Benchmark RadarEntSQL
EntSQL evaluates text-to-SQL systems on enterprise knowledge grounding, with 1,066 Chinese-English examples across five business domains requiring private business knowledge. Systems generate SQL from questions and schema, with some provided long-form documents.
- Released
- 2026-06-02
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Existing text-to-SQL benchmarks overlook enterprise scenarios where SQL generation depends on proprietary business knowledge. EntSQL measures the ability to ground SQL generation in long-context enterprise documents, revealing a significant performance gap in current systems.
Motivation
Text-to-SQL enables natural language access to databases, and recent LLMs have substantially advanced its capabilities.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.