Querying Databases with Function Calling (arXiv 2502.00032)
This forum entry indexes an academic paper from the curated reading list on Querying Structured Information.
Metadata
| Field | Content | |-------|---------| | Title | Querying Databases with Function Calling | | Authors | Connor Shorten, Charles Pierse, Thomas Benjamin Smith, Karel D'Oosterlinck, Tuana Celik, Erika Cardenas, et al. (12 authors) | | Published | January 2025 (arXiv listing: 2502.00032) | | Source | https://arxiv.org/abs/2502.00032 | | Type | Academic paper | | Section | Querying Structured Information |
About the paper
The paper examines how databases can be queried via LLM function calling, placing itself at the intersection of information retrieval and structured data access in the LLM era. Rather than translating natural language into SQL (the Text-to-SQL paradigm), function calling lets a model invoke database-native APIs and query operators directly, which can simplify integration with vector databases and hybrid search systems.
> Note: The original forum post is a stub — the paper's abstract text was not reproduced. Specific methods, experiments, and quantitative results should be verified against the original PDF at arxiv.org/abs/2502.00032.
Related entries cross-referenced in the list
- A Survey of Text-to-SQL in the Era of LLMs
- A Survey on Employing Large Language Models for Text-to-SQL Tasks
- Large language model for table processing: a survey (Jan 2025, Frontiers)
- Next-Generation Database Interfaces: A Survey of LLM-based Text-to-SQL
- Latency: cascade retrieval with early stopping; cache hot queries
- Quality: validate offline gains against online user satisfaction
- Safety: source whitelisting and output filtering for open retrieval
- Cost: route to smaller models, use sparse+dense hybrid retrieval
- Original paper: Querying Databases with Function Calling, arXiv:2502.00032
Context from the list's commentary
The surrounding analysis frames this work within the evolution of neural information retrieval — from BM25 and dense retrieval to agentic search, where "whether to retrieve, how many times, and which tools to call" becomes part of the model's action space. Engineering checklist items highlighted for practitioners deploying LLM-database integrations:
Takeaways by audience
1. Researchers: reproduce core baselines; check whether statistical significance and compute costs are reported. 2. Engineers: evaluate integration cost of pluggable modules (encoders, retrievers, planners) into existing stacks. 3. Product managers: focus on user-perceivable benefits (latency, answer trustworthiness) rather than offline metrics alone.