New framework assesses LLM-based social simulation model capacity

Researchers have introduced the Model Capacity Frontier (MCF), a framework designed to assess the empirical performance of Large Language Model (LLM)-supported agent-based models (ABMs). MCF characterizes the attainable performance across various configurations, represented by an estimated Pareto frontier. This framework was demonstrated using a crowd simulation of a university commencement event with approximately 12,000 agents, comparing a conventional crowd model with successive LLM-supported architectural extensions. The study revealed that initial LLM integration improved flow direction but negatively impacted temporal occupancy and spatial distribution, while subsequent refinements enhanced temporal occupancy and recovered spatial performance. AI

IMPACT Provides a framework for systematically developing and evaluating LLM-based agent-based models, guiding architectural improvements for better empirical fidelity.

RANK_REASON The cluster contains a research paper detailing a new framework for LLM-supported social simulation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.MA (Multiagent) →

• alphaXiv
• CatalyzeX Code Finder for Papers
• DagsHub
• Gotit.pub
• Hugging Face
• Influence Flower
• Large Language Model
• Model Capacity Frontier
• Pareto frontier
• ScienceCast

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →