Learning to Retrieve: Dual-Level Long-Term Memory for Text-to-SQL Agents

arXiv:2606.00547 · cs.CL · Submitted 2026-05-30 · Read on arXiv

cs.CL

Submitted: 2026-05-30

Updated: 2026-10-13

Code: https://github.com/YiboWANG214/MERIT

Project page: https://bird-interact.github.io/7

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Text-to-SQL — 3× in this paper · explained in 3 episodes
• Benchmark — 2× in this paper · explained in 2 episodes
• Reinforcement Learning — 1× in this paper · explained in 4 episodes

Sources

• Mem0: Building Production-Ready AI Agents with Scalable Long-Term Memory
• Process Reward Models for LLM Agents: Practical Framework and Directions
• Memp: Exploring Agent Procedural Memory
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
• BIRD-INTERACT: Re-imagining Text-to-SQL Evaluation for Large Language Models via Lens of Dynamic Interactions
• GPT-4o System Card
• RAP: Retrieval-Augmented Planning with Contextual Memory for Multimodal LLM Agents
• SimpleMem: Efficient Lifelong Memory for LLM Agents
• MemoChat: Tuning LLMs to Use Memos for Consistent Long-Range Open-Domain Conversation
• OpenAI GPT-5 System Card
• AgentPRM: Process Reward Models for LLM Agents via Step-Wise Promise and Progress
• SQLNet: Generating Structured Queries From Natural Language Without Reinforcement Learning
• Memory-R1: Enhancing Large Language Model Agents to Manage and Utilize Memories via Reinforcement Learning
• Qwen3 Technical Report
• On the Structural Memory of LLM Agents
• MemRL: Self-Evolving Agents via Runtime Reinforcement Learning on Episodic Memory
• Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learning
• Memento: Fine-tuning LLM Agents without Fine-tuning LLMs

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering