Memory Architectures for Multi-Turn Text-to-SQL: A Benchmark and Empirical Study

Researchers introduced EnterpriseMem-Bench, a multi-turn Text-to-SQL benchmark of 300 sessions and 1,400 queries built from three enterprise domains, to evaluate how five frontier AI models handle memory across conversational turns. Testing revealed that all models, including GPT-5 mini and Claude Sonnet 4.6, collapsed to zero execution accuracy by Turn 3 in stateless conditions, while working memory dominated performance improvements over more complex memory architectures. The study also found a generational regression in Claude Sonnet 4.6, which underperformed its predecessor by 17-33 percentage points on SEC EDGAR data, and that reasoning caused error distributions to become mono-modal, with all non-correct turns producing wrong-result errors.

arXiv:2605.26394v1 Announce Type: new Abstract: Multi-turn Text-to-SQL is central to enterprise analytics yet remains predominantly evaluated in single-turn settings. We introduce EnterpriseMem-Bench, a multi-turn Text-to-SQL benchmark of 300 sessions and 1,400 turns built programmatically from three enterprise domains BIRD financial, SEC EDGAR, Northwind , with deterministic ground truth and per-turn memory-critical annotation. We evaluate five frontier models -- GPT-5 mini, GPT-5.2, Claude Sonnet 4.5, Sonnet 4.6, and Opus 4.6 -- across five memory conditions enabling a three-way ablation isolating working-memory window size, episodic retrieval, and semantic augmentation as independent effects. All Claude models are evaluated with extended thinking enabled to maintain parity with GPT reasoning models. We introduce the Memory Benefit Score MBS as a per-turn diagnostic metric. Four findings emerge: 1 stateless multi-turn Text-to-SQL collapses to zero execution accuracy by Turn 3 across all five models, even under reasoning; 2 memory-architecture complexity does not monotonically improve accuracy -- working memory dominates, and additional components produce model- and dataset-dependent effects from +14 to -16 percentage points; 3 Claude Sonnet 4.6 underperforms Sonnet 4.5 by 17-33pp on SEC EDGAR across conditions, a generational regression persisting under reasoning; 4 under reasoning, Claude error distributions become mono-modal -- every non-correct turn is a wrong-result error. We release the benchmark, agent, and evaluation code.