I am an undergraduate at HKUST (RMBI + MATH + AI), currently on exchange at Stanford as part of the Interdisciplinary Honors Program. My research sits at the intersection of mathematical finance, stochastic analysis, and machine learning — with a focus on building rigorous theoretical foundations for world models applied to quantitative finance.

"Factor models learn correlations. World models learn causation. The difference is everything."

I am the founder of Alpha Flow, a research initiative developing mathematical frameworks that unify stochastic differential equations, mean-field game theory, and generative modeling for financial applications. My background traces from Physics Olympiad training through the Tsinghua Physics Talent Program to quantitative research at HKUST and Stanford.

Research Interests

World Models Market dynamics, latent state prediction, E-Game-C framework
Mathematical Finance SDEs, Lévy processes, stochastic control, mean-field games
Generative Models Diffusion models, flow matching, transformers, VAEs
Reinforcement Learning Model-based RL, HJB equations, PPO / DDPG
LLM Evaluation Hallucination detection, financial NLP, SEC filing analysis
Quantitative Finance Alpha mining, limit order books, backtesting metrics

Papers & Projects

Mathematical Framework for World Models in Quantitative Finance
HongJin HE · arXiv preprint, July 2026 Preprint
Introduces the E-Game-C architecture, replacing recurrent networks with mean-field game equilibrium operators for market dynamics modeling. Presents seven original theorems unifying stochastic decomposition with world model theory. Connects controllability to alpha generation via a novel SDE framework.
LLM Hallucination in Financial Disclosures
HongJin HE · Under peer review, 2026 Under Review
Analyzed 5,552 U.S. firms' SEC filings across 3,600+ model responses. Identifies D-type hallucination — where models fabricate confident answers when no source information exists. GPT-4o showed a 54.8% improvement over GPT-3.5 in Prompt Score under controlled prompting conditions.
Diffusion-Flow Alpha Mining
HongJin HE · 2025–2026 Project
Diffusion-flow alpha factor generation achieving 3× improvement over random baseline (Mean IC: 0.148 vs ~0.05). Generates low-correlation alpha distributions suitable for portfolio construction. PyTorch implementation, <5 min training on Colab T4.
Stanford AI Labs Ecosystem
HongJin HE · 2026 Project
Nine interactive visualizations mapping Stanford's AI research landscape across 2018–2026 — collaboration networks, research trend clustering, and lab ecosystem structure. Built during Stanford CS exchange.

Education

2025–2026
Stanford University
Interdisciplinary Honors Program · CS Exchange
2024–2029
HKUST
B.Sc. · Risk Management & Business Intelligence, Mathematics, AI
2023
Tsinghua University
Physics Talent Program · National Physics Olympiad rank 92