UWSpace

UWSpace is the University of Waterloo’s institutional repository for the free, secure, and long-term home of research produced by faculty, students, and staff.

Depositing Theses/Dissertations or Research to UWSpace

Are you a Graduate Student depositing your thesis to UWSpace? See our Thesis Deposit Help and UWSpace Thesis FAQ pages to learn more.

Are you a Faculty or Staff member depositing research to UWSpace? See our Waterloo Research Deposit Help and Self-Archiving pages to learn more.

Photo by Waterloo staff

Communities in UWSpace

Select a community to browse its collections.

Recent Submissions

  • Thumbnail Image
    Item type: Item ,
    Democratizing and Modernizing Information Access: From Open Rerankers to Scalable RAG Evaluation
    (University of Waterloo, 2025-12-09) Pradeep, Ronak
    Modern information access increasingly relies on complex pipelines involving large language models (LLMs), fundamentally changing how users interact with information, from sophisticated multi-stage retrieval pipelines to end-to-end retrieval-augmented generation (RAG) systems. While these advancements enhance user experience, they also introduce significant challenges. The research community's growing reliance on proprietary, black-box models for key tasks like document reranking creates barriers to innovation and reproducibility (the Component Challenge). Furthermore, progress is hampered by the lack of a shared, standardized ecosystem for executing and measuring information access systems (the Benchmarking Challenge). Finally, the generative nature of RAG systems makes them fundamentally harder to evaluate than traditional systems that return document lists; new methodologies are required to assess factual accuracy and completeness in a reliable, scalable manner (the Evaluation Challenge). We argue that progress depends on the synergistic development of open, high-effectiveness system components and the reliable, scalable evaluation frameworks necessary to assess them. This thesis addresses these challenges through a narrative arc that begins with pushing existing paradigms to their limits. But, given that frontier models dominate today's landscape, we find a pressing need for new open-source solutions. We begin by analyzing the dominant supervised ranking paradigm, developing multi-stage pipelines that demonstrated high effectiveness but also highlighted inherent complexity and cost. Subsequently, we conducted a systematic exploration of model backbones, loss functions, and negative mining strategies to squeeze effectiveness gains from supervised pointwise cross-encoders. Next, we continue with a large-scale empirical study on the newly evolving generative retrieval paradigm, which revealed its scalability limitations on large, real-world collections. This portion culminates in the final contribution to the Component Challenge: RankZephyr, an open-source 7B-parameter listwise reranker. By leveraging a carefully designed instruction distillation curriculum, RankZephyr matches and often surpasses the effectiveness of much larger proprietary models like GPT-4. It provides the community with a powerful, transparent, and accessible zero-shot reranking module, breaking the dependence on black-box systems for this critical task. All methods described have broad community adoption, and our models and evaluation frameworks continue to support ongoing research efforts across open-source IR and RAG development. With powerful open components in hand, the focus shifts to benchmarking. To address the Benchmarking Challenge, this work introduces Ragnarök, a reusable, end-to-end RAG framework designed to standardize how retrieval-augmented generation systems are constructed and assessed. Serving as the backbone for the TREC 2024 Retrieval-Augmented Generation Track, Ragnarök provides the research community with a shared experimental platform, critical data resources, and reproducible and effective baselines. By encapsulating the full RAG pipeline — from retrieval and grounding to generation and scoring — within a single, transparent framework, TREC 2024 Retrieval-Augmented Generation Track and Ragnarök enable reproducible experimentation at scale. This not only ensures fair comparisons across diverse approaches but also establishes a foundation for cumulative progress in open-domain information access research, where previously ad hoc and non-replicable setups have often impeded reliable evaluation. Building on this infrastructure, the thesis then directly tackles the Evaluation Challenge by introducing the AutoNuggetizer framework. This framework refactors the classic and well-studied nugget-based evaluation methodology for the modern era of LLMs. By automating the evaluation of the recall of the information nugget in RAG responses and validating the approach at scale in TREC 2024 Retrieval-Augmented Generation Track, this work provides a reliable and scalable methodology to measure the quality of generative information access systems. In summary, this thesis contributes to the field of information access by exploring the limits of existing retrieval and ranking paradigms, developing powerful open-source components for modern information access systems, and creating the frameworks and methodologies required to benchmark and evaluate them. The contributions include a comprehensive analysis of supervised ranking and generative retrieval paradigms, an open-source state-of-the-art listwise reranker (RankZephyr), a standardized framework for RAG benchmarking (Ragnarök), and a scalable methodology for evaluating generative systems (AutoNuggetizer). Together, this thesis addresses the three core challenges identified at the outset, providing the community with both the tools to build effective systems and the methodologies to assess them rigorously. The widespread adoption of these artifacts by researchers and practitioners already underscores their tangible impact and utility in driving the field forward. In the future, on the reranking front, we would like to build faster, more efficient rerankers that can reason over the texts and generalize to several domains. On the benchmarking front, we will expand tasks to capture "deep research" information needs that demand multi-hop reasoning and query decomposition. On the evaluation front, we hope to extend the AutoNuggetizer methodology to several tasks that go beyond web retrieval, into other domains like biomedical texts and conversational question answering.
  • Thumbnail Image
    Item type: Item ,
    Real-Time Assessment of Pain and Physical Activity in Knee Osteoarthritis: The Moderating Roles of Self-Efficacy, Fear of Movement, and Locus of Control
    (University of Waterloo, 2025-12-09) Almaw, Rachel
    Knee osteoarthritis (OA) is a highly prevalent chronic musculoskeletal condition and growing health concern in Canada. Characterized by loss of articular cartilage in the joint and fluctuating pain, knee OA is a leading contributor to physical and psychosocial disability (Steinmetz et al., 2023). By 2041, knee OA will affect over 10 million Canadians, driven by an aging and obese population (Sharif et al., 2017). Physical activity (PA) is recommended for knee OA management to reduce pain; but pain is commonly cited as a barrier to engagement (Bannuru et al., 2019). The links between PA exposure and pain are unclear. Current assessment tools cannot capture the temporal relationship between pain and PA in knee OA, as each dynamically influences the other throughout the day. This dynamic relationship is further complicated by psychosocial factors, including fear of movement, self-efficacy, and locus of control, which are associated with altered pain perception and PA behaviour in knee OA (Bayrak & Alkan, 2025). Yet, their roles in shaping the temporal relationship between pain and PA remain unclear in real-world contexts. Objectives and Hypothesis The primary objective of this study was to determine whether psychosocial factors (fear of movement, self-efficacy, and locus of control) moderate the temporal relationship between pain and subsequent PA in knee OA. The secondary objective examined the reverse temporal association: whether these psychological factors moderated the effect of PA on subsequent pain. It was hypothesized that (1) pain would reduce subsequent PA, with higher self-efficacy and internal locus of control weakening this negative effect and higher fear of movement strengthening the association; and (2) PA would increase subsequent pain, with higher self-efficacy and internal locus of control weakening this positive effect and higher fear of movement strengthening the association. Methods This prospective cohort study used ecological momentary assessment (EMA) over nine days. Fifty community-dwelling adults who had symptomatic knee OA consistent with American College of Rheumatology clinical criteria participated. Exclusion criteria included other forms of arthritis, neurological conditions, active cancer treatment, non-nociceptive pain, or lack of a compatible smartphone for EMA.Participants reported knee pain intensity, duration, and analgesic use four times daily using an EMA approach, via scheduled prompts delivered using a smartphone application (m-Path). Participant could also initiate prompts, unscheduled, to report pain flares. A pain composite score was calculated using z-score standardization of pain intensity and duration with participant-specific means and standard deviations, providing a novel within-person approach to characterizing knee OA pain. PA was continuously tracked using accelerometers (Actigraph wGT3X-BT) during waking hours. Psychosocial factors including fear of movement, self-efficacy, and locus of control were assessed through validated questionnaires at baseline. The primary objective examined whether the pain composite score influenced subsequent step count in the 90 minutes after EMA prompts, and whether this association was moderated by psychosocial factors (fear of movement, self-efficacy, locus of control). Generalized linear mixed models with Tweedie distributions and random intercepts examined these relationships, adjusting for age, gender, analgesic use, study day, prompt number, and prompt type. The secondary objective examined whether step count in the 90 minutes before EMA prompts influenced subsequent pain intensity, and whether this association was moderated by the same psychosocial factors. Linear mixed-effects models with random intercepts examined these relationships, adjusting for the same covariates. Results Complete data were available for 49 participants (64.8±7.2 years, 77% women) (1 accelerometer failure). These 49 participants reported moderate pain (KOOS Pain: 65.3±15.6), high daily activity (mean 11,661 steps/day), and mixed pain patterns (92% experienced both constant and intermittent pain). Average EMA compliance was 90%, with participants completing an average of 32 prompts over the study period. Valid accelerometer data was available for 96.6% of monitoring days. Fear of movement moderated the association between within-person pain composite score and subsequent step count (β = -0.0144, 95% CI [-0.0277, -0.0010], p = 0.035). Individuals with higher fear of movement showed a 4.9% greater step reduction per 1-SD increase (3.4 points on a 24-point scale) in fear of movement score. Self-efficacy for pain management also moderated this association (β = 0.0059, 95% CI [0.0018, 0.0100], p = 0.005). Individuals with higher self-efficacy showed a 6.8% smaller step count reduction per 1-SD increase (11.9 points on a 50-point scale) in self-efficacy score. Locus of control showed no moderation effect. In the reverse temporal direction, internal locus of control demonstrated a weak positive interaction with within-person step count on subsequent pain intensity (β = 0.000012, 95% CI [0.000002, 0.000022], p = 0.014). Individuals with higher internal locus of control experienced slightly greater pain intensity following higher step counts compared to those with lower internal locus of control. Self-efficacy for pain management demonstrated a significant main effect on subsequent pain intensity (β = -0.028740, 95% CI [-0.052042, -0.005439], p = 0.019). Individuals with higher self-efficacy reported lower pain intensity regardless of step count. Fear of movement showed no moderation or main effect. Discussion This study, using EMA and objective accelerometry, provides preliminary evidence that psychosocial factors moderate the temporal pain-PA relationship in adults with symptomatic knee OA. Those with higher self-efficacy for managing pain showed 6.8% smaller step reductions following greater pain; while those with higher fear of movement showed 4.9% greater reductions in step count after experiencing greater pain. In the reverse temporal direction, higher step counts were associated with increased subsequent pain intensity. Internal locus of control statistically moderated this association, though the interaction was very small and of uncertain clinical significance. Real-time assessment captured substantial within-person variability in pain and physical activity, including participant-initiated reporting of pain flares (77.6% of participants). The small interactions likely reflect considerable heterogeneity due to unmeasured contextual environmental factors. These findings suggest that self-efficacy, fear of movement, and locus of control may influence how individuals with knee OA respond to pain in daily life. While larger and more diverse samples are needed to establish clinical significance, these results suggest that self-management interventions targeting these factors and personalized pacing strategies during pain-free periods may help individuals with knee OA maintain activity despite pain fluctuations.
  • Thumbnail Image
    Item type: Item ,
    Spatial Ecology and Abundance of Turtles in Grundy Lake Provincial Park
    (University of Waterloo, 2025-12-09) Ross, Mackenzie
    Across the globe biodiversity is being lost at an unprecedented and unsustainable rate. The loss of biodiversity has direct and negative impacts on food production and other critical ecosystem services, to such a degree that we are currently experiencing a mass extinction event. Turtles are a group of vertebrates whose slow life histories and high juvenile mortality rates make them especially vulnerable to environmental changes. Anthropogenic climate change, habitat loss, and vehicle mortality are just a handful of factors responsible for an estimated 60% of turtle species being extinct or threatened with extinction. In Ontario there are currently eight native species of freshwater turtle, each are classified federally or provincially as at-risk species. Ontario represents the northern range periphery of these turtles presenting unique challenges including shorter active seasons and harsher winter weather. Grundy Lake Provincial Park is a protected area east of Georgian Bay Mnidoo Gamii in central Ontario. The three most common turtle species found in the park are the Blanding’s turtle, painted turtle and snapping turtle. Grundy Lake Provincial Park may be a stronghold for the globally endangered Blanding’s turtle, because of this in Chapter 2 we examined the spatial ecology of Blanding’s turtles at Grundy Lake Provincial Park by evaluating (1) habitat use, (2) home range and space use areas, (3), habitat selection, (4) minimum daily distance travelled, and (5) travel corridor identification. We found that both sex and behavioural season significantly impacted the spatial ecology of Blanding’s turtles, especially females during the nesting season. We identified travel corridors and areas of high-use in the park, highlighting areas important for habitat connectivity and nesting migrations. In Chapter 3, we examined population abundance and nesting distribution of all three species of freshwater turtle; our goal was to (1) estimate population abundance, (2) estimate somatic growth rate, (3) determine areas of high nest density, (4) evaluate carapace length distribution, and (5) estimate biomass and density. We estimated an abundance of 183 adult female painted turtles, 100 adult Blanding’s turtles and 63 adult female snapping turtles. Across species, somatic growth rate was greater in juveniles compared to adults. There was evidence of sexual size dimorphism in painted and Blanding’s turtles. Biomass and density estimates were highest for snapping turtles despite a low population estimate, emphasizing their role in the ecosystem. The turtle population in Grundy Lake Provincial Park represents an important stronghold near their northern range limit and offers a critical opportunity for long-term research in a protected area.
  • Thumbnail Image
    Item type: Item ,
    Classification Results for Intersective Polynomials With No Integral Roots
    (University of Waterloo, 2025-12-08) Banks, Nicolas
    In this thesis, we algebraically classify strongly intersective polynomials - polynomials with no integer roots but with a root modulo every positive integer - of degree 5--10. In particular, we compute a list of possible Galois groups of such polynomials. We also prove constraints on the splitting behaviour of ramified primes (i.e. primes that ramify in a splitting field of the polynomial). In the process, we show that intersectivity can be thought of as a property of a Galois number field, together with its set of subfields of specified degrees. This was achieved with characterisations of Berend-Bilu and Sonn, the latter of which we also generalise. Implementations in SageMath and GAP are provided. We also utilise Hensel's Lemma and other standard results on the local behaviour of simple field extensions.
  • Thumbnail Image
    Item type: Item ,
    Hardware-Assisted Defenses for Data Integrity and Confidentiality
    (University of Waterloo, 2025-12-08) ElAtali, Hossam
    The increasing complexity of modern computing systems and their exposure to the internet expose sensitive data to a range of security threats from remote adversaries. Bugs in software can lead to run-time attacks that gain direct access to sensitive data in memory, compromising its integrity and confidentiality. Furthermore, hardware and/or compiler optimizations can introduce data-dependent behavior that expose sensitive data to side-channel leakage, even in the absence of software bugs, breaking confidentiality. As business needs evolve, different usage scenarios, such as outsourced computation, have gained popularity, making the task of protecting data integrity and confidentiality more complex. This dissertation investigates how the integrity and confidentiality of sensitive data at run-time can be efficiently preserved through hardware-assisted mechanisms. I consider a range of usage scenarios and threat models, from protecting data sent to remote servers for outsourced computation by untrusted code, to protecting data processed locally from other vulnerable or malicious parts of the system. Specifically, this dissertation addresses: 1. how to efficiently protect data confidentiality against side-channel leakage with negligible overheads. Existing solutions to side-channel leakage suffer from significant overheads, making their deployment difficult in situations where performance is critical. I address this problem with CacheSquash, a software-transparent hardware mechanism to effectively harden against transient side-channel attacks such as Spectre and Meltdown with near-zero overheads. 2. How to combine protections against both direct access and side channels. I propose BliMe, a novel architecture that relies on remote attestation, taint-tracking and hardware-enforced data obliviousness to protect sensitive data processed by untrusted code in an outsourced computation setting. 3. For integrity, I propose PBI, a novel hardware primitive that enables efficient memory protection for sandboxing and in-process isolation, thereby safeguarding both data confidentiality and integrity. 4. Finally, I address how to efficiently combine memory safety and side-channel protection mechanisms for data integrity and confidentiality. For this, I propose BLACKOUT, a hardware-software extension to CHERI that enforces data-oblivious computation on sensitive data, and inherits the memory safety properties of CHERI, all while introducing minimal overheads. The proposed solutions confirm that hardware-assisted mechanisms can indeed be used to efficiently protect data at run-time, both from direct access and side-channel leakage. I conclude my dissertation with promising directions for future work.