STBench: Assessing the Ability of Large Language Models in Spatio-Temporal Analysis

TL;DR

This paper introduces STBench, a comprehensive benchmark dataset with 13 tasks and over 60,000 QA pairs to evaluate large language models' capabilities in understanding and reasoning about spatio-temporal data.

Contribution

The paper develops a new benchmark dataset, STBench, to systematically assess LLMs' spatio-temporal understanding across four key dimensions, addressing limitations of prior evaluations.

Findings

LLMs perform well in knowledge comprehension and reasoning tasks

Potential for improvement with prompting and fine-tuning techniques

Assessment of 13 different large language models

Abstract

The rapid evolution of large language models (LLMs) holds promise for reforming the methodology of spatio-temporal data mining. However, current works for evaluating the spatio-temporal understanding capability of LLMs are somewhat limited and biased. These works either fail to incorporate the latest language models or only focus on assessing the memorized spatio-temporal knowledge. To address this gap, this paper dissects LLMs' capability of spatio-temporal data into four distinct dimensions: knowledge comprehension, spatio-temporal reasoning, accurate computation, and downstream applications. We curate several natural language question-answer tasks for each category and build the benchmark dataset, namely STBench, containing 13 distinct tasks and over 60,000 QA pairs. Moreover, we have assessed the capabilities of 13 LLMs, such as GPT-4o, Gemma and Mistral. Experimental results reveal that existing LLMs show remarkable performance on knowledge comprehension and spatio-temporal reasoning tasks, with potential for further enhancement on other tasks through in-context learning, chain-of-though prompting, and fine-tuning. The code and datasets of STBench are released on https://github.com/LwbXc/STBench.