Prompt Engineering Guidance for Conceptual Agent-based Model Extraction using Large Language Models

TL;DR

The paper presents a structured, QA-driven prompt engineering pipeline to extract ABM specifications from conceptual models in JSON form, enabling human readability and auto-code generation by LLMs. It defines modular prompts (covering model aim, agent sets, environment, and execution) and explicit JSON schemas to address data diversity, execution order, and boundaries. By prescribing QA instructions and a strict JSON output format, the approach aims to improve accuracy, reproducibility, and automation in converting conceptual ABMs into implementable simulations. The work advances practical ABM automation by providing a clear, machine-consumable information extraction workflow that can feed code-generation processes in LLM ecosystems.

Abstract

This document contains detailed information about the prompts used in the experimental process discussed in the paper "Toward Automating Agent-based Model Generation: A Benchmark for Model Extraction using Question-Answering Techniques". The paper aims to utilize Question-answering (QA) models to extract the necessary information to implement Agent-based Modeling (ABM) from conceptual models. It presents the extracted information in formats that can be read by both humans and computers (i.e., JavaScript Object Notation (JSON)), enabling manual use by humans and auto-code generation by Large Language Models (LLM).

Figures (1)

• Figure 1: List of prompts and their target information