A field guide to the LLM-to-SQL research explosion

What it does This repository is the official companion to several major database and data-engineering surveys—most notably a TKDE'25 paper and a VLDB'24 tutorial—that attempt to map the post-LLM Text-to-SQL landscape. It collects papers, slides, and benchmark evaluations, then organizes them into a structured lifecycle covering model architectures, data scarcity, evaluation metrics, and error analysis.

The interesting bit Rather than dumping links into an awesome-list, the authors impose a five-level taxonomy of challenges, from solved problems to a five-year vision, and break every solution into three pipeline modules: pre-processing, translation, and post-processing. It is essentially an attempt to turn a gold-rush of research into a curriculum.

Key highlights

• Anchored to peer-reviewed surveys and tutorials at TKDE, VLDB, and SIGMOD, not just preprints.
• Splits the field into four lifecycle stages—Model, Data, Evaluation, and Error Analysis—with dedicated chapters for core modules.
• Tracks recent and upcoming work from major venues including ICML, SIGMOD, AAAI, ICDE, and KDD.
• Provides slides and links to associated code repositories for the surveyed papers.
• Explicitly maps how target users and capabilities shift across four eras of language-model development.

Caveats

• The repository is primarily a curated documentation resource; despite the Python classification, no executable framework or code is visible in the README.
• The analytical depth lives in linked PDFs and external chapters, so expect a structured reading list rather than a self-contained textbook.

Verdict Researchers and practitioners who need to orient themselves in the post-2023 Text-to-SQL literature will save hours here. If you are looking for a drop-in library to translate natural language into SQL, this is not it.