Southwestern University of Finance and Economics
Bachelor of Economics
About Me
I am a sophomore studying Economics at the Southwestern University of Finance and Economics (SWUFE), advised by Prof. Hong Zou. I spent Spring 2026 as a visiting student at UC Berkeley. Currently, I am a research intern working with Kehang Zhu (Harvard) on LLM-based market simulation, a project advised by Prof. John Horton (MIT Sloan).
I use computational methods to study economic questions. My main focus is using LLM agents to simulate economic behavior, which makes it possible to ask how markets, rules, and institutions shape what people do. These are questions that are often hard to test with human subjects alone. I have also worked with NLP methods on policy text, building a structured database of China’s consumption-promotion policies from over 5 million government documents.
I am still in the early stages of my research journey and actively exploring these areas. If your interests intersect with computational social science, LLM agents, or text-as-data, I’d love to connect. You can find my CV here.
Education
Publications
PersonaForge: Psychology-Grounded Dual-Process Architecture for Personality-Consistent Role-Playing Agents
Findings of the ACL: ACL 2026
Large Language Models excel at role-playing but struggle to maintain consistent personalities across extended multi-turn interactions. We propose PersonaForge, combining (1) a three-layer personality architecture grounded in psychological theory and (2) a dual-process generation mechanism inspired by cognitive science. We test two falsifiable claims: Claim 1 (Orthogonality): Psychology-grounded dimensions (Big Five + Defense Mechanisms) provide more orthogonal constraints than natural language descriptions, reducing long-dialogue drift. Claim 2 (Integration Necessity): High-dimensional personality constraints create “production interference” requiring a cognitive workspace (Inner Monologue) to resolve—removing it degrades performance below simpler baselines. Experiments on 88 characters demonstrate: (1) +19.4% personality consistency (PC) with human correlation r=0.82, (2) reduced drift over 50-turn conversations (6.3% vs. 24.8% baseline), and (3) +64.7% defense mechanism manifestation. External validation on RoleBench confirms generalization (73.2% win-rate, drift 8.4% vs. 20.4%). Selective dual-process activation achieves 96% of full-system performance with only 13.4% token overhead. Human evaluation confirms more authentic and psychologically coherent character behaviors.
@inproceedings{tong-zou-2026-personaforge,
title = "{P}ersona{F}orge: Psychology-Grounded Dual-Process Architecture for Personality-Consistent Role-Playing Agents",
author = "Tong, Jizhou and
Zou, Sirui",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-acl.386/",
doi = "10.18653/v1/2026.findings-acl.386",
pages = "7845--7874",
ISBN = "979-8-89176-395-1"
}Research Experience
Generative Market Simulation
Building a prediction market where LLM agents trade through the same interface a human trader would, to study which strategies emerge and how prices form.
Supervised by Kehang Zhu (Harvard), advised by Prof. John Horton (MIT Sloan)
Decoding China’s Consumption Policy
A two-stage text-as-data pipeline over government documents (mainly 2006–2024), measuring which policy instruments are used, at what intensity, and for whom.
Project lead, advised by Prof. Hong Zou and Dr. Han Xiao (SWUFE)
Basic Public Service Equalization and Common Prosperity
A city-year panel of 280 cities (2012–2024) asking whether unequal access to public services widens the urban–rural income gap.
Research assistant, advised by Prof. Wei Fan (SWUFE)
Projects
ScrollWeaver & Soulverse
A multi-agent engine that turns novels and worldbuilding documents into interactive, co-creatable worlds. Its personality architecture was written up as PersonaForge, our Findings of ACL 2026 paper.
Team lead · Global 3rd Prize, 2025 Soul AI Agent Competition
Magic Brush
An AI drawing tutor that walks children through a sketch step by step with real-time feedback.
Bronze Prize (Top 30), Doubao × Intel “Make Ideas Real” AI App Challenge