SOCIA-EVO: Automated Simulator Construction via Dual-Anchored Bi-Level Optimization

2026-04-19

AI Summary

This work proposes the SOCIA-EVO framework to address optimization instability in long-horizon large language model agents during automated simulator construction, a challenge arising from context drift and entanglement between structural and parametric errors. The approach enforces structural plausibility through static blueprints and decouples structure search from parameter calibration via bilevel optimization. Additionally, it incorporates a self-maintained strategy playbook that dynamically prunes ineffective strategies by integrating Bayesian-weighted retrieval with execution feedback. This enables hypothesis-driven evolution of simulators whose statistical distributions align closely with real-world observational data, substantially enhancing the stability, fidelity, and verifiability of long-horizon simulations.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Multiagent Systems: Agent-Based Simulation and Emergent Behavior
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems

Abstract

Automated simulator construction requires distributional fidelity, distinguishing it from generic code generation. We identify two failure modes in long-horizon LLM agents: contextual drift and optimization instability arising from conflating structural and parametric errors. We propose SOCIA-EVO, a dual-anchored evolutionary framework. SOCIA-EVO introduces: (1) a static blueprint to enforce empirical constraints; (2) a bi-level optimization to decouple structural refinement from parameter calibration; and (3) a self-curating Strategy Playbook that manages remedial hypotheses via Bayesian-weighted retrieval. By falsifying ineffective strategies through execution feedback, SOCIA-EVO achieves robust convergence, generating simulators that are statistically consistent with observational data. The code and data of SOCIA-EVO are available here: https://github.com/cruiseresearchgroup/SOCIA/tree/evo.

Problem

• automated simulator construction
• distributional fidelity
• contextual drift
• optimization instability
• structural and parametric errors

Innovation

• dual-anchored optimization
• bi-level optimization
• simulator construction
• Bayesian-weighted retrieval
• distributional fidelity