SpaRAGi: Spatial Inference using Retrieval-Augmented Generation

ACL ARR 2025 February Submission3155 Authors

Abstract: The advent of large language models (LLMs) has enabled powerful applications across several domains such as science, healthcare, finance, and law. However, LLMs are challenged when asked domain-specific questions. In particular, the spatial knowlege and spatial inference capabilities of LLMs are limited. Our goal is to enhance their accuracy for queries that reason about spatial data. To this end, we leverage the emerging Retrieval Augmented Generation (RAG) paradigm via which LLMs can enrich their context using external data, during inference. We present a framework that i) extracts context from a geospatial database regarding the spatial relations between entities, and ii) retrieves the relevant context to a query at inference time, forwarding it to the LLM to enhance its accuracy. Overall, our framework sets the ground for the use of spatial knowledge retrieval techniques for improving the effectiveness of LLMs.

Paper Type: Long

Research Area: Generation

Research Area Keywords: retrieval-augmented generation

Contribution Types: Model analysis & interpretability, NLP engineering experiment, Approaches to low-resource settings, Approaches low compute settings-efficiency, Data resources

Languages Studied: english

Submission Number: 3155