mlsanityData to 14 Oct 2026

From Test-Time Scaling to Reusable Memory: Measuring Crystallization in Text-to-SQL

arXiv 2608.07213 ↗First posted 7 Aug 2026cs.CLcited by 0

On the map: LLM reasoning → Knowledge- and tool-augmented reasoning

Abstract (arXiv)

Test-time scaling can correct difficult text-to-SQL queries, but the extra computation is normally discarded after each answer. Systems increasingly retain verified repair episodes, yet evaluations still report one end-to-end score. It cannot distinguish replay on recurring questions from help on unseen questions, or identify the responsible memory choice. We call measuring this future value the crystallization problem. Our controlled evaluation holds the single-shot solver fixed and varies one memory choice at a time. We separately measure replay, cross-question retention, and held-out same-database transfer. On BIRD, storing verified corrected queries improves held-out first-attempt accuracy by 4.34 percentage points. This gain captures 44.4% of the accuracy headroom provided by on-demand repair on the same questions. Controlled interventions identify database-specific content as the main operating ingredient. Reliable verification and broader retrieval coverage yield supported gains; richer formats and elaborate retrievers do not. Open-source code, evaluation artifacts, and reproduction instructions are available at https://github.com/ai-jiaqian/text-to-sql-memory-crystallization.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Knowledge- and tool-augmented reasoning, one of 17 directions in LLM reasoning. Builds on 30 of its 66 references that are in the map. First posted on arXiv 7 Aug 2026.

Builds on

30 of 66 references.

Closest ancestors · 4

Same direction, newest first

  1. 2026Memo-SQL: Structured Decomposition and Experience-Driven Self-Correction for Training-Free NL2SQL15 Jan
  2. 2025Arctic-Text2SQL-R1: Simple Rewards, Strong Reasoning in Text-to-SQL22 May
  3. 2025Reward-SQL: Boosting Text-to-SQL via Stepwise Execution-Aware Reasoning and Process-Supervised Rewards7 May
  4. 2024The Death of Schema Linking? Text-to-SQL in the Age of Well-Reasoned Language Models14 Aug

Other references · 26

  1. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,986
  2. 2023MemGPT: Towards LLMs as Operating Systemscited by 1,389
  3. 2025ReasoningBank: Scaling Agent Self-Evolving with Reasoning Memorycited by 284
  4. 2025Evaluating Memory in LLM Agents via Incremental Multi-Turn Interactionscited by 254
  5. 2025A Survey on Test-Time Scaling in Large Language Models: What, How, Where, and How Well?cited by 223
  6. 2024CHESS: Contextual Harnessing for Efficient SQL Synthesiscited by 169
  7. 2023C3: Zero-shot Text-to-SQL with ChatGPTcited by 164
  8. 2024RSL-SQL: Robust Schema Linking in Text-to-SQL Generationcited by 52
  9. 2025XiYan-SQL: A Novel Multi-Generator Framework For Text-to-SQLcited by 40
  10. 2024E-SQL: Direct Schema Linking via Question Enrichment in Text-to-SQLcited by 35
  11. 2025DeepEye-SQL: A Software-Engineering-Inspired Text-to-SQL Frameworkcited by 28
  12. 2026AgentSM: Semantic Memory for Agentic Text-to-SQLcited by 14
14 more
  1. 2026Arming Data Agents with Tribal Knowledgecited by 7
  2. 2026Retrieval-Augmented LLM Agents: Learning to Learn from Experiencecited by 7
  3. 2026Causal Intervention-Based Memory Selection for Long-Horizon LLM Agentscited by 6
  4. 2024Leveraging Prior Experience: An Expandable Auxiliary Knowledge Base for Text-to-SQLcited by 5
  5. 2026Beyond Experience Retrieval: Learning to Generate Utility-Optimized Structured Experience for Frozen LLMscited by 5
  6. 2025SEED: Enhancing Text-to-SQL Performance and Practical Usability Through Automatic Evidence Generationcited by 4
  7. 2025RubikSQL: Lifelong Learning Agentic Knowledge Base as an Industrial NL2SQL Systemcited by 4
  8. 2026Learning to Retrieve: Dual-Level Long-Term Memory for Text-to-SQL Agentscited by 4
  9. 2026MemTrace: Probing What Final Accuracy Misses in Long-Term Memorycited by 4
  10. 2026RoboPhD: Self-Improving Text-to-SQL Through Autonomous Agent Evolutioncited by 3
  11. 2025ORANGE: An Online Reflection ANd GEneration framework with Domain Knowledge for Text-to-SQLcited by 1
  12. 2025Continual Learning of Domain Knowledge from Human Feedback in Text-to-SQLcited by 1
  13. 2026EGREFINE: An Execution-Grounded Optimization Framework for Text-to-SQL Schema Refinementcited by 1
  14. 2026TAHOE: Text-to-SQL with Automated Hint Optimization from Experiencecited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Correct Now, Insufficient Later: Auditing Update Sufficiency in Context Compression
  2. 2025Agentar-Scale-SQL: Advancing Text-to-SQL through Orchestrated Test-Time Scaling
  3. 2026LSREP: A Longitudinal State-Replay Protocol for Evaluating Conversational Memory, with ICE v2 as an Audited Local-First Architecture
  4. 2026Causal Episodic Memory for Feedback-Driven Agent Repair
  5. 2026LIMIT: Less Is More for Instruction Tuning in Text-to-SQL
  6. 2026Storage Is Not Memory: A Retrieval-Centered Architecture for Agent Recall
  7. 2026Auditing Long-Term Memory Evaluation: Repeated Judging, Reader Variation, and Negative Controls
  8. 2025Rethinking Agentic Workflows: Evaluating Inference-Based Test-Time Scaling Strategies in Text2SQL Tasks