Global ETD Search
Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.
Results
Showing 1 to 20 of 36 for “"Reliable AI"”.
-
Towards Safe And Reliable AI Models
… applications have become indispensable in our daily lives. Emerging advanced AI models, from general-purpose Large Language Models (LLMs) such as ChatGPT and Deepseek to domain-specific systems such as autopilot for autonomous vehicles, are changing the way we work and live by increasing …
-
Building Reliable AI under Distribution Shifts
… where distribution shifts—differences between training and deployment data—can significantly impact their reliability. These shifts affect models in multiple ways, leading to degraded generalization, fairness collapse, loss of robustness, and new safety vulnerabilities. This dissertation …
-
Ontology-Grounded Knowledge Graphs for Reliable AI Documentation
… professionally while introducing unsupported claims, omitting key details, or subtly shifting what the source supports, forcing costly line-by-line verification. This verification burden limits adoption in high-stakes settings, especially clinical documentation, where time savings are often …
-
Trust through Transparency: Towards Reliable AI for All
… and uneven ways, from image classifiers failing to recognize an otter out of water, to LLMs being nearly 3 times worse at recalling facts about Somalia than Sweden. In this dissertation, I’ll detail interpretability techniques to scalably illuminate and efficiently intervene on discovered …
-
Leveraging Metaplasticity and Physical Constraints for Reliable AI
… extreme flexibility, provided by millions of trainable parameters. However, this flexibility comes at the expense of making deep learning models prone to overfitting their training data. As a result, models become vulnerable to noise-based adversarial attacks, develop rigid decision boundaries, …
-
SCREENING OUT-OF-DISTRIBUTION INPUTS FOR RELIABLE AI PREDICTIONS
… such as Deep neural networks (DNNs) across domains, there is significant interest in deploying these models in real-world AI systems. DNNs are, however, known to produce incorrect predictions on inputs from scenarios that are less likely to occur according to their training distribution. This …
-
TOWARDS RELIABLE AI UNDER DISTRIBUTION SHIFTS: A DATA-CENTRIC PERSPECTIVE
… often rely on spurious correlations in the training data, leading to performance degradation and unreliability when processing inputs under distribution shifts. This thesis systematically studies the robustness to distribution shifts for ML models from a data-centric perspective. First, we …
-
Toward reliable AI: Practical and certified defenses for adversarially robust vision systems
… achieved remarkable success across numerous domains, yet its vulnerability to adversarial perturbations, as first revealed in 2013, has significantly limited its applicability in critical and safety-sensitive applications. Beyond the practical concerns, this phenomenon is intrinsically …
-
Building reliable AI assistive technologies: a comprehensive process for inclusive and transparent co-design
Artificial Intelligence (AI) is used to improve assistive technologies, but these systems can fail in various ways. Despite being designed for individuals, many technologies lack user engagement from design to evaluation. This raises the question for users: “Can I trust this technology to perform …
-
Scaling Real-world Simulation for Safe Autonomy
… artificial intelligence with physical systems aims to empower robots with superintelligence, enabling them to perform diverse tasks in real-world settings. However, current AI-based autonomous systems excel mainly in controlled setups and struggle in new, unstructured, and extreme conditions. …
-
Language Modeling in Finance: Evaluation, Extraction, and Economic Insight
… opportunities and challenges for high-stakes domains such as finance. This work positions finance as both a critical application area and a rigorous diagnostic lens for evaluating the reasoning, bias, and reliability of Large Language Models (LLMs). It uncovers structural limitations of …
-
Helping Women Help Themselves: Women Entrepreneurs and the Displaced Labour of Responsible Generative AI
Generative Artificial Intelligence (GenAI) is increasingly promoted as a means of narrowing the gender gap in entrepreneurship, yet women adopt these tools at lower rates than men, and existing research explains this gap largely through deficits in confidence, skills, and access. This study shifts …
-
Glioma Segmentation in MRI using a 3D Hybrid U-NET with Adaptive Self-Attention and Multi-Modal Fusion
… MRI modalities such as T1, T1Gd, T2, and FLAIR provide complementary information for identifying clinically important tumor sub-regions, including enhancing tumor (ET), tumor core (TC), and whole tumor (WT). However, many existing segmentation methods do not fully preserve modality-specific …
-
ENHANCING TRUSTWORTHINESS AND SAFETY IN FOUNDATION MODELS
… neighbor image retrieval, ensuring resistance against adversarial manipulation. We further propose adversarial weight perturbation to improve the generalization of graph neural networks under adversarial conditions, and develop a law of robustness beyond Isoperimetry to establish a new …
-
Comparing human and generative AI annotation performance: Understanding linguistic features of teacher talk in classroom discussions
Generative artificial intelligence (AI) proves to be a powerful tool which has been implemented in many fields; however, it often lacks the specialized knowledge necessary to be successfully incorporated into fields such as teacher education and training. Using discourse analysis, three annotators …
-
Towards interpretable and controllable machine learning models via logic reasoning
… distribution shifts. Addressing these issues remains at the center of developing trustworthy ML systems for real-world applications. Our research focuses on providing a principled solution to these issues through logic reasoning formalism. Specifically, we study the fundamental technique of …
-
Design, optimization, development and explanation of artificial intelligence algorithms, capable of running on distributed big data systems, based on multidimensional and complex datasets
… and advancements in Artificial Intelligence (AI) have redefined data-driven decision-making across industries. However, managing complex, high-dimensional, and multi-frequency datasets remains challenging, particularly in finance and cloud computing sectors where data variability, real-time …
-
Toward robust and fair ultrasound classification with clinical context
… image analysis, yet its clinical translation remains limited by three persistent challenges: imperfect supervision, lack of clinical contextualization, and poor generalization under subpopulation shift. This thesis addresses these challenges through a unified research program focused on …
-
Artificial Intelligence for Resilience in Smart Grid Operations
… is defined as the ability to harden the system against and quickly recover from high-impact, low-frequency events. The introduction of two-way flows of information and electricity in the smart grid raises concerns of cyber-physical attacks. Proliferated penetration of renewable energy sources such …
-
Hierarchies, search, and generative models in sequential decision-making
Intelligent agents trained using deep reinforcement learning (DRL) or imitation learning (IL) can achieve or even surpass expert-level performance in various controlled settings. However, translating this success to real-world tasks remains difficult, especially in high-stakes domains such as …
Page 1 of 2