Economic World Models

Building AI-system blueprints for agentic economies, policy sandboxes, and self-correcting economic twins.

Project
AI agents Economic world models Agentic economies Policy simulation Sim-to-real twins
Research team
Economic World Models compare physical-world transitions with agent-generated economic-world transitions

Economic decisions are usually made in the real world first and evaluated afterward. Economic World Models invert that workflow: they ask how we can build computable economies where heterogeneous agents observe, reason, act, interact, adapt, and co-evolve with markets and institutions before policies, strategies, or AI agents are deployed at scale.

The project develops an implementation-oriented systems blueprint for EconWM systems. It connects economic discipline with AI-system design, treating an economy as a generative engine whose state transitions are produced from inside the modeled world. Agents form beliefs, choose actions, alter the aggregate state, and then respond to the world they helped create.

The central goal is to make economic worlds buildable, testable, and alignable: useful as policy sandboxes for human decision-makers, training environments for economic agents, planning engines for interventions, and safety testbeds for emergent risks such as manipulation, collusion, instability, and cascading failure.

What The Project Builds

Agentic economic sandboxes

Simulation environments where households, firms, banks, investors, regulators, and AI agents can interact under explicit objectives, constraints, information, and rules.

A six-level capability ladder

A taxonomy from fixed rule-based agent worlds to adaptive agents, LLM-based autonomous agents, self-evolving agents, evolving institutions, and sim-to-real economic twins.

Alignment and evaluation loops

Mechanisms for comparing simulated outcomes with real economic evidence, diagnosing deviations, and correcting agents, mechanisms, and transition dynamics over time.

Six capability levels of Economic World Model systems

Systems Blueprint

The proposed EconWM architecture has four connected layers: economic agents, economic environments, agent-environment co-evolution, and real-world alignment. This makes the project more than a survey of AI agents in economics: it is a build plan for economic digital twins that can reason, simulate, learn, and stay empirically grounded.

Core components of an Economic World Model

Research Team

The work brings together researchers across Shenzhen Loop Area Institute, CUHK-Shenzhen School of Data Science, the University of Hong Kong, and Nanyang Technological University.

Selected Publications

From Economic Agents to Agentic Economies: A Systems Blueprint for Economic World Models

Jiale Han, Xiang Li, Jing Qian, Wenyuan Gu, Pin Gao, Ye Luo, Hongyuan Zha, Dacheng Tao, Benyou Wang, and Lin William Cong. Working Paper.

Paper
TwinMarket: A Scalable Behavioral and Social Simulation for Financial Markets

Yuzhe Yang, Yifei Zhang, Minghao Wu, Kaidi Zhang, Yunmiao Zhang, Honghai Yu, Yan Hu, and Benyou Wang. NeurIPS 2025; Best Paper, ICLR 2025 Financial AI Workshop.

Paper
UCFE: A User-Centric Financial Expertise Benchmark for Large Language Models

Yuzhe Yang, Yifei Zhang, Yan Hu, Yilin Guo, Ruoli Gan, Yueru He, Mingcong Lei, Xiao Zhang, Haining Wang, Qianqian Xie, Jimin Huang, Honghai Yu, and Benyou Wang. Findings of NAACL 2025.

Paper
FinBen: A Holistic Financial Benchmark for Large Language Models

Qianqian Xie, Weiguang Han, Zhengyu Chen, Ruoyu Xiang, Xiao Zhang, Yueru He, Mengxi Xiao, Dong Li, Yongfu Dai, Duanyu Feng, Yijing Xu, Haoqiang Kang, Ziyan Kuang, Chenhan Yuan, Kailai Yang, Zheheng Luo, Tianlin Zhang, Zhiwei Liu, Guojun Xiong, Zhiyang Deng, Yuechen Jiang, Zhiyuan Yao, Haohang Li, Yangyang Yu, Gang Hu, Huang Jiajia, Xiaoyang Liu, Alejandro Lopez-Lira, Benyou Wang, Yanzhao Lai, Hao Wang, Min Peng, Sophia Ananiadou, and Jimin Huang.

Paper
Open-FinLLMs: Open Multimodal Large Language Models for Financial Applications

Jimin Huang, Mengxi Xiao, Dong Li, Zihao Jiang, Yuzhe Yang, Yifei Zhang, Lingfei Qian, Yan Wang, Xueqing Peng, Yang Ren, Ruoyu Xiang, Zhengyu Chen, Xiao Zhang, Yueru He, Weiguang Han, Shunian Chen, Lihang Shen, Daniel Kim, Yangyang Yu, Yupeng Cao, Zhiyang Deng, Haohang Li, Duanyu Feng, Yongfu Dai, VijayaSai Somasundaram, Peng Lu, Guojun Xiong, Zhiwei Liu, Zheheng Luo, Zhiyuan Yao, Ruey-Ling Weng, Meikang Qiu, Kaleb E Smith, Honghai Yu, Yanzhao Lai, Min Peng, Jian-Yun Nie, Jordan W Suchow, Xiao-Yang Liu, Benyou Wang, Alejandro Lopez-Lira, Qianqian Xie, Sophia Ananiadou, and Junichi Tsujii.

Paper
No Language Is an Island: Unifying Chinese and English in Financial Large Language Models, Instruction Data, and Benchmarks

Gang Hu, Ke Qin, Chenhan Yuan, Min Peng, Alejandro Lopez-Lira, Benyou Wang, Sophia Ananiadou, Jimin Huang, and Qianqian Xie.

Paper

From Papers to Systems

The Economic World Models page is the umbrella. Under it, TwinMarket provides a concrete financial market sandbox, MicroVerse explores scientific micro-world simulation, and the curated Awesome Econ World Models repository keeps the paper trail organized. Together they turn the idea of "agents in an environment" into a program about evolving worlds: social worlds, market worlds, biological worlds, and eventually policy-facing digital twins.