APS: Bias-Controlled Adaptive Prototype Simulation for Population-Scale LLM Agents

2026-05-19

AI Summary

This work addresses the computational bottleneck in large-scale LLM-based agent simulations, where costs grow linearly with population size and simulation rounds. To overcome this, the authors propose the Adaptive Prototype Simulation (APS) framework, which constructs localized response surfaces using core prototypes, tail singletons, and shadow auditing agents to drastically reduce online LLM invocations. APS integrates recursive oracle allocation, residual propagation estimation, and dynamic budgeting, augmented by a bias-control mechanism that leverages shadow auditing corrections and tail-protection routing to preserve high-fidelity population-level response distributions. Evaluated on a ten-million-agent opinion dynamics simulation, APS achieves a 381.1× reduction in computational cost while maintaining a final-round Jensen–Shannon divergence of only 0.094, substantially outperforming baseline methods under equivalent computational budgets.

Abstract

LLM-agent simulation offers a flexible computational tool for studying population response trajectories that depend on scenario events, memory, demographics, and evolving social context. However, full multi-round simulation scales linearly with both population size and horizon, requiring every agent to query the LLM at every round. We propose Adaptive Prototype Simulation (APS), a framework that reframes scalable LLM-based simulation as a recurrent oracle-allocation problem. APS retains the designated LLM as the online transition oracle while querying adaptive core prototypes, selected singleton-tail agents, and shadow-audit agents. Prototype responses induce local response surfaces for nearby agents, reducing online LLM calls without replacing the underlying transition model. To control approximation bias, shadow-audit residual correction estimates propagation residuals for aggregate correction and future budget allocation, while tail-protected singleton routing directly queries selected isolated, heterogeneous, or high-curvature regions that are vulnerable to smoothing. Theoretically, we treat APS as an estimator for full-scale high-precision individual social simulation and decompose its errors into prototype-coverage error, shadow-audit residual-correction error, local-propagation bias, and temporal context mismatch. Under the reported protocols, APS gives lower reference-aligned distributional discrepancy than scale-oriented and same-budget baselines while reducing online LLM calls, with ablations and compact robustness checks diagnosing the main bias-control mechanisms. In a 10M-agent, multi-round public-opinion simulation, APS achieves a 381.1-fold reduction over full simulation, with reference-aligned final-round JSD of 0.094 against the corresponding full-LLM reference.

Problem

• LLM-agent simulation
• population-scale simulation
• computational scalability
• approximation bias
• online LLM calls

Innovation

• Adaptive Prototype Simulation
• LLM-agent simulation
• bias-controlled approximation
• shadow-audit correction
• scalable social simulation