Hi, I’m Huaiyu Jia (贾怀宇), a third-year Ph.D. student in the FinTech Thrust at The Hong Kong University of Science and Technology (Guangzhou), advised by Dr. Shuo Sun. Before that, I received my bachelor’s degree from the School of Information and Software Engineering at University of Electronic Science and Technology of China, where I had the opportunity to work with Prof. Dajiang Chen.
My research interests lie primarily in AI for Blockchain (AI4Blockchain). In particular, I am interested in combining artificial intelligence with blockchain systems to analyze complex on-chain data, understand on-chain behaviors, and enable more intelligent and reliable on-chain decision-making and interactions. I am also broadly interested in AI agents, on-chain finance, and autonomous blockchain systems.
Beyond research, I actively participate in blockchain and AI hackathons. I particularly enjoy turning research ideas into working systems, especially prototypes that connect AI agents with on-chain data, payments, and executable blockchain workflows.
🔥 News
- 2026.10: TxSum accepted at EMNLP 2026 (October 24–29). Paper
- 2026.09: AlphaOpsBench: Benchmarking End-to-End Alpha Strategy Operationalization in Prediction Markets. Paper
- 2026.08: Towards Event-Aware Forecasting in DeFi accepted at KDD 2026 (August 9–13). Paper
- 2026.08: AlphaForgeBench, a benchmark that asks language models to write executable trading strategies, at KDD 2026 (August 9–13). Paper
- 2026.04: Unlocking the Forecasting Economy: A Suite of Datasets for the Full Lifecycle of Prediction Market: [Experiments & Analysis]. Paper / Project
- 2024.11: FS2M published in the International Journal of Web Information Systems. Paper
📝 Publications
AlphaOpsBench: Benchmarking End-to-End Alpha Strategy Operationalization in Prediction Markets
Huaiyu Jia, Mingxuan Zhao, Jincheng Gao, Zifan Peng, Wentao Zhang, Siguang Li, Shuo Sun
- Tests whether a language model can turn a coarse prediction-market idea into an auditable executable program.
- Built on 581 source-preserving strategy records and a lifecycle-scale Polymarket dataset, comparing direct generation with a staged design-then-code protocol.
Unlocking the Forecasting Economy: A Suite of Datasets for the Full Lifecycle of Prediction Market: [Experiments & Analysis]
*Huaiyu Jia, *Luofeng Zhou, Wentao Zhang, Lin William Cong, Siguang Li, Shuo Sun
| Paper | Code | Project |
- A continuously synchronized Polymarket lifecycle dataset from October 2020 onward: 3.29 million markets, 1.90 billion fills, and 21 million oracle events.
- Experiments cover NBA calibration, CPI expectation reconstruction, and resolution prediction. The public interface is PolyMonitor. Huaiyu Jia and Luofeng Zhou contributed equally.
Towards Event-Aware Forecasting in DeFi: Insights from On-chain Automated Market Maker Protocols
Huaiyu Jia, Jiehshun You, Yizhi Luo, Jingyu Liu, Shuo Sun
| Paper | Code |
- 8.9 million annotated on-chain events from Pendle, Uniswap v3, Aave, and Morpho, with transaction type and block-height timestamps.
- An uncertainty-weighted loss that reduces event-time error while keeping event-type accuracy.
AlphaForgeBench: Benchmarking End-to-End Trading Strategy Design with Large Language Models
Wentao Zhang, Mingxuan Zhao, Jincheng Gao, Jieshun You, Huaiyu Jia, Yilei Zhao, Bo An, Shuo Sun
| Paper | Code |
- Evaluates language models as quantitative researchers who write executable alpha factors and trading strategies.
- Separates financial reasoning from brittle step-by-step trade execution, then backtests the generated strategies.
TxSum: User-Centered Ethereum Transaction Understanding with Micro-Level Semantic Grounding
Zifan Peng, Jingyi Zheng, Yule Liu, Huaiyu Jia, Qiming Ye, Jingyu Liu, Xufeng Yang, Mingchen Li, Qingyuan Gong, Xuechao Wang, Xinlei He
- A token-flow explanation task for complex Ethereum transactions, built from user interviews.
- MATEX, a multi-agent framework that checks explanations against raw traces before showing them to users.
From Anonymous Wallets to Behavioral Agents: A Calibrated Policy Test Bed for DAO Governance
Peizhe Li, Huaiyu Jia, Liang Zhang, Shuo Sun
- DAO-SIM builds heterogeneous behavioral agents from anonymous on-chain transfers, then simulates token flows with profile-level super agents and wallet-level child agents.
- On more than 3 million Uniswap transactions, it reproduces the whale effect, path dependence, and the rich-get-richer effect, and tests counterfactual governance policies.
FS2M: Fuzzy Smart IoT Device Pairing Protocol via Speak to Microphone
Huaiyu Jia, Dajiang Chen, Zhidong Xie, Zhiguang Qin
International Journal of Web Information Systems, 2025
- Pairs nearby IoT devices from ambient speech, so devices that hear a similar environment can agree on a key.
- Instantiates the pairing with an elliptic-curve asymmetric fuzzy encapsulation, AFEM-ECC.
🛠 Projects
- PolyMonitor. A live workspace on the Polymarket lifecycle data: prices, fills, oracle events, and macro context. Code
- AlphaForgeBench. Benchmark and backtest harness for language models that synthesize executable trading strategies. Paper
- Deep AMM Events. Code and preprocessed data for event-aware forecasting on four AMM protocols. Poster / Paper
- Polymarket Data Engine. Research-grade infrastructure for reconstructing the Polymarket lifecycle from market creation and fill-level trading to oracle resolution.
- Polymarket Execution Simulator. A market-microstructure simulator for studying execution, liquidity, order-book dynamics, and paper-to-live fidelity on Polymarket.
- Polymarket Replay Lab. A reproducible historical replay environment for fill-level, trade-only, and L2 order-book experiments with execution modeling and oracle-aware settlement.
- Polymarket Smart Money. Evidence-based research on Polymarket wallets, from historical qualification and confirmed trades to verified analysis of market evidence.
🏆 Hackathon Projects
-
POT-402 Gateway — Portaldot Online Mini Hackathon S1.
A Portaldot-native HTTP 402 gateway that turns native POT transfers into verifiable access receipts for pay-per-call APIs. The prototype demonstrates a complete challenge → payment → receipt verification → premium API unlock flow, and exposes the payment proof as a reusable primitive for downstream services. -
OmniYield — LI.FI Vibeathon.
A multi-agent framework for cross-chain yield monitoring and execution that combines on-chain market sensing, risk-adjusted yield selection, and LI.FI MCP/quote routing. The system separates sensing, decision-making, and execution so that cross-chain rebalancing decisions can be explained and previewed before transactions are broadcast. -
SkillTip — Hackathon Galactica — WDK / Tipping Bot track.
An autonomous reward agent that discovers reusable AI skills, evaluates their utility and reuse signals, and settles rewards in Sepolia USDT through Tether WDK. It combines policy-constrained AI judging, a self-custodial treasury, pending-claim handling, and an auditable on-chain payout ledger. -
AutoScholar — Stacks x402 hackathon project.
A research molbot network that quotes premium research workflows, issues machine-readable x402 payment challenges, verifies Stacks testnet settlement, and releases both human-readable research dossiers and structured handoff packets for downstream agents. -
PolySignal — Reactive Network hackathon.
An on-chain whale-alert path for Polymarket. A Polygon trade is relayed to a Sepolia origin contract, evaluated by a Reactive Contract on Lasna, and recorded on a Sepolia destination contract without a centralized keeper. -
OH-ARP — Octant hackathon.
A Uniswap v4 hook that routes stablecoin liquidity sitting outside the active range into an Octant-style vault, returning most of the yield to liquidity providers and sending a share to public goods.
📖 Education
- 2024.09 - Present, Ph.D. in FinTech, The Hong Kong University of Science and Technology (Guangzhou).
- 2020.09 - 2024.06, B.Eng., School of Information and Software Engineering, University of Electronic Science and Technology of China.