Selected work

Representative Papers

Uniform-Loss Automated Market Making for Prediction Markets

Jul 19, 2026

This work addresses a key limitation of existing automated market makers (AMMs) in prediction markets: while they bound the worst-case total loss for subsidy providers, they offer no control over how this loss is distributed across price and time. Building upon the loss-versus-rebalancing (LVR) framework, the paper introduces the first uniform-loss AMM, whose instantaneous LVR is proportional to the pool value and independent of the current price. By establishing a bidirectional correspondence between winning martingales and pricing functions, and integrating dynamic liquidity management, the proposed mechanism enables on-demand shaping of the expected cumulative loss trajectory. The authors theoretically prove the existence of uniform-LVR pricing functions under general winning martingales and validate the approach through representative examples, offering a novel tool for controlling loss distribution in AMM design.

0 citationsRead paper

Volatility in Prediction Markets: A Structural Approach

Jul 09, 2026

Traditional ARCH/GARCH models struggle to capture the distinctive volatility dynamics in binary prediction markets, where prices represent bounded probabilities, exhibit a fixed expiration date, and yield binary payoffs. This work proposes the first volatility model that integrates economic structural mechanisms: it combines a Wright–Fisher process to characterize the convergence of uncertainty as contracts approach expiry and a Glosten–Milgrom order-flow mechanism to capture volatility induced by informed trading. The resulting structural framework not only yields interpretable volatility measures but also reveals fundamental differences in information arrival patterns between economic and sports-related contracts. Empirical analysis using large-scale Kalshi data demonstrates that the proposed structural variables significantly outperform standard GARCH specifications, with a hybrid structural-GARCH model achieving the best predictive performance and exhibiting strong cross-category generalization capabilities.

0 citationsRead paper

EVMbench: Evaluating AI Agents on Smart Contract Security

Mar 05, 2026

This work addresses the critical risk of substantial blockchain asset losses due to smart contract vulnerabilities by introducing EVMbench, the first end-to-end evaluation benchmark tailored for AI agents. Built upon 117 real-world vulnerabilities and a local Ethereum execution environment, the framework programmatically assesses AI capabilities across the full spectrum of vulnerability detection, repair, and exploitation. Integrating the Ethereum Virtual Machine (EVM), a curated vulnerability dataset, on-chain state validation, and AI-driven code generation models, EVMbench demonstrates empirically that state-of-the-art AI agents can autonomously discover and exploit vulnerabilities in realistic settings. The entire suite—including tasks, code, and tooling—is open-sourced to foster ongoing security evaluation and research in this domain.

0 citationsRead paper

Measuring CEX-DEX Extracted Value and Searcher Profitability: The Darkest of the MEV Dark Forest

Jul 17, 2025

This paper addresses the challenge of quantifying centralized exchange–decentralized exchange (CEX-DEX) arbitrage value extraction on Ethereum, for which off-chain CEX trading data is typically required. We propose the first empirical framework that relies solely on on-chain data, eliminating dependence on external CEX feeds. By enhancing heuristic identification techniques, we conduct a large-scale analysis of 19 months of Ethereum transaction data, identifying 7.2 million arbitrage transactions. We estimate that 19 prominent searchers collectively extracted $233.8 million in value, with the top three accounting for 75% of the total. Our study provides the first empirical evidence of extreme concentration in the arbitrage market and demonstrates deep coordination between searchers and block builders—significantly influencing transaction ordering and profit distribution. These findings offer critical insights into the economic structure of maximal extractable value (MEV), highlight systemic centralization risks, and establish a methodological foundation for governance design in decentralized protocols.

0 citationsRead paper

Does Your Blockchain Need Multidimensional Transaction Fees?

Apr 21, 2025

Blockchain systems employing unidimensional transaction fees (e.g., Ethereum’s gas mechanism) suffer from imbalanced utilization of multidimensional resources—computation, storage, and bandwidth—leading to throughput bottlenecks. Method: We propose the first modeling and quantitative analysis framework for transaction pricing under multidimensional resource constraints. We formulate multidimensional fee optimization as a zero-sum game and prove it is NP-hard to achieve a k-dimensional α-approximation. Under resource capacity scaling, we establish a rigorous feasibility comparison criterion and derive, for the first time, theoretical bounds on the approximation ratio of unidimensional gas pricing relative to true multidimensional resource consumption. Contribution/Results: Our analysis shows that multidimensional pricing can significantly improve system throughput—by up to several-fold under typical workloads—and that this gain is precisely characterized by the α parameter. The framework provides protocol designers with theoretically grounded, quantitative tools for navigating the complexity–performance trade-off in fee-market design.

0 citationsRead paper

Recent publications

Latest Papers

Uniform-Loss Automated Market Making for Prediction Markets

Jul 19, 2026

This work addresses a key limitation of existing automated market makers (AMMs) in prediction markets: while they bound the worst-case total loss for subsidy providers, they offer no control over how this loss is distributed across price and time. Building upon the loss-versus-rebalancing (LVR) framework, the paper introduces the first uniform-loss AMM, whose instantaneous LVR is proportional to the pool value and independent of the current price. By establishing a bidirectional correspondence between winning martingales and pricing functions, and integrating dynamic liquidity management, the proposed mechanism enables on-demand shaping of the expected cumulative loss trajectory. The authors theoretically prove the existence of uniform-LVR pricing functions under general winning martingales and validate the approach through representative examples, offering a novel tool for controlling loss distribution in AMM design.

0 citationsRead paper

Volatility in Prediction Markets: A Structural Approach

Jul 09, 2026

Traditional ARCH/GARCH models struggle to capture the distinctive volatility dynamics in binary prediction markets, where prices represent bounded probabilities, exhibit a fixed expiration date, and yield binary payoffs. This work proposes the first volatility model that integrates economic structural mechanisms: it combines a Wright–Fisher process to characterize the convergence of uncertainty as contracts approach expiry and a Glosten–Milgrom order-flow mechanism to capture volatility induced by informed trading. The resulting structural framework not only yields interpretable volatility measures but also reveals fundamental differences in information arrival patterns between economic and sports-related contracts. Empirical analysis using large-scale Kalshi data demonstrates that the proposed structural variables significantly outperform standard GARCH specifications, with a hybrid structural-GARCH model achieving the best predictive performance and exhibiting strong cross-category generalization capabilities.

0 citationsRead paper

EVMbench: Evaluating AI Agents on Smart Contract Security

Mar 05, 2026

This work addresses the critical risk of substantial blockchain asset losses due to smart contract vulnerabilities by introducing EVMbench, the first end-to-end evaluation benchmark tailored for AI agents. Built upon 117 real-world vulnerabilities and a local Ethereum execution environment, the framework programmatically assesses AI capabilities across the full spectrum of vulnerability detection, repair, and exploitation. Integrating the Ethereum Virtual Machine (EVM), a curated vulnerability dataset, on-chain state validation, and AI-driven code generation models, EVMbench demonstrates empirically that state-of-the-art AI agents can autonomously discover and exploit vulnerabilities in realistic settings. The entire suite—including tasks, code, and tooling—is open-sourced to foster ongoing security evaluation and research in this domain.

0 citationsRead paper

Measuring CEX-DEX Extracted Value and Searcher Profitability: The Darkest of the MEV Dark Forest

Jul 17, 2025

This paper addresses the challenge of quantifying centralized exchange–decentralized exchange (CEX-DEX) arbitrage value extraction on Ethereum, for which off-chain CEX trading data is typically required. We propose the first empirical framework that relies solely on on-chain data, eliminating dependence on external CEX feeds. By enhancing heuristic identification techniques, we conduct a large-scale analysis of 19 months of Ethereum transaction data, identifying 7.2 million arbitrage transactions. We estimate that 19 prominent searchers collectively extracted $233.8 million in value, with the top three accounting for 75% of the total. Our study provides the first empirical evidence of extreme concentration in the arbitrage market and demonstrates deep coordination between searchers and block builders—significantly influencing transaction ordering and profit distribution. These findings offer critical insights into the economic structure of maximal extractable value (MEV), highlight systemic centralization risks, and establish a methodological foundation for governance design in decentralized protocols.

0 citationsRead paper

Does Your Blockchain Need Multidimensional Transaction Fees?

Apr 21, 2025

Blockchain systems employing unidimensional transaction fees (e.g., Ethereum’s gas mechanism) suffer from imbalanced utilization of multidimensional resources—computation, storage, and bandwidth—leading to throughput bottlenecks. Method: We propose the first modeling and quantitative analysis framework for transaction pricing under multidimensional resource constraints. We formulate multidimensional fee optimization as a zero-sum game and prove it is NP-hard to achieve a k-dimensional α-approximation. Under resource capacity scaling, we establish a rigorous feasibility comparison criterion and derive, for the first time, theoretical bounds on the approximation ratio of unidimensional gas pricing relative to true multidimensional resource consumption. Contribution/Results: Our analysis shows that multidimensional pricing can significantly improve system throughput—by up to several-fold under typical workloads—and that this gain is precisely characterized by the α parameter. The framework provides protocol designers with theoretically grounded, quantitative tools for navigating the complexity–performance trade-off in fee-market design.

0 citationsRead paper