Asian Institute of Digital Finance
PhD Research & Publications
Meet our PhD researchers and explore their work across artificial intelligence, financial technology, sustainable finance, blockchain, risk and quantitative methods.
2021 Intake
4 researchers
Tan Xue Wen
Supervisor: Stanley Kok
Research interests: Explainable AI, large language models, AI for social good and computational linguistics
Li Shuping
Research interests: Default prediction, fraud detection and natural language processing
Lu Chung I
Supervisor: Julian Sester
Research interests: Machine learning, deep learning, reinforcement learning and portfolio optimisation
Zhou Zhou
Supervisor: Allaudeen Hameed
Research interests: Empirical asset pricing, FinTech, climate and sustainable finance
2022 Intake
6 researchers
Anne Elizabeth Lee Jin
Research interests: Green finance, sustainable finance and asset pricing
Kenneth See Dehui
Supervisor: Xiaofan Li
Research interests: CBDCs, DeFi, systemic financial risk and strategic incentives
Luo Bingqiao
Supervisor: Bingsheng He
Research interests: Cryptocurrency, graph analytics, data mining and fraud/risk analysis
Ong Kang Leng Fabian
Supervisor: Johan Sulaeman
Research interests: ESG, Web3, blockchain and digital communities
Shimin Zhang
Supervisor: Daniel Rabetti
Research interests: Blockchain, Web3, NFTs, machine learning, NLP and causal inference
Xie Haoyu
Supervisor: Julian Sester
Research interests: Stochastic control, mean field games and machine learning
2023 Intake
5 researchers
Gao Wenhan
Supervisor: Ying Chen
Research interests: Text mining, financial forecasting, generative language models, thematic modelling and explainable machine learning
Dennis Thumm
Supervisors: Ying Chen, Johan Sulaeman and Ling Feng
Research interests: Causality, complexity, differentiable computing, language models and probabilistic machine learning
Keane Ong Wei Yang
Supervisors: Erik Cambria, Gianmarco Mengaldo and Paul Pu Liang
Research interests: Socially intelligent AI, multimodal foundation models, explainable AI and financial narratives
Woon Jiahui
Supervisor: Johan Sulaeman
Research interests: Data science, machine learning and risk management
Zhang Dawei
Supervisor: Johan Sulaeman
Research interests: InsurTech, decentralised finance, blockchain, big data and machine learning
2024 Intake
4 researchers
Ahmed Syalabi Seet
Supervisor: Johan Sulaeman
Research interests: AI for sustainability, sustainable AI and semi-supervised learning
Zhao Weibo
Research interests: Cybersecurity, artificial intelligence and financial forecasting
Zhou Yongqi
Supervisor: Ying Chen
Research interests: Big data, machine learning, natural language processing and financial forecasting
Zhengxi Qian
Supervisor: Johan Sulaeman
Research interests: Machine learning, asset pricing, behavioural economics and finance, and policy analysis
2025 Intake
4 researchers
Boonthicha Saejia
Research interests: Artificial intelligence, blockchain, cybersecurity, large language modelling and RegTech
Nguyen Thi Hoa
Supervisor: Ying Chen
Research interests: Time-series forecasting, reinforcement learning, machine learning in finance and complex systems
Yuncong Liu
Research interests: AI-driven financial research, reinforcement learning for finance, quantitative trading and alpha-signal mining
Qi Yihan
Supervisor: Julian Sester
Research interests: Machine learning, natural language processing, asset pricing and dynamic risk management
Student Publications
| Student | Publication | Journal / Conference | Date |
|---|---|---|---|
| Keane Ong | OmniSapiens: A Foundation Model for Social Behavior Processing via Heterogeneity-Aware Relative Policy Optimization Summary: OmniSapiens uses a new reinforcement-learning method to balance uneven training signals across heterogeneous behavioural data. It improves performance and generalisation across psychological and social-behaviour tasks while producing more consistent, interpretable reasoning. | ICML 2026 | Jul 2026 |
| Keane Ong | Human Behavior Atlas: Benchmarking Unified Psychological and Social Behavior Understanding Summary: Human Behavior Atlas unifies more than 100,000 text, audio and visual samples covering affect, cognition, pathology and social processes. Models trained on the benchmark outperform existing multimodal LLMs and transfer more effectively to previously unseen behavioural datasets. | ICLR 2026 | Apr 2026 |
| Luo Bingqiao | Multi-Chain Graphs of Graphs: A New Approach to Analyzing Blockchain Datasets Summary: The work releases cross-chain datasets that represent transactions within each token as local graphs and relationships between tokens as a global graph. This hierarchy enables research on link prediction, anomaly detection and token classification. | NeurIPS | Dec 2024 |
| Luo Bingqiao | ORDER: Optimal Routing with Path Indexing in Exchange Graph Summary: ORDER addresses optimal routing across fragmented financial exchange networks. Its hierarchical bucket path index, lazy computation and adaptive controller identify high-quality routes under changing exchange rates, liquidity conditions and capacity constraints. | SIGMOD | 2026 |
| Keane Ong | Towards Robust ESG Analysis Against Greenwashing Risks: Aspect-Action Analysis with Cross-Category Generalization Summary: The A3CG dataset links sustainability claims to concrete actions and tests models across ESG categories. This supports more transparent analysis that is less vulnerable to vague, exaggerated or selectively presented corporate sustainability reporting. | ACL Main Conference | Jul 2025 |
| Keane Ong | Deriving Strategic Market Insights with Large Language Models: A Benchmark for Forward Counterfactual Generation Summary: The paper introduces Fin-Force, a benchmark that evaluates how well LLMs generate plausible future market scenarios from financial news. Its experiments reveal current limitations and establish a foundation for scalable, decision-oriented forward counterfactual analysis. | EMNLP 2025 (Oral) | Oct 2025 |
| Luo Bingqiao | RICH: Real-time Identification of Negative Cycles for High-efficiency Arbitrage Summary: RICH combines colour-coding, dynamic programming, graph encoding and reduction to find profitable negative cycles without exhaustive traversal. On real-world cryptocurrency and foreign-exchange data, it is up to 32.69 times faster than existing methods. | PVLDB | 2025 |
| Keane Ong | Explainable Natural Language Processing for Corporate Sustainability Analysis Summary: This study outlines how lexical, semantic and syntactic NLP, combined with interpretable and faithful explanations, can help analysts process complex sustainability disclosures while reducing data ambiguity, limited analyst capacity and subjective bias. | Information Fusion | Jan 2025 |
| Gao Wenhan | Innovation Value Discrepancy and Its Role in Shaping Firms' Short-Term Gains and Sustainable Growth Summary: Using 2.5 million patents from 7,506 US-listed firms, the study shows that scientifically strong but initially under-recognised innovations are associated with sustained long-term growth, while economically prominent but scientifically weaker innovations tend to deliver less persistent gains. | International Review of Financial Analysis | Sep 2026 |
| Qi Yihan | Empirical Analysis of the Model-Free Valuation Approach: Hedging Gaps, Conservatism, and Trading Opportunities Summary: Historical option data show that model-free super-hedging is only marginally more conservative than the industry-standard Heston model. The measured hedging gap also supports a profitable trading strategy with explicit pathwise downside protection. | Quantitative Finance (forthcoming) | Forthcoming, 2026 |
| Tan Xue Wen & Kenneth See | ScamGPT-J: Inside the Scammer's Mind, A Generative AI-Based Approach Toward Combating Messaging Scams Summary: ScamGPT-J replicates scammer tactics and generates likely responses in real time. By comparing incoming messages with simulated scam dialogue, it helps users recognise suspicious interactions instead of relying solely on conventional detection and blocking. | ICIS | Dec 2024 |
| Kenneth See | PSSimPy: A Design Science Approach to Constructing and Implementing a Large-Value Payment System Simulator Summary: PSSimPy is a Python simulator for analysing the dynamic behaviour of large-value payment systems. Its design emphasises modularity, accessibility and data privacy, making it adaptable for testing blockchain and distributed-ledger innovations. | ICIS | Dec 2024 |
| Kenneth See | The Satoshi Laundromat: A Review on the Money Laundering Open Door of Bitcoin Mixers Summary: This review evaluates methods for tracing mixed Bitcoin transactions and identifies a significant AML gap: decentralised mixers using off-chain coordination and randomised fees can make tainted funds especially difficult to follow. | Journal of Financial Crime | 2024 |
| Kenneth See | Simulating Blockchain Applications in Large-Value Payment Systems through Agent-Based Modeling Summary: An interactive agent-based simulator lets policymakers test blockchain designs for large-value payment systems, including atomic delivery-versus-payment and tokenised incoming payments. The scenarios expose trade-offs among settlement delays, liquidity efficiency and credit use. | AAMAS Demo Track | Jun 2025 |
| Tan Xue Wen | The Shape of Reasoning: Topological Analysis of Reasoning Traces in Large Language Models Summary: The paper applies topological data analysis to the geometry of LLM reasoning traces. A compact set of topological features predicts reasoning quality more effectively than conventional graph metrics, offering a scalable signal for automated evaluation and reinforcement learning. | ICML 2026 Workshop | 2026 |
| Yuncong Liu | FinGPT: Enhancing Sentiment-Based Stock Movement Prediction with Dissemination-Aware and Context-Enriched LLMs Summary: FinGPT enriches financial-news sentiment analysis with dissemination breadth, contextual data and explicit instructions. The resulting instruction-tuned model improves short-term stock-movement prediction accuracy by 8% over existing approaches. | AAAI 2025 Workshop | Dec 2024 |
| Zhou Zhou | The Dark Side of Geographically Dispersed Information: Evidence from Lockdown of Subsidiaries Summary: Using pandemic lockdowns as a natural experiment, the study finds that limiting analysts' access to nearby subsidiaries improves forecast accuracy, price informativeness and liquidity, suggesting that salient local signals can sometimes distort firm-level information processing. | Journal of Accounting and Economics (R&R) | 2024 |