http://www.rssboard.org/rss-specification 720 https://escholarship.org/uc/ucsd_etd/rss Recent eScholarship items from UC San Diego Electronic Theses and Dissertations Sun, 30 Aug 2026 04:52:54 +0000 https://escholarship.org/uc/item/7n09z9qt Extrachromosomal DNA (ecDNA) is associated with aggressive tumor biology, but its role in tumor immune escape remains incompletely understood. We combined ecDNA annotations, RNA-seq data, somatic mutations, HLA presentation metrics, tumor purity estimates, and immune deconvolution results from a TCGA pan-cancer cohort. Our goal was to determine if ecDNA-positive tumors are immune-cold and to understand the mechanism behind this phenotype. Once adjusted for cancer type and tumor purity, tumors with ecDNA showed decreased cytolytic activity, a weaker CD8 signal, and reduced inferred cytotoxic immune infiltration, and these effects became more severe with higher ecDNA loads. Conversely, ecDNA-positive tumors did not reveal significant decreases in PHBR-based antigenicity metrics or strong binder fractions, which contradicts the notion of a broad loss of neoantigens. Although HLA-A and B2M showed selective decreases, the broader MHC-I transcriptional program was not coordinately suppressed.... https://escholarship.org/uc/item/7n09z9qt Sat, 22 Aug 2026 00:00:00 +0000 Iyer, Krithika https://escholarship.org/uc/item/95j6b20x In technology-mediated environments, perceptual information is filtered, transformed, or augmented, requiring observers to infer social, physical, and agentive states from perceptual evidence. I characterize this challenge as representation-reality alignment (RRA): The process through which humans coordinate perception, action, and inference to evaluate how perceptual representations correspond to underlying reality and to guide behavior accordingly. Across three chapters, I examine how children and adults interpret mediated information when perceptual input provides ambiguous or indirect evidence about the world.Chapter 1 asks how children aged 4 to 6 years infer a video chat partner’s visual access from the self-view. Children increasingly used the self-view as evidence of another person’s visual perspective with age; explicit reasoning emerged earlier than spontaneous action-based use. Chapter 2 asks how children aged 3 to 6 years judge whether perceptual features in augmented... https://escholarship.org/uc/item/95j6b20x Wed, 12 Aug 2026 00:00:00 +0000 Lin, Chaolan https://escholarship.org/uc/item/7k2661dg Scientific applications are increasingly using neural networks (NNs) at the edge as a fundamental tool for advancing fields such as particle physics and materials science. As the scientific instruments used in these experiments become more advanced, they produce a lot more data (e.g., 40 TB/s) than before. As a result, scientists are relying on edge NNs, which have more capabilities than traditional algorithms, to process the data. To process data quickly enough, these scientific edge NNS have unique requirements. They must (1) be heavily quantized and (2) execute fully on chip. Even more so, these scientific NNs often operate in high radiation environments (1000× that of space). My thesis focuses on using hardware-software co-design to create efficient, fault-tolerant computer architectures for NNs that execute fully on chip so that they meet the strict latency and throughput requirements laid out by these scientific experiments. I defend the following thesis statement: On-chip... https://escholarship.org/uc/item/7k2661dg Wed, 12 Aug 2026 00:00:00 +0000 Weng, Olivia https://escholarship.org/uc/item/3rs0w1cq The sixth-generation (6G) cellular ecosystem is evolving toward integrated sensing and communication (ISAC), where the same infrastructure supports both high-rate data exchange and environmental perception. Vehicular networks are among the most demanding use cases: autonomous navigation demands ultra-reliable connectivity and centimeter-level positioning accuracy, yet conventional beam training protocols incur prohibitive overhead and global navigation satellite system (GNSS)-based localization fails in urban canyons. This dissertation develops a hybrid model/data-driven framework addressing these challenges through three contributions. The first introduces a passive radar-aided framework for multiuser millimeter wave (mmWave) link configuration. A roadside unit senses existing automotive frequency modulated continuous wave (FMCW) transmissions from multiple vehicles via a mixing filter bank with constant false alarm rate (CFAR) detection, isolating individual vehicle signals... https://escholarship.org/uc/item/3rs0w1cq Wed, 12 Aug 2026 00:00:00 +0000 Chen, Yun https://escholarship.org/uc/item/1986x5w0 “I don’t know if you noticed, but when you perform, there’s crickets in the audience” - a professor who-shall-not-be-named once said to me at a Spring conference. Ouch. The infamous Conference Room 144. On the outside – completely unassuming serenity. On the inside - a bloodbath.The professor continued: “You ask too many questions. Stop trying to think. Stop trying to learn this.” My first thought: “Sorry, I’m not supposed to learn at an educational institution? How does that make sense?” Little did I know that precisely this piece of advice was the catalyst to my growth as an artist at UCSD. I spent most of my first year asking questions - clarifying questions, questions that asked for permission, validation, and questions that in truth were new, inventive ways for me to apologize for mistakes I hadn’t even made yet. I held onto the idea of getting it right so tightly that most of the time I actually missed the answers to those questions. But guided by the ever-so-blunt professor’s... https://escholarship.org/uc/item/1986x5w0 Tue, 11 Aug 2026 00:00:00 +0000 Nguyen, Mai Lan https://escholarship.org/uc/item/18p9h6g3 Though activist groups in the San Diego-Tijuana area have long organized for Palestinian liberation, the continuous bombardment of Gaza by Israel and its resulting humanitarian crisis in the wake of October 7, 2023, has influenced a surge in transnational advocacy actions, oftentimes evoking comparisons or connections between the U.S.-Mexico border and borders surrounding Gaza and the West Bank. Analyzing events for Palestinian liberation and/or border advocacy between October 2023 and July 2024, this thesis reimagines the application of “linked fate” through a spatially removed, transnational lens motivated by interlocking systems and mechanisms of power. I propose that border sites offer tangible and theoretical links by which geographically disconnected communities may conclude that what happens in one impacts the other, clarifying how activists in the San Diego-Tijuana borderlands at times draw “linked fate” solidarity relationships through four common narratives: Directly,... https://escholarship.org/uc/item/18p9h6g3 Sat, 8 Aug 2026 00:00:00 +0000 Dorais, Brandi Nikole https://escholarship.org/uc/item/8092b9f3 To address the youth mental health crisis, youth must be engaged in improvement efforts. However, there is limited research on how youth come to participate in this work and what conditions enable and constrain their efforts. In this yearlong case study, I examined a high school student-led mental health advocacy committee situated in a large urban U.S. district. Interviews (N=31), documents, and field notes were analyzed using a five-cycle inductive and deductive coding process, guided by social cognitive theory and the critical youth empowerment framework.The first research question asked what shapes students’ pathways into mental health advocacy. Findings reveal seven elements that form the preliminary Proximity-Based Model of Youth Mental Health Advocacy Development: a prosocial orientation, experience related to mental health concerns, empathic concern, disillusionment and frustration with inadequate systems, moral conviction and urgency, self-efficacy, and opportunity.The... https://escholarship.org/uc/item/8092b9f3 Fri, 7 Aug 2026 00:00:00 +0000 Levine, Rebecca https://escholarship.org/uc/item/77w9v0gj Beyond compartmentalization and protection, cellular membranes are highly dynamic and specialized structures that perform molecular transport and serve as platforms for protein localization and modulation of function. The lipid composition of membranes defines their identity, material properties, interacting partners, and cellular functions. While bilayer lipids establish the dominant fluid lamellar topology, nonbilayer lipids destabilize uniform lipid packing and introduce elastic curvature stress that enables morphological changes, membrane fission and fusion, and protein interaction. Cells expend considerable energy to establish specific lipid compositions for specialized membrane compartments, and a growing body of literature has provided evidence for tight homeostatic regulation of lipid composition and biophysical properties. Through combinations of headgroups, linkages, and acyl chain chemistry, eukaryotic cells produce more than 2000 species of lipids, many of whose... https://escholarship.org/uc/item/77w9v0gj Fri, 7 Aug 2026 00:00:00 +0000 Milshteyn, Daniel https://escholarship.org/uc/item/5gc98387 The local-to-global principle—the idea that global structure is determined by, or can be certified from, local constraints—appears across mathematics, theoretical computer science, and physics under different disguises. This thesis studies two manifestations of this principle: in quantum error-correcting codes, where local parity checks certify the integrity of the global encoded information, and in the entanglement bootstrap, where local reduced density matrices determine the global structure of quantum phases of matter.Part I constructs quantum low-density parity check (qLDPC) codes. We first build a family of quantum locally testable codes (qLTCs) achieving linear dimension, near-linear distance, and near-constant soundness, by generalizing square-complex qLDPC codes to higher-dimensional cubical chain complexes. We then extend the cubical construction to the sheaf cochain complexes, on which we define Poincaré duality and a cup product. Poincaré duality clarifies the duality... https://escholarship.org/uc/item/5gc98387 Fri, 7 Aug 2026 00:00:00 +0000 Lin, Ting-Chun https://escholarship.org/uc/item/26t2h37b Modern energy systems are increasingly shaped by flexible demand, distributed storage, and building infrastructure. These systems generate rich data and require fast, adaptive decisions, but they are also governed by physical laws, safety constraints, and strategic behavior that are difficult to capture with generic black-box learning. This creates a disconnect between the structure-rich models used in classical energy systems analysis and the flexible learning pipelines favored in modern data-driven control. Methods that ignore known structure often suffer from poor sample efficiency, weak interpretability, and unreliable constraint handling, while purely model-based approaches struggle with hidden objectives and computational scale. The goal of this thesis is to bridge this disconnect by developing structured learning and control methods that integrate optimization, physics, and control-theoretic structure directly into trainable decision-making architectures.First, the dissertation... https://escholarship.org/uc/item/26t2h37b Fri, 7 Aug 2026 00:00:00 +0000 Bian, Yuexin https://escholarship.org/uc/item/91c9h60g As antenna array apertures in the next-generation multiple-input-multiple-output (MIMO) wireless systems scale up, it is necessary to ensure reasonable scaling in the hardware cost and power consumption. In the past decade, a hybrid architecture was proposed which balanced the beamforming functionality across digital precoding and analog precoding with phase shifters, switches, or true time delays. This evolution from a fully digital architecture to a hybrid architecture enabled expanding the array aperture beyond the number of radio frequency chains. For next-generation cellular, to ensure scalability to even larger apertures, we propose the tri-hybrid MIMO paradigm which replaces the conventional antennas in MIMO with more capable reconfigurable antennas. This dissertation investigates two types of reconfigurable antennas, namely hybrid parasitic arrays and dynamic metasurface antennas (DMAs).In the first contribution, we use a circuit-theoretic model for parasitic arrays which... https://escholarship.org/uc/item/91c9h60g Thu, 6 Aug 2026 00:00:00 +0000 Deshpande, Nitish https://escholarship.org/uc/item/8wt2985z The interleukin-17 (IL-17) family comprises a group of cytokines involved in host defense against pathogens, autoimmune diseases and anti-tumor immunity, making IL-17 signaling pathways promising therapeutic targets. Among the six IL-17s found in human, Interleukin-17 D (IL-17D) is the least studied. Previous work from our lab has demonstrated that IL-17D is regulated by the NRF2 pathway and contributes to tumor and virus surveillance through recruitment of natural killer cells. Compared to wild-type (WT) mice, IL-17D knockout (KO) mice exhibit reduced resistance to carcinogen-induced tumorigenesis, viral and bacterial infection, and experimental colitis.As the function of IL-17D appears to be context specific, and a recent study demonstrated IL-17D may promote atopic dermatitis development through DDX5 inhibition in keratinocytes, this dissertation explored the roles of IL-17D in the skin biology. In mouse psoriasis-like and atopic dermatitis-like skin inflammation models, IL-17D... https://escholarship.org/uc/item/8wt2985z Thu, 6 Aug 2026 00:00:00 +0000 Wang, Lu https://escholarship.org/uc/item/7kf2p0v2 The glacial fjords of Kalaallit Nunaat (Greenland) are critical connectors between the Greenland Ice Sheet (GrIS) and the global ocean—delivering freshwater from the melting ice sheet to the ocean and channeling deep, warm ocean waters back toward the ice. The GrIS has been losing mass at an accelerating rate since at least 1990 due to unabated rising global air temperatures from industrial human activities. This increasing addition of fresh glacial meltwater to the salty ocean has cascading local to global impacts. A challenge for improving projections of ice loss and ocean changes is the representation of local fjord processes in largescale models. Addressing this and improving fjord parameterizations requires coordinated, accessible, long-term datasets spanning multiple fjords and timescales—essential, also, to furthering multidisciplinary fjord ecosystem understanding. While fjord dynamics play an outsized role in determining future global sea level rise and ocean circulation... https://escholarship.org/uc/item/7kf2p0v2 Thu, 6 Aug 2026 00:00:00 +0000 Roth, Aurora https://escholarship.org/uc/item/6c8854xf This dissertation consists of three chapters in macroeconomics and finance. The first two chapters focus on how monetary policy shocks affect lending by depository institutions. The third chapter explores how conglomerates develop in Brazil.Chapter 1 revisits the commercial and industrial (C&I) loan puzzle: standard models predict bank lending should fall when interest rates rise, yet empirical results find the opposite for C&I loans. Allowing contractionary and expansionary monetary policy shocks to affect lending differently, I show the contractionary puzzle disappears: C&I lending falls as theory predicts while C&I lending still puzzlingly declines after expansionary shocks. The decline is concentrated in the top 1% of lenders, and firm-level evidence is consistent with large firms (which tend to use large banks) substituting away from C&I loans and refinancing debt with corporate bonds at the lower market rates.Chapter 2 compares credit union and commercial... https://escholarship.org/uc/item/6c8854xf Thu, 6 Aug 2026 00:00:00 +0000 Yee, Steven https://escholarship.org/uc/item/67r7b9cr On a modular curve of arbitrary congruence level, we introduce the notion of a “Makdisi symbol”, a device that simultaneously gives a moduli-friendly coordinate system while also having an elegant Hecke theory. Concretely, these are cuspidal projections of products of two weight 1 Eisenstein series, and their study was originally pioneered by the work of Kamal Khuri-Makdisi. We show that a certain subclass of symbols, the “invertible Makdisi symbols”, yield precisely the eigenforms of rank zero; combining this with the moduli interpretation, we obtain a systematic and relatively efficient algorithm to determine the rational points on a modular curve, so long as the curve has a rank zero eigenform. The algorithm is p-adic in nature, based on the method of Chabauty-Coleman, and does not require finding any equations for the curve. https://escholarship.org/uc/item/67r7b9cr Thu, 6 Aug 2026 00:00:00 +0000 Xu, Christopher https://escholarship.org/uc/item/5tk886jq Computer graphics algorithms are efficient, accurate, and predictive at forward simulation. Photo-realistic images can be generated from precise physical quantities that describe the geometry of objects in the scene, their reflectance properties, and the lighting conditions. The inverse-rendering problem is to invert these algorithms and recover the scene description from the induced images, recovering the cause from the effect. This work addresses inverse rendering of scene geometry via analysis-by-synthesis. The method starts with an initial hypothesis of the geometry, renders an image, compares it with a reference, and uses the error to update the hypothesis. There are three crucial components to this approach: (a) suitable mathematical representations to describe the geometry, (b) mathematical tools and algorithms that estimate gradients for a rendering algorithm, and (c) objective functions to compare estimates with reference images. The design and analysis of these three... https://escholarship.org/uc/item/5tk886jq Thu, 6 Aug 2026 00:00:00 +0000 Mehta, Ishit https://escholarship.org/uc/item/50491433 Quantum computing has emerged as a transformative frontier of computation. In recent years, quantum hardware has scaled at an unprecedented rate. As this momentum continues, the central challenge is shifting upward in the stack—from hardware-level feasibility toward system-level scalability. This thesis will focus on quantum computer architecture and compiler, introducing the challenges and opportunities to efficiently harness device capabilities and lower the demands on hardware technology, thereby accelerating timelines for practical quantum advantage. https://escholarship.org/uc/item/50491433 Thu, 6 Aug 2026 00:00:00 +0000 Zhang, Hezi https://escholarship.org/uc/item/4hh707c0 Advances in high-throughput sequencing and spatial profiling technologies have enabled unprecedented characterization of tumor biology; however, most approaches interrogate tumor-intrinsic programs, the tumor microenvironment, and systemic immunity in isolation, limiting mechanistic understanding of therapeutic response. In this dissertation, we develop an integrative, multi-scale framework to define tumor ecosystem states that capture the coordinated interactions between malignant cells, stromal compartments, and immune populations across molecular, spatial, and systemic levels. Using bulk RNA sequencing of triple-negative breast cancer (TNBC) tumors, we identify three reproducible tumor ecosystem states (subtypes) defined by distinct combinations of epithelial, immune, and stromal programs. These subtypes are associated with differential survival outcomes and are governed by subtype-specific regulatory networks, including interferon-driven immune activation, stromal fibrosis... https://escholarship.org/uc/item/4hh707c0 Thu, 6 Aug 2026 00:00:00 +0000 mesrizadeh, zahra https://escholarship.org/uc/item/3n48v781 Understanding physiological flows is essential to the advancement of medical science. This dissertation investigates two physiologically relevant flow problems and their applications through a combination of theoretical modeling and experimental analysis. The first investigation concerns cerebrospinal-fluid flow in the central nervous system, with particular emphasis on the in vitro experimental characterization of the slow fluid-particle drift induced by cardiac-driven motion in the spinal canal. The study shows that Time-Spatial Labeling Inversion Pulse (Time-SLIP), a non-contrast MRI technique, can be used to measure mean Lagrangian drift in a simplified phantom. It also identifies potential limitations associated with signal quality, spatial resolution, and post-processing methodology, which may complicate future clinical applications of the method. The second topic concerns blood flow in the main pulmonary artery. This problem is first addressed through the development of... https://escholarship.org/uc/item/3n48v781 Thu, 6 Aug 2026 00:00:00 +0000 Campos, Obed Armando https://escholarship.org/uc/item/3dd44049 This dissertation consists of three chapters examining the design and impacts of safety net programs in the United States. Chapter 1 examines how access to the Supplemental Nutrition Assistance Program (SNAP) has affected the health and economic outcomes of low-income families in the modern era. Exploiting state-level expansions under Broad-Based Categorical Eligibility (BBCE)—one of the largest SNAP reforms since the 2000s—I use a difference-in-differences design with multiple large-scale data sources on program usage and population health. The expansion increased SNAP participation and generated spillovers to Medicaid and WIC enrollment. Increased resources improved food security and health care utilization. Although short-term health gains are modest, they are more pronounced among non-White populations, including reductions in low birth weight and obesity. Notably, I find no significant effects on employment status and only a slight decline in hours worked. These findings... https://escholarship.org/uc/item/3dd44049 Thu, 6 Aug 2026 00:00:00 +0000 Shin, Jaeyeon https://escholarship.org/uc/item/2z0390w5 A variety of renewable energy sources are needed to support the modern electrical grid. Lithium metal batteries and fuel cells are high-energy-density, high-power technologies that are needed to meet the demand for immediate power delivery, such as electrified vehicles or portable electronics. The lifetime and power output of these technologies can be improved via the efficient usage of their material components. This dissertation investigates the mechanisms and morphologies of metal plating and stripping inside two relevant clean energy technologies. Firstly, the conformal plating of a Ni/P alloy on a high-aspect ratio porous oxide substrate was studied for solid oxide fuel cells (SOFCs). The fabrication of a porous conductive interface improves the power output of thin-film SOFCs. Secondly, the effect of low temperature on the stripping of lithium metal anodes was studied. Improved homogeneity of the stripped interface and the spontaneous formation of faceted crystals under... https://escholarship.org/uc/item/2z0390w5 Thu, 6 Aug 2026 00:00:00 +0000 Hopkins, Emma https://escholarship.org/uc/item/0tn784q0 This dissertation examines the experiences of myself and other academic laborers in dialogue with the experiences of our families that labored within the immigrant agricultural economy. This research is particularly timely with the gutting of Diversity Equity, and Inclusion programs across industries and institutions, alongside the exploitative racialized labor that has existed before, during and after. This dissertation uses an intergenerational examination of labor to reflect on how to navigate this current moment in the higher education. Building off previous research, my multi-disciplinary dissertation engages with the history of farmwork and academic labor through the frameworks of critical race theory and labor studies from the theoretical vantage of everyday life. This is put in dialectic with the results of an ongoing neo-liberalization of the University. “Afterlives” emerges as a key concept to track the ongoing salience of the lived and affective experiences of the farmworkers... https://escholarship.org/uc/item/0tn784q0 Thu, 6 Aug 2026 00:00:00 +0000 Hernandez, Keven https://escholarship.org/uc/item/9bb404b2 An orbit equivalence between two free p.m.p. actions of countable groups is Shannonif the partition generated by the associated cocycles have finite Shannon entropy. The main result of this dissertation is that if one of the groups is finitely generated, amenable and non- virtually-cyclic, or has a w-normal subgroup that is infinite and non-virtually-cyclic, or has an amenable quotient that is infinite and non-virtually-cyclic and the action factors onto an action of the quotient, then the two actions have the same Rokhlin entropy. This extends a result of Austin and results of Kerr and Li to the Rokhlin entropy setting. The proof uses a new notion of dissipable orbit equivalence, witch is a symmetric condition that yields the equality of entropy. https://escholarship.org/uc/item/9bb404b2 Wed, 5 Aug 2026 00:00:00 +0000 Chen, Qingyuan https://escholarship.org/uc/item/9b00s7jz Renormalization group (RG) fixed points are a particular class of systems that govern the universal properties of phases and phase transitions of many-body systems. States at RG fixed points usually have simple local entanglement properties from which one can derive (i.e. entanglement-bootstrap) the universal structural properties of phases and critical points. This dissertation collects a series of works on 2+1D chiral gapped phases within the Entanglement Bootstrap formalism. One long-term goal is to establish an operator bulk/edge correspondence from entanglement, thereby connecting the operator algebras of bulk chiral gapped systems with those of edge conformal field theories (CFTs). Moreover, a local entanglement property that is commonly shared by a family of fixed points can be used as a condition to enumerate all such fixed points. This dissertation includes one work in this direction, where we attempt to enumerate all 1+1D CFTs in a small Hilbert space. https://escholarship.org/uc/item/9b00s7jz Wed, 5 Aug 2026 00:00:00 +0000 Li, Xiang https://escholarship.org/uc/item/8nt9d39q The academia-industry gap is a longstanding divide between the expectations of early-career software developers and students' ability to meet those expectations upon graduating from an undergraduate CS program. This gap persists due to stark differences between university coursework and professional development: software developers contribute to large, existing code bases while students create projects from scratch. As professional developers increasingly adopt GenAI tools, this gap is only widening. To address this gap, this dissertation focuses on the programming processes students use to create software. The first part of this dissertation contributes to our understanding of students’ programming processes. I’ll uncover how students engage in incremental development for small projects, program comprehension in large code bases, and AI collaboration for software development tasks. The second part of this dissertation demonstrates that by measuring these processes, we can... https://escholarship.org/uc/item/8nt9d39q Wed, 5 Aug 2026 00:00:00 +0000 Shah, Anshul https://escholarship.org/uc/item/7z796252 Generative audio systems have emerged as a burgeoning paradigm for creative multimedia, with an expanse of possibilities for people to interact with and create sonic artifacts. This innovation has been driven largely by continuous-domain diffusion models, which despite their high quality and performance are generally cumbersome, rigid generative systems that lack the ability for flexible expression in reasonable time. In this proposal, I will outline our work in advancing the core modeling capabilities of such systems and building a future of interactive generative audio, which falls along two main thrusts (1) improving the controllability of generative audio systems for diverse axes of creation and (2) improving the efficiency of generative audio systems to enable real-time interaction. https://escholarship.org/uc/item/7z796252 Wed, 5 Aug 2026 00:00:00 +0000 Novack, Zachary Emerson https://escholarship.org/uc/item/7z1089cf Non-linear Random Matrix Theory (RMT) has recently emerged as a powerful paradigm for the theoretical understanding of deep learning theory. Throughout recent works, a universality principle, the Gaussian Equivalence Theorem (GET), has become an indispensable tool allowing for the behavior of complex nonlinear neural networks to be understood through tractable linear kernel models. This thesis contributes to this emerging field, first by using the GET universality principle to derive a novel scaling law in Neural Tangent Kernel (NTK) regression, and second by studying the implications of this idealized linear equivalence on a high-dimensional nonlinearly separable dataset. The main contributions can be summarized as follows. We study finite-width NTK regression under a dual-power law model, in which the training data covariance spectrum exhibits a power-law decay and the labels are generated with deterministic signal whose source-exponent measures the difficulty of the... https://escholarship.org/uc/item/7z1089cf Wed, 5 Aug 2026 00:00:00 +0000 Cranston, Collin https://escholarship.org/uc/item/769753w5 Plasmonic nanoparticles are powerful sensing platforms whose optical properties emerge from the spatial relationships between particles. Controlling these relationships—how particles find one another, how closely they pack, and whether they coalesce—unlocks tunable optical responses for applications ranging from pathogen detection to photoacoustic imaging. This dissertation establishes a mechanistic framework for peptide-directed nanoparticle assembly into hierarchical fractal architectures via diffusion-limited aggregation (DLA) and demonstrates how that understanding enables the design of advanced plasmonic biosensors.I first demonstrate that peptides can drive gold and silver nanoparticle aggregation into a colorimetric reporter for oral pathogens. Enzyme-activatable prodrugs selectively target Candida albicans and Porphyromonas gingivalis through protease-triggered antimicrobial peptide release, coupling therapeutic activity to peptide-induced nanoparticle aggregation, with... https://escholarship.org/uc/item/769753w5 Wed, 5 Aug 2026 00:00:00 +0000 Amer, Lubna https://escholarship.org/uc/item/75t6j634 Pressure-sensitive adhesives form bonds under light pressure without heat, solvent, or chemical curing, making them useful in applications ranging from tapes and packaging to biomedical devices and flexible electronics. Their performance depends on a difficult balance of properties: they must be compliant enough to conform to surfaces and form intimate contact, yet mechanically stable enough to resist debonding, creep, and failure under different loading conditions. This dissertation investigates liquid crystal elastomers as a platform for pressure-sensitive adhesion across time scales, loading modes, and material architectures. Liquid crystal elastomers combine polymer network elasticity with the orientational order of liquid crystalline mesogens, enabling soft elasticity, strain-induced alignment, and deformation-dependent stiffening. First, loosely crosslinked liquid crystal elastomers were studied under sustained tensile creep and compared with polydimethylsiloxane networks... https://escholarship.org/uc/item/75t6j634 Wed, 5 Aug 2026 00:00:00 +0000 Chambers, Robert Joseph https://escholarship.org/uc/item/6gn756vz Semidiurnal alongshore currents on the Southern California Bight (SCB) inner shelf vary on scales of O (10 km) in a manner the progressive surface tide does not explain. Using a seven-year ADCP record on the 33 m isobath at three sites (Point Loma, La Jolla, Cardiff) spanning 35 km alongshore, we attribute this variability to a northward-propagating, superinertial mode-1 coastal trapped wave (CTW) with wavelength 39–49 km, seasonally modulated by stratification. Applying a harmonic skill score to the same record in two projections yields SS90 values of 18–22% for the depth-resolved baroclinic velocity and 53–57% for the depth-averaged quasi-barotropic velocity—a 35 percentage point gap. SUNTANS simulations reproduce ?2-forced CTW propagation across the SCB at phase speeds of 0.78–0.94 m/s and wavelengths of 35–42 km, bracketing the observations. The depth-averaged coastal internal tide at this site is harmonically predictable at skill levels the depth-resolved projection... https://escholarship.org/uc/item/6gn756vz Wed, 5 Aug 2026 00:00:00 +0000 Okun, Kevin Michael https://escholarship.org/uc/item/6813x7r3 Over the last twenty years, from generative AI to predictive tabular models, neural networks have become the dominant paradigm for machine learning. Despite their ubiquity, we do not have a clear understanding of the fundamental mechanisms that drive their success. In particular, the process by which they learn task-relevant features, and therefore outperform other model classes, has remained unknown. This dissertation establishes a unifying mechanism for feature learning: iterated estimation of the Average Gradient Outer Product (AGOP).The first contribution introduces the Neural Feature Ansatz (NFA), a structural identity connecting the AGOP of a trained neural network to the features encoded in its weights. We give experimental evidence that the NFA holds across architectures, establishing that AGOP captures the features learned by trained neural networks. Because the AGOP is well-defined for any differentiable predictor, we use this observation to design the Recursive Feature... https://escholarship.org/uc/item/6813x7r3 Wed, 5 Aug 2026 00:00:00 +0000 Beaglehole, Daniel https://escholarship.org/uc/item/5dv078j9 We prove a homological analog of a spectral sequence computing the continuous group cohomology of a p-adic Lie group. We describe a cap product map, which gives the spectral sequence the structure of a differential graded module over the multiplicative cohomological spectral sequence. We then turn to studying the Hochschild cohomology of the category of smooth complex representations of a reductive p-adic group, via the Bernstein decomposition. We exhibit some results on Noetherianness, identifying a sufficient condition on Bernstein blocks which yields Noetherian Hochschild cohomology. We also investigate the principal block for G=GL_n(F) specifically, by understanding the Iwahori-Hecke algebra as a smash product. https://escholarship.org/uc/item/5dv078j9 Wed, 5 Aug 2026 00:00:00 +0000 Xue, Christopher https://escholarship.org/uc/item/4z1243v9 A fundamental result in representation theory of complex semisimple Lie algebras is the Borel--Weil--Bott theorem which constructs irreducible representations of a Lie algebra in terms of cohomology of certain homogeneous bundles on projective varieties. A generalization of this theorem for Lie superalgebras is not known and this dissertation studies aspects of this problem in the context of algebraic supergeometry. Using new methods from commutative algebra and invariant theory, we completely calculate the sheaf cohomology of the structure sheaf for five infinite families of partial flag supervarieties. We show that these cohomology rings are tensors products of the singular cohomology of an ordinary partial flag variety and the Tor-groups of a new family of varieties that we call compositional varieties. A compositional variety is a generalization of a determinantal variety obtained by imposing rank conditions on compositions of maps; we study geometric properties of these... https://escholarship.org/uc/item/4z1243v9 Wed, 5 Aug 2026 00:00:00 +0000 Pal, Abhik https://escholarship.org/uc/item/4h45w3n6 Bacterial adaptations are shaped by their unique genome sequences as well as the transcriptional regulatory network (TRN) that governs gene expression. Advancement in sequencing technologies has resulted in the rapid accumulation of both RNA-sequencing and whole genome sequencing data, providing the opportunity to obtain deeper insight into bacterial systems. With the rapid increase in sequencing data, there is a need for scalable data analytic tools that enable the extraction of biologically meaningful information. In this dissertation, we employ non-negative matrix factorization (NMF) and independent component analysis (ICA)-based approaches to decompose bacterial pangenomes and TRNs, respectively, and gain insight into bacterial diversity and evolution. We first applied NMF to the pangenome of the Pseudomonas aeruginosa species to identify co-occurring gene sets across groups of strains (“phylons”), and investigated the genetic basis for differential traits across the species.... https://escholarship.org/uc/item/4h45w3n6 Wed, 5 Aug 2026 00:00:00 +0000 Bajpe, Heera https://escholarship.org/uc/item/26n9r99r School leaders committed to justice-oriented equity work must navigate organizational systems that both enable and constrain their efforts. This mixed-methods, multiple-case study employs Critical Social Network Analysis (CSNA) to examine how leaders’ social positioning within their organizational contexts shapes their perceptions of opportunity and constraint around equity work. Data were collected in two school districts in Southwestern border states. Across the cases, I analyze whole-network survey data from 85 administrators and conduct in-depth interviews with 28 leaders.The dissertation is organized as three articles. In Article 1, I examine how leaders define and operationalize equity in the post-Covid political climate. Findings indicate that many leaders’ equity efforts are locally focused on instructional practice and relational support rather than broader social transformation. Examining equity within specific organizational problem spaces surfaces potential tensions... https://escholarship.org/uc/item/26n9r99r Wed, 5 Aug 2026 00:00:00 +0000 Trautman, David Albert Rodes https://escholarship.org/uc/item/2679x6r4 An additive action is an effective action of Gn a on an n-dimensional variety with an open orbit, and an additive variety is a normal variety admitting an additive action. In the first half of this dissertation, we study additive actions on projective space, which are classified by certain local Artinian algebras, and we provide a geometric explanation for the unexpected appearance of some moduli spaces within this classification. We also describe the G2 a -equivariant Sarkisov program for smooth additive surfaces. In the second half, motivated by the fact that not all additive varieties are Mori dream spaces, we present some conjectures about extremal rays in the Kleiman–Mori cone of additive varieties, and we prove them in specific cases. https://escholarship.org/uc/item/2679x6r4 Wed, 5 Aug 2026 00:00:00 +0000 Castellano Macias, Felipe https://escholarship.org/uc/item/2382j4bs This dissertation examines how cross-sectional methods can be used to study the transmission of macroeconomic shocks, combining methodological contributions with applications to fiscal policy, financial intermediation, and business cycle dynamics. In Chapter 1, I study cross-sectional designs that exploit heterogeneous exposure to aggregate shocks for estimating partial-equilibrium elasticities in macroeconomics. I show that these designs can fail to identify partial-equilibrium elasticities when exposure to the shock of interest is correlated with exposure to aggregate variables that move in general-equilibrium. I develop a test for this identification failure and propose a new decomposition method that leverages both cross-sectional and time-series variation to address it. Applying the method to estimate US cross-sectional fiscal multipliers, I find that accounting for heterogeneous exposure to monetary policy reduces the estimated two-year multiplier from 1.5 to 1. ... https://escholarship.org/uc/item/2382j4bs Wed, 5 Aug 2026 00:00:00 +0000 Donaldson, Maria Paula https://escholarship.org/uc/item/1x28z9z2 Large language models (LLMs) have achieved remarkable progress on a wide range of tasks by learning from vast collections of human-written text. This training paradigm instills in LLMs the ability to produce fluent, contextually appropriate language—but it also imposes a fundamental constraint: the model learns to imitate the outcomes of human reasoning, as recorded in text, rather than the cognitive processes of reasoning itself. As a result, LLMs face persistent difficulties with tasks that require structured exploration, grounded symbolic computation, or deliberate planning—capabilities that, in humans, operate partly outside of language.This dissertation explores three distinct approaches to move language model reasoning beyond the boundaries imposed by language imitation. First, we reframe LLM reasoning as a planning problem. Reasoning via Planning (RAP) repurposes the same language model as both a reasoning agent and a world model, enabling structured exploration of the... https://escholarship.org/uc/item/1x28z9z2 Wed, 5 Aug 2026 00:00:00 +0000 Hao, Shibo https://escholarship.org/uc/item/0rz4x93g In this dissertation, I investigate how people determine the meaning of their own preferences and choices, as well as those of others, especially when those choices are driven by multiple causes such as internal conflicts, social expectations, or concerns about social evaluation. I approach this question through three lines of research: (1) the psychology of second-order preference conflicts, (2) causal inferences about others’ preferences, and (3) strategic signaling under social evaluation. Chapter 1 examines how choices can mask second-order preference conflicts: conflicts between the preferences people have and the preferences they wish to have. I propose a psychological framework for understanding these conflicts and provide initial evidence for an empirical measure of these conflicts. Chapter 2 explores how people infer others’ preferences from choices involving default options. I propose a causal reasoning account in which people reason backward from an observed choice... https://escholarship.org/uc/item/0rz4x93g Wed, 5 Aug 2026 00:00:00 +0000 Liu, Xingyu (Shirley) https://escholarship.org/uc/item/0c88h83c Mitochondria are essential organelles within a eukaryotic cell, with their own genome (mtDNA) responsible for cellular energy production through oxidative phosphorylation. The mtDNA-encoded mitochondrial transfer RNAs (mt-tRNAs), despite comprising only a small fraction of the mitochondrial genome, account for a disproportionate number of disease-associated mutations. Many of these mutations disrupt mitochondrial gene expression and lead to severe mitochondrial disorders. However, the molecular mechanisms by which mt-tRNA sequence variation affect transcription and RNA maturation are insufficiently characterized. Understanding how these mutations alter the mt-tRNA life cycle is key to identify early points of dysfunction and for development of targeted therapeutic strategies. The goal of this dissertation is to define how mt-tRNA sequence variation influences transcriptional pausing on mt-tRNA encoding sequences and mt-tRNA processing, to establish mechanistic links between these... https://escholarship.org/uc/item/0c88h83c Wed, 5 Aug 2026 00:00:00 +0000 Munozvilla, Jubilee Haddasah https://escholarship.org/uc/item/04r4n6fp This dissertation examines the role of Catholicism in shaping educational reform, political conflict, and gendered authority in Chile between 1964 and 1973. Scholarship has often treated the Catholic Church as either a cohesive institutional actor or as a reactive force in the face of political polarization. By contrast, this project argues that Catholicism functioned as an internally fragmented political field, in which clergy, lay organizations, educators, and students mobilized shared theological frameworks to advance competing visions of social and political change. Focusing on the reformist period of Christian Democratic governance and the subsequent Popular Unity years, this dissertation traces how education became a central arena in which these conflicts unfolded. Drawing on archival sources from the Archive of the Archbishopric of Santiago, educational policy documents of the era, Catholic periodicals, and university publications, it examines debates over agrarian reform,... https://escholarship.org/uc/item/04r4n6fp Wed, 5 Aug 2026 00:00:00 +0000 Gibbens, Holly Marie https://escholarship.org/uc/item/6wv5h9tt There have been several proposed methods for artificial intelligence systems. This thesis focuses on two approaches: library learning (neurosymbolic) and reasoning-based language models. We begin by proposing a dual-system for artificial intelligence -- where the system requires both knowledge and composition -- before introducing our two key settings that we frame through this dual-system. First, the symbolic domain is formally defined with program synthesis before highlighting an ambitious strategy in library learning. This is followed by original work studying a novel neurosymbolic approach to the ARC-AGI benchmark. Then, we introduce the neural domain -- where we frame reasoning as search and follow with work that studies how post-training data influences downstream reasoning behavior in chess. This thesis concludes by solidifying a unified framework for artificial intelligence that is based on identified similarities between the principled, theoretical system proposed in... https://escholarship.org/uc/item/6wv5h9tt Mon, 3 Aug 2026 00:00:00 +0000 Dionisopoulos, Lucas https://escholarship.org/uc/item/3bc7p568 Many modern problems in data science involve data that is fundamentally measure-valued. In these situations, natural questions arise about how one should perform standard tasks from machine learning and numerical analysis. The tools of optimal transport point to intuitive analogs of these tasks by viewing the measures themselves as single points on the Wasserstein manifold. Then, by computing optimal transport maps from a fixed reference, we can embed each measure into a common linear space (the tangent space of the Wasserstein manifold at the fixed reference), which allows us to apply out-of-the-box methods on the embeddings. When the data is static, these methods can be used to build classifiers, generate new measures from particular classes, efficiently compute clusters, and produce visualizations via dimension reduction. In this dissertation, we extend this technique of ``linearized optimal transport'' to the case where our measure-valued data is evolving in time. In this... https://escholarship.org/uc/item/3bc7p568 Thu, 30 Jul 2026 00:00:00 +0000 Karris, Nicholas https://escholarship.org/uc/item/0f3869xt This dissertation consists of three chapters on healthcare access and demand for healthcare in low-income countries.Chapter 1 studies the impact of expanding access to facility birth on early-life mortality and subsequent healthcare use in Malawi. Using an exogenous increase in travel costs to health facilities, I find that only high-risk pregnancies with access to high-quality care experience significant survival gains. The systematic mismatch between high-risk pregnancies and low-quality facilities at the last mile helps explain the limited mortality reductions from access expansion. Exposure to low-quality facilities also decreases future healthcare visits, suggesting that expanding access to poor service quality can backfire by discouraging engagement with formal care.In Chapter 2, we leverage the 5x increase in medical colleges in India between 1980 and 2020 to measure the impact of college openings in a district on access to care, take-up of care, and health outcomes. We... https://escholarship.org/uc/item/0f3869xt Thu, 30 Jul 2026 00:00:00 +0000 Budillon, Fulvia https://escholarship.org/uc/item/1x1244ss Physical therapists use standardized outcome measures to assess patient function and guide clinical decisions, yet adoption has been limited by fragmented workflows, scoring complexity, and limited automated interpretation. Existing digital tools are embedded in proprietary electronic health records, limited to specific measures, subscription-based, or reliant on server-side processing. This thesis presents PhysioMeter, a free, open-source, client-side web application for scoring and interpreting physical therapy outcome measures. Developed in collaboration with licensed physical therapists, PhysioMeter implements a declarative Measurements, Calculations, and Interpretations pipeline in YAML configuration files, allowing measures to be added or modified without changes to the application code. All computation and storage occur within the clinician's browser; no patient data are transmitted to a remote server. It is also installable as a Progressive Web App that runs offline... https://escholarship.org/uc/item/1x1244ss Wed, 29 Jul 2026 00:00:00 +0000 Ji, Daniel https://escholarship.org/uc/item/7f99f7c0 Winds and tides force oceanic motion at many different scales from basin scale overturning to centimeter scale turbulence. The physical mechanisms that transfer kinetic energy and momentum between these scales allow the global climate to be influenced by processes that are too small to be resolved by models. Understanding the physics behind these transfers of energy and momentum holds the key to parameterizing their impact in models. Observations allow us to understand the physics on a deeper level, and continue to see the ocean from new perspectives as technology evolves. This thesis uses observations to illuminate the physical mechanisms by which internal wave breaking transfers energy and momentum between scales, and the human mechanisms that allow those observations to exist. In Chapter 1 we use observations to show the pathway of kinetic energy cascade from internal tides to turbulence in frequency-space. Chapter 2 presents observations and ray tracing of near-inertial wave... https://escholarship.org/uc/item/7f99f7c0 Mon, 27 Jul 2026 00:00:00 +0000 Bellerjeau, Charlotte https://escholarship.org/uc/item/74z2v6qk Multimodal large language models (MLLMs) have demonstrated impressive capabilities across vision-language tasks, yet fundamental challenges remain in understanding their internal multimodal processing, aligning their behavior with human preferences, and enabling faithful reasoning. This dissertation addresses these challenges through three lines of work spanning multimodal understanding, alignment, and reasoning. We begin with multimodal understanding, where we develop an information-theoretic framework that traces how visual information is encoded, compressed, and shaped by textual instructions across MLLM layers. Building on these insights, we introduce CoMMIT, a coordinated instruction tuning approach that resolves the learning imbalance between visual encoders and the backbone LLM, and MDGD, which preserves visual knowledge during fine-tuning by decoupling modality-specific gradient updates. We then turn to multimodal alignment, proposing OCEAN, which models chain-of-thought... https://escholarship.org/uc/item/74z2v6qk Mon, 27 Jul 2026 00:00:00 +0000 Wu, Junda https://escholarship.org/uc/item/6pw5c2xq Recent breakthroughs in generative AI have produced a new generation of powerful music synthesis models; the dominant paradigm to emerge is text-to-music generation, in which complete musical pieces are synthesized from natural language descriptions. While technically impressive, this paradigm does not always reflect how real-life musicians actually work. Music-making rarely begins with a sentence — it is a practice that spans orchestrating individual parts, separating and analyzing musical textures, interpreting notation, accompanying other performers, and ultimately engaging in real-time musical interaction. This dissertation presents four works that collectively explore these dimensions of music-making through the lens of diffusion models. The works presented in this dissertation trace a coherent arc through the musical process as a practitioner may experiences it. The first addresses the structural understanding and creation of music — how individual tracks relate and... https://escholarship.org/uc/item/6pw5c2xq Mon, 27 Jul 2026 00:00:00 +0000 Karchkhadze, Tornike https://escholarship.org/uc/item/6dm19691 This dissertation presents three works composed during my six years as a PhD student in Composition at the University of California, San Diego, and lays out a theoretical and philosophical groundwork for understanding my electroacoustic and electronic sonic practice. I specifically unpack sonic worlding, posthuman listening, and cyborgian entanglements as analytical and theoretical lenses for understanding my compositional work, and as a roadmap for future creative engagements and further development of these ideas. All three use Rosi Braidotti’s notion of posthumanism, which I seek to demonstrate is fertile ontological and aesthetic framework for grounding my creative practice, analyzing my works, and situating them in dialogue with current discourse. The theory–practice cartography I develop here contextualizes three electroacoustic pieces: Machine Spectre, for chamber orchestra and electronics; INTRA for cello, live electronics and lights; and noocambriidæ, for multichannel... https://escholarship.org/uc/item/6dm19691 Mon, 27 Jul 2026 00:00:00 +0000 Whiteley, Ess https://escholarship.org/uc/item/3gb5676m Deploying autonomous robots in safety-critical settings remains difficult, as they must operate in uncertain, dynamic environments while satisfying safety requirements, especially near humans. Value-function-based safety filters offer a principled way to deploy robots in these settings: a robot may pursue a task, while the runtime filter intervenes only when necessary to prevent failures. Despite their theoretical appeal, these methods are difficult to use in complex autonomy stacks. First, constructing safety value functions that represent the constraints required for autonomous deployment is non-trivial. Second, formal verification methods are often conservative or scale poorly with the dimensionality of the system. Third, standard safety filter formulations often encode safety solely as obstacle avoidance. Fourth, existing methods target narrow operational domains, limiting use in uncertain real-world conditions. This dissertation addresses these challenges through... https://escholarship.org/uc/item/3gb5676m Mon, 27 Jul 2026 00:00:00 +0000 Tonkens, Sander https://escholarship.org/uc/item/3g5981hk Neutrino rest mass enables left-handed states to “flip” to right-handed states and vice versa. In-medium effects can enhance the probability for such spin-flip. We demonstrate that a full many-body calculation of this process in neutrino-dense environments can lead to spin-flip probabilities that exceed by orders of magnitude those calculated with mean-field treatments. We study simple configurations with a few neutrinos in well-defined momentum states, for which we show that the helicity conversion enhancement is connected to many-body momentum exchange. Such an effect would therefore be missed in a calculation that considers only forward processes. We speculate on the potential astrophysical implications of these results and the range of applicability of our calculation and its limitations. https://escholarship.org/uc/item/3g5981hk Mon, 27 Jul 2026 00:00:00 +0000 Xu, Yiheng https://escholarship.org/uc/item/3fx6r9xv This dissertation consists of three chapters on the local economic effects of immigrant composition. Across settings, I study how the demographic, socioeconomic, and cultural characteristics of migrants shape urban growth, public-good provision, and housing markets. Chapter one studies internal migration and urban growth in Africa. I examine how migrant inflow size and birthplace diversity jointly affect city-level economic activity in rapidly urbanizing countries. I construct a subnational panel of internal migration across Africa and develop a nonlinear shift-share instrument that separately identifies shocks to migrant levels and birthplace composition. I find that cities receiving migrants from more diverse birthplaces experience lower short-run growth but long-run urbanization gains. Greater diversity is associated with higher ethnic conflict, but also with faster structural transformation. The chapter shows that migrant composition can generate effects distinct from... https://escholarship.org/uc/item/3fx6r9xv Mon, 27 Jul 2026 00:00:00 +0000 Gray, Andre https://escholarship.org/uc/item/3023w2pd Fluid dynamics on curved and evolving surfaces arises throughout soft matter physics, biology, and morphogenesis, where geometry and topology fundamentally shape flow behavior, yet classical fluid theories remain largely restricted to simple stationary geometries. This thesis develops a unified geometric and variational framework for fluid mechanics on arbitrary surfaces using coordinate-free differential geometry, Onsager’s variational principle, and structure-preserving discretization. We first reformulate the evolving surface Stokes equations into an invariant variational form that avoids tangent–normal decomposition and naturally yields stable numerical schemes on surfaces of arbitrary shape and topology. We then derive a vorticity-based formulation of the Navier–Stokes equations on curved surfaces with nontrivial topology, orientability, and boundary curvature, revealing curvature-dependent effects in viscous forces, vortex dynamics, harmonic modes, and vorticity generation.... https://escholarship.org/uc/item/3023w2pd Mon, 27 Jul 2026 00:00:00 +0000 Zhu, Cuncheng https://escholarship.org/uc/item/2wt4b2rc Understanding how emotional states shape action – and how action, in turn, influences emotional experience – is a central challenge in human clinical neuroscience. These interactions emerge through coordinated activity across brain networks involved in emotion, motor control, and autonomic regulation and are disrupted across many neurodevelopmental and psychiatric conditions. Tourette syndrome (TS), a neurodevelopmental disorder characterized by unwanted motor and vocal tics that are influenced by emotional arousal, autonomic state, and premonitory urges, provides a particularly informative model system for investigating these interactions. Using TS as a model system, this dissertation combines functional neuroimaging and autonomic physiology to examine how functional brain network organization and bodily state interact across multiple levels of analysis. First, I establish a normative foundation by characterizing the reproducibility and developmental organization of whole-brain... https://escholarship.org/uc/item/2wt4b2rc Mon, 27 Jul 2026 00:00:00 +0000 Ali, Sana https://escholarship.org/uc/item/1gk282qk “Causal diversity,” as an umbrella term, refers to the diversity found within causation and related subject matters. This dissertation investigates four specific topics on which considerations about causal diversity may shed new light: philosophical approaches to theorizing about causation, background conditions, causal explanations, and the causal grounds of Simpson’s paradox. First, the dissertation distinguishes between two kinds of philosophical accounts of causation: grand theories and middle-range theories. Grand theories of causation treat causation primarily as a homogeneous category and aim to improve our philosophical understanding of causation at a highly general level. By contrast, middle-range theories of causation treat causation primarily as a heterogeneous category, with each theory aiming to improve our philosophical understanding about a specific type of causal relation. The two kinds of theory complement each other in providing a comprehensive philosophical... https://escholarship.org/uc/item/1gk282qk Mon, 27 Jul 2026 00:00:00 +0000 Cai, Weixin https://escholarship.org/uc/item/0q55m60j This dissertation studies how transportation costs and frictions shape economic activity within and across cities. Across three essays, it examines how commuting costs, transit networks, and trade frictions affect where people live and work, how firms and households connect to markets, and how public policy changes welfare.The first chapter develops a quantitative urban model in which private and public transit jointly determine commuting costs within a city. We investigate how private providers’ decisions shape commuting costs considering complementarities with the public network, and the welfare gains and spatial consequences of policies that directly shift prices such as fare regulation and subsidies. The mechanisms underscore that the endogenous response of the private sector and network-wide cost interactions are central to understand the effects of transit interventions.The second chapter, coauthored with Leticia Juárez and Román Zárate, studies optimal transit networks... https://escholarship.org/uc/item/0q55m60j Mon, 27 Jul 2026 00:00:00 +0000 Mosqueda Juarez, Jordan https://escholarship.org/uc/item/05k2f5sw Programming environments have enabled complex problem solving for a wide variety of contexts, but the gap between a user's mental model and the program's underlying state always exists. The complexities of programming contexts, such as GUI development, AI-powered code generation, and non-linear data exploration, have exacerbated the gap. This dissertation argues that live and responsive programming environments serve as a critical cognitive framework that maintains human agency across these diverse contexts. By providing continuous, executable feedback situated in the corresponding contexts, liveness does more than simply accelerate development; it allows users to externalize their reasoning, validate program behavior in real time, and maintain control.To that end, this dissertation explores how liveness can achieve these goals in several distinct settings, with different user populations. First, via a system named Unfold, I demonstrate how liveness eases the debugging process... https://escholarship.org/uc/item/05k2f5sw Mon, 27 Jul 2026 00:00:00 +0000 Huang, Ruanqianqian (Lisa) https://escholarship.org/uc/item/4nq2098j Many problems arising in optimization, game theory, statistics, and machine learning are nonconvex and difficult to solve globally. Local methods can be useful, but they may fail to identify global solutions, certify global optimality, or characterize full solution sets. Moment-SOS relaxations provide a global framework for polynomial optimization through semidefinite programming and algebraic positivity certificates. This dissertation studies how this framework can be adapted to provide global methods for structured problems that are not simply standard polynomial optimization problems. The first part studies generalized Nash equilibrium problems with quasi-linear constraints, where each player's objective is polynomial and each player's constraints are linear in that player's own strategy and polynomial in the other players' strategies. A generalized Nash equilibrium is a strategy tuple in which each player's strategy is a global minimizer of that player's optimization... https://escholarship.org/uc/item/4nq2098j Fri, 24 Jul 2026 00:00:00 +0000 Choi, Jiyoung https://escholarship.org/uc/item/0h83k5z7 As classrooms become more diverse, teachers are increasingly positioned as agents of positive social change. Yet this expectation often assumes teacher agency rather than examining whether, how, and under what conditions teachers are able to act in support of equity. This question is especially important in India, where the Right to Education Act has reserved 25\% of entry-level seats in private schools for students from Economically Weaker Sections (EWS) and disadvantaged groups since 2009. While the policy has expanded access to private schooling for more than 22 million EWS students, access alone does not guarantee meaningful inclusion. This mixed-methods, three-paper dissertation examines how teachers in two private schools in Delhi understand and enact agency with EWS students. Grounded in ecological theories of teacher agency, it uses an explanatory sequential design to work with 143 teachers. The first study maps variation in teachers' perceived agency and identifies... https://escholarship.org/uc/item/0h83k5z7 Fri, 24 Jul 2026 00:00:00 +0000 Tripathi, Tarang https://escholarship.org/uc/item/8095n2tw For decades, path tracing has been the gold standard for image synthesis. Artists design a virtual scene, place lights appropriately, and path tracing simulates lighting by sampling random light paths throughout the scene. Many advanced techniques are often utilized to enable simulation of complex lighting effects such as caustics. Although path tracing has been the focus of research for so long, interactive path tracing has largely been elusive until much more recently, when GPUs began to support hardware ray tracing. Still, even after ray tracing hardware has received multiple generations of improvements, practical real-time path tracing implementations are limited to relatively simple lighting conditions, and struggle with more complex lighting effects often seen in offline renderers.This dissertation outlines several key improvements which enable real-time renderers to efficiently handle much more complex lighting conditions, including effects such as caustics.We first introduce... https://escholarship.org/uc/item/8095n2tw Thu, 23 Jul 2026 00:00:00 +0000 Hedstrom, Trevor James https://escholarship.org/uc/item/7q03k6fx The eyes—the ‘windows to the soul’—are the primary apparatus of vision. Pupils adjust to regulate the light intake, while eye movements bring relevant visual information to the fovea. However, recent work indicates that the eyes not only receive sensory input but also reflect content held “in mind” with working memory (WM). For example, during memory maintenance, the size of the pupil adapts to the brightness feature of what is remembered, and the gaze veers toward the location of remembered stimuli. While these findings highlight a potential role for ocular signals in WM, it is unclear whether these signals are mere retrospective sensory echoes, or whether they are goal-oriented and flexible to WM task demands. The sensorimotor recruitment model of WM holds that perceptual processing regions engaged during visual perception are recruited during WM maintenance. Crucially, the nature and strength of recruitment are modulated by the task demands. Therefore, if the ocular WM... https://escholarship.org/uc/item/7q03k6fx Thu, 23 Jul 2026 00:00:00 +0000 Dong, Yueying https://escholarship.org/uc/item/5522h19w This dissertation investigates the normative significance of asymmetrical power and its implications for equality, democracy, and responsibility. Across five body chapters, I argue that understanding what is objectionable about such hierarchies is essential to understanding what equality demands of our institutions. “Moral Equality, Relational Equality, and Democracy” defends the view that democracy is constitutive of a society of equals. Against recent critics who argue that unequal distributions of political power are compatible with moral equality, I show that their conception of moral equality is implausibly permissive, and that a more defensible conception entails that inequalities in political power damages our status as equals. “Relational Equality and the Parallel Case for Workplace Democracy” extends this line of reasoning into the economic sphere. I argue that if democracy is necessary to secure equality at the state level, then parallel reasoning grounds... https://escholarship.org/uc/item/5522h19w Thu, 23 Jul 2026 00:00:00 +0000 Chipp-Miller, Aaron David https://escholarship.org/uc/item/4k74d0tc This dissertation studies how expectations and information shape macroeconomic policy and inference about aggregate dynamics. The first chapter studies the impact of interest rate disagreement between the Federal Reserve and financial markets on the transmission of U.S. monetary policy. I show that these expectation gaps are central to monetary transmission: when disagreement is high, the effectiveness of conventional policy is dampened through substantially weaker effects on output. The results challenge explanations based on a systematic Fed information advantage and instead point to disagreement as a state variable for monetary transmission. The second chapter, coauthored with Edoardo Briganti, develops a high-frequency cross-sectional approach to identifying military news shocks. We show that regressions of defense contractors' cross-sectional stock returns, as implied by asset pricing theory, allow us to identify changes in expected defense spending. The resulting... https://escholarship.org/uc/item/4k74d0tc Thu, 23 Jul 2026 00:00:00 +0000 Amodeo, Francesco https://escholarship.org/uc/item/4cn0q96j This dissertation develops numerical methods for PDEs and stochastic dynamics in which the main computational difficulty comes from anisotropy, singular probability measures, nonlocal operators, or heavy-tailed data. The unifying idea is to choose the numerical representation from the structure of the continuous operator: anisotropic neighborhoods for non-divergence elliptic equations, radial integral kernels for fractional Fokker–Planck equations, and learned generators for continuous-time policy evaluation.The first part studies linear elliptic equations in non-divergence form. These equations arise in stochastic control and fully nonlinear PDEs, but their lack of variational structure makes standard finite element approaches delicate. We introduce a monotone meshfree finite difference method based on nonlocal relaxation and local ?1 optimization. The method constructs sparse positive stencils on point clouds and gives an improved stencil-width guarantee in the near-degenerate... https://escholarship.org/uc/item/4cn0q96j Thu, 23 Jul 2026 00:00:00 +0000 Ye, Qihao https://escholarship.org/uc/item/3sg7g7sz Over time, oceanographic research techniques are transforming. Autonomous platforms are taking on roles once reserved for ship-based campaigns, instruments are shrinking, and costs of electronic components are falling. Adapting to the changing landscape requires reliable power at sea and affordable, field-ready sensing. This dissertation advances both.Part one addresses energy autonomy for small oceangoing platforms through the development and field validation of a wave energy converter (WEC) model. A lumped-parameter framework driven by Morison-equation forcing is validated in a laboratory setting with an idealized representative of a heave plate WEC. The model is then validated against two field deployments of the Floating Wave Energy Receiver (FloWER), a field going heave-plate WEC. The FloWER was deployed twice offshore of La Jolla, CA where its dynamics were recorded via on-board electronics. Heave plate motion was reproduced with a median ? 2 of 0.65 and a mean Willmott... https://escholarship.org/uc/item/3sg7g7sz Thu, 23 Jul 2026 00:00:00 +0000 Walker, Hannah https://escholarship.org/uc/item/0xv3p3g5 Reconfigurable networks that use optical circuit switches are slowly paving their way into datacenters due to their ability to offer high bandwidth at low cost and power. However, the majority of these reconfigurable networks are aimed at conventional datacenter clusters where the traffic patterns are unpredictable. While dynamically reconfigurable network topologies have shown promise for general datacenter workloads, in this work, I investigate the extent to which they may benefit machine learning (ML) collectives. Several algorithms have been proposed to implement collectives on static networks (of different topologies), but less well studied has been the interaction between these algorithms and reconfigurable network fabrics. Therefore, I seek to quantify the potential benefit that a reconfigurable network layer can bring to the large-scale collectives that underpin large DNN training. Following this, I study the extent of fragmentation in ML torus clusters. In multi-tenant... https://escholarship.org/uc/item/0xv3p3g5 Thu, 23 Jul 2026 00:00:00 +0000 Athapathu, Dilini Rukshani https://escholarship.org/uc/item/06n032vw How can artificial agents learn to make decisions from high-dimensional sensory input? This question is central to building robots and embodied systems that operate under raw, noisy observations in diverse environments. Reactive policies trained end-to-end tend to overfit, fail under distribution shift, and rarely transfer across tasks or embodiments. This dissertation argues that scalable visuomotor control requires control-oriented representations and predictive models, and shows that temporal difference learning provides a unifying principle for building such world models. It develops this claim through three connected contributions. The first examines how learned representations shape robustness and generalization in visual reinforcement learning. The second reframes world models as latent predictive models whose value lies in supporting accurate decisions rather than reconstructing observations, leading to the model-based reinforcement learning methods TD-MPC and TD-MPC2.... https://escholarship.org/uc/item/06n032vw Thu, 23 Jul 2026 00:00:00 +0000 Hansen, Nick https://escholarship.org/uc/item/9h49s65k This dissertation consists of three chapters, all using experiments to study human decision making in economic environments. Chapter 1 uses an in-person laboratory experiment to examine how erroneous statistical inference and strategic misprediction impact people's choices to interact with or exit from environments characterized by potential manipulation. Using a novel adversarial bandit game, I find that while subjects exhibit errors in Bayesian updating in directly elicited beliefs, providing subjects with correct posteriors does not change aggregate exit behavior. Moreover, I find no evidence for non-normative strategic prediction. Together, these findings suggest that people neither over-exit nor under-exit on average from potentially manipulated environments once risk preferences are accounted for. Chapter 2 uses an in-person laboratory experiment to study the role of communication in bargaining, specifically, how claiming reserve prices ("explicit messaging'')... https://escholarship.org/uc/item/9h49s65k Wed, 22 Jul 2026 00:00:00 +0000 Cizeski, Jay https://escholarship.org/uc/item/7c62t91r 3D humans have become an increasingly important component in applications such as virtual reality, augmented reality, gaming, and e-commerce. As the quality and accessibility of 3D human representations continue to improve, there is a growing demand for methods that can not only reconstruct realistic 3D humans, but also understand and edit them in a semantically meaningful manner. However, despite the growth of publicly available 3D assets, annotated 3D human data remains scarce. Consequently, existing approaches can only segment and edit fixed categories, restricting their flexibility and applicability in real-world scenarios. In this dissertation, we investigate methods for semantic understanding, reconstruction, and editing of 3D humans that reduce the dependence on scarce annotated 3D human data.The first part of this dissertation focuses on semantic understanding of 3D humans through text-driven and open-vocabulary segmentation methods. We first propose a language-driven... https://escholarship.org/uc/item/7c62t91r Wed, 22 Jul 2026 00:00:00 +0000 Suzuki, Keito https://escholarship.org/uc/item/5mr8k59g Working in Ronchetti's topological-groupoid model $\HH^*_{\rm gpd}(G,K)$ of the derived Hecke algebra, we show that the anti-involution constructed by Ollivier--Schneider via Yoneda composition is realized simply by the coordinate swap $F(x,y) \mapsto F(y,x)$. Second, we lift this groupoid model to the chain level: we construct an explicit differential graded algebra $\A\cp$ built from equivariant maps from the action groupoid $\underline{G}^2$ into the endomorphism complex of an injective resolution of $k$, prove it is a unital associative DGA under a convolution product defined via restriction and corestriction, establish a $k$-algebra isomorphism $H^*(\A\cp) \cong \HH^*_{\rm gpd}(G,K)$.We also generalize a theorem of Ardakov--Schneider on the Bernstein center to the category $\Mod_k(G,\psi)$ of smooth representations with a fixed central character $\psi : C \to k^\times$ on a closed central subgroup $C \leq Z(G)$. For a locally pro-$p$ group $G$ with no proper open centralizers... https://escholarship.org/uc/item/5mr8k59g Wed, 22 Jul 2026 00:00:00 +0000 Zhao, Zehong https://escholarship.org/uc/item/4m24g967 The analysis of complex data sampled from geometric domains is a central challenge for applications ranging from geospatial analysis to chip design. Modern applications often require solving geometric problems repeatedly over massive datasets. Learning-based methods offer a way to exploit recurring structure and produce fast approximations that can be integrated into larger data-driven pipelines. However, such approaches to these problems are particularly challenging as geometric data often exhibits intrinsic invariances that must be preserved by the model. The goal of this thesis is to develop effective learning-based approaches tailored for approximating solutions to geometric problems. I address this along two directions: first, by designing methods for metric approximations and second, by developing neural architectures that approximate solutions for families of geometric problems. In the first direction, I introduce two data-driven data structures for approximating... https://escholarship.org/uc/item/4m24g967 Wed, 22 Jul 2026 00:00:00 +0000 Chen, Samantha https://escholarship.org/uc/item/3h67340k Echoes of Sappho; Tending to Wounds of War and Loss in North American Literature explores the literary reception history of Sappho in North America. This project relies upon a self-reflexive sociological method of research, inspired by the injunction made by second wave feminists to “Start with the Self,” as recurrently echoed by contemporary post-critical philosophers and the self-reflexive sociologists. Tethering an inquiry into the history of Sapphic Reception in North America to personal and self-oriented methodological constraints, while also giving weight to contemporary debates in the fields of literary reception theory, philosophies of history and presentism, this project demonstrates an alternate framework for historiographic scholarship, one which embraces the social foundations of scholarly writing to theorize reception, literature, and historiography. Relying upon the work and historical situatedness of contemporary poets Makenzie Read, Eloise Klein Healy, and Sarah... https://escholarship.org/uc/item/3h67340k Wed, 22 Jul 2026 00:00:00 +0000 Read, Makenzie Rose https://escholarship.org/uc/item/08b880hr Island wakes play an important role in redistributing momentum, heat, tracers, and energy in the ocean. This dissertation investigates the dynamics and variability of island wakes and coastal upwelling around Mahé, Seychelles, an island located on the shallow Seychelles Plateau in the Indian Ocean, using a combination of field observations and numerical simulations. The regional circulation is strongly shaped by the seasonally reversing monsoon, with eastward flow during the northwest monsoon and westward to northwestward flow during the southeast monsoon. Superimposed on this seasonal variability are energetic near-inertial oscillations, which introduce substantial unsteadiness into the flow. First, we examine drag characteristics associated with the island wake. By decomposing form drag into barotropic and baroclinic components, the analysis links the isopycnal displacement and momentum extraction by the island. Results show that near-inertial oscillations substantially... https://escholarship.org/uc/item/08b880hr Wed, 22 Jul 2026 00:00:00 +0000 Chen, Hsin-yi https://escholarship.org/uc/item/2q6757k6 Visual signals like rendered images, radiance fields, vector graphics, and the outputs of shader programs are smooth in most places but exhibit sharp changes at object boundaries, material transitions, and decision boundaries in the programs that compute them (e.g., where control flow branches on conditionals, step functions, or min/max). These discontinuities are central to perception, yet they pose fundamental difficulties for gradient-based pipelines that assume smoothness. This thesis is organized around three pillars of gradient-based visual computing: how we represent such signals, how we optimize their parameters to fit observations, and how we differentiate through the integrals that arise in the forward process. In each setting, standard methods break down at discontinuities; we show that treating discontinuities as first-class yields sharper representations, optimizers that preserve edges, and derivatives that correctly include the boundary term.In representation, neural... https://escholarship.org/uc/item/2q6757k6 Fri, 17 Jul 2026 00:00:00 +0000 Belhe, Yash Anand https://escholarship.org/uc/item/9pc0f57g We provide a complete characterization of the unirationality of universal moduli spaces of semistable bundles on curves, and study their intersection theory over hyperelliptic curves. We compute the Chow rings of the universal moduli spaces of semistable rank two bundles over genus two curves, extend these computations to the stack of all bundles, and present a guiding question for higher genera and ranks in the form of a conjecture. The injectivity part of the conjecture is proven for rank two bundles over hyperelliptic curves of arbitrary genus with odd degree. These computations further allow us to compute the Chow rings of universal marked hyperelliptic Jacobians. https://escholarship.org/uc/item/9pc0f57g Thu, 16 Jul 2026 00:00:00 +0000 Saha, Shubham https://escholarship.org/uc/item/8d88h3kh We determine the τ n -torsion in the first 5 lines of the E2 page of the C-motivic Adams spectral sequence using the techniques of Burklund-Xu. In particular, every element in this range is either τ 1 -torsion or τ -free. We also show that τ n -torsion elements can appear only in Adams filtration at least 2n + 2 and give further evidence of a possible 3n bound. https://escholarship.org/uc/item/8d88h3kh Thu, 16 Jul 2026 00:00:00 +0000 Benson, Jordan https://escholarship.org/uc/item/892295tz Infants learn to share attention with caregivers long before they can speak, yet the development of this capacity remains poorly understood, in part because almost everything we know comes from controlled laboratory studies of Western, middle-class families using scripted cues that bear little resemblance to the behaviors that caregivers seem to actually use at home. This dissertation addresses that gap, across two parallel lines of inquiry.The first concerns the development of joint attention itself. The results of Chapter 2 show that attention-following in an methodologically advanced laboratory paradigm emerges slowly and incrementally across the first year, with no reliable following before 7 months, and gradual growth thereafter. In addition, Chapter 3 shows that laboratory and home settings yield systematically different developmental pictures. The same infants follow attention substantially more following at home than in the lab, developmental trajectories diverge across... https://escholarship.org/uc/item/892295tz Thu, 16 Jul 2026 00:00:00 +0000 Tang, Yueyan https://escholarship.org/uc/item/605701j1 Aspiring autocrats often face substantial opposition at the subnational level, where locally elected officials use their access to resources and public platforms to resist democratic backsliding. How do central governments respond to this threat? This dissertation argues that antipluralist parties engage in a systematic strategy of fiscal recentralization—reducing subnational fiscal autonomy—to starve local opponents of the resources necessary for national contestation.I implement a multi-method three stage research design, primarily focusing on democratic backsliding in Poland, 2015-2023. First, leveraging a novel dataset of Polish municipal finances and elections (2004–2024), I implement a close-elections regression discontinuity design. Municipalities narrowly electing a PiS mayor see significant annual increases in autonomous revenue and decreases in transfer dependency, indicating selective protection from fiscal recentralization. I also find evidence that opposition-held... https://escholarship.org/uc/item/605701j1 Thu, 16 Jul 2026 00:00:00 +0000 Ntounias, Theodoros https://escholarship.org/uc/item/58z7j1d3 While optimal control for finite-dimensional nonlinear systems has been studied extensively theoretically, few general-purpose numerical tools, especially for model order reduction and approximation, have been developed that are as straightforward to use as the methods available for linear systems. In the interest of developing scalable, general-purpose computational methods, we approach these topics for weakly nonlinear systems using local polynomial approximations. First, we consider the polynomial-polynomial regulator (PPR), a generalization of the famous linear-quadratic regulator (LQR) optimal control problem, but for polynomial dynamics and polynomial cost functions. Leveraging tensor structure through the use of a Kronecker product representation, we obtain local approximations to the value function and optimal control law that are guaranteed to be superior to the widely-used linear approximations locally. Combined with novel low-rank approximation techniques, our... https://escholarship.org/uc/item/58z7j1d3 Thu, 16 Jul 2026 00:00:00 +0000 Corbin, Nicholas https://escholarship.org/uc/item/4cv9427z Cancer is often characterized as a disease of evolution, where a cell has evolved to prioritize its own fitness without regard for the organism. However, the process of evolution predates cancer initiation. My work uses models informed by data to study this precancerous evolution and equip clinicians and patients with the tools needed to reduce cancer risk. First, we developed a method to estimate the growth rate of clones in blood using single-cell whole genome sequencing data. I applied this method to published data and showed that higher growth rates were associated with progression to cancer. Where possible, I validated these estimates using alternative methods. I built an R package, cloneRate, to allow others to quickly and easily estimate growth rates. Structured data in electronic health records represents a tiny fraction of a patient’s medical history. To address this, I used large language models to extract clinically relevant information from unstructured... https://escholarship.org/uc/item/4cv9427z Thu, 16 Jul 2026 00:00:00 +0000 Johnson, Brian Daniel https://escholarship.org/uc/item/0wc4m9q4 Hamilton-Jacobi-Bellman (HJB) equations are nonlinear partial differential equations that arise naturally from the dynamic programming principle in optimal control applications. In most applications, the HJB equations cannot be solved analytically. Rather, the solutions can often only be solved approximately by numerical methods, which nevertheless become prohibitively expensive for high dimensional systems. This thesis research presents some efficient numerical methods and related developments for such problems using staticization and duality representations.For deterministic optimal control problems, the statical-quadratic duality is used to transform the Hamiltonian in the HJB equation, by which the original optimal control problem is converted to a statical control problem with linear dynamics, with the introduction of auxiliary control processes corresponding to the nonlinearity in the dynamics.This approach is also used for stochastic applications, for which the time-dependent... https://escholarship.org/uc/item/0wc4m9q4 Thu, 16 Jul 2026 00:00:00 +0000 Zheng, Yifei https://escholarship.org/uc/item/07c3n7c6 This dissertation aims to contribute a new perspective of legal criticism to the field of Victorian Literature and argues that female ghosts depicted in nineteenth century narratives were metaphorically critical of gendered legal injustice in British law. I will also be contending that the metaphorical ghost figure is a posthuman entity that provides insight on how bodies are excluded and othered, and that this construction predates the technological posthuman cyborg discussed in Donna Haraway and Katherine Hayles’ work. Literary ghosts will be shown to encompass a variety of posthuman qualities in Victorian Literature, including their simultaneous corporeal and incorporeal forms, their creation from being othered in life, and hauntings that enact resistance to social limitations. Each of the narratives that will be closely analyzed throughout this dissertation share the common factor of depicting female ghosts whose lives were affected by a flawed legal system that did not protect... https://escholarship.org/uc/item/07c3n7c6 Thu, 16 Jul 2026 00:00:00 +0000 Miller, Kellie https://escholarship.org/uc/item/9vs81969 Targeted protein degradation strategies such as proteolysis targeting chimeras (PROTACs) and molecular glue degraders have transformed both basic research and drug discovery. These small molecule strategies harness cellular machinery to ubiquitinate and degrade a protein of interest. This technology offers the opportunity to control protein fate which can help us uncover and tackle diseases. While a very promising strategy, there are several barriers to successfully applying it in practice including a lack of robust ligands, intensive empirical optimization, and poor physicochemical properties. In this dissertation, I outline solutions to overcome these challenges including a chemical genetic approach to degrade protein aggregates, chemical tuning the kinetics of PROTACs to accelerate discovery of active PROTACs, and a fatty acid prodrug method to aid in transport across the blood-brain barrier. https://escholarship.org/uc/item/9vs81969 Tue, 14 Jul 2026 00:00:00 +0000 Gadbois, Gillian Elizabeth https://escholarship.org/uc/item/9n72p7z4 This dissertation examines a central paradox of authoritarian rule: how regimes expand access to the legal system without allowing it to become an independent political constraint. I argue that this balance relies heavily on frontline legal intermediation. In authoritarian settings, legality is produced not just by formal rules, but by the everyday actors who translate them into practice. These intermediaries broaden access, channel grievances, and increase state capacity, while also restricting when and how legal access translates into practical empowerment.Through three empirical studies set in China, this project traces how lawyers and institutions mediate citizen-state interactions. A nationwide audit of the 12348 legal-aid hotline—involving 3,200 randomized calls across 302 cities—shows that legal institutions often grant recognition while withholding practical leverage. Rather than excluding politically sensitive or socially disadvantaged callers outright, the system provides... https://escholarship.org/uc/item/9n72p7z4 Tue, 14 Jul 2026 00:00:00 +0000 Zu, Ziwen https://escholarship.org/uc/item/6pj4k2dp In the era of artificial intelligence (AI), foundation models have become engines for producing text, code, images, and complex design artifacts through human–AI interaction. This paradigm shift transforms not only how content is created but also changes the security landscape. As AI systems are increasingly deployed in critical domains such as software engineering, hardware design, scientific discovery, and digital media, ensuring the trustworthiness of both models and their generated artifacts becomes a fundamental challenge. Key concerns include ownership attribution, provenance tracking, authenticity verification, misuse prevention, and regulatory compliance across the lifecycle. However, achieving trustworthy and secure AI at scale is inherently difficult due to the massive volume of generated content, the need for real-time interaction, and the risk that artifacts can be adversarially modified and redistributed. Effective protection mechanisms must, therefore, be scalable,... https://escholarship.org/uc/item/6pj4k2dp Tue, 14 Jul 2026 00:00:00 +0000 Zhang, Ruisi https://escholarship.org/uc/item/3m344766 This dissertation is dedicated to the theoretical study of exotic quantum critical phenomena, primarily using quantum field theory techniques. Its core consists of three parts, each devoted to a different quantum critical system. The first part examines a continuous field theory describing an SU(N) critical spin chain coupled to a gapless dissipative bath. A renormalization group analysis reveals a rich phase diagram with multiple fixed points, one of which corresponds to an exotic dissipative critical phase. The second part investigates a lattice of single-impurity multichannel Kondo models, where analytical and numerical results reveal the existence of a ``dimensionality-changing'' quantum phase transition between dynamically decoupled zero-dimensional non-Fermi liquids and a one-dimensional valence-bond-solid phase. Finally, the third part focuses on quantum phase transitions involving symmetric mass generation (SMG), for which a controlled perturbative approach is developed.... https://escholarship.org/uc/item/3m344766 Tue, 14 Jul 2026 00:00:00 +0000 Martin, Simon https://escholarship.org/uc/item/2g82x3d7 This thesis focuses on state estimation in nonlinear systems, leveraging Bayesian frameworks, optimization algorithms, and differentiable programming. We apply both sequential and variational data assimilation methods which integrate prior information, dynamical models, and observations to estimate ``posterior'' distributions. The central objective is to estimate a system state from partial, noisy observations and to constrain or improve this process through the inclusion of domain knowledge. The techniques employed draw upon numerical methods, Bayesian inference, control theory, geophysical fluid dynamics, climate science, and software engineering. The application areas in these studies include autonomous underwater vehicle (AUV) navigation, tropical ocean state estimation, and differentiable modeling of the Earth system. Chapter one develops a sequential estimation algorithm to localize small AUVs using side-scan sonar observations. The algorithm is novel in both its... https://escholarship.org/uc/item/2g82x3d7 Tue, 14 Jul 2026 00:00:00 +0000 Davenport, Ellen https://escholarship.org/uc/item/20b677zz From fighting over toys to competing for prestigious college admissions, resource allocation has been at the core of individual lives and human society throughout history. My dissertation examines how humans develop economic cognition and behavior through the lens of social distribution. Across three chapters, I analyzed this question at two levels: how children understand and weigh economic mental states of individuals in reward distribution (Chapter 1), and how individuals reason about broader socioeconomic institutions, including meritocracy and egalitarianism (Chapters 2 and 3). Findings show that sensitivity to profit motive emerges in middle childhood and affects children’s reward allocation decisions (Chapter 1), that reasoning about social compensation systems develops from childhood through adulthood (Chapter 2), and that by adolescence, distributional preferences become tied to self-interest (Chapter 3). Together, these studies trace the developmental trajectories through... https://escholarship.org/uc/item/20b677zz Tue, 14 Jul 2026 00:00:00 +0000 Shao, Shuai https://escholarship.org/uc/item/1c32v5c5 One of the central problems in Arithmetic Statistics is to count number field extensions of fixed degree over an arbitrarily fixed base field such that the Galois group of the Galois closure is isomorphic to a prescribed Galois group and the absolute norm of the discriminant is bounded above. A general conjectural framework for the asymptotic behavior of such counting functions was formulated by Malle.This dissertation focuses on the case in which the Galois group is of the form C2 ≀H, where H is a transitive permutation group. Malle’s prediction in this setting was first established in the case C2 ≀ C2 ∼= D4 by Cohen, Diaz y Diaz, and Olivier [CDyDO02]. Subsequently, Klüners [Klü12] proved the expected asymptotic formula for general H under mild assumptions. In Chapter 3, we provide a detailed account of the progress in this direction.The primary contribution of this dissertation is the derivation of improved power-saving error terms for counting extensions with Galois group... https://escholarship.org/uc/item/1c32v5c5 Tue, 14 Jul 2026 00:00:00 +0000 Chakraborty, Arijit https://escholarship.org/uc/item/09m9n8pm The ultraproduct construction in the context of operator algebras (specifically tracial von Neumann algebras) was first implicitly introduced by Wright in 1954 and has since been used to great effects in both the theory of von Neumann algebras and C*-algebras. Its great success is owed to at least three reasons: 1. First, by considering an operator algebra as sitting within its ultrapower, one may consider various asymptotic phenomena that may not be available within the algebra itself. This includes, e.g., central sequence algebras and asymptotic freeness phenomena. These asymptotic phenomena allow one to obtain structural properties of the operator algebra in question by first passing to a larger algebra within the ultrapower witnessing certain asymptotic behaviors and then back to the original algebra in an approximate fashion; 2. Second, ultraproducts are intrinsically related to model theory. Model-theoretic results such as Łoś's theorem, Keisler-Shelah theorem,... https://escholarship.org/uc/item/09m9n8pm Tue, 14 Jul 2026 00:00:00 +0000 Gao, David https://escholarship.org/uc/item/02h9v7m8 Reinforcement learning (RL) has emerged as a powerful framework for sequential decision-making under uncertainty, with significant impact across a wide range of modern applications, including robotics, recommendation systems, autonomous driving, and modern large language model (LLM) training. At its core, RL studies how an agent interacts with an environment over time, making decisions based on observed states in order to maximize cumulative rewards. This interaction is typically modeled as a Markov decision process (MDP), where the agent must balance exploration and exploitation while learning from potentially limited and noisy feedback. In this thesis, we study reinforcement learning under two practical constraints: adaptivity and privacy. In the first part, we study reinforcement learning under constraints on policy switching costs, motivated by practical scenarios where frequent changes in policies or actions are expensive or undesirable. We consider a formulation of... https://escholarship.org/uc/item/02h9v7m8 Tue, 14 Jul 2026 00:00:00 +0000 Qiao, Dan https://escholarship.org/uc/item/7vf8c2sw Fish chorusing, in which many fish call simultaneously for extended periods, is often associated with reproduction. It therefore provides insight into spawning times, habitat use, biodiversity, and ecological change. Through passive acoustic monitoring (PAM), we can non-invasively and continuously monitor sound-producing fish. Here, I leverage PAM across California National Marine Sanctuaries and the Southern California Bight to develop automated fish chorus detection methods and to investigate how fish chorusing is distributed across space and time, what drives chorusing, and how chorusing responds to ocean warming.In Chapter 1, I developed SoundScape Learning, an unsupervised automated method that separates fish chorus from soundscape, through integration of randomized robust principal component analysis, unsupervised clustering, and a neural network classifier. Applied across 14 recording locations off central and southern California, SoundScape Learning successfully detected... https://escholarship.org/uc/item/7vf8c2sw Mon, 13 Jul 2026 00:00:00 +0000 Kim, Ella Bea https://escholarship.org/uc/item/7b89441b Modern machine learning models, despite their power, often lack formal guarantees of reliability. This thesis addresses this challenge by studying the reliability of machine learning algorithms through the complementary lenses of replicability and testability. First, we address the challenge of creating verifiable learning algorithms through the framework of testable learning. Standard distribution-specific learning algorithms provide guarantees only when the underlying data distribution meets certain assumptions, which may not hold in practice. Testable learning strengthens these guarantees by requiring an algorithm to either output a correct hypothesis or reject the data when it detects that the distributional assumptions are violated. We develop novel and efficient testable learners for fundamental tasks, including robust parameter estimation and learning halfspaces. Our work overcomes key limitations of prior methods, providing the first polynomial-time testable learners... https://escholarship.org/uc/item/7b89441b Mon, 13 Jul 2026 00:00:00 +0000 Liu, Sihan https://escholarship.org/uc/item/4d7539nj Building general-purpose humanoid robots that can perform diverse tasks in unstructured environments is a central goal of robotics. This dissertation addresses three core challenges toward this vision: whole-body control, scalable data collection, and data-efficient learning for humanoid manipulation. We begin by developing learning-based whole-body control for humanoid robots. We present ExBody, a reinforcement learning framework that enables expressive upper-body motions by leveraging large-scale human motion capture data, while maintaining robust locomotion. Building on this control foundation, we introduce Open-TeleVision, an immersive teleoperation system that provides stereoscopic visual feedback, enabling operators to intuitively collect high-quality manipulation demonstrations. We then address the integration of locomotion and manipulation through Mobile-TeleVision, which combines predictive motion priors with decoupled upper and lower body control, allowing... https://escholarship.org/uc/item/4d7539nj Mon, 13 Jul 2026 00:00:00 +0000 Cheng, Xuxin https://escholarship.org/uc/item/4bb7h4kv Understanding a scene from a single image calls for structured 2D predictions; when the goal is 3D, models must also infer coherent geometry and semantics despite occlusion and ambiguity. This dissertation connects structured 2D modeling to unified panoptic 3D reconstruction and compositional generative 3D models aligned with the input image.Chapter 2 presents Text Spotting Transformers (TESTR), a unified Transformer with a shared encoder and dual decoders that jointly localize curved scene text and recognize characters by regressing control points for polygonal or Bézier-curve boundaries. The design forgoes separate detection-and-recognition stages, region-of-interest pooling, and heavy heuristic post-processing, and includes box-to-polygon refinement; TESTR reaches state-of-the-art text spotting on challenging benchmarks.Chapter 3 introduces Uni-3D, a universal model for panoptic 3D scene reconstruction. A depth-aware Transformer jointly predicts panoptic segmentation and monocular... https://escholarship.org/uc/item/4bb7h4kv Mon, 13 Jul 2026 00:00:00 +0000 Zhang, Xiang https://escholarship.org/uc/item/3p9135g1 Targeted protein degradation (TPD) has recently gained widespread interest as both a novel therapeutic strategy and a useful tool in biomedical research. TPD modalities, such as bifunctional proteolysis targeting chimeras (PROTACs) and monovalent molecular glue degraders (MGDs), coopt the endogenous ubiquitin-proteasome system to induce the degradation of a target protein. Though TPD is a promising nascent chemical strategy, rationally designing a molecule to degrade specific targets remains challenging, particularly for MGDs. In this dissertation, I present work that develops a MGD targeting an understudied transcription factor, ZBTB11, validates ZBTB11 depletion as a potential strategy to overcome resistance to KRAS inhibition in pancreatic cancer, and defines the structure-activity-relationships underlying degradation of this novel target. https://escholarship.org/uc/item/3p9135g1 Mon, 13 Jul 2026 00:00:00 +0000 Tran, Nathan Luong https://escholarship.org/uc/item/08r3s7dc In many neural networks, different parameter values can yield the same loss, often due to underlying symmetries in the parameter space. We introduce a general framework for continuous symmetries based on equivariance in activation functions, revealing a new set of nonlinear, data-dependent symmetries. Using these symmetries, we derive topological properties of the minima and identify conserved quantities in gradient flows. As a practical application, we present an algorithm called symmetry teleportation, which leverages parameter symmetries to search loss level sets for points with desired properties. This approach leads to improvements in both convergence and generalization. We also discuss empirical approaches to find symmetries and uncover more structures in neural networks. https://escholarship.org/uc/item/08r3s7dc Mon, 13 Jul 2026 00:00:00 +0000 Zhao, Bo https://escholarship.org/uc/item/9vt129wn Vision restoration technologies offer the potential to cure blindness in over 200 million patients suffering from irreversible vision loss due to retinal degenerative diseases. While emerging therapies such as optoelectronic prostheses and optogenetic gene therapies show promise for vision restoration, they still require significant development to achieve robust visual outcomes. A notable complication is that retinal plasticity following degeneration may impact the quality of restored vision. Here we used subretinal electrical stimulation to examine changes in retinal circuit function. While the presence of increased spontaneous activity is well documented following degeneration, the spatiotemporal characteristics are poorly understood. We found that spontaneous activity is highly correlated across time and space in the degenerated retina. Despite this, stimulus-evoked receptive fields retain high spatiotemporal fidelity due to normal surround-mediated inhibition, suggesting... https://escholarship.org/uc/item/9vt129wn Thu, 9 Jul 2026 00:00:00 +0000 Carleton, Maya Alexandra Christine https://escholarship.org/uc/item/8jn1m31j The structural integrity of marine structures depends on the performance of their pile foundations, particularly landside (seismic) piles, that resist lateral earthquake forces. Their comparatively short, unbraced height above the mudline contributes to their high lateral stiffness, concentrating cyclic inelastic demands at the pile-to-deck (top) connection and the in-ground plastic hinge. In performance-based seismic design, strain demands at these hinge regions are obtained from nonlinear analyses and evaluated against limits defined in ASCE/COPRI 61-14.However, the code-prescribed in-ground hinge strain limits have been widely questioned because they lack clear experimental or analytical justification. Although more stringent limits have been proposed based on in-air testing of steel tubes and pipes, these studies do not reflect key contributors to marine pile response, including soil confinement and combined axial and lateral loading. This discrepancy highlights the need for... https://escholarship.org/uc/item/8jn1m31j Thu, 9 Jul 2026 00:00:00 +0000 Richards, Theresa L https://escholarship.org/uc/item/9gj551mz This thesis studies inverse problems for hyperbolic equations and Lorentzian manifolds, with the goal of recovering coefficients of a PDE or the geometry of an unknown region from boundary measurements. It develops microlocal and geometric methods adapted to Lorentzian settings. The first part concerns inverse problems for hyperbolic PDEs, including nonlinear and partial data problems; the second part concerns geometric rigidity questions for Lorentzian manifolds using time separation and scattering data.We first study a bounded time inverse scattering problem for a semilinear Dirac equation in (0, T)×R 3 with smooth nonlinearity F(x, z). We prove that the solution map from initial data at t = 0 to the solution at time t = T determines F(x, z) for x ∈ R 3 and |z| ≤ M, under natural a priori knowledge of low-order derivatives of F at z = 0. A key ingredient is a collision simulation mechanism that compensates for the absence of a physical boundary. Next, we analyze the restricted... https://escholarship.org/uc/item/9gj551mz Mon, 6 Jul 2026 00:00:00 +0000 Yi, Yuchao