Institution profile

Queen's University

Academic institutionnorthamerica · ca
Official website
Research library384linked papers
Opportunities0open roles
Selected work

Representative Papers

AIDev: Studying AI Coding Agents on GitHub

Feb 09, 2026

This study addresses the scarcity of large-scale, real-world data on AI-powered coding agents and their impact on software engineering practices. To bridge this gap, we present AIDev, the first large-scale dataset capturing the adoption and collaboration patterns of five prominent AI coding agents—Codex, Devin, Copilot, Cursor, and Claude Code—in authentic development environments. Constructed via GitHub API and multi-source tracing techniques, AIDev comprises 932,791 pull requests generated by these agents across 110,000 repositories and 72,000 developers, along with complete interaction contexts—including comments, reviews, commits, and linked issues—for 33,596 high-star projects. This dataset establishes a foundational empirical resource for investigating AI adoption behaviors, human–AI collaboration dynamics, and productivity effects in software development.

5 citationsRead paper

Real-Time Fast Marching Tree for Mobile Robot Motion Planning in Dynamic Environments

May 29, 2023IEEE International Conference on Robotics and Automation

This paper addresses the challenge of real-time motion planning for mobile robots in dynamic environments. We propose the Real-Time Fast Marching Tree (RT-FMT) algorithm, which integrates FMT* with the RT-RRT* framework to establish a hybrid planning paradigm combining sampling-based exploration, incremental tree growth, and real-time rewiring. Our key contribution is a novel dynamic tree rewiring mechanism that enables the tree root to track the robot’s pose in real time while actively avoiding dynamic obstacles. RT-FMT supports coordinated generation of local and global paths, multi-goal path reuse, and prioritized execution of local trajectories. Simulation results demonstrate that, compared to RT-RRT*, RT-FMT reduces path execution cost by 12%–28%, decreases average arrival time by 19%, and significantly enhances both dynamic obstacle avoidance capability and planning timeliness.

3 citationsRead paper

Monitoring Timed Properties (Revisited)

Jun 29, 2022International Conference on Formal Modeling and Analysis of Timed Systems

This work addresses the challenge of online monitoring for real-time systems, where temporal properties are specified in Metric Interval Temporal Logic (MITL) and recognized by Timed Büchi Automata (TBA). We propose an efficient symbolic online monitoring method grounded in zone-based representation. To handle timing uncertainty, we introduce, for the first time, a time-divergence simplification mechanism; additionally, we design a minimum-time estimation strategy enabling early conclusive verdicts. Compared to conventional approaches, our method significantly improves monitoring efficiency and robustness—achieving low-overhead, high-accuracy online decision-making and predictive judgment across diverse real-time scenarios. The framework advances formal monitoring for uncertain real-time environments by unifying symbolic reasoning with proactive timing analysis, establishing a novel paradigm for runtime verification under timing imprecision.

3 citationsRead paper

RedDebate: Safer Responses through Multi-Agent Red Teaming Debates

Jun 04, 2025arXiv.org

Identifying unsafe behaviors in large language models (LLMs) faces challenges including high-cost human evaluation, poor generalizability of single-model evaluators, limited scalability of existing automated methods, and rapid degradation of supervised signals. Method: This paper proposes a fully automated multi-agent debate framework integrating adversarial debate with red-teaming, augmented by a long-term memory module that continuously accumulates and retrieves safety knowledge—enabling self-identification and self-correction without human intervention. Contribution/Results: It is the first work to achieve end-to-end automated synergy between multi-agent debate and red-teaming; its memory-enhanced iterative optimization significantly improves robustness. On the HarmBench benchmark, the debate mechanism alone reduces unsafe responses by 17.7%; incorporating long-term memory further boosts mitigation to over 23.5%.

2 citationsRead paper

A large-scale exploratory study on the proxy pattern in Ethereum

Jun 04, 2024Empirical Software Engineering

This study systematically investigates the deployment landscape, upgrade practices, and security risks of proxy patterns in Ethereum. To address the lack of large-scale empirical analysis, we propose a multidimensional methodology integrating static analysis, dynamic call-graph reconstruction, on-chain transaction tracing, and cluster-based statistical analysis. Applying this approach to over one million smart contracts, we conduct the first comprehensive, network-wide proxy pattern census. Our findings reveal that 37.2% of upgradeable contracts exhibit high-risk proxy configurations, and we identify 127 real-world exploited proxy vulnerabilities. We further characterize three novel risk categories: proxy misuse, logic flaws, and governance failures. As a key contribution, we release ProxyBench—the first open-source benchmark dataset of proxy contracts—designed to support security assessment, formal verification, and tool evaluation for upgradeable smart contracts.

2 citationsRead paper
Recent publications

Latest Papers