Digital Commons Network^™

Purifai: Detecting And Fixing Search-Induced Distortions In Web-Augmented Llms, Guoqing Wang, Zhao Zhang, Zeyu Sun, Xiaofei Xie, Yizhou Chen, Yanchao Tan, Dan Hao

Research Collection School Of Computing and Information Systems

As Large Language Models (LLMs) increasingly serve as interfaces for proprietary data (e.g., enterprise knowledge bases, legal statutes), ensuring their fidelity to trusted internal information is paramount. While integrating real-time web search can enhance model utility, it introduces a critical vulnerability: the ingestion of conflicting, misleading, or hallucinated content from the open web can override the model's adherence to its verified internal knowledge. We define this failure mode as search-induced distortion, a significant risk in high-stakes domains where the internal knowledge base serves as the absolute ground truth.To address this challenge, we present PurifAI, a proactive, model-agnostic, cache-level purification system …

Verbalizing Lightgcn: Direct Learning Of Textual Representations From User-Item Interaction Graph Via Llms, Manh-Khanh Ngo Huu, Hady Wirawan Lauw

Research Collection School Of Computing and Information Systems

In this work, we propose VerbaLightGCN, a novel LLM-based recommendation framework that integrates the semantic understanding of LLMs with user-item interaction modeling. Traditional collaborative filtering (CF) models typically embed user and item IDs into a latent space to capture interaction signals. However, pretrained LLMs cannot natively interpret these learned embeddings. To bridge this gap, VerbaLightGCN adopts a CF-as-text paradigm, in which collaborative signals are encoded in textual form and directly learned from the user–item interaction graph, and are then combined with semantic information to construct user and item profiles that function as latent embeddings. Inspired by LightGCN, our method retains …

Generation-Augmented Video Corpus Moment Retrieval, Mingjin Kuai, Qianyin Xiao, Juncheng Li, Jin Peng, Lizi Liao, Wei Ji

Research Collection School Of Computing and Information Systems

Video Corpus Moment Retrieval (VCMR) requires models to efficiently retrieve and precisely locate specific moments relevant to natural language queries within a massive, untrimmed video corpus. However, existing discriminative approaches typically rely on shallow visual-textual feature matching mechanisms, which often struggle to capture fine-grained semantic differences. To address this limitation, we propose Video-GAR, a novel framework that reframes the conventional retrieval task from superficial matching to generative understanding, positing that the capability for query reconstruction evidences deep semantic comprehension. Specifically, Video-GAR orchestrates three synergistic components: To overcome the computational efficiency bottleneck, we construct a Bi-Mamba backbone that leverages the linear …

Variational Speculative Decoding: Rethinking Draft Training From Token Likelihood To Sequence Acceptance, Xiandong Zou, Jianshu Li, Jing Huang, Pan Zhou

Research Collection School Of Computing and Information Systems

Speculative decoding accelerates inference for (M)LLMs, yet a training-decoding discrepancy persists: while existing methods optimize single greedy trajectories, decoding involves verifying and ranking multiple sampled draft paths. We propose Variational Speculative Decoding (VSD), formulating draft training as variational inference over latent proposals (draft paths). VSD maximizes the marginal probability of target-model acceptance, yielding an ELBO that promotes high-quality latent proposals while minimizing divergence from the target distribution. To enhance quality and reduce variance, we incorporate a path-level utility and optimize via an Expectation-Maximization procedure. The E-step draws MCMC samples from an oracle-filtered posterior, while the M-step maximizes weighted likelihood using …

Train In Vain: Functionality-Preserving Poisoning To Prevent Unauthorized Use Of Code Datasets, Yuan Xiao, Yuchen Chen, Jiaming Wang, Wei Song, Jun Sun, Shiqing Ma, Yanzhou Mu, Juan Zhai, Chunrong Fang, Jin Song Dong, Zhenyu Chen

Research Collection School Of Computing and Information Systems

The widespread availability of large-scale code datasets has accelerated the development of code large language models (CodeLLMs), raising concerns about unauthorized dataset usage. Dataset poisoning offers a proactive defense by reducing the utility of such unauthorized training. However, existing poisoning methods often require full-dataset poisoning and introduce transformations that break code compilability. In this paper, we introduce FunPoison, a functionality-preserving poisoning approach that injects short, compilable weak-use fragments into executed code paths. FunPoison leverages reusable statement-level templates with automatic repair and conservative safety checking to ensure side-effect freedom, while a type-aware synthesis module preserves type correctness, suppresses static-analysis warnings, and …

Itimo: An Llm-Empowered Synthesis Dataset For Travel Itinerary Modification, Zhuoxuan Huang, Yunshan Ma, Hongyu Zhang, Hua Ma, Zhu Sun

Research Collection School Of Computing and Information Systems

Addressing itinerary modification is crucial for enhancing the travel experience as it is a frequent requirement during traveling. However, existing research mainly focuses on fixed itinerary planning, leaving modification underexplored due to the scarcity of shape need-to-modify itinerary data. To bridge this gap, we formally define the itinerary modification task and propose a general pipeline to construct the corresponding dataset, namely iTIMO. This pipeline frames the generation of shape need-to-modify itinerary data as an intent-driven perturbation task. It instructs large language models to perturb real-world itineraries using three operations: REPLACE, ADD, and DELETE. Each perturbation is grounded in three intents: …

Unmasking Twitter Bots: An Applied Machine Learning Approach, Rayane El Raba’A, Layal Abu Daher

BAU Journal - Science and Technology

The rapid growth of social networks has led to increased challenges, such as fraud, cyberbullying, and the spread of automated accounts (bots). Detecting anomalies within these networks is essential to maintaining security and trust. This study explored machine learning algorithms: Random Forest, XGBoost, Support Vector Machine (SVM), and Logistic Regression for anomaly detection in social networks, specifically focusing on Twitter bot identification, By applying AI-driven data mining techniques to a dataset of 37,438 Twitter bot accounts dataset, the research evaluates the effectiveness of these models in detecting unusual patterns. XGBoost achieved the highest accuracy (84.9%), with an ROA_AUC of 0.87, …

Enhanced E-Commerce Recommendation Experience With Collaborative Sentiment Analysis And Ranked Content-Based Filtering, Sandi El Zein, Farah Hamze, Amani Merhi, Lynn Noureddine, Lama Affara

BAU Journal - Science and Technology

The rise of online shopping has made informed purchasing decisions increasingly difficult, as consumers face an overwhelming number of product choices and struggle to manually evaluate specifications and user reviews. This paper presents an AI-powered, sentiment-aware product recommendation tool that effectively aligns user preferences with real-world customer feedback from online reviews. The proposed system utilized Instruct ABSA deep learning models for feature extraction and DeBERTa-v3 for sentiment analysis to turn reviews into interpretable scores that would nominate optimal products. An interactive Rasa-based chatbot interface, TopPickAI, was developed to give a seamless user experience, educate users on product features, and conversationally …

Generative Endurance Logic: An Axiomatic Framework For Reasoning About Outcome-Generating Objects Under Constraints, Hafiz Burhan Ul Haq, Muhammad Nauman Irshad

Neutrosophic Systems with Applications

This paper introduces Generative Endurance Logic (GEL), a formal framework for studying objects through the outcomes they can produce. In many cases, an object cannot be judged only by a fixed truth value, score, or utility value. A rule, model, action, or strategy may behave well in one situation but fail when the context changes or when small perturbations occur. GEL addresses this issue by treating each object as a generator of outcomes. Each object a is linked to a generation map Ga:X×Ω→Y, where X is the context space, Ω is the …

An Interactive Multi-Objective Programming Approach For Optimizing Fully Trapezoidal Spherical Fuzzy Linear Programming Problem With Application, Sultan S. Alodhaibi, Hissah Ibrahim Almuzini, Hamiden Abd El-Wahed Khalifa

Neutrosophic Systems with Applications

In this paper, a linear programming framework with completely uncertain parameters is investigated by employing trapezoidal spherical fuzzy numbers (TrSFNs). The proposed formulation incorporates a spherical fuzzy (SF) decision environment in which the optimization process simultaneously maximizes the degree of positive membership while minimizing the corresponding neutral and negative membership degrees. By utilizing the concept of the α -cut associated with TrSFNs, the original fully fuzzy linear programming problem is transformed into an interval-valued linear programming model with confidence levels. To rank and compare the resulting interval objective values, an interval ordering approach based on the decision maker's preferences—considering the …

Interval-Valued Neutrosophic Dombi Bonferroni Mean Aggregation Operators In Medical Diagnosis And Sustainable Energy, Maryam Faisal, Muhammad Nadeem, Muhammad Kamran

Neutrosophic Systems with Applications

Medical diagnosis is one of the most difficult fields in which decisions must be made due to the fact that medical information often has characteristics of uncertainty, incompleteness, imprecision and even contradiction. Traditional aggregation and decision-making methods are often not well suited to such complexities, and may result in less reliable diagnostic outcomes. In order to overcome these drawbacks, the authors propose a new approach using a novel representation of Interval-Valued Neutrosophic Sets (IVNSs), the Dombi operational laws, and Bonferroni Mean (BM) aggregation operators. The proposed framework is specifically aimed at coping with uncertainty, indeterminacy and falsity all at once …

Exploiting Uncertainty Of Computational Methodology In Optimizing User Interface In Human-Computer Interaction, Nada Mohamed, Alshaimaa A. Tantawy

Neutrosophic Systems with Applications

Human-computer interaction (HCI) evaluation and optimization of user interfaces (UIs) constitute a complex multi-criteria decision-making challenge, marked by conflicting evaluation dimensions, subjective expert judgments, and inherent uncertainty in user experience assessment. Traditional evaluation approaches, such as heuristic expert reviews and user satisfaction surveys, rely on sharp, binary classifications that fail to capture the gradual and overlapping nature of human cognitive and affective states. This limitation necessitates a more robust uncertainty-aware methodology that can model the true complexity of HCI evaluation. This paper proposes a hybrid mathematical model that integrates various Multi-Criteria Decision Making (MCDM) techniques of Entropy, and Simple Additive …

Recursive Neutrosophic Superhypergraphs With Illustrative Applications, Takaaki Fujita, Ajoy Kanti Das, Suman Das, Sankar Prasad Mondal, Volkan Duran

Neutrosophic Systems with Applications

Finite hypergraphs generalize ordinary graphs by permitting each hyperedge to join any nonempty set of vertices, and thus provide a natural model for truly multiway interactions. To represent hierarchical and multi-layer structure, SuperHyperGraphs iterate the powerset operation so that set-valued entities created at one level can be treated as vertices at higher levels. Independently, recursive hypergraphs allow edge recursion: an edge may contain not only vertices but also lower-level edges, yielding nested (and possibly self-referential) incidence controlled by a specified recursion depth. In this work we introduce and axiomatize Recursive Neutrosophic SuperHyperGraphs, a unified framework that combines vertex …

A Scoping Review Of Sycophancy In Large Language Models: Operational And Theoretical Recognition, Kallen Zhou, Manning Littlejohn, Isabella Garrard

Endeavors: Mississippi State Undergraduate Research Journal

As large language models (LLMs) usage grows across different domains, sycophancy, the tendency for output to align with users, is increasingly being recognized as a primary issue arising from applying LLMs into critical areas. Current research has provided a variety of theoretical definitions, mitigation techniques, and quantification for sycophancy. However, there is little to no consistency across different papers. This scoping review seeks to connect different works on LLM sycophancy by identifying themes in theoretical definitions, measurement methods, and inducement techniques of sycophancy. By analyzing 26 papers (preprints, conference proceedings, and journal articles) from arXiv, ACL Anthology, and Scopus, this …

Building Trustworthy Information Systems: A Unified Framework For Comparative Risk Detection, Parisa Momeni

USF Tampa Graduate Theses and Dissertations

Risk detection in large scale information systems increasingly depends on heterogeneous data generatedby both centralized and distributed ecosystems. While centralized systems provide curated and validated reports, distributed environments produce large-scale and real-time observational evidence. Existing computational approaches analyze these ecosystems in isolation, limiting systematic comparison of risk repre-sentations across heterogeneous sources.

This dissertation presents a unified computational framework for comparative risk detection across centralized and distributed information systems. The framework provides a domain independent methodology for transforming heterogeneous risk reporting data into comparable multidimensional representations. To enable interpretable comparison of heterogeneous risk distributions, this work introduces the Geometric Overlap Score …

Language Models For Oncology Clinical Text: How Model Architecture And Data Strategies Shape Tumor Phenotype Extraction And Disease Progression Detection, Thanh Duong

USF Tampa Graduate Theses and Dissertations

The rapid growth of electronic health records (EHRs) has created new opportunities to apply machine learning to clinical data.However, a large portion of important clinical information is still stored in unstructured text, such as pathology reports, radiology reports, and longitudinal clinical notes.These documents contain key details about tumor characteristics, diagnoses, treatments, and patient outcomes.Extracting structured and useful information from this text is challenging due to complex medical language, varied document formats, and the need to combine information across multiple reports over time.This dissertation studies how language models can be designed and adapted to better extract and use oncology-specific information from …

Leveraging Spatial Statistics For Domain Adaptation Of Vision Language Models In Medical Vqa, Himanshu Raj

Master’s Dissertations

Recent advances in Vision–Language Models (VLMs) have demonstrated strong performance in Medical Visual Question Answering (Medical VQA) task. Although they perform very well within their domains, these models often experience issues with their generalization ability on unknown clinical distribution data because of different imaging technologies and patient groups used in various medical facilities. Generalization problems faced by these models make their practical application in the field of VLM-based medical VQA systems rather difficult. To overcome this limitation we proposed our method named Spatial Semantics Aware Domain Adaptation (SSADA), which is an integrated framework that combines both finetuning and prompt-based in-context …

Artificial Intelligence For Deep Earth Science: Key Challenges, Major Application Scenarios And Development Pathways, Qingyun Di, Liang Zhao, Yikang Zheng, Zhi Geng, Zhichao Yu, Xiaocai Shan, Chao Li, Zhiyao Xu, Pengfei Lv

Bulletin of Chinese Academy of Sciences (Chinese Version)

Deep Earth science is central to understanding Earth’s internal architecture and the coupled evolution of its major spheres, while also underpinning energy security, the supply of critical mineral resources, and resilience to major geohazards. Nevertheless, the advancement of deep Earth science is currently hindered by insufficient in situ observations under extreme conditions, the difficulty of integrating multi-source heterogeneous data, and the limited capability to model complex multiphysics coupling processes. Recent advances in artificial intelligence offer a potential route beyond these limitations. By integrating data-driven learning with physical and geological understanding, AI is reshaping deep Earth science from empirical interpretation to …

American Sign Language Recognition And Analysis Using Deep Learning, Saurabh Kumar Soni

Master’s Dissertations

In this work I build a system that recognizes isolated American Sign Language (ASL) words, and I use it to ask one fairly direct question: when training data is scarce, is it better to look at the video pixels or at the geometry of the signer’s body? To find out, I train two very different models on exactly the same clips. The first is appearance-based. Every frame is run through standard preprocessing and a ResNet50 backbone pre-trained on ImageNet, which turns it into a 2048-dimensional feature vector, and a Bidirectional LSTM then reads that sequence over time. The second model …

Enhanced Embedding For Multimodal Medical Visual Question And Answering, Akash Suna

Master’s Dissertations

Visual Answering of questions in the field of Medical which is called as (VqA) has grown as a dominant area of research that fuse processing of natural language and vision of computer often known as CV or NLP to assist in medical decision-making. However, effective multimodal fusion between medical images and clinical questions remains a significant challenge. This thesis examines the application of the Perceiver IO architecture as an efficient multimodal aggregator for medical VQA. The work has been carried out in multiple directions. First, a classification-based framework is developed by combining Vision Transformer (ViT) and ClinicalBERT alongside a Perceiver …

Efficient Multimodal Foundation Model Tuning For Hallucination Mitigation, Fei Zhao

ETDs from 2020-2029

Over the past few years, multimodal foundation models have achieved remarkable progress in perception and understanding. However, two challenges limit their reliability: (1) dependence on offline training, which in most real-world settings requires large volumes of labeled data and, as a result, hinders the model’s ability to adapt to new data or domains; (2) weak cross-modal grounding, which often leads to hallucinated content generation, producing descriptions that are linguistically fluent but inconsistent with the input visual evidence. This dissertation frames hallucination mitigation as an outcome of transitioning from fixed learning (static, offline fine-tuning) to adaptive, feedback-driven lifelong learning. By incorporating …

Development Of An Adaptive Pelican Crossing Model Using Fuzzy Logic In Mixed Traffic Conditions, Manazil Adam, Andyka Kusuma, R. Jachrizal Sumabrata

Smart City

Traffic management at at-grade pedestrian crossing facilities (pelican crossings) in highly populated areas, such as the Universitas Indonesia Station, faces significant inefficiency challenges. During peak hours, the fixed-time system is frequently disabled and replaced with subjective manual control by security personnel, thereby triggering irregular stop-and-go cycles and a high accumulation of vehicle delays. This study aims to develop a hybrid adaptive control model integrating Computer Vision, Genetic Algorithm (GA), and Fuzzy Logic to optimize intersection performance under mixed traffic conditions. The research methodology begins with the extraction of traffic and pedestrian characteristic data, calculated manually through recorded field observations. This …

Design And Evaluation Of A Code-Switching-Aware Multilingual Conversational Ai System Using Advanced Rag Architectures, Ashutosh Juvale

Master’s Dissertations

Conversational artificial intelligence has become the primary interface through which hundreds of millions of users in India seek information and customer support. Yet the way these users actually write and speak is fundamentally at odds with the monolingual assumptions baked into most retrieval and generation systems: they code-switch, fluidly mixing one or more of the twenty-two scheduled languages of India with English, frequently typing Indic words in the Roman script ("mera refund kab tak aayega"). Standard Retrieval-Augmented Generation (RAG) pipelines silently fail on such input — the retriever returns off-topic passages because the query and the knowledge base live in …

A Comprehensive Review Of The A* Algorithm: Evolution, Applications, And Future Trends In Path Planning, Saleel H. Abood, Hussein M. H. Al-Khafaji, Mohanned M. H. Al-Khafaji

Journal of Soft Computing and Computer Applications

Despite being a fundamental problem to autonomous robotics and intelligent navigation systems, path planning is still a challenge. The A^* algorithm is often used among search-based techniques for optimal search performance, as it's a tradeoff of computation. The above techniques have been developed for various applications as many versions of A^* Dynamic A^* (D^*), D^* Lite, Hybrid A^*, and Anytime A^* are suggested to deal with dynamic environments, real-time constraints, and kinematic restrictions. This paper comprehensively and structurally reviews the A^* algorithm and its major extensions, encompassing historical development, methodological …

Developing A Model To Generate More Digital Data Of Indian Languages For Multilingual Applications, Arya Bagde

Master’s Dissertations

Most of India’s scheduled languages remain critically under-served by language technology because parallel (translated) text — the raw material that modern multilingual systems depend on — is extremely scarce. Back-translation can synthesise such data automatically, but its quality varies enormously, and unfiltered synthetic data can be worse than no data at all. This dissertation develops a framework that generates synthetic parallel data for four low-resource Indian languages spanning three language families and four scripts — Assamese (Indo-Aryan, Bengali script), Bodo (Tibeto-Burman, Devanagari), Manipuri (Tibeto-Burman, Bengali script) and Santali (Austroasiatic, Ol Chiki)—and introduces CASCADE, a learned multi-signal quality gate that scores …

Load Profile Analysis And Forecasting For Rural Mini Grids In Uganda, Prossy Mutesi, Santos L. Kihwele, Emmanuel S. Matee

Tanzania Journal of Science

Accurate load forecasting is essential for the reliable and cost-effective operation of rural mini grids, where constrained generation capacity and high penetration of renewable energy resources require well-informed operational decisions. This study examines electricity demand characteristics and forecasting performance for the Buzaami and Ssenyondo mini grids in Uganda, with particular focus on diurnal load profiles, peak demand behavior, and seasonal variability. 2022 operational data show extended peak demand from early morning to late evening, driven by socio-economic activities that strain resource scheduling and reliability management. To address these challenges, the study evaluates and compares Long Short-Term Memory (LSTM) networks, fuzzy …

A Switch-Point-Aware Contrastive Approach To Sentiment Analysis Of Hinglish Code-Mixed Text, Prasant Kumar Sahoo

Master’s Dissertations

With the increasing use of social media in non-English-speaking regions, especially in India, people often use Romanized Hindi and English together in their online communication. In a single sentence, they frequently mix Romanized Hindi and English, creating code-mixed text. However, most multilingual transformer models are pre-trained primarily on monolingual data. As a result, NLP systems face challenges when processing code-mixed text, as a single word may be fragmented into meaningless subword pieces, making it difficult for the model to capture its semantic meaning accurately. In this dissertation, we propose a parameter efficient neural architecture consisting of three main components to …

Automatic Glossing In Under-Resourced Languages: Case Studies In Bribri And Cook Islands Māori, Carter D. Anderson

Linguistics Undergraduate Senior Theses

Interlinear glossing is a major task in Indigenous language documentation. In this paper, I explore how effectively two Large Language Models, ByT5 and Gemini 2.5 Flash, can produce interlinear glossed text. I also examine how prompting an LLM with different types of information (dictionary entries, other training samples, and translations) can augment model performance. I apply these models to two under-resourced Indigenous languages: Bribri, which is morphologically complex from Costa Rica, and Cook Islands Māori, which has a simpler morphology and is from the Cook Islands in the Pacific Ocean. ByT5 exhibits much better performance when glossing Cook Islands Māori …

Beyond The Best Prompt: A Coverage View Of Multilingual Reasoning, Harshiv Mistry

University Honors Theses

Multilingual LLMs reason more accurately in English than in other languages, and recent work links part of this gap to reasoning behavior: native-language traces contain fewer cognitive behaviors (verification, backtracking, subgoal setting, backward chaining) that support effective problem solving. We test whether prompting for these behaviors at inference time narrows the gap, across seven conditions varying chain-of-thought, instruction and reasoning language, and cognitive-behavior descriptions, on two models, three languages. We find that English-scaffolded reasoning is the strongest single strategy on both models, closing the Hindi gap on Qwen, though the explicit scaffold's value over plain chain-of-thought is model-dependent. Beyond aggregate …

The Intersection Between Mindfulness And Cybersecurity: A Tool To Reduce Burnout And Improve Operational Effectiveness, Ivo Ricardo Dias Rosa

Journal of Cybersecurity Education, Research and Practice

Abstract: This paper offers a conceptual discussion of how mindfulness, understood as present moment awareness and deliberate attention regulation, can support cybersecurity professionals. Drawing on a narrative synthesis of workplace mindfulness, burnout, and high pressure decision making literature, we map plausible self regulation mechanisms to typical cyber defense tasks. Rather than presenting new empirical data, we develop an explanatory framework linking attention, reactivity, and recovery to decision quality, team communication, and adherence to incident playbooks. We focus on two connected outcomes: reducing burnout in roles with sustained cognitive and emotional demands, and improving operational effectiveness during critical situations such as …