Other Computer Engineering Commons^™

Are Large Language Models Safe? A Vulnerability Analysis Of Generated Source Code, James Richards-Perhatch, Mitchell Milander, James M. Halvorsen, Assefaw Gebremedhin

Military Cyber Affairs

The increasing complexity of software and demands for rapid deployment have pushed the software industry to rely more on large language models (LLMs) in developing source code. However, as this technology is still relatively recent, questions can arise about the safety of the generated code. This paper presents an analysis of seven LLMs with respect to the presence of vulnerabilities within source code. Our findings show that LLMs are more likely to produce vulnerable web applications than vulnerable C programs, and that vulnerabilities are more likely to occur when program size and complexity increases.

Artificial Sense-Making Dataset, Jason A. Bengtson, John Sandstrom, Nathan Camp

NMSU Library: Datasets

No abstract provided.

Enhancing Non-Player Character Dialogue In Video Gages: An Evaluation Of Large Language Model-Generated Responses, Lam P. Quach

Master's Theses

As video games increasingly emphasize narrative depth and player immersion, the quality of Non-Player Character (NPC) dialogue has become crucial for creating engaging gaming experiences. This thesis investigates the potential of Large Language Models (LLMs) to generate high-quality NPC dialogue by comprehensively evaluating four state-of-the-art models: Gemma 3 27B, Mistral 7B, QWEN 2.5, and LLAMA 3.1. The study employs a mixed-methods approach, combining human evaluation (N=50 participants) with AI-based assessment across five key benchmarks: coherence, personality expression, engagement, style/tone appropriateness, and overall quality. Participants evaluated 32 dialogue samples (8 per model) generated for a fantasy game context featuring two distinct …

A Predictive Framework For Early Detection And Personalised Monitoring Of Parkinson’S Disease Using Artificial Intelligence And Large Language Models, Priyadharshini S

Theses and Dissertations

Parkinson’s Disease (PD) is a multifaceted and progressive neurodegenerative disorder that presents a spectrum of motor and non-motor symptoms. Early and accurate diagnosis is essential for effective disease management and improved patient outcomes, yet remains clinically challenging due to symptom overlap and diagnostic limitations. This thesis proposes a comprehensive and interpretable artificial intelligence (AI)-driven diagnostic framework that aims to transform the early detection, personalised monitoring, and treatment recommendation process for PD. The proposed solution integrates deep learning, radiomics, evolutionary optimisation, and large language models (LLMs), ensuring a highly accurate and clinically adaptable system.

The research begins by analysing T2-weighted 3D …

Traceai: Intelligent Distributed Tracing Using Large Language Models, Mihir Dhirajlal Satra

Master's Projects

Distributed systems are difficult to trace using traditional methods due to the scale of data volume and complexity, and they usually require a lot of manual analysis. TraceAI tries to solve these problems by integrating Large Language Models with the tracing tools to automatically enhance the trace data evaluation. The project aims to provide an AI-driven solution for monitoring and understanding the flow of requests across services, anomaly detection, root cause analysis and performance optimization. It can thus automate finding out systems problems using LLMs thereby carrying out large scale trace data analysis. Anticipated results from the effort will be …

Phishing Detection Using Continual Learning And Large Language Models, Gopi Prajeev Battula

Master's Projects

Adaptive phishing detection remains crucial as the nature of cyber-attacks changes over time, which renders static models obsolete. This project extends phishing detection through the implementation of continual learning approaches, namely Elastic Weight Consolidation (EWC) and Learning Without Forgetting (LWF) with RoBERTa, a Large Language Model (LLM) and compares the results of these approaches against GPT-4o-mini, another LLM. Our approach begins with fine-tuning RoBERTa on multiple phishing datasets to establish an effective baseline. EWC is then implemented to preserve vital model parameters based on their importance measured by the Fisher Information Matrix, while LWF uses knowledge distillation to retain prior …

Disease Diagnosis Using Rag Llm With Smart Prompt Engineering, Qadeerullah Syed

Master's Projects

Although recent trends indicate that LLMs outperform traditional methods in solving complex problems with enhanced reasoning, there has been barely any progress in replicating the quality of diagnoses like those of actual human doctors. The identification of an accurate diagnosis with thorough reasoning is still a significant challenge, even with advanced AI models. The process of performing accurate diagnosis remains challenging due to a lack of transparency in state-of-the-art models existing today, a lack of explanation in the diagnosis process, an emphasis on results rather than reasoning, and a lack of foundational knowledge in models, along with limited exploration of …

Galora: A Lightweight Graph-Aware Llm Framework For Node Classification On Text-Attributed Graphs, Mayur Choudhary

Master's Projects

With the exponential rise of language models (LMs) and their potential to understand semantic relationships, large LMs are being used across a wide range of applications. Text-attributed graphs (TAGs) are one notable example where LLMs can be combined with Graph Neural Networks (GNNs) to enhance node classification results. TAGs associate textual content with each node and are commonly seen in various domains such as social networks, citation graphs, recommendation systems, etc. Effectively modeling TAGs would enable deeper insights into different aspects of the graph and improve decision-making in relevant domains. We present GaLoRA, a parameter-efficient framework to integrate structural information …

Extraction Of A Knowledge Graph Of Biomedical Relationships, Brian Tran

Master's Projects

Rapid release in biomedical literature poses a challenge in linking information. This thesis aims to extract data from expanding datasets to identify and form meaningful relationships between biomedical entities. Large language models (LLMs) enable us to learn at a rapid pace. Creation of LLms from scratch are impractical. This thesis aims to collect a small dataset, containing biomedical papers, and use it to train large language models (LLMs) to extract entities from the text and learn the relationships between these entities. The experiment will be divided into two stages and utilize EU-ADR and ChemProt dataset. Starting with named entity recognition …

Domain-Specific Graph Rag Pipelines: Optimized Approaches For Building Efficient Personal Knowledge Repositories, Omkar Yadav

Master's Projects

Managing personal data, including notes, calendar events, to-do lists, and other personal information, has become increasingly complex and challenging. In response to this issue, I propose a framework using RAG that enables a large language model (LLM) to efficiently query this data without requiring training on the personal data itself. Conventional retrieval systems, including those leveraging vector-based retrievalaugmented generation (RAG), are effective at handling basic queries but struggle to deliver coherent global abstractions, integrate diverse knowledge sources, and account for temporal nuances. This research explores domain-specific Graph based RAG frameworks that incorporate a knowledge graph to better model relationships, thereby …

Enhancing Code Review Automation With Large Language Models Using Qlora Fine-Tuning And Rags, Sumukh Naveen Aradhya

Master's Projects

In this technological era where Artificial Intelligence and Machine Learning are revolutionizing various domains, Large Language Models (LLMs) are emerging as a very powerful tool. In the software development lifecycle, it is extremely important to have reliable code reviews to ensure security and maintain code quality. This project aims to survey various existing methodologies to aid creation of efficient code review automation agents and also research on ways to make this process more efficient. Parameter Efficient Fine-Tuning (PEFT) methodologies such as LoRA and QLoRA have been explored with an additional focus on a hybrid model that combines adaptive QLoRA with …

Prompt Engineering Principles For Generative Ai Use In Extension, Paul A. Hill, Lendel K. Narine, Aubree L. Miller

Journal of Extension

The prevalence of Generative AI (GenAI) and Large Language Models (LLMs) is increasing rapidly. For Extension professionals, the utilization of prompt engineering is key to leveraging GenAI and LLMs effectively. Prompt engineering involves crafting prompts that elicit desired LLM responses. This article discusses prompt engineering principles, providing examples and guidance. The application of prompt engineering in Extension is explored, showcasing its potential to enhance programs, deliver personalized advice, engage audiences, and disseminate research-based information. By learning prompt engineering skills, Extension professionals can harness the power of GenAI and LLMs, enhancing their ability to address complex challenges in the 21st century.

Sdebuddy - Code Documentation Using Large Language Models, Nischay Nagendra

Master's Projects

In this fast developing world of software development, it is crucial to maintain the quality of code and the developers’ productivity. This can be done effectively with good code documentation. SDEBuddy uses the latest generation of Large Language Models (LLMs) and finetuning procedures to create code documentation. In this project, state-of-the-art models such as Llama2 and Llama3 are employed to mimic the behavior of the given code and produce documentation. Such models are tuned for various programming languages and documentation formats using LoRA and QLoRA fine-tuning approaches. These models are evaluated in terms of the BLEU score, ROUGE score and …

Opinion Graphs Construction For Reviews Using Transfer Learning And Large Language Models, Yichen Lin

Master's Projects

With the rapid development of the Internet, reading online reviews before making a purchase, booking a hotel, or making a restaurant reservation has become a part of daily life. Customers often consider reviews as crucial supplementary information before making decisions on how to spend their money. However, reading many reviews to gain helpful information takes time and effort. This project proposes a new method OpinionGraphGenerator that aims to create opinion graphs from hotel reviews to reduce the high volume of text in reviews while preserving essential insights. In an opinion graph, vertices are semantically similar opinions, where each opinion consists …

Optimizing Web Design Code Generation: A Comparative Study Of Finetuning, Pretrained Models, And Rag (Retrieval Augmented Generation), Srinivas Rao Chavan

Master's Projects

Website Creation is revolutionized by automated code generation, reducing the development effort, speeding the production process, and ensuring consistency in design. Automated web design code generation has emerged as a transformative tool bridging the gap between design and development. In this research, a website design tool is developed and used to create visual layouts, exporting them as JSON designs. These JSON outputs were then transformed into textual prompts, optimized using established HCI principles and UI/UX rules to ensure consistency, visual hierarchy, aesthetics and minimalistic design, accessibility, user-friendly navigation and flexibility. These generated prompts were fed into large language models for …

Enhancing Environmental Health And Safety: Fine-Tuning Large Language Models For Domain-Specific Applications, Mohammad Adil Ansari

Master's Projects

This study aims to simplify Environmental Health and Safety (EHS) by leveraging the power of Large Language Models (LLMs). In this research, we focus on fine-tuning three LLMs — LLaMA, Mistral, and Falcon — using PEFT techniques such as QLoRA and SFT, to address domain-specific needs such as safety compliance, incident reporting, and knowledge dissemination. Our research methodology involves fine-tuning each LLM model on a custom dataset compiled from various regulatory agencies, supplemented by targeted web scraping and manual collection of questionnaires to capture and enrich the models with the latest regulations and guidelines. This study aims to compare the …

Employing Large Language Models And Retrieval Augmented Generation For Enhanced Predictive Flexibility In Cancer Mortality Prediction, Mridang Kejriwal

Master's Projects

Today, cancer is a major health risk to thousands of people, and there are over a two-hundred different types of cancer. Luckily, over the past several years, the outcomes and survival rates have increased, all thanks to machine learning, specifically Recurrent Neural Networks (RNN) and Long Short-Term memory (LSTM) networks. However, the current prognostic models don’t allow healthcare professionals to adapt the variables to mimic all the different features of every type of cancer, resulting in a model that works but is not as accurate as it could be. This study explores improving the accuracy and adaptability of the current …

Leveraging Large Language Models For Transforming Student Information Into Actionable Data, Sree Hari Karri

Master's Projects

Admission season places significant demands on university committees, necessitating the review of vast arrays of documents to assess students’ competence. This project advances the development of an automated system designed to streamline this process by evaluating application materials such as Letters of Recommendation (LoRs), Statements of Purpose (SoPs), and resumes. Utilizing a variety of advanced Natural Language Processing (NLP) techniques, the system compares the performance of several Large Language Model (LLM) approaches. It also experiments with different data handling strategies, including the use of vector stores versus traditional context-based processing, to optimize model efficiency and accuracy. Special attention is given …

Exploring The Use And Misuse Of Large Language Models (Llms), Hezekiah Paul D. Valdez

Master's Projects

Large Language Models (LLMs) have quickly gone from simple rule-based systems to complex knowledge bases capable of tackling many different tasks across a variety of fields. What began as an exercise in human-computer interaction has become the basis for artificial intelligence in a variety of mediums. When attached to larger systems, LLMs become generative assistants that can perform highly on human proficiency assessments and other benchmark skill assessments. This increase in proficiency has led these systems to be deployed in fields such as cybersecurity, business, and programming to help improve productivity and efficiency. However, such a wide availability has allowed …

Llamatalk: Empowering Conversations With Retrieval-Augmented Generation, Aravind Rokkam

Master's Projects

This research report talks about the implementation and a comparative study of Llama 7B model’s fine-tuning technique and Retrieval Augmented Generation (RAG) capabilities in the context of creating a reliable AI therapist. This study focuses on training these models using diverse datasets consisting of doctor-patient conversations predominantly addressing general health issues. Using a technique like fine-tuning within the Llama 7B model, the project focuses on training the model with a diverse dataset comprising doctor-patient interactions primarily addressing general health concerns. Additionally, carefully organized mental health dataset from HOPE dataset, ensuring the bot's responsiveness to mental health inquiries. Through integration with …

Fine-Tuning Large Language Models For Folder Structure Generation, Likhith Nemani

Master's Projects

Starting a new project is a significant challenge in the software development world. Building a new project skeleton and configurations will require vast amounts of time and effort. This project aims to overcome the difficulty presented by this challenge using advanced large language models, specifically fine-tuning LLMs. Our initial focus with the implementation is to use the powerful capabilities of advanced modern models to simplify and accelerate the complicated process of getting new projects started. The solution process begins with a user posting a README file to a predetermined repository. This README file then is used as a source for …

Nuancenet: Comparative Analysis Of Ai In Complex Language Interpretation For Disaster Detection, Pavan Koushik Kommuri

Master's Projects

Disaster Detection using Twitter content is critical for emergency response, but accurately identifying relevant tweets remains challenging due to nuances, informal language, and emotional expressions. This paper presents a comparative analysis between traditional Machine Learning models, Deep Learning models and Large Language Models (LLM) for classifying disaster vs. non-disaster tweets. While existing works have applied pattern recognition and dataset-specific learning, LLMs with their deeper understanding of linguistics and semantics can potentially handle the complexities of tweets more effectively. This study leverages LLMs including Llama2, Mistral, and Falcon, Open AI GPT 3.5, hypothesizing their superior contextual comprehension will excel in tweets …

Prediction Of 2024 Indian Pm Election Results Using Sentiment Analysis On Twitter Data, Surabhi Gupta

Master's Projects

This sentiments analysis study presents a methodical approach to predict the 2024 Indian Prime Minister Election. Data collected spanning from 2020 to 2023 from Twitter using hashtags such as IndianPMElection2024 and on topics such as the revocation of the special status of Jammu and Kashmir, the Farm Bill, and the Digital India initiative, form the core of this research. We utilized a combination of sentiment extraction tools-namely, the NLP Town's Bidirectional Encoder Representations from Transformers (BERT)-based multilingual uncased sentiment model, Valance Aware Dictionary for Sentiment Reasoning (VADER), and TextBlob. Additionally, we used a well-established machine learning model Naive Bayes, deep …