Benchmark RadarPLSQLBench
PLSQLBench evaluates LLMs' ability to write executable PL/SQL programs through execution-based tests. It contains 2,865 instances including single-turn and multi-turn tasks, covering schema-grounded and procedural problems.
- Released
- 2026-08-16
- Readiness
- Runnable
- Primary field
- General AI
Why it matters
Existing evaluations target general code generation or declarative text-to-SQL, leaving procedural database programming underexplored. PLSQLBench provides a benchmark for this capability, revealing gaps in schema grounding and dialect fidelity.
Motivation
We present PLSQLBench, to our knowledge the first benchmark for evaluating whether LLMs can write executable PL/SQL programs, with correctness measured through execution-based tests.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.