PhysicsAgentABM: Physics-Guided Generative Agent-Based Modeling

TL;DR

PhysicsAgentABM rethinks generative ABMs by performing inference at the population level within semantically meaningful clusters, using a symbolic-neural fusion to produce calibrated transition priors. The ANCHOR clustering mechanism, driven by LLMs, creates behaviorally coherent abstractions that adapt to regime shifts, while a decoupled agent-level realization preserves heterogeneity. Across epidemiology, finance, and social diffusion, the framework achieves improved event-time accuracy and calibration, validated through rolling-window evaluations and case studies such as the Singapore COVID-19 circuit breaker. The approach offers a scalable, interpretable, and uncertainty-aware paradigm for simulating complex systems under partial observability with real-world impact in public health, economics, and information dynamics.

Abstract

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.

Figures (13)

• Figure 1: Overview of PhysicsAgentABM Architecture. Inference is performed at the cluster level via symbolic and neural pathways with uncertainty-aware fusion, followed by stochastic agent-level realization. ANCHOR enables behaviorally coherent abstraction.
• Figure 2: ANCHOR Overview. An overview of our clustering mechanism.
• Figure 3: ANCHOR cross-contextual clusters (epidemiology).ANCHOR identifies four distinct clusters among 1,000 agents based on joint semantic and cross-contextual behavioral responses. We report normalized averages of the top-5 dominant behavioral motifs per cluster, capturing coordinated shifts in response intensity, isolation, and compliance across contexts, with signed values indicating motif direction and strength. See Appendix Section \ref{['supp-sec:anchor']} for detailed motif descriptions and interpretation.
• Figure 4: SEIRD dynamics under rolling-window forecasting. Infection trajectories for a 1,000-agent COVID-19 simulation comparing Rule-ABM, neural and LLM baselines with our model.
• Figure 5: Population-level market belief dynamics inferred by the agent ecosystem.(A) Probability mass assigned to the realized market sentiment regime (S&P 500) over time. Daily belief estimates fluctuate with market volatility(light blue line), while the 7-day moving average reveals a stable, coherent belief trajectory. Compared to a rule-based ABM, our model maintains higher and more persistent alignment with realized regimes, indicating superior collective belief formation under uncertainty. (B) Our model recovers a balanced regime distribution consistent with the realized market structure, while the rule-based ABM exhibits regime bias.