AgoraSim: A Hybrid Agent-Based Modeling Framework

2026-07-07

AI Summary

Current large language model (LLM) agent simulations are often misinterpreted as social predictions and lack comparability with explicit models of social dynamics. This work proposes a hybrid agent-based modeling (ABM) framework that automatically translates textual or multimodal inputs into editable ABM configurations, enabling collaborative simulation among LLMs, vision-language models, rule-based, stochastic, and custom API-driven agents. By unifying structured decision-making objects across agent types, the framework facilitates cross-model behavioral comparison, trajectory auditing, and hypothesis testing. Integrated with a Python SDK, command-line interface, and local user interface, the system supports the identification of critical scenarios requiring empirical validation. To our knowledge, this is the first systematic approach to integrate multimodal foundation models with classical ABM for controllable social simulation.

Abstract

LLM-agent simulations make natural-language social scenarios easy to instantiate, but their outputs can be overread as predictions and are often difficult to compare with explicit social dynamics. We present AgoraSim, a hybrid agent-based modeling framework for scenario-oriented social reaction analysis. AgoraSim resolves textual or multimodal artifacts into editable ABM configurations, runs ratio-controlled populations that mix LLM, vision-language, custom-endpoint, random, and classical agents, and compares the same scenario against matched classical reference dynamics. All agents emit a shared structured decision object, enabling common action spaces, interaction protocols, metrics, and audit records. Exposed through a local UI, Python SDK/CLI, and REST API, AgoraSim helps users inspect scenario trajectories, compare modeling assumptions, and identify cases that warrant empirical validation.

Problem

• LLM-agent simulation
• social dynamics
• agent-based modeling
• scenario analysis
• model comparison

Innovation

• hybrid agent-based modeling
• LLM-agent simulation
• structured decision object
• multimodal scenario parsing
• classical reference dynamics