PhysicsAgentABM: Physics-Guided Generative Agent-Based Modeling
This work provides a new paradigm for scalable and calibrated simulation with LLMs, which is significant for researchers and practitioners in fields like public health, finance, and social sciences who need to model complex multi-agent systems.
This paper introduces PhysicsAgentABM, a new generative agent-based modeling framework that addresses the limitations of LLM-based multi-agent systems and classical ABMs. It achieves this by shifting inference to behaviorally coherent agent clusters, using state-specialized symbolic agents and a multimodal neural transition model, and then allowing individual agents to stochastically realize transitions. The method reduces LLM calls by 6-8 times and shows consistent gains in event-time accuracy and calibration across various domains.
Large language model (LLM)-based multi-agent systems enable expressive agent reasoning but are expensive to scale and poorly calibrated for timestep-aligned state-transition simulation, while classical agent-based models (ABMs) offer interpretability but struggle to integrate rich individual-level signals and non-stationary behaviors. We propose PhysicsAgentABM, which shifts inference to behaviorally coherent agent clusters: state-specialized symbolic agents encode mechanistic transition priors, a multimodal neural transition model captures temporal and interaction dynamics, and uncertainty-aware epistemic fusion yields calibrated cluster-level transition distributions. Individual agents then stochastically realize transitions under local constraints, decoupling population inference from entity-level variability. We further introduce ANCHOR, an LLM agent-driven clustering strategy based on cross-contextual behavioral responses and a novel contrastive loss, reducing LLM calls by up to 6-8 times. Experiments across public health, finance, and social sciences show consistent gains in event-time accuracy and calibration over mechanistic, neural, and LLM baselines. By re-architecting generative ABM around population-level inference with uncertainty-aware neuro-symbolic fusion, PhysicsAgentABM establishes a new paradigm for scalable and calibrated simulation with LLMs.