Digital Commons Network^™

Reducing Class-Wise Performance Disparity Via Margin Regularization, Beier Zhu, Kesen Zhao, Jiequan Cui, Qianru Sun, Yuan Zhou, Xun Yang, Hanwang Zhang

Research Collection School Of Computing and Information Systems

Deep neural networks often exhibit substantial disparities in class-wise accuracy, even when trained on class-balanced data—posing concerns for reliable deployment. While prior efforts have explored empirical remedies, a theoretical understanding of such performance disparities in classification remains limited. In this work, we present Margin Regularization for performance disparity Reduction (MR2 ), a theoretically principled regularization for classification by dynamically adjusting margins in both the logit and representation spaces. Our analysis establishes a margin-based, class-sensitive generalization bound that reveals how per-class feature variability contributes to error, motivating the use of larger margins for “hard” classes. Guided by this insight, MR2 optimizes …

Llmqua: Practical Backdoor Injection On Large Language Model Quantization, Xiangxiang Chen, Peixin Zhang, Jun Sun, Jin Song Dong, Wenhai Wang, Jingyi Wang

Research Collection School Of Computing and Information Systems

Quantization is widely used to enable local deployment of large language models (LLMs) on resource-constrained devices. Recent work (e.g., QuRA) shows quantization can be exploited via rounding manipulation to implant backdoors. However, such an attack has been evaluated only on small models and does not directly apply to LLMs due to three key constraints: (1) limited poisoning data from small, task-agnostic calibration sets; (2) layer-wise quantization restricting adversarial access to global representations; and (3) lack of gradient access in quantization pipelines, blocking gradient-based attacks.We propose LLMQuA, a practical quantization-phase backdoor attack tailored to the LLM setting. LLMQuA (i) injects backdoors …

Prompting Frameworks For Large Language Models: A Survey, Xiaoxia Liu, Jingyi Wang, Jun Sun, Xiaohan Yuan, Guoliang Dong, Peng Di, Wenhai Wang, Dongxia Wang

Research Collection School Of Computing and Information Systems

Since the launch of ChatGPT, a powerful AI Chatbot developed by OpenAI, large language models (LLMs) have made significant advancements in both academia and industry, bringing about a fundamental engineering paradigm shift in many areas. While LLMs are powerful, it is also crucial to best use their power where “prompt” plays a core role. However, the booming LLMs themselves, including excellent APIs like ChatGPT, have several inherent limitations: (1) temporal lag of training data, and (2) the lack of physical capabilities to perform external actions. Recently, we have observed the trend of utilizing prompt-based tools to better utilize the power …

Who You Explain To Matters: Learning By Explaining To Conversational Agents With Different Pedagogical Roles, Zhengtao Xu, Junti Zhang, Anthony Tang, Yi-Chieh Lee

Research Collection School Of Computing and Information Systems

Conversational agents are increasingly used in education for learning support. An application is “learning by explaining”, where learners explain their understanding to an agent. However, existing research focuses on single roles, leaving it unclear how different pedagogical roles influence learners’ interaction patterns, learning outcomes and experiences. We conducted a between-subjects study (N=96) comparing agents with three pedagogical roles (Tutee, Peer, Challenger) and a control condition while learning an economics concept. We found that different pedagogical roles shaped learning dynamics, including interaction patterns and experiences. Specifically, the Tutee agent elicited the most cognitive investment but led to high pressure. The Peer …

Bridging Bug Localization And Issue Fixing: A Hierarchical Localization Framework Leveraging Large Language Models, Jianming Chang, Xin Zhou, Lulu Wang, David Lo, Bixin Li

Research Collection School Of Computing and Information Systems

Automated issue fixing is a critical task in software debugging and has recently garnered significant attention from academia and industry. However, existing fixing techniques predominantly focus on the repair phase, often overlooking the importance of improving the preceding bug localization phase. As a foundational step in issue fixing, bug localization plays a pivotal role in determining the overall effectiveness of the entire process. To enhance the precision of issue fixing by accurately identifying bug locations in large-scale projects, this paper presents BugCerberus, the first hierarchical bug localization framework powered by three customized large language models. First, BugCerberus analyzes intermediate representations …

Fingerprinting Voice Commands Of Vpn-Protected Smart Speakers, Xiaoguang Guo, Keyang Yu, Qi Li, Dong Chen

Computer Science Faculty Research and Publications

Extensive recent research has shown that it is surprisingly easy to infer Amazon Alexa voice commands over their network traffic data. To prevent these traffic analytics (TA)-based inference attacks, smart home owners are considering deploying virtual private networks (VPNs) to safeguard their smart speakers. In this work, we design a new machine learning-powered attack framework—VoiceAttack that could still accurately fingerprint voice commands on VPN-encrypted voice speaker network traffic. We evaluate VoiceAttack under 5 different real-world settings using Amazon Alexa and Google Home. Our results show that VoiceAttack could correctly infer voice command sentences with a Matthews Correlation Coefficient (MCC) of …

Phase 5 - Post Implementation, Anna Pugerud

Distinguished Student Scholarship Collection

This paper presents a post‑implementation analysis and system design plan for digitizing animal health records within the Animal Health Services (AHS) department. The project addresses inefficiencies, data inaccuracies, and physical storage limitations associated with paper‑based documentation by proposing the adoption of a Software as a Service (SaaS) digital records solution. Through stakeholder interviews, observations, and document analysis, the project identifies functional, non‑functional, and system requirements necessary to support streamlined reporting, treatment tracking, and regulatory compliance. The analysis outlines feasibility across technical, economic, and organizational dimensions, demonstrating that the proposed system can reduce documentation time, improve response times, and lower operational …

A View Under The Hood: Duquesne Kline's Law And Computing Program, Wesley M. Oliver, Katherine L.W. Norton, Martin Mckown, David Horrigan

West Virginia Law Review

No abstract provided.

Low-Complexity Structured Neural Networks And Their Usage In Image And Signal Processing, Adam Kuzmicki

Doctoral Dissertations and Master's Theses

Conventional neural networks face significant challenges due to high computational costs, large parameter counts, and reliance on backpropagation, which restricts their application in resource-constrained and real-time settings. To address these challenges, this thesis proposes three structured neural network (NN) architectures grounded in the theories of sparse and self-contained factorizations of transforms, with applications to image compression, reconstruction, classification, encryption, and also adaptive wideband multi-beam beamforming. The first neural network architecture, named DCTrix-Net, replaces conventional spatial con- volution with highly sparse factorization of the discrete Cosine transform (DCT) complemented by Toeplitz-structured weight initialization, achieving at least 97% FLOP reduction over CNNs, …

Ai Interpretability In Healthcare Communication, Ananya Jeyappragash

Dartmouth College Master’s Theses

Artificial intelligence has increasingly been adopted in healthcare, largely for specialized tasks and under significant human oversight. The use of large black-box systems raises important concerns about transparency in high-stakes environments such as clinical decision-making. Clinical communication is fundamentally human-centered, and failures in judgment can have serious consequences for patient care. Overestimating the reasoning abilities of large language models may lead to undue trust in fabricated or “hallucinated” outputs, while rejecting AI-assisted tools altogether may preserve inefficient workflows and contribute to missed or delayed diagnoses. These concerns reflect a broader tradeoff between accuracy and interpretability: although more complex models may …

The Energy And Environmental Footprint Of Ai, Michael P. Vandenbergh, Ethan I. Thorpe, Jonathan M. Gilligan

Michigan Journal of Environmental & Administrative Law

Artificial intelligence (AI) has the potential to create major economic and social benefits, but also to rapidly escalate electricity demand and its associated environmental impacts. Information availability has been a cornerstone of environmental law for half a century, and this Article argues that providing information to individual, corporate, and other users about the electricity demand and environmental impacts of AI can reduce those impacts without delaying development of the technology. Little is known about how different large language models (LLMs) compare on these metrics, though. To address whether users have access to the information necessary to address this shortcoming, the …

Match-A-Fit, Brianna Mendoza, Adan Diaz De Leon, Pedro Jacobo, Juan Marco Saca Dada

Posters - 2026

Welcome to Match-a-Fit! Match-a-Fit is an iOS application that allows the user to create a digital closet by uploading images of their clothing items. With AI, the program can generate outfits based on the digital closet, the time, and the occasion. Match-a-Fit’s purpose is designed to help users who struggle to get ready, run out of time, or can’t decide on an outfit, by easily generating outfit options based on the occasion.

Maddenlite, Sergio Pena

Presentations - 2026

Problem •“What If” scenarios impossible to test accurately •Commercial games rely on arcade physics •Spreadsheets lack visual engagement

Motivation •Passion for football analytics •Desire to simulate cross-era matchups •Apply math models to real-world sports data

Solution •Python based simulation engine using historical play-by-play data •Simulates outcomes based on probability

A.I.R.E., Laurene Robinson

Presentations - 2026

•Cybersecurity analysts rely on reverse engineering to understand suspicious software. •Ghidra can surface decompiled code, but it does not fully explain function purpose, behavioral meaning, or analyst priority. •When symbols are stripped and context is weak, analysts must still reconstruct intent manually from low-level output. •That process is Time-consuming , complex and , operationally costly

Harvest Scanner, Alexander Murphy

Posters - 2026

In current times, people can find themselves at the whims of markets and may be spending more than they realize or want to on regular, everyday goods. New tools can help users keep track of the goods they are paying for. Harvest Scanner was developed to scan and track local grocery prices from stores using their publicly available website information. It was developed in Python using PyQt5 for GUI. The database is stored as an SQL file with Python using SQLite engine. Users will be able to view local grocery prices in a database interface (GUI). There are many features …

A.I.R.E. - Ai-Assisted Reverse Engineering, Laurene Robinson

Posters - 2026

Reverse engineering plays a vital role in cybersecurity by helping analysts examine unknown binaries, investigate malware, identify vulnerabilities, and better protect sensitive systems. However, once a program is compiled and stripped, the meaningful names that describe its behavior are lost, leaving behind generic function labels like FUN_00401a30. Analysts must then manually interpret decompiled code, trace call chains, and infer program behavior function by function, which is slow and mentally demanding on large binaries. To address this challenge, this project introduces A.I.R.E., a local Ghidra extension that extracts contextual evidence from stripped functions and uses a locally hosted language model to …

Topshelf, Ayden Jay Soliz

Posters - 2026

With so many great video games releasing each year, it becomes challenging to keep up with the latest. Players find it difficult to maintain an updated list of future games to play, and many existing online trackers have become too complicated to use. TopShelf is designed to be a simple video game backlogging website that will track games for the player. By connecting to an online video game database API, users can add/drop games from their personal list and enable tracking and receive emails for platform releasing. Gamers can leave all the tracking and updates responsibilities to TopShelf

Lego®-Based Cubesats For Space Outreach, Stefan Brandle, Joshua Brown, Evan Smith, Drew Nye, Sydney Reddy, Twidy Kwae, Kaiya Grzesiak, Matthew Roderer, John Hetz

Faculty-Led Student Projects in CSE

Taylor University Computer Science and Engineering, in collaboration with NearSpace Education, has initiated a multi-year effort to enhance space-awareness by offering K-12 students career-impacting opportunities to build and fly satellites that have a LEGO®-robotics-based payload. This effort is intended as an outreach in the spirit of the ThinSat project (2017-2021) supported by Virginia Space, Twiggs Space Lab, Orbital ATK, NearSpace Launch, and the NASA Wallops Flight Facility, among others. We hope to receive permission to name the satellite LEGO-SAT-0.

Next-Generation Democratic Cyber Statecraft - Balancing The Signal: Shutdown Shocks And Democratic Digital Governance, Scott M. Di Panni

School of Public Policy Capstones

This paper develops Next-Generation Democratic Cyber Statecraft (NG-DCS), a unified strategic doctrine for democratic governments to contest the cognitive domain against authoritarian adversaries. Drawing on twenty-six years of cross-national panel data (1999–2024) spanning 213 countries, game-theoretic modeling, and qualitative case analysis, the paper establishes three interconnected empirical and theoretical foundations. First, cross-national OLS regression across 160+ countries demonstrates that regime type is the dominant structural determinant of internet freedom (R²=0.615, β=2.513, p< 0.001), explaining more than twice the variance attributable to per-capita wealth (R²=0.268). Democratic governance, not economic development, produces open digital environments. Second, a two-way fixed effects (TWFE) difference-in-differences study exploiting government-ordered internet shutdowns as discrete policy interventions finds that digital restrictions causally degrade V-Dem governance quality by 0.21–0.38 standard deviations (p< 0.001 across all specifications). Treatment effects are immediate (β=−0.302 at k=0) and persist through five post-treatment years (β=−0.246 at k=+5), indicating structural rather than transitory governance damage. Parallel trends validation (p=0.352) and Callaway–Sant’Anna heterogeneity-robust estimation (ATT=−0.230, SE=0.077) support causal identification. Instrumental variable triangulation (2SLS β=−0.949, p=0.005) confirms that simultaneity was attenuating, not inflating, the primary estimates. Third, formal game-theoretic analysis reveals that the current U.S.–adversary equilibrium is (Restrain, Escalate)—the risk-dominant but Pareto-inferior outcome of a Stag Hunt structure. China, Russia, North Korea, and Venezuela each occupy structurally distinct positions (Stackelberg commitment, asymmetric two-level, autarky, and reactive trigger, respectively), requiring differentiated doctrinal responses rather than a uniform strategic playbook. Generative AI and algorithmic governance are shown to accelerate cognitive vulnerability by collapsing influence operation costs and exploiting engagement-optimized platform architectures that systematically degrade deliberative capacity in democratic populations.

Watch Me Watch: Reaction Videos As A Social Form Of Online Video Engagement, Rodney Okyere, Carlos Augusto Bautista Isaza, Jaehoon Pyon, Ihudiya Finda Ogbonnaya-Ogburu, Shuo Niu, Sang Wong Lee

Computer Science

Reaction videos (RVs) are surging in popularity, emerging as a distinctive facet of participatory culture on modern social video-sharing platforms such as YouTube, TikTok, and Twitch. This study aims to explore not only the motivations and engagement patterns of viewers towards RVs, but also the underlying nature of the virality and community-building phenomena that this sub-genre of video content fosters. We conducted 16 semi-structured interviews with individuals who identified as regular viewers of reaction videos to gain a deeper understanding of how they discover reaction videos (RQ1), the values that drive their motivations for viewing RVs (RQ2), and the ways …

Be Responsible In Your Answers! Monitoring Out-Of-Domain Behaviors In Domain-Specific Llms, Boquan Li, Chenzhe Lou, Zhe Ren, Peixin Zhang, Zirui Fu, Jun Sun, Yaowen Zheng

Research Collection School Of Computing and Information Systems

Large Language Models (LLMs) have accelerated the rapid development of chatbot web applications in various domains, such as coding, biomedicine and psychology. Compared to general LLMs like ChatGPT, domain-specific LLMs require a greater sense of responsibility. For instance, if a programming LLM casually answers medical or psychological questions, it not only misleads the public but also poses legal risks. This highlights new demands for monitoring and preventing such irresponsible behaviors. Existing efforts attempt to monitor LLMs from multiple aspects, such as lying, jailbreaks, and toxic content, while overlooking out-of-domain behaviors. In this work, we propose an innovative LLM domain monitoring …

Architecture-Agnostic Test-Time Adaptation Via Backprop-Free Embedding Alignment, Xiao Ma, Young D. Kwon, Pan Zhou, Dong Ma

PhD Student’s Publications Collection

Test-Time Adaptation (TTA) adapts a deployed model during online inference to mitigate the impact of domain shift. While achieving strong accuracy, most existing methods rely on backpropagation, which is memory and computation intensive, making them unsuitable for resource-constrained devices. Recent attempts to reduce this overhead often suffer from high latency or are tied to specific architectures such as ViT-only or CNN-only. In this work, we revisit domain shift from an embedding perspective. Our analysis reveals that domain shift induces three distinct structural changes in the embedding space: translation (mean shift), scaling (variance shift), and rotation (covariance shift). Based on this …

Scalable Multi-Task Low-Rank Model Adaptation, Zichen Tian, Antoine Ledent, Qianru Sun

PhD Student’s Publications Collection

Scaling multi-task low-rank adaptation (LoRA) to a large number of tasks induces catastrophic performance degradation, such as an accuracy drop from 88.2% to 2.0% on DOTA when scaling from 5 to 15 tasks. This failure is due to parameter and representation misalignment. We find that existing solutions, like regularization and dynamic routing, fail at scale because they are constrained by a fundamental trade-off: strengthening regularization to reduce inter-task conflict inadvertently suppresses the essential feature discrimination required for effective routing. In this work, we identify two root causes for this trade-off. First, uniform regularization disrupts inter-task knowledge sharing: shared underlying knowledge …

Ethical Issues Of The Advancement Of Artificial Intelligence, Primo G. Moreno

Augustana Center for the Study of Ethics Essay Contest

Artificial intelligence has rapidly become integrated into our modern society, allowing for new streamlined methods to reign supreme. These methods raise concerns about control, privacy, and the shifting boundaries between human and machine decision-making. This paper examines how AI has developed, how it is currently being used, and the ethical dilemmas that arise as societies integrate it into systems such as healthcare, law enforcement, and businesses. Through the expansion of historical, technical, and philosophical perspectives, this paper argues that society must prioritize having set guidelines and systems for monitoring newly created AI systems. Without them, the implementation of AI models …

Semantic Entanglement In Vector-Based Retrieval: A Formal Framework And Context-Conditioned Disentanglement Pipeline For Agentic Rag Systems, Nick Loghmani

iSchool - All Scholarship

Retrieval-Augmented Generation (RAG) systems deployed in agentic environments depend on the geometric properties of vector representations to retrieve contextually appropriate evidence for autonomous reasoning. When source documents conflate multiple topics within contiguous text regions, standard vectorization pipelines produce embedding spaces in which semantically distinct content occupies overlapping geometric neighborhoods — a condition we term semantic entanglement. This paper formalizes semantic entanglement as a model-relative measure of cross-topic overlap, defines an Entanglement Index (EI) as a quantitative proxy, and argues that higher EI is associated with reduced attainable Top-K retrieval precision under cosine similarity retrieval. We introduce the Semantic Disentanglement …

Penforge: On-The-Fly Expert Agent Construction For Automated Penetration Testing, Huihui Huang, Jieke Shi, Junkai Chen, Ting Zhang, Yikun Li, Chengran Yang, Eng Lieh Ouh, Lwin Khin Shar, David Lo

Research Collection School Of Computing and Information Systems

Penetration testing is essential for identifying vulnerabilities in web applications before real adversaries can exploit them. Recent work has explored automating this process with Large Language Model (LLM)-powered agents, but existing approaches either rely on a single generic agent that struggles in complex scenarios or narrowly specialized agents that cannot adapt to diverse vulnerability types. We therefore introduce PenForge, a framework that dynamically constructs expert agents during testing rather than relying on those prepared beforehand. By integrating automated reconnaissance of potential attack surfaces with agents instantiated on the fly for context-aware exploitation, PenForge achieves a 30.0% exploit success rate (12/40) …

Stacked From One: Multi-Scale Self-Injection For Context Window Extension, Wei Han, Pan Zhou, Shuicheng Yan

Research Collection School Of Computing and Information Systems

The limited context window of contemporary large language models (LLMs) remains a primary bottleneck for their broader application across diverse domains. Although continual pre-training on long-context data offers a straightforward solution, it incurs prohibitive data acquisition and computational costs. To address this challenge, we propose SHAREDLLM, a novel framework based on multi-grained context compression and query-aware information acquisition. SHAREDLLM comprises two stacked short-context LLMs: a lower model serving as a compressor and an upper model acting as a decoder. The lower model compresses long inputs into compact, multi-grained representations, which are then forwarded to the upper model for context-aware processing. …

Challenges In Synchronous And Remote Collaboration Around Visualization, Matthew Brehmer, Maxime Cordeil, Christophe Hurter, Takayuki Itoh, Wolfgang Büschel, Mahmood Jasim, Arnaud Prouzeau, David Saffo, Lyn Bartram, Sheelagh Carpendale, Chen Zhu-Tian, Andrew Cunningham, Anthony Tang, Samuel Huron, Masahiko Itoh, Arpit Joshi, Kiyoshi Kiyokawa, Hideaki Kuzuoka, Bongshin Lee, Guillermo Molina León

Research Collection School Of Computing and Information Systems

We characterize 16 challenges faced by those investigating and developing remote and synchronous collaborative experiences around visualization. Our work reflects the perspectives and prior research efforts of an international group of 29 experts from across human-computer interaction and visualization sub-communities. The challenges are anchored around five collaborative activities that exhibit a centrality of visualization and multimodal communication. These activities include exploratory data analysis, creative ideation, visualization-rich presentations, joint decision making grounded in data, and real-time data monitoring. The challenges also reflect the changing dynamics of these activities in the face of recent advances in extended reality (XR) and artificial intelligence …

Thinktank-Me: A Multi-Expert Framework For Middle East Event Forecasting, Haoxuan Li, He Chang, Yunshan Ma, Yi Bin, Yang Yang, See-Kiong Ng, Tat-Seng Chua

Research Collection School Of Computing and Information Systems

Event forecasting is inherently influenced by multifaceted considerations, including international relations, regional historical dynamics, and cultural contexts. However, existing LLM-based approaches employ single-model architectures that generate predictions along a singular explicit trajectory, constraining their ability to capture diverse geopolitical nuances across complex regional contexts. To address this limitation, we introduce ThinkTank-ME, a novel Think Tank framework for Middle East event forecasting that emulates collaborative expert analysis in real-world strategic decision-making. To facilitate expert specialization and rigorous evaluation, we construct POLECAT-FOR-ME, a Middle East–focused event forecasting benchmark. Experimental results demonstrate the superiority of multi-expert collaboration in handling complex temporal geopolitical forecasting …

Discrete Diffusion For Bundle Construction, Teng Tu, Ai Li, Yunshan Ma, Shuo Xu, Xiaohao Liu, Haokai Ma, Liang Pang, Tat-Seng Chua

Research Collection School Of Computing and Information Systems

As a central task in product bundling, bundle construction aims to select a subset of items from large item catalogs to build an entire bundle or, more practically, complete a partial bundle. Existing methods often rely on the sequential construction paradigm that predicts items one at a time, nevertheless, this paradigm is fundamentally unsuitable for the essentially unordered bundles. In contrast, non-sequential methods model a bundle as a set, but still face two dimensionality curses: the combinatorial space grows exponentially with both bundle length and catalog size. Accordingly, we identify two technical challenges: 1) how to effectively and efficiently model …