Zhenyu Qi

Mr. Zhenyu Qi

orcid

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
18 September 2025 in Automated Software Engineering
Verified
Journal Article
Open access
PDF file
ZQZhenyu Qi
Zhenyu Qi
HLHaotang Li
Haotang Li
HQHao Qin
Hao Qin
KPKebin Peng
Kebin Peng
SHSen He
Sen He
XQXue Qin
Xue Qin
As the Virtual Reality (VR) industry expands, the need for automated GUI testing is growing rapidly. Large Language Models (LLMs), capable of retaining information long-term and analyzing both visual and textual data, are emerging as a potential key to deciphering the complexities of VR’s evolving user interfaces. In this paper, we conduct a case study to investigate the capability of using LLMs, particularly GPT-4o, for field of view (FOV) analysis in VR exploration testing. Specifically, we validate that LLMs can identify test entities in FOVs and that prompt engineering can effectively enhance the accuracy of test entity identification from $$\varvec{41.67\%}$$ to $$\varvec{71.30\%}$$ . Our study also shows that LLMs can accurately describe identified entities’ features with at least a $$\varvec{90\%}$$ accuracy rate. We further find out that the core features that effectively represent an entity are color, placement, and shape. Furthermore, the combination of the three features can especially be used to improve the accuracy of determining identical entities in multiple FOVs with the highest F1-score of $$\varvec{0.70}$$ . Additionally, our study demonstrates that LLMs are capable of scene recognition and spatial understanding in VR with precisely designed structured prompts. Finally, we find that LLMs fail to label the identified test entities, and we discuss potential solutions as future research directions.

27 October 2024
Verified
Conference Paper
Open access
WHWenjie Huang
Wenjie Huang
JCJinfu Chen
Jinfu Chen
HCHuashan Chen
Huashan Chen
ZQZhenyu Qi
Zhenyu Qi
XYXiaojia Yang
Xiaojia Yang
SHSen He
Sen He
Virtual reality (VR) offers unprecedented user experiences across diverse areas. However, few studies have focused on code clone in VR software, though it is a significant factor in the emergence and spread of security vulnerabilities in open-source software. In this work, we conduct the first extensive quantitative empirical analysis of code clone in open-source VR software. We create a dataset of 83 open-source VR projects sourced from GitHub and propose to use multiple metrics to facilitate quantitative analysis. Using the NiCad code clone detection tool, we provide a thorough analysis of code clone in our dataset to tackle five meticulously crafted research questions. Our analysis of the experimental results reveals various insights, enhancing our understanding of VR software and paving the way for future research endeavors.

30 April 2026 in Animals
Verified
Journal Article
Open access
PDF file
HLHaotang Li
Haotang Li
ZQZhenyu Qi
Zhenyu Qi
TATanvir Ahmed
Tanvir Ahmed
SZShungeng Zhang
Shungeng Zhang
SHSen He
Sen He
GLGuoming Li
Guoming Li
High litter moisture content (LMC) in poultry houses is a primary driver of footpad dermatitis, elevated ammonia emissions, and bacterial proliferation. These conditions directly compromise broiler welfare and productivity. Existing monitoring methods, including oven-drying, contact-based sensors, and near-infrared spectroscopy, suffer from invasiveness, single-point limitation, or surface-only measurement. This study investigates ultra-wideband (UWB) impulse radar as a non-contact sensing modality for estimating the LMC of cedar wood shaving bedding under controlled laboratory conditions. A four-phase experimental program was conducted. Phases 1–3 characterized signal–moisture relationships across 0–50% LMC, manure simulant contamination, and bedding structural changes (loose, compacted, caked). Phase 4 tested whether UWB radar can estimate litter LMC when a stationary broiler body obstructs the beam under combined contamination and structural conditions. A progressive feature engineering approach and an SVC-gated mixture-of-experts regression architecture were used to address each confounding factor. Full technical details are provided in the Methods Section. Under clean conditions, the baseline model achieved R2=0.97 and RMSE = 2.48% LMC. Under combined realistic conditions (manure contamination, caked bedding, centered carcass), the full pipeline achieved R2=0.91 and RMSE = 4.53% LMC, with 98.8% bird detection accuracy from the radar signal alone. These laboratory findings suggest that the UWB radar can sense litter moisture through a stationary broiler body. The results support its potential as the sensing core of a non-contact monitoring system for precision poultry farming.

24 July 2025 in arXiv
Verified
Preprint
Open access
PDF file
YXYifan Xu
Yifan Xu
JCJinfu Chen
Jinfu Chen
ZQZhenyu Qi
Zhenyu Qi
HCHuashan Chen
Huashan Chen
JWJunyi Wang
Junyi Wang
SHSen He
Sen He
Virtual Reality (VR) has emerged as a transformative technology across industries, yet its security weaknesses, including vulnerabilities, are underinvestigated. This study investigates 334 VR projects hosted on GitHub, examining 1,681 software security weaknesses to understand: what types of weaknesses are prevalent in VR software; when and how weaknesses are introduced; how long they have survived; and how they have been removed. Due to the limited availability of VR software security weaknesses in public databases (e.g., the National Vulnerability Database or NVD), we prepare the first systematic dataset of VR software security weaknesses by introducing a novel framework to collect such weaknesses from GitHub commit data. Our empirical study on the dataset leads to useful insights, including: (i) VR weaknesses are heavily skewed toward user interface weaknesses, followed by resource-related weaknesses; (ii) VR development tools pose higher security risks than VR applications; (iii) VR security weaknesses are often introduced at the VR software birth time.

2 February 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
MHMuhid Hassan
Muhid Hassan
SGSujan Ghimire
Sujan Ghimire
MMMuntasir Mamun
Muntasir Mamun
HLHaotang Li
Haotang Li
ZQZhenyu Qi
Zhenyu Qi
Tomas CernyTomas Cerny
Tomas Cerny
Context: Deploying microservices at scale is difficult due to volatile demand and strict latency. AI promises to automate scaling, placement, and rollback decisions.Objective: We synthesize evidence on AI techniques for microservice deployment, reported metrics, and system impacts.Method: We followed Kitchenham and PRISMA guidelines. We searched major libraries, dual-screened records, applied quality criteria, and snowballed. Eligible studies were extracted and narratively synthesized.Results: Across studies, AI methods outperform heuristic baselines for prediction, autoscaling, and placement. Reported gains include lower latency, fewer SLA violations, and higher resource efficiency. Reinforcement learning and sequence models are most common, but evaluation designs and baselines vary, limiting comparability. Reproducibility, energy, and cost reporting remain uneven.Conclusions: AI-enabled deployment improves prediction, autoscaling, and placement, yielding lower latency, fewer SLA violations, and higher efficiency. However, heterogeneous workloads, metrics, and baselines limit comparability. Future studies should adopt standardized benchmarks and jointly report tail latency, SLO breaches, energy, and cost across cloud settings to strengthen evidence and guide practice.

30 September 2026 in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Verified
Journal Article
Open access
YFYue Feng
Yue Feng
ZWZi Wang
Zi Wang
HCHuashan Chen
Huashan Chen
ZQZhenyu Qi
Zhenyu Qi
JZJiyue Zhao
Jiyue Zhao
SHSen He
Sen He
When a specific voice rises above the cacophony of a crowded room, how does your perceptual system filter out competing sounds and turn toward that single talker? In this paper, we first conduct a formative study examining this Cocktail Party Effect, uncovering that the closed-loop auditory-perception-response patterns are significant, discriminable, and serve as a unique biometric signature. Based on this observation, we present CocktailAuth, a novel authentication prototype for head-worn devices, implemented and evaluated on commercial VR platforms. Unlike existing methods that typically rely on disruptive explicit inputs, vulnerable static biometrics, or shallow external motion patterns, CocktailAuth exploits the deep-seated, perception-driven head movements naturally exhibited during selective listening. We design four auditory scenarios to elicit these perceptual responses and construct a comprehensive feature framework characterizing both static statistical descriptors and dynamic time series. We evaluate CocktailAuth with a VR prototype in a study of 50 participants. Under a leave-one-session-out evaluation protocol, CocktailAuth achieves an EER of 4.64% and a BAC of 95.36%, and further improves to an EER of 2.07% and a BAC of 97.98% when aggregating multiple samples. By leveraging these auditory perception-driven mechanisms, CocktailAuth offers strong resistance to mimicry attacks and provides a natural, secure, and hardware-friendly solution for ubiquitous wearable devices.