About

Wenqi Zhou, Ph.D., is a tenured Associate Professor of Information Systems and Technology at the Palumbo Donahue School of Business at Duquesne University. He holds the PwC Alumni Faculty Fellowship in IST and was previously awarded the Inaugural David Warco Faculty Fellowship (2021-2024) in recognition of his exemplary scholarship and contributions to higher education teaching. His research interests include studying AI, online user-generated contents, Healthcare Information Systems and economics, information disclosure and transparency, and tech ethics issues related to e-commerce and social media platforms. Zhou is affiliated with the Center for Computational Analysis of Social and Organizational Systems (CASOS) and IDearS Center at Carnegie Mellon University, as well as the Grefenstette Center at Duquesne University. He received his Ph.D. from George Washington University School of Business. Zhou has been actively involved in academic service, serving as Conference Chair for the Workshop on e-Business in 2025 and 2026, an Associate Editor of Information and Management, and an Area Editor of AI and Management Science of Computational and Mathematical Organizational Theory. He is also the founder and faculty director of the IST mentorship program at Duquesne, which connects students with senior managers and CIOs in the region. Zhou teaches undergraduate courses in introductory information systems and business intelligence, as well as graduate courses in behavioral analytics and AI for business. His work has been published in numerous reputable journals, and he has received multiple awards and grants recognizing his research excellence and teaching contributions.

Research topics

• Artificial Intelligence
• Natural Language Processing
• Machine Learning
• Computer Science
• Data Mining
• Mathematics
• Engineering
• Geography

Selected publications

• Impact of pressure on the transmission performance of underwater wireless power transfer systems

  Results in Engineering · 2026-02-04

  articleOpen access

  This paper systematically investigates the effects of pressure on the performance of underwater wireless power transfer (WPT) systems by modeling the changes in coupler parameters and their impact on system power and efficiency. A comprehensive theoretical framework is developed to establish quantitative relationships between pressure variations and the coupler’s relative permeability, self-inductance, and mutual inductance, as well as system performance metrics. To ensure precision, four key shape factors are introduced, and an interior-point optimization-based fitting method is proposed for parameter determination. The results indicate that with every doubling of pressure, the mutual inductance and self-inductance of the coupler decrease by approximately 0.76% and 0.43%, respectively, while system power increases by 1.6% and efficiency exhibits a marginal decrease of 0.02%. The developed model demonstrates high reliability and precision, achieving coupler parameter errors below 1% and efficiency errors under 7%, thereby providing a solid basis for the evaluation and optimization of underwater WPT systems under varying pressure conditions.

  PublisherDOI

• DebugLM: Learning Traceable Training Data Provenance for LLMs

  arXiv (Cornell University) · 2026-03-18

  preprintOpen access

  Large language models (LLMs) are trained through multi-stage pipelines over heterogeneous data sources, yet developers lack a principled way to pinpoint the specific data responsible for an observed behavior. This lack of observability reduces debugging to reactive patching and makes failures prone to recur under distribution shift or subsequent model updates. To address this limitation, we propose DebugLM, a framework that equips LLMs with built-in data provenance, enabling them to explicitly trace the origins of their behaviors to specific training data sources. Specifically, the model learns to associate its responses with unique provenance tags that indicate the responsible dataset, empowering developers to precisely identify where undesirable behaviors are learned. Building on this capability, DebugLM further supports targeted test-time remediation, enabling developers to selectively trigger targeted refusal for specified data sources without retraining or modifying model parameters. Experiments demonstrate that DebugLM provides accurate behavior tracing in multi-stage training pipelines and effective test-time remediation while preserving the general utility of the model.

  PublisherDOI

• An Online Monitoring Method for Measuring Ice Thickness Based on Rotating Ring Electrode Array

  IEEE Transactions on Instrumentation and Measurement · 2026-01-01

  article1st authorCorresponding

  Severe icing can compromise the safe and stable operation of power grids, making the monitoring of ice on transmission and distribution equipment a prerequisite for effective prevention and mitigation of grid-related ice disasters. This study presents a novel approach for measuring ice thickness using a rotating ring electrode array based on the capacitive effect of ice. Theoretical models and finite element simulations demonstrated that optimizing the electrode structure specifically increasing the electrode coverage, outer diameter, and the number of electrode units significantly enhanced signal strength and sensitivity, while increasing electrode height improved penetration depth. Electrode thickness was found to have a negligible effect. Field experiments under natural conditions confirmed the strong, nonlinear relationship between measured capacitance and ice thickness, and also validated the effectiveness of the device. By integrating a multilayer perceptron (MLP) model with key measurement parameters, high-precision prediction of ice thickness was achieved with the regression model demonstrating a coefficient of determination up to 0.92. The developed device and method enable real-time, remote, and non-destructive monitoring of ice accretion on transmission lines. These advances not only overcome the limitations of traditional monitoring technologies but also provide essential technical support for enhancing the resilience and operational safety of power grids in regions subject to severe icing.

  PublisherDOI

• Printing Technology-Enabled Intelligent Packaging for Food Quality and Safety Monitoring: A Review

  Food and Bioprocess Technology · 2026-02-27

  article
  PublisherDOI

• DebugLM: Learning Traceable Training Data Provenance for LLMs

  ArXiv.org · 2026-03-18

  articleOpen access

  Large language models (LLMs) are trained through multi-stage pipelines over heterogeneous data sources, yet developers lack a principled way to pinpoint the specific data responsible for an observed behavior. This lack of observability reduces debugging to reactive patching and makes failures prone to recur under distribution shift or subsequent model updates. To address this limitation, we propose DebugLM, a framework that equips LLMs with built-in data provenance, enabling them to explicitly trace the origins of their behaviors to specific training data sources. Specifically, the model learns to associate its responses with unique provenance tags that indicate the responsible dataset, empowering developers to precisely identify where undesirable behaviors are learned. Building on this capability, DebugLM further supports targeted test-time remediation, enabling developers to selectively trigger targeted refusal for specified data sources without retraining or modifying model parameters. Experiments demonstrate that DebugLM provides accurate behavior tracing in multi-stage training pipelines and effective test-time remediation while preserving the general utility of the model.

  PublisherOA PDF

• Research on aerodynamic parameter prediction of ice-covered conductors based on multi-model fusion

  Journal of Physics Conference Series · 2026-05-01

  articleOpen access

  Abstract Conductor galloping, induced by asymmetric ice accretion, remains a critical threat to the structural integrity and operational stability of power grids. Quantifying the aerodynamic characteristics of ice-covered conductors is essential for elucidating galloping mechanisms, assessing risk, and optimizing mitigation strategies. Traditional approaches, wind tunnel testing, and Computational Fluid Dynamics (CFD) provide high-fidelity aerodynamic parameters, but they are frequently constrained by prohibitive costs, extended experimental cycles, and intensive computational requirements, hindering real-time application. To address these limitations, this study leverages data-driven machine learning to model the nonlinear, high-dimensional aerodynamic behaviour of iced conductors. We developed an efficient ML-based prediction platform designed to provide rapid aerodynamic parameter estimation. This framework provides a robust technical foundation for early-warning systems and intelligent maintenance of transmission-line galloping risks.

  PublisherOA PDFDOI

• Experimental Analysis of Droplet Impact on Superhydrophobic Surfaces under the Influence of Electric Field

  2025-10-17

  article

  This work investigates how externally applied electric fields regulate droplet impact dynamics on superhydrophobic surfaces, motivated by the need to mitigate icing in power and aerospace systems. We develop and calibrate an experimental platform that controls droplet size, impact velocity, substrate temperature, ambient humidity, and electric field strength, and we use high-speed imaging to classify impact outcomes. We map regime transitions from rebound-dominated to breakup- and emission-dominated morphologies with increasing field strength, quantify changes in spreading and retraction metrics, and identify empirical thresholds for secondary breakup. We further show that electric fields influence charge redistribution and the balance between inertia and capillarity, expanding the accessible parameter space for morphology control and delaying icing onset under subcooling. These findings provide actionable guidance for designing icephobic superhydrophobic surfaces and for electrically assisted icing mitigation.

  PublisherDOI

• Code Execution as Grounded Supervision for LLM Reasoning

  ArXiv.org · 2025-06-12

  preprintOpen access

  Training large language models (LLMs) with chain-of-thought (CoT) supervision has proven effective for enhancing their reasoning abilities. However, obtaining reliable and accurate reasoning supervision remains a significant challenge. We propose a scalable method for generating a high-quality CoT supervision dataset by leveraging the determinism of program execution. Unlike existing reasoning dataset generation methods that rely on costly human annotations or error-prone LLM-generated CoT, our approach extracts verifiable, step-by-step reasoning traces from code execution and transforms them into a natural language CoT reasoning. Experiments on reasoning benchmarks across various domains show that our method effectively equips LLMs with transferable reasoning abilities across diverse tasks. Furthermore, the ablation studies validate that our method produces highly accurate reasoning data and reduces overall token length during inference by reducing meaningless repetition and overthinking.

  PublisherOA PDFDOI

• Federated Learning via TEE-Based Dual-Branch Architecture and Interaction-Aware Pruning

  Lecture notes in computer science · 2025-11-15

  book-chapter1st authorCorresponding
  PublisherDOI

• Numerical study on stretch extinction mechanism of NH3/H2/air laminar counterflow premixed flames

  Fuel Processing Technology · 2025-08-27 · 2 citations

  articleOpen access1st author

  Ammonia and hydrogen are two most promising carbon-free fuels emerging in recent years, and their co-combustion is well recognized as an efficient approach to solve the issues associated with ammonia's poor combustion behaviors. This study emphasizes fundamentally the combustion properties, particularly the stretch-induced extinction limit as well as the underlying physical mechanism of the NH 3 /H 2 /air laminar counterflow premixed flames by carrying out simulations with detailed fuel chemistry and transport models. The results demonstrate that hydrogen addition significantly extends the ammonia flame extinction strain rate, with the equivalence ratio corresponding to the maximum extinction strain rate shifting toward leaner stoichiometry as hydrogen addition increases. The combination of thermal, chemical, and transport effects of hydrogen enhances the NH 3 /H 2 premixed flame stability. More specifically, the contribution of thermal effect to extinction prevails under the fuel-rich condition, decreasing with the decrement of equivalence ratio. The effective Lewis number of the premixture is responsible for the distinct thermal effect response behaviors in fuel-lean condition compared with the stoichiometric and rich conditions. By comparatively analyzing chemical kinetics and flame structure between the strongly-stable and near-extinction flames, it elucidates the governing chemical pathways and critical radical interactions responsible for the NH 3 /H 2 stretched premixed flame extinction. • H 2 preferential diffusion shifts peak extinction strain rate to leaner stoichiometry. • Preferential diffusion/stretch interplay causes nonlinear T max - κ response at lean ϕ . • H 2 thermal effect grows with equivalence ratio, dominating at rich condition. • H/OH branching reactions with disparate temperature sensitivities govern extinction. • Lean flame with higher temperature is more stable than stoichiometric/rich flames.

  PublisherDOI

Frequent coauthors

• Muhao Chen

  University of California, Davis

  103 shared

• Fei Wang

  20 shared

• Zekun Li

  18 shared

• Yao‐Yi Chiang

  17 shared

• Xiang Ren

  15 shared

• Bryan Hooi

  13 shared

• Hoifung Poon

  11 shared

• David Held

  11 shared

Education

• Ph.D., Information Systems & Technology

  Duquesne University

• M.S., Information Systems & Technology

  Duquesne University

• B.S., Information Systems & Technology

  Duquesne University

Awards & honors

• Inaugural David Warco Faculty Fellowship (2021-2024)
• Duquesne School of Business Dean's Award for Excellent in Re…
• Paluse Faculty Research Grant (2022-2023)
• Grefenstette Center for Ethics in Science, Technology, and L…
• Best Paper Award at Workshop on e-Business (2018)