Synthetic Text-to-SQL
Gretel AI
105,851 synthetic text-to-SQL records spanning dozens of domains and SQL task types.
Generated with Gretel Navigator: each record includes a schema, a natural-language question, the SQL answer, and an explanation. It is split into 100,000 training and 5,851 test records, about 23 million tokens. It was designed to cover more domains and SQL features than WikiSQL or Spider.
Access
- Cost
- FreeNo charge.
- License
- Open licenseApache 2.0
- Sign-up
- No sign-up
- How you get it
- API and Bulk downloadParquet, JSON Lines
What's in it
- One row is
- One row per question with database context and SQL
- Covers
- Not geographic
- Years
- 2024
- Updated
- One-time release
- Size
- 105,851 records (100,000 train, 5,851 test), about 23M tokens
Quality
Broad domain coverage and a clear generation process, but entirely synthetic and not run against real databases.
- Schemas and data are invented
- SQL may not run on real database engines
- Can differ in style from real-world questions
Connects to
- Spider: Human-written, cross-domain text-to-SQL benchmark for real-world comparison
- WikiSQL: Earlier, simpler text-to-SQL set
Questions it could answer
- Which SQL constructs are rare in synthetic versus human-written sets?
- How does model accuracy change by complexity label?