- Yichen Xu (Wrote it) · Shipei Lin (Wrote it) · Chenhao Ma (Wrote it)
- What we read
- We have read only the description the publisher gives. No quotations until someone reads the source itself.
- Found by
- The Radar, run of November 4, 2026, via
arxiv-ai-safety-query(tier 1: the source’s own channel or an official record) - Cite this
- arXiv, “Text-to-SQL Models Make Choices Without Saying So”, November 3, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/arxiv-2026-11-03-text-to-sql-unspoken-choices