Benchmark RadarAI BENCHMARK PROFILE
EnterpriseMem-Bench
EnterpriseMem-Bench is a multi-turn Text-to-SQL benchmark with 300 sessions and 1,400 turns across three enterprise domains, featuring deterministic ground truth and per-turn memory-critical annotations.
- Released
- 2026-05-25
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
It addresses the lack of multi-turn evaluation in Text-to-SQL, providing insights into memory architecture effects and model performance degradation.
Motivation
Multi-turn Text-to-SQL is central to enterprise analytics yet remains predominantly evaluated in single-turn settings.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.