Digital Commons Network^™

Enhancing Cybersecurity Strategies Through Automated Cti Extraction, Risk Prioritization, And Privacy-Conscious Information Sharing, Spencer Rian Massengale

Theses and Dissertations

Cybersecurity operations require the ability to collect and analyze large amounts of cyber threat intelligence (CTI) to assess risks and formulate defensive strategies against emerging threats. This task has become increasingly complex due to the rapid evolution of cyber threats and the growing volume of unstructured, natural-language CTI sources. The scale of data and analysis needed to utilize CTI effectively far exceeds humans' manual capacity, especially for organizations with limited resources. This research focuses on leveraging Large Language Models (LLMs) and machine learning techniques to enhance CTI extraction, risk assessment, and data sharing. We utilized LLMs to automate the extraction …

Sd-Weat: Towards Robustly Measuring Bias In Input Embeddings For Artificial Intelligence Language Models, Magnus Gray

Theses and Dissertations

Artificial intelligence (AI) is rapidly transforming industries and markets, from healthcare to entertainment, revolutionizing decision-making processes. However, as AI grow more influential, they also risk amplifying existing biases, potentially leading to harmful consequences. Recent advancements in large language models (LLMs), such as GPT-4 and Llama, have heightened concerns about bias in natural language processing (NLP) tasks, driving the need for robust methods to detect and mitigate bias. Current approaches, such as the Word Embedding Association Test (WEAT) and its sentence-level extension the Sentence Encoder Association Test (SEAT) often fall short in capturing the nuances of biases in the input embeddings …

On The Benefits Of Directness In Virtual Characters For Motivational Interviews, Michael O'Mahony, Cathy Ennis, Robert Ross

Conference papers

Understanding the factors influencing successful engagement with Embodied Conversational Agents (ECAs) remains a significant challenge. This understanding could be used to personalise agents to users to improve interactions. Some studies have shown that simulating personalities in healthcare agents can improve effectiveness and engagement. However, it is not yet well understood how variations of agent personality can be leveraged to improve user engagement with Motivational Interviewing (MI) ECAs. Specifically how the balance between agent warmth and directness can be controlled in an MI agent to improve likeability and engagement. We conducted an online Wizard-of-Oz (WoZ) mediated study of two variants of …

Using Llms To Establish Implicit User Sentiment Of Software Desirability, Sherri Weitl-Harms, John D. Hastings, Jonah Lum

Research & Publications

This study explores the use of LLMs for providing quantitative zero-shot sentiment analysis of implicit software desirability, addressing a critical challenge in product evaluation where traditional review scores, though convenient, fail to capture the richness of qualitative user feedback. Innovations include establishing a method that 1) works with qualitative user experience data without the need for explicit review scores, 2) focuses on implicit user satisfaction, and 3) provides scaled numerical sentiment analysis, offering a more nuanced understanding of user sentiment, instead of simply classifying sentiment as positive, neutral, or negative.

Data is collected using the Microsoft Product Desirability Toolkit (PDT), …

Video Game Development 3.0: Ai-Driven Collaborative Co-Creation, Jay Ratican, James Hutson

Faculty Scholarship

The evolution of game development has transitioned from manual coding (Software 1.0) to data-driven Artificial Intelligence (AI) (Software 2.0), and now to a more advanced stage—video game development 3.0. This phase is characterized by AI-driven processes leveraging large language models (LLMs), neural networks, and other AI techniques that autonomously generate code, content, and narratives. This paper explores the foundational technologies underpinning this paradigm shift, including customizable AI modules, dynamic asset creation, and intelligent non player characters (NPCs) that adapt to player interactions. It also highlights the integration of AI with emerging technologies like Virtual Reality (VR), Augmented Reality (AR), and …

Towards Unified Multimodal Editing With Enhanced Knowledge Collaboration, Kaihang Pan, Zhaoyu Fan, Juncheng Li, Qifan Yu, Hao Fei, Siliang Tang, Richang Hong, Hanwang Zhang, Qianru Sun

Research Collection School Of Computing and Information Systems

The swift advancement in Multimodal LLMs (MLLMs) also presents significant challenges for effective knowledge editing. Current methods, including intrinsic knowledge editing and external knowledge resorting, each possess strengths and weaknesses, struggling to balance the desired properties of reliability, generality, and locality when applied to MLLMs. In this paper, we propose UniKE, a novel multimodal editing method that establishes a unified perspective and paradigm for intrinsic knowledge editing and external knowledge resorting. Both types of knowledge are conceptualized as vectorized key-value memories, with the corresponding editing processes resembling the assimilation and accommodation phases of human cognition, conducted at the same semantic …

Enhancing Password Security And Memorability Using Machine Learning And Linguistic Patterns, Jared Wise

LSU New Orleans Theses and Dissertations

In the digital age, text-based passwords remain a primary method for securing online accounts. Yet, users frequently face a dilemma between creating passwords that are easy to remember and sufficiently secure against cyberattacks. This research introduces an approach to password generation that bridges this gap by utilizing linguistic patterns, particularly song lyrics, to develop highly secure and naturally memorable passwords. Using large lyric datasets gained from web scrapes from popular song lyric websites (AZ Lyrics, Genius), features are extracted from a corpus of over 5 million lyrics using sentence structure and natural language processing in a novel way. In using …

Ali-Agent: Assessing Llms’ Alignment With Human Values Via Agent-Based Evaluation, Jingnan Zheng, Han Wang, Tai D. Nguyen, An Zhang, Jun Sun, Tat-Seng Chua

Research Collection School Of Computing and Information Systems

Large Language Models (LLMs) can elicit unintended and even harmful content when misaligned with human values, posing severe risks to users and society. To mitigate these risks, current evaluation benchmarks predominantly employ expertdesigned contextual scenarios to assess how well LLMs align with human values. However, the labor-intensive nature of these benchmarks limits their test scope, hindering their ability to generalize to the extensive variety of open-world use cases and identify rare but crucial long-tail risks. Additionally, these static tests fail to adapt to the rapid evolution of LLMs, making it hard to evaluate timely alignment issues. To address these challenges, …

Adapting Deep Learning Models For Downstream Web Tasks: Multimodal Models, Task Development, Agent Adaptation, Graham Annett

Boise State University Theses and Dissertations

This dissertation presents a framework for the development of deep learning models tailored for dynamic web tasks, leveraging generalized pre-trained multimodal transformers. A task generation framework, applied to multiple web datasets, is introduced, facilitating instruction fine-tuning of models for executing multi-step web workflows. This approach enhances the adaptability of pre-trained models to a spectrum of novel web tasks, which is vital for the reliable operation of web agents.

Moreover, this work proposes an encoding schema extending the Decision Transformer, which advances the adaptability of these models for downstream tasks through targeted modality tokenization, thereby broadening their practical applicability. These enhancements …

Towards General Conceptual Model Editing Via Adversarial Representation Engineering, Yihao Zhang, Zeming Wei, Jun Sun, Meng Sun

Research Collection School Of Computing and Information Systems

Since the rapid development of Large Language Models (LLMs) has achieved remarkable success, understanding and rectifying their internal complex mechanisms has become an urgent issue. Recent research has attempted to interpret their behaviors through the lens of inner representation. However, developing practical and efficient methods for applying these representations for general and flexible model editing remains challenging. In this work, we explore how to leverage insights from representation engineering to guide the editing of LLMs by deploying a representation sensor as an editing oracle. We first identify the importance of a robust and reliable sensor during editing, then propose an …

Chain Of Preference Optimization: Improving Chain-Of-Thought Reasoning In Llms, Xuan Zhang, Chao Du, Tianyu Pang, Qian Liu, Wei Gao, Min Lin

Research Collection School Of Computing and Information Systems

The recent development of chain-of-thought (CoT) decoding has enabled large language models (LLMs) to generate explicit logical reasoning paths for complex problem-solving. However, research indicates that these paths are not always deliberate and optimal. The tree-of-thought (ToT) method employs tree-searching to extensively explore the reasoning space and find better reasoning paths that CoT decoding might overlook. This deliberation, however, comes at the cost of significantly increased inference complexity. In this work, we demonstrate that fine-tuning LLMs leveraging the search tree constructed by ToT allows CoT to achieve similar or better performance, thereby avoiding the substantial inference burden. This is achieved …

User Acceptance Of Advice By Ai Agents: Expectation-System Fit Perspective, Jingyuan Cai, Fiona Fui-Hoon Nah

Research Collection School Of Computing and Information Systems

Algorithms have increasing influence on our daily decisions, especially when the recommendations are presented by human-like AI agents. This study applies the Theory of Effective Use to investigate how the fit between the user’s role expectation for an AI agent and the agent’s interaction style impacts AI advice adoption. We proposed a new concept termed Perceived Expectation-System Fit (PESF) and empirically examined its impact on user perceptions and advice acceptance. We found that low PESF reduces advice acceptance by diminishing cognitive and affective trust in the AI agent. Furthermore, increased algorithm transparency increases PESF's impact on decision-making. Our findings provide …

Lova3 : Learning To Visual Question Answering, Asking And Assessment, Henry Hengyuan Zhao, Pan Zhou, Difei Gao, Bai Shou, Mike Zheng Shou

Research Collection School Of Computing and Information Systems

Question answering, asking, and assessment are three innate human traits crucial for understanding the world and acquiring knowledge. By enhancing these capabilities, humans can more effectively utilize data, leading to better comprehension and learning outcomes. Current Multimodal Large Language Models (MLLMs) primarily focus on question answering, often neglecting the full potential of questioning and assessment skills. Inspired by the human learning mechanism, we introduce LOVA3 , an innovative framework named “Learning tO Visual question Answering, Asking and Assessment,” designed to equip MLLMs with these additional capabilities. Our approach involves the creation of two supplementary training tasks GenQA and EvalQA, aiming …

Supervised Generative Adversarial Networks For Time Series Generation In Embedding Space, Mohammadreza Eskandarinasab

All Graduate Theses and Dissertations, Fall 2023 to Present

Time series data, such as weather forecasts, stock market trends, or heart rate monitors, plays a vital role in many areas of our lives. However, creating realistic synthetic time series data for research and testing purposes has been a significant challenge due to limitations in existing methods, which often struggle with accuracy and consistency. In this study, we developed two new approaches to generate high-quality time series data more effectively. The first method introduces a dual-feedback system that helps the model learn and replicate real data patterns more accurately by providing guidance at different stages of the learning process. The …

Towards Comprehensive And Interpretable Video Understanding, Khoa Vo

Graduate Theses and Dissertations

Video understanding is a critical domain in computer vision, focusing on analysis of sequential visual data to extract meaningful spatiotemporal information for tasks such as action recognition, video captioning, video retrieval, and temporal action localization, etc. Despite significant advancements with spatio-temporal convolutional neural networks and attention-based video models, current methods face limitations, including inadequate representation of main actors, lack of fine-grained modeling of relevant objects, and limited interpretability.
This thesis addresses these challenges by proposing novel approaches that enhance video understanding through modeling interactions among entities (actors and objects) and between entities and the environment, while improving interpretability in the …

Enhancing Low-Resource Language Performance In Multilingual Large Language Models, Mingqi Li

All Dissertations

The large language models play an important role in many natural language tasks. However, training these models requires large amounts of data, which is not available for many languages. A noticeable performance gap exists between English and other languages, with low-resource languages showcasing this gap prominently. Therefore, it becomes imperative to improve large language models for low-resource languages. To address these challenges, we developed knowledge distillation and strategic prompt-learning, and attention alignment methods to improve the representation capabilities of large language models for low-resource language, and then enhanced their performance in downstream tasks.

In our first study, we developed a …

Collaboration! Towards Robust Neural Methods For Routing Problems, Jianan Zhou, Yaoxin Wu, Zhiguang Cao, Wen Song, Jie Zhang, Zhiqi Shen

Research Collection School Of Computing and Information Systems

Despite enjoying desirable efficiency and reduced reliance on domain expertise, existing neural methods for vehicle routing problems (VRPs) suffer from severe robustness issues – their performance significantly deteriorates on clean instances with crafted perturbations. To enhance robustness, we propose an ensemble-based Collaborative Neural Framework (CNF) w.r.t. the defense of neural VRP methods, which is crucial yet underexplored in the literature. Given a neural VRP method, we adversarially train multiple models in a collaborative manner to synergistically promote robustness against attacks, while boosting standard generalization on clean instances. A neural router is designed to adeptly distribute training instances among models, enhancing …

Reevo: Large Language Models As Hyper-Heuristics With Reflective Evolution, Haoran Ye, Jiarui Wang, Zhiguang Cao, Federico Berto, Chuanbo Hua, Haeyeon Kim, Jinkyoo Park, Guojie Song

Research Collection School Of Computing and Information Systems

The omnipresence of NP-hard combinatorial optimization problems (COPs) compels domain experts to engage in trial-and-error heuristic design process. The long-standing endeavor of design automation has gained new momentum with the rise of large language models (LLMs). This paper introduces Language Hyper-Heuristics (LHHs), an emerging variant of Hyper-Heuristics that leverages LLMs for heuristic generation, featuring minimal manual intervention and open-ended heuristic spaces. To empower LHHs, we present Reflective Evolution (ReEvo), a generic searching framework that emulates the reflective design approach of human experts while far surpassing human capabilities with its scalable LLM inference, Internet-scale domain knowledge, and powerful evolutionary search. Evaluations …

Comparative Analysis Of Multi-Criteria Techniques In Neutrosophic Environment And Their Applications To Economic Condition Assessment, Asmaa Elsayed, Mai Mohamed

Neutrosophic Systems with Applications

In economic decision-making, the challenge of evaluating multiple, often conflicting criteria necessitates advanced Multi-Criteria Decision-Making (MCDM) techniques. Traditional methods can struggle with the inherent uncertainty, ambiguity, and imprecision of real-world data. This paper addresses these challenges by investigating the effectiveness of various MCDM techniques within neutrosophic environments, with a particular focus on the Criteria-wise Alternatives Ranking and Correlation Analysis for Composite Scoring (CARCACS) method. Neutrosophic sets, which incorporate truth, falsity, and indeterminacy, provide a robust framework for addressing the vagueness and inconsistencies found in economic indicators such as GDP growth, employment levels, inflation rates, trade balances, investment activity, and government …

Unified Generative And Discriminative Training For Multi-Modal Large Language Models, Wei Chow, Juncheng Li, Kaihang Pan, Qifan Yu, Hao Fei, Zhiqi Ge, Shuai Yang, Siliang Teng, Hanwang Zhang, Qianru Sun

Research Collection School Of Computing and Information Systems

In recent times, Vision-Language Models (VLMs) have been trained under two predominant paradigms. Generative training has enabled Multimodal Large Language Models (MLLMs) to tackle various complex tasks, yet issues such as hallucinations and weak object discrimination persist. Discriminative training, exemplified by models like CLIP, excels in zero-shot image-text classification and retrieval, yet struggles with complex scenarios requiring fine-grained semantic differentiation. This paper addresses these challenges by proposing a unified approach that integrates the strengths of both paradigms. Considering interleaved image-text sequences as the general format of input samples, we introduce a structure-induced training strategy that imposes semantic relationships between input …

Self-Supervised Fine-Tuning For Neural Expert Finding, Budhitama Subagdja, Dan Sanchari, Ah-Hwee Tan

Research Collection School Of Computing and Information Systems

Expert finding systems allow ones to find individuals who have expertise in specific fields or domains. Traditional expert finding are mostly based on topic modeling or keyword search methods that are limited in their capability to encode contextual knowledge from natural language. To address the limitation, this paper presents Neural Expert Finder (NEF), a novel method that takes a transfer learning approach based on transformer encoder networks to leverage the rich seman-tic and syntactic patterns of language encoded in pre-trained language models (PLMs). We propose a self-supervised learning approach utilizing contrastive training using both positive and automatically generated negative samples …

Generative Artificial Intelligence In Business Higher Education: A Focus Group Study, Xuenan Huo, Keng Siau

Research Collection School Of Computing and Information Systems

This research investigates the opportunities and challenges of integrating generative artificial intelligence (GenAI) into business higher education, drawing insights from an asynchronous focus group research study with doctoral students who serve dual roles as both learners and educators. Key opportunities identified through thematic analysis include knowledge acquisition, intelligent co-ideation, supportive augmentation, and personalized learning. Challenges identified include AI trustworthiness, cognitive dependency, human value, policy and instruction, assessment integrity, and identity management. This study clarifies GenAI’s specific role in business education and provides practical insights for effectively integrating GenAI to enhance learning outcomes and address emerging challenges. An analysis theory on …

Multi-Criteria Decision-Making Approach Based On Correlation Coefficient For Multi-Polar Interval-Valued Neutrosophic Soft Set, Hamza Naveed, Saalam Ali

Neutrosophic Systems with Applications

The correlation coefficient between two factors is crucial in statistical computation, indicating the extent and evolution of the appropriate link. The precision of applicability evaluations frequently relies on the thoroughness and caliber of data obtained from a certain dataset. Statistical research sometimes entails data marked by intrinsic trade-offs and uncertainty. This study seeks to present m-polar interval-valued neutrosophic soft sets (mPIVNSSs) through the integration of m-polar fuzzy sets with interval-valued neutrosophic soft sets. The suggested mPIVNSS structure is a significantly generalized version of m-polar neutrosophic soft sets and serves as a substantial extension of interval-valued neutrosophic soft sets. In this …

De-Identification Is Not Enough: A Comparison Between De-Identified And Synthetic Clinical Notes, Atiquer Rahman Sarkar, Yao-Shun Chuang, Noman Mohammed, Xiaoqian Jiang

Faculty, Staff and Student Publications

For sharing privacy-sensitive data, de-identification is commonly regarded as adequate for safeguarding privacy. Synthetic data is also being considered as a privacy-preserving alternative. Recent successes with numerical and tabular data generative models and the breakthroughs in large generative language models raise the question of whether synthetically generated clinical notes could be a viable alternative to real notes for research purposes. In this work, we demonstrated that (i) de-identification of real clinical notes does not protect records against a membership inference attack, (ii) proposed a novel approach to generate synthetic clinical notes using the current state-of-the-art large language models, (iii) evaluated …

Digital Humanities: Using Computational Methods On Literature To Understand Human-Water Relations, Ariel Yang

Cybersecurity Undergraduate Research Showcase

By using computational techniques to analyze literature, deeper insights can be gained into human-water relationships across different historical and cultural contexts. Natural Language Processing (NLP) and other data science methods can explore applications of traditional ecological knowledge (TEK) and underlying emotions or beliefs in literature to help understand sustainability. Protecting this sensitive cultural data through ethical applications can further secure future implementations of policies, urban planning, and environmental relationships.

Developing Linguistic Patterns To Mitigate Inherent Human Bias In Offensive Language Detection, Toygar Tanyel, Besher Alkurdi, Serkan Ayvaz

Turkish Journal of Electrical Engineering and Computer Sciences

With the proliferation of social media, there has been a sharp increase in offensive content, particularly targeting vulnerable groups, exacerbating social problems such as hatred, racism, and sexism. Detecting offensive language use is crucial to prevent offensive language from being widely shared on social media. However, the accurate detection of irony, implication, and various forms of hate speech on social media remains a challenge. Natural language-based deep learning models require extensive training with large, comprehensive, and labeled datasets. Unfortunately, manually creating such datasets is both costly and error-prone. Additionally, the presence of human-bias in offensive language datasets is a major …

A Systematic Review Of The Effects Of Ai-Assisted Moderation On Individuals And Groups, Zehui Yu, Lukas Otto, Dennis Assenmacher, Claudia Wagner

Human-Machine Communication

This review paper provides a conceptualization of AI-assisted content moderation with various degrees of autonomy and summarizes experimental evidence for how different levels of automation in content moderation and related losses of autonomy affect individuals and groups. Our results show that current research predominantly focuses on individuallevel effects, necessitating a shift toward understanding the impact on groups. The study highlights gaps in exploring different levels of AI-assisted moderation interventions and misalignments of different conceptualizations that make comparing research results difficult. The discussion underscores the prevailing emphasis on harmful content removal and advocates for investigating more constructive moderation techniques, emphasizing the …

Designing Customized Loss Functions For Training Deep Neural Networks, Ali Pourramezan Fard

Electronic Theses and Dissertations

This dissertation explores the critical role of loss functions in enhancing the predictive performance of deep machine learning models. Loss functions are an integral element of all the ongoing advances we witness daily in this domain. I design custom loss functions and their impacts on various machine learning tasks, particularly in computer vision.

In the first stage of my research, I aim to improve the prediction performance of deep learning models by providing them with more precise feedback associated with task requirements. This led me to create the concept of assistive loss functions. My first proposed loss function, inspired by …

Vision-Language Integration For Enhanced Locomotion Mode Prediction, Ehsan Ahmadi

LSU Master's Theses

Wearable exoskeletons offer significant potential in enhancing human mobility in industrial environments. However, their adaptability to dynamic, task-intensive settings presents challenges, especially in accurately predicting locomotion modes such as ladder climbing, stair navigation, low-space movement, and obstacle navigation. This research proposes a multimodal framework that integrates visual data and speech commands to improve locomotion mode prediction in unpredictable environments. Multimodal data was collected using smart glasses, capturing both the user’s perspective (field-of-view, FOV) and voice during locomotion tasks. State-of-the-art models—CLIP, ImageBind, and GPT-4o—process these visual and linguistic inputs to predict locomotion activities. The models were evaluated in zero-shot and fine-tuned …

Dynamic Knowledge Elicitation: Leveraging Student Feedback For Improved Language Model Distillation, Reuven Muller

Master's Theses

Large Language Models (LLMs) have significantly advanced the field of natural language processing but remain resource-intensive and impractical for many organizations. Specialist models offer a viable alternative, often developed through Knowledge Distillation (KD) techniques. However, traditional KD methods rely on predefined static datasets to elicit knowledge from the teacher model, failing to dynamically address the weaknesses of the student model during training. This research introduces two novel methods for adaptive knowledge elicitation: Feedback-Driven Question Generation and Agent-Based Targeted Question Generation. These methods iteratively expand the training dataset based on the student model’s performance, leveraging a teacher model to generate targeted …