http://www.rssboard.org/rss-specification 720 https://escholarship.org/uc/ucb_etd/rss Recent eScholarship items from UC Berkeley Electronic Theses and Dissertations Sun, 11 Oct 2026 11:25:18 +0000 https://escholarship.org/uc/item/9qc931tq This dissertation includes two works of the Tri-laboratory Effort in Nuclear Data (TREND), a collaboration among Lawrence Berkeley National Laboratory, Los Alamos National Laboratory, and Brookhaven National Laboratory that undertakes the challenges in accelerator-based isotope production. The experiments include the continuation of a long-standing campaign of cross section measurements via stacked-target irradiations, and a venture into the realm of in-beam y-ray spectroscopy, rarely employed in the field of isotope production.Motivated by the demand of 103Pd for brachytherapy, characterization of the nat Ag(p,x) reactions between incident proton energy of 40 MeV and 200 MeV was performed. The experimental methods and results of stacked-target irradiations on targets of silver, copper, and nickel are discussed, with an additional focus on the underlying complications in secondary-particle-induced reactions and the usage of monitor reactions. Experimental measurements made were... https://escholarship.org/uc/item/9qc931tq Tue, 8 Sep 2026 00:00:00 +0000 Lee, Yun-Hsuan "Abby" https://escholarship.org/uc/item/8mk7n44z Microtubule-based motors play essential roles in intracellular organization and cell division by transporting cargos and generating force along microtubules. Kinesin and dynein are microtubule motors that move toward the plus and minus ends of microtubules, respectively. These motors are highly coordinated to enable bidirectional cargo transport and precise spatial organization within cells. However, the molecular mechanisms that regulate kinesin and dynein activity and coordination remain elusive.In my doctoral work, I addressed this question using biochemical reconstitution and single-molecule imaging. First, I studied the activation and regulation of kinesin and dynein by the mitochondrial adaptor protein TRAK. I showed that TRAK activates dynein and enhances kinesin activation in vitro. In addition, I found that TRAK adaptors can recruit kinesin and dynein simultaneously, and these complexes, in which both motors are associated, move exclusively to the plus end at kinesin... https://escholarship.org/uc/item/8mk7n44z Thu, 3 Sep 2026 00:00:00 +0000 Aslan, Merve https://escholarship.org/uc/item/8gv560vq Theological trends in the seventeenth and eighteenth centuries influenced the development of prose fiction generally and the sentimental novel in particular. In 1667, Latitudinarian clergyman Simon Patrick claimed the Eucharist was like a “History or Romance” because of its ability to evoke sympathy—a unique way of putting the common notion that one ought to meditate on the life of Christ and become Christ-like. Sympathy then is morally good insofar as it causes one to become Christ-like and become morally good like Him. Pierre Hadot and Peter Sloterdijk have emphasized the capacity of philosophy’s literary form to inculcate spiritual practices; I argue that early novelists understood their work as inculcating a practice of sympathy that could be directed toward Christ. Clarissa not only alludes to but also adapts Jeremy Taylor’s 1650 Holy Living and Holy Dying, a representative of the imitatio tradition of Thomas à Kempis. However, the novel is Janus-faced, simulating secularity... https://escholarship.org/uc/item/8gv560vq Thu, 3 Sep 2026 00:00:00 +0000 Manno, Matthew https://escholarship.org/uc/item/8dn7b1q3 Active matter, composed of agents that consume energy at the microscopic scale, exhibits a wealth of collective behaviors that transcend traditional equilibrium thermodynamics. An example of this behavior is motility-induced phase separation (MIPS), where self-propelled particles assemble into dense and dilute phases in the absence of attractive interactions. While the bulk phase behavior of MIPS is becoming increasingly well-understood, the interfaces separating these active phases remains a subject of significant debate. Surprising observations, such as the measurement of negative surface tensions and the emergence of unique ”bubbly” phase separation in two dimensions, challenge our traditional understanding of interfaces. This thesis is devoted to resolving these outstanding questions surrounding active interfaces. We first develop a generalized capillary-wave theory for active systems by deriving an equation of motion for active interfaces and extracting out a non-equilibrium... https://escholarship.org/uc/item/8dn7b1q3 Thu, 3 Sep 2026 00:00:00 +0000 Langford, Luke Anthony https://escholarship.org/uc/item/7sg5v89j Quantum simulation using ultracold fermions in optical lattices provides valuable insights for strongly correlated many-body phenomena in regimes challenging for classical simulations. In this dissertation, we report the design and construction of an apparatus capable of producing quantum degenerate Bose and Fermi gases. The atoms are then loaded into hexagonal optical superlattices created using two commensurate wavelengths. The apparatus lays the foundation for studying interacting fermions in geometrically frustrated lattices.We also report our experimental results on the characterization of the quantum geometry of Bloch band structures. Using rubidium Bose-Einstein condensates in the honeycomb lattice, we were able to measure the quantum distance around band touching points using parallel transport. Additionally, by simulating optical excitations using noninteracting fermions in a periodically modulated lattice, we observe quasimomentum-dependent optical selection rules, which... https://escholarship.org/uc/item/7sg5v89j Thu, 3 Sep 2026 00:00:00 +0000 Chang, Shao-Wen https://escholarship.org/uc/item/7mm1907b This dissertation studies methodological challenges in modern causal inference with a particular focus on the trustworthiness and robustness of estimation methods. While recent advances in machine learning and statistical methodology have vastly expanded the scope and toolkit of causal analysis, many questions remain unresolved. In practice, researchers often make causal conclusions despite unobserved counterfactuals, possible unmeasured confounding, as well as covariate imbalance in observational studies. First, unlike supervised prediction problems, since causal estimands such as treatment effects are never directly observed, model choice and uncertainty quantification become substantially more difficult in the absence of ground truth against which competing methods can be compared. Second, the assumption of no unobserved confounding is implausible in practice and existing sensitivity analysis methods seek to quantify how violations of this assumption could affect causal conclusions.... https://escholarship.org/uc/item/7mm1907b Thu, 3 Sep 2026 00:00:00 +0000 Huang, Yaxuan https://escholarship.org/uc/item/71s6q75j In robotic manipulation, action chunking, a technique that predicts and executes a sequence of actions rather than one at a time, is a powerful tool for capturing noisy and non-Markovian behavior in human data or prior experience. While action chunking policies have seen lots of successes in imitation learning, they are typically trained with supervised learning on human demonstration data that are often costly to collect. Reinforcement learning (RL) offers a promising alternative by enabling robots to autonomously collect data and continuously self-improve from a well-specified reward function, but existing RL methods rely on simple policy classes that often struggle to capture the multi-modality in prior data. In this dissertation, we discuss algorithmic and theoretical foundations for scalable RL with action chunking policies. We start by presenting practical algorithms for optimizing flow-matching policies in Q-learning. Then, we establish a theoretical framework where we... https://escholarship.org/uc/item/71s6q75j Thu, 3 Sep 2026 00:00:00 +0000 Li, Qiyang https://escholarship.org/uc/item/6t79d0n4 The works investigate reliability and performance enhancement for ultra-scaled advanced CMOS and memory devices through experimental characterization, physical modeling, and device simulation. The study first examines random telegraph noise (RTN) in advanced transistors under cryogenic operation, revealing the strong dependence of carrier trapping dynamics on temperature and bias conditions. A physics-based RTN modeling framework, combined with experimental results, is developed to predict RTN behavior in nanoscale transistors. In addition, the impact of ultra-thin aluminum incorporation within high-k metal gate stacks (HfO₂/TiN) is investigated. Electrical characterization and material analyses reveal that aluminum modifies the effective work function through oxygen scavenging from the TiON layer. The reliability trade-offs associated with Al incorporation in the metal gate are further clarified. Next, to mitigate the side effects of Al-incorporated gates, an oxygen insertion... https://escholarship.org/uc/item/6t79d0n4 Thu, 3 Sep 2026 00:00:00 +0000 Lee, Dasom https://escholarship.org/uc/item/6pg4r0h9 The search for next-generation spintronic devices with ultrafast operation, low power consumption, and fully electrical functionality is a key frontier in condensed-matter physics and materials science, driven by the demand for energy-efficient and high-speed information technologies. Devices of antiferromagnetic (AFM) quantum materials are a promising platform owing to the fact that AFM do not produce stray magnetic fields making them viable to be more tightly packed, they possess ultrafast dynamics, and are robust to external magnetic perturbations. Intercalated transition-metal dichalcogenides (I-TMDs) provide a tunable platform for exploring electrically driven switching in low-dimensional materials. By inserting magnetic transition-metal ions into the van der Waals gaps of layered dichalcogenides, intercalation stabilizes long-range magnetic order and complex spin textures that are highly sensitive to composition and symmetry. Previous studies in FexNbS2, showed that electrical... https://escholarship.org/uc/item/6pg4r0h9 Thu, 3 Sep 2026 00:00:00 +0000 Rodriguez, Josue https://escholarship.org/uc/item/6842j5qr Modern vision models face a fundamental scaling challenge: real-world images and videos contain far more visual information than current models can process since they treat every pixel equally and encode all of them uniformly. High-resolution images contain small text, distant objects, and fine local details, while long videos multiply this burden across time. This thesis studies efficient methods and designs to scale vision models to extremely high spatiotemporal fidelity, inspired by the selective nature of human vision where high-fidelity processing is allocated unevenly across space, scale, and time.The first part of the thesis studies whether scaling vision models to higher fidelity is necessary. I show that Scaling on scales (S2 ), a naive approach of scaling up visual resolution by running frozen vision models on tiles of larger images, can already match or exceed the benefit of scaling model size across classification, dense prediction, multimodal language-model benchmarks,... https://escholarship.org/uc/item/6842j5qr Thu, 3 Sep 2026 00:00:00 +0000 Shi, Baifeng https://escholarship.org/uc/item/6273394p We partially extend the results of [8] by constructing Hasse invariants on good reductions of proper abelian type Shimura varieties, and using them to associate Galois representations to systems of Hecke eigenvalues. This is made possible by (1) using the Grifths bundle, as described by Goldring in [7], as a generalization of the Hodge line bundle, and (2) using the abelian type Ekedahl-Oort stratifcation constructed by Shen and Zhang in [26] and independently by Imai, Kato, and Youcis in [10]. Our main contribution is utilizing these constructions together under the main ideas of [8]. https://escholarship.org/uc/item/6273394p Thu, 3 Sep 2026 00:00:00 +0000 Gonzales, Sean https://escholarship.org/uc/item/5rb610hj Peridynamics is a nonlocal reformulation of continuum mechanics aiming at modeling fracture and damage in solids. Unlike the finite element method which relies on a predefined mesh, peridynamics is established upon interactions among discrete material points within a given domain. There are two types of peridynamic formulations: bond-based and state-based. Bond-based peridynamics follows a ``bottom-up'' philosophy, where the bond-level mechanical interactions are determined first and their aggregated results emerge as the material behavior in macro-scale. In contrast, state-based peridynamics adopts a ``top-down'' approach where the macroscopic material response is first defined and subsequently distributed to bond-level interactions. This dissertation advances both formulations through new theoretical developments and practical engineering applications. The foundation of this work is the establishment of a unified bond-associated nonlocal deformation gradient. Based on this,... https://escholarship.org/uc/item/5rb610hj Thu, 3 Sep 2026 00:00:00 +0000 Hu, Xuan https://escholarship.org/uc/item/4wx445qh Modern vision and language models are powerful because they are not trained from scratch for every task. They are pre-trained on large, diverse datasets, adapted through post-training, and increasingly deployed as agents that use tools, feedback, and test-time computation. Robots, however, do not yet benefit from the same scaling paradigm. Robot data is expensive to collect, tied to specific hardware and environments, and difficult to obtain at the scale required for general pre-training. As a result, many robot learning systems remain data-hungry, task-specific, and brittle under changes in objects, scenes, embodiments, and task instructions. This dissertation explores how principles from foundation model training and deployment can be adapted to robot learning, with the goal of building robot systems that are more efficient, scalable, and generalizable.In this dissertation, I study how to make robot learning both efficient and scalable by approaching the problem along three... https://escholarship.org/uc/item/4wx445qh Thu, 3 Sep 2026 00:00:00 +0000 Fu, Letian https://escholarship.org/uc/item/4dh5g860 Rotating shear flows are fundamental to the dynamics of numerous natural and engineered systems, from protoplanetary accretion disks to aircraft wake vortices. This dissertation investigates the multi-faceted dynamics of hydrodynamic stability in these systems, bridging the fundamental theory of wave-wave resonances in incompressible environments with the global wave-mean flow interactions inherent to rotating, stratified fluids. Part I investigates the weakly nonlinear stability of incompressible columnar vortices, demonstrating that the triadic resonance of wave modes is governed by a set of hydrodynamic selection rules. Employing a multi-scale perturbation analysis, we prove that resonant interactions between smooth neutral modes are strictly conservative and confined to the Manley—Rowe relations. Using wave pseudoenergy within a large-k WKBJ framework, we show that the selection rules topologically prohibit intrinsic instability. Consequently, the breakdown of a columnar... https://escholarship.org/uc/item/4dh5g860 Thu, 3 Sep 2026 00:00:00 +0000 Wang, Jinge https://escholarship.org/uc/item/2z57p08m Modern foundation models have made rapid progress in language, vision, and reasoning. However, many real-world tasks remain difficult because they require more than mapping a single input to a single output: they require decomposing ambiguous goals, coordinating multiple specialized systems, grounding intermediate representations in the visual world, and allocating computation adaptively across subtasks. This thesis studies how task decomposition with multi-agent systems can make generative and reasoning models more controllable, scalable, and effective. I approach this question through two complementary settings. First, I show how large language models can serve as planners and controllers for visual generation, translating complex text prompts into structured spatial and temporal representations, using feedback to correct generation errors, and enabling finer control over images and videos. This visual part also includes Describe Anything, which produces detailed localized descriptions... https://escholarship.org/uc/item/2z57p08m Thu, 3 Sep 2026 00:00:00 +0000 Lian, Long https://escholarship.org/uc/item/2gx2w1vr In this thesis I use limb-reduced reptiles as a focal group to explore several largely ignored areas of inquiry in paleontology, herpetology, comparative anatomy, and macroecology. The convergence on limb-reduced or limbless body plans adapted for a burrowing lifestyle is a major theme in the evolution of squamate reptiles, but its implications for the anatomy, evolution, biogeography, and long-term success of clades are far from fully understood. Here, I develop novel techniques for understanding fossilization potential of extant species (Chapter 1), and understanding the evolution of surface patterns (Chapter 3), and apply these to limbless squamates to understand what information about their evolution might actually be available via the fossil record, and how scalation patterns may reflect their evolution and ecology. I address a major shortfall in the current understanding of the diversity of limbless squamates by constructing the first densely-sampled phylogeny of dibamid... https://escholarship.org/uc/item/2gx2w1vr Thu, 3 Sep 2026 00:00:00 +0000 Krone, Isaac Weston https://escholarship.org/uc/item/2256c9qb Transition metal hydrides are being researched as moderators for high temperature micro reactor sutilizing HALEU fuels. Their high densities of hydrogen allow for high moderating efficiency per unit volume compared to other high temperature moderators such as graphite. Zirconium hydride is of particular interest due to its low thermal neutron absorption cross-section and its well-established supply chain. However, hydrogen retention is a key concern for the operation of these reactors. In particular, the effects of radiation on hydrogen retention and phase stability are not well understood. While understanding impact of radiation on the mechanical properties is also of high importance. Phase pure δ-ZrHx and ε-ZrHx were fabricated using Zircaloy-4 as a precursor alloy. These samples were thoroughly characterized using a variety of X-Ray Diffraction (XRD) and electron microscopy techniques. In-situ TEM ion irradiations were conducted using 1MeV Kr ions, while ex-situ ion irradiations... https://escholarship.org/uc/item/2256c9qb Thu, 3 Sep 2026 00:00:00 +0000 Parkison, Darren https://escholarship.org/uc/item/0x8697p0 This dissertation focuses on developing systematic frameworks to understand and improve the reasoning capabilities of large language models (LLMs) in two axes: reliability and efficiency. In terms of reliability, we study how LLMs learn parametric knowledge during training, and how they combine separate atomic knowledge to deduce new conclusions during test time. We develop theoretical frameworks and use out-of-context reasoning as a concrete lens to analyze model behavior, providing both theoretical explanations and empirical evidence that LLMs can hallucinate or fail to generalize systematically. In terms of efficiency, we develop novel paradigms for test-time scaling to improve LLMs' capabilities and reduce inference cost. We focus on latent space reasoning, particularly chain-of-continuous-thought, demonstrating its theoretical advantages where continuous thoughts can maintain a superposition of multiple solutions and thus enable implicit parallel thinking.... https://escholarship.org/uc/item/0x8697p0 Thu, 3 Sep 2026 00:00:00 +0000 Zhu, Hanlin https://escholarship.org/uc/item/0g2863pg Of the three domains of life, bacteria are by far the most ancient and most diverse.Bacteria thus have evolved creative and varied solutions to different biochemical, structural, and cellular problems. One representation of this diversity is the catalog of different intracellular compartments that bacterial species produce to store excess nutrients, make biochemical reactions more efficient, and/or coordinate cellular processes. The number of bacterial organelles identified has greatly expanded in the last few decades, owing to the development of high-resolution imaging technology. However, many questions regarding the cell biology of these organelles still remain unanswered. How does the bacterial cell regulate the production and function of their organelles? What mechanisms are in place to control the spatial distribution of organelles, and are there any similarities to those in eukaryotic systems? In species that produce multiple types of compartments, is there any coordination... https://escholarship.org/uc/item/0g2863pg Thu, 3 Sep 2026 00:00:00 +0000 Ra, Yein https://escholarship.org/uc/item/9z7588pg Noncoding sequences coordinate gene regulation through transcription factor binding and haplotype-specific expression, but translating regulatory sequence into quantitative predictions at the level of individual molecules and individual people remains challenging. This dissertation addresses two aspects of that problem using genomic deep learning, with a shared focus on how data representation shapes model behavior. The first project asks whether representing DNA as both nucleotide sequence and local structure changes what a transcription-factor binding model can learn and explain. In Chapter 2, I develop DeepShape, a convolutional neural network that augments one-hot DNA sequence with five DNA structural attributes, including minor groove width, propeller twist, helical twist, roll, and electrostatic potential, to predict transcription factor (TF) binding across 919 regulatory targets in 148 cell types. Shape features contribute prediction accuracy independently of sequence... https://escholarship.org/uc/item/9z7588pg Wed, 2 Sep 2026 00:00:00 +0000 Keivanfar, Ryan Luca https://escholarship.org/uc/item/9z64r3cm Since 2020, researchers have documented a concerted sociopolitical effort to eliminate diversity, equity, and inclusion (DEI) programs in higher education. These attacks have led to state legislatures dismantling DEI programs and infrastructure at several universities. In this case study of an intraorganizational DEI-centered Community of Practice (DCoP) at a private university in California, I investigated how the participation of eight of its senior higher education leaders informed and influenced their perspectives and practices in ways that advanced equity-focused organizational change efforts. I drew data from semistructured interviews of the eight participants along with document analysis of texts they generated or that were on the university websites. Expanding on current higher education literature on DEI, communities of practice (CoPs), and organizational change management (OCM), I found that senior higher education leaders use DCoPs for professional information sharing,... https://escholarship.org/uc/item/9z64r3cm Wed, 2 Sep 2026 00:00:00 +0000 Soriano-Bilal, Mohammed https://escholarship.org/uc/item/9wj626jb Modern diffusion models are famous for their sample quality. However, viewing these models only as samplers risks overlooking their rich internal representations. Much of the progress in computer vision was facilitated by transferring features from image classifiers, yet the features of diffusion models, which have been scaled to far larger datasets, remain unexamined. If their representations were better understood, these generative models could be repurposed for new capabilities, such as discriminative tasks or controllable generation. In this thesis, we show that diffusion models indeed contain rich representations that can be reused in new tasks and domains. We start in Chapter 2 by designing a general-purpose feature extractor for diffusion models, enabling an image generator to achieve state-of-the-art performance on the discriminative task of semantic correspondence. Then, in Chapter 3, we use this same extractor for controllable generation, allowing diffusion models... https://escholarship.org/uc/item/9wj626jb Wed, 2 Sep 2026 00:00:00 +0000 Luo, Grace https://escholarship.org/uc/item/9w06684s Effective security depends on understanding behavior: how systems communicate, how users interact, and how malicious actors deviate from expected patterns. That understanding rests on two pillars: observability, which surfaces useful signals, and detection, which uses those signals to identify activity. Yet translating these pillars into practical defenses poses significant hurdles, from extracting meaningful structure from raw activity to building detectors that remain reliable amid noise, change, and adversarial pressure. In this dissertation, we develop AI-assisted methods that make security analysis more practical, robust, and deployable across modern internet systems. For observability, we introduce Matryoshka, a system that automatically generates deterministic, semantically-aware parsers for heterogeneous security logs, and GGFAST, a framework that automatically extracts discriminative structure from network traffic to build fast, interpretable classifiers, including in... https://escholarship.org/uc/item/9w06684s Wed, 2 Sep 2026 00:00:00 +0000 Piet, Julien Thomas https://escholarship.org/uc/item/9sj5h376 Building a machine learning model is an iterative process: train, evaluate, identify failures, and improve. Benchmarks once anchored this loop, but as models have improved and deployment has expanded, evaluation has become the bottleneck. Metrics fail to capture what practitioners care about, failures are hard to diagnose without manual inspection, and preference signals from users are noisy and hard to decompose. Model development is now constrained not by capability, but by our ability to measure and understand model behavior.To understand your model, you must understand your data. By systematically analyzing the data that flows through models (their outputs, the real-world conversations users have with them, and the training examples that shaped them) we can characterize what people actually use models for, identify where and why models fail, and understand what drives their behavior.We develop a suite of methods that operationalize this view across the model development lifecycle.... https://escholarship.org/uc/item/9sj5h376 Wed, 2 Sep 2026 00:00:00 +0000 Dunlap, Lisa https://escholarship.org/uc/item/9sg796rf A graph G = (V (G), E(G)) is an ordered pair consisting of a finite vertex set V (G) and an edge set E(G) of unordered pairs of distinct vertices. Given a graph G, a graph H is an induced subgraph of G if V (H) ⊆ V (G) and E(H) is exactly the set of edges among V (H) inherited from G. A class A of graphs is hereditary if A contains all induced subgraphs of graphs in A. Hereditary classes are particularly interesting because each can be characterized by a set of forbidden induced subgraphs. Where a class A is not hereditary, we define the hereditary subclass of A to be the largest hereditary class contained within A, and the hereditary closure to be the smallest hereditary class containing A. In this dissertation, I study the hereditary subclasses and closures of important non-hereditary graph classes, defined by relationships among invariants.Given a graph, we can color its vertices such that no two adjacent vertices are the same color. If such a coloring exists with k colors,... https://escholarship.org/uc/item/9sg796rf Wed, 2 Sep 2026 00:00:00 +0000 Whitman, Rebecca https://escholarship.org/uc/item/9rk0g3rq Adeno-associated virus (AAV) mediated gene therapies for inherited retinal degenerations require precise targeting of the outer retina, particularly in the retinal pigment epithelium (RPE), which plays an essential role in maintenance of the light-sensitive neuron, photoreceptors. Mutations in RPE-expressed genes, including RPE65, lead to blinding diseases such as Leber congenital amaurosis type 2 (LCA2). While subretinal injection achieves optimal RPE transduction, it is an invasive, inpatient surgical procedure associated with iatrogenic damage, including retinal detachment. Intravitreal (IVT) surgery is a far less invasive, outpatient procedure but has previously been shown to have limited penetration of the inner limiting membrane (ILM) of the retina and thus poor outer retina transduction. The goal of this study is to identify capsid variants capable of overcoming this barrier. Here, high-throughput screenings of libraries of over 1.28x109 capsids from different AAV serotypes... https://escholarship.org/uc/item/9rk0g3rq Wed, 2 Sep 2026 00:00:00 +0000 Singh, Sonali https://escholarship.org/uc/item/9qr32359 This dissertation studies how behavioral forces shape household financial decision-making. Across three chapters, we combine administrative credit-panel data, structural modeling, and a controlled laboratory experiment to show that departures from the frictionless, fully-informed benchmark of standard models are not peripheral: they govern how households respond to debt relief, how they form beliefs about asset returns, and how they invest in financial knowledge. Taken together, the chapters argue that the behavioral channel is central to understanding, and designing policy around, household finance. The first chapter examines the long-run consequences of temporary debt relief, using the federal student loan repayment pause from 2020 to 2023 as a natural experiment. Exploiting variation in loan eligibility through an instrumental variables strategy applied to the University of California Consumer Credit Panel (UC-CCP), we show that the pause initially reduced delinquency and raised... https://escholarship.org/uc/item/9qr32359 Wed, 2 Sep 2026 00:00:00 +0000 Lyu, Junru https://escholarship.org/uc/item/9q51g4n7 This dissertation is concerned with a variety of problems in quantum computing and the spectral theory of quantum-mechanical channels and Hamiltonians. We begin with the problem of quantum gate synthesis. The challenge is to find a fixed, finite set S of quantum gates such that, given a quantum circuit component U acting on n qubits and target precision ε, there is a short, efficiently computable composition of gates from S that approximates U to precision ε. This problem has already led to a rich literature in the case n=1 with deep connections to modern number theory. In this dissertation we present the state-of-the-art algorithm for a single qubit, achieving length 7 log (1/ε). The techniques used include the lattice structure of certain dense subgroups of PU(2) and Lenstra's algorithm for integer programming in convex domains. The next topic considered is scientific computing algorithms intended to run on a quantum computer. Specifically, we study a framework for... https://escholarship.org/uc/item/9q51g4n7 Wed, 2 Sep 2026 00:00:00 +0000 Stier, Zachary https://escholarship.org/uc/item/9pd2k2vc Large language models (LLMs) have become remarkably proficient at generating coherent, high-quality responses. However, they often struggle to produce diverse outputs, especially when multiple, equally plausible answers exist. This limitation becomes critical in areas like adversarial testing, search, and synthetic data generation, where generating distinct yet valid responses is essential. This thesis opens with a discussion of the timely need for addressing the diversity of language models. I’ll then introduce SimpleStrat, a lightweight method that leverages the model itself to automatically stratify the solution space. By performing stratified sampling, we can improve diversity and coverage without sacrificing quality. I’ll also describe how we measure resampling diversity using CoverageQA, and discuss how these ideas extend to creative and open-ended tasks. To extend these ideas into the weight space, I’ll discuss Stylus, which navigates the quality-diversity tradeoff by leveraging... https://escholarship.org/uc/item/9pd2k2vc Wed, 2 Sep 2026 00:00:00 +0000 Wong, Justin https://escholarship.org/uc/item/9nm0t5t6 Electric field noise produced by the surface of ion trap electrodes reduces the fidelity of quantum computing operations. Despite decades of investigation its microscopic origins remain unclear.This thesis investigates the origins of surface electric field noise by examining its dependence on trap temperature. A set of Al-Cu surface Paul traps are fabricated together and sent to MIT Lincoln Labs (MITLL) and the University of California, Berkeley (UCB). MITLL measures noise from 7 K to room temperature above a first trap and UCB measures noise from room temperature to 590 K above a second trap. The noise measurements from the two experiments are combined to study the noise dependence over the full temperature range from 7-590 K. This is the largest temperature range ever studied using a single trap design and connects the temperature ranges used in all previous studies.We find qualitatively different behaviors at low and high temperatures: from 7-430 K we find that the noise follows... https://escholarship.org/uc/item/9nm0t5t6 Wed, 2 Sep 2026 00:00:00 +0000 Saarel, Benjamin Vickers https://escholarship.org/uc/item/9mr3g8r4 Molten actinide containing halide salts are used to fuel molten salt nuclear reactors. Before these reactors can be deployed, it is necessary to characterize the thermophysical properties of these fuel salts for fuel qualification, licensing, and technological applications on the front end (supply chain), reactor operations (maintenance), and back end (end of life) of their fuel cycles. To connect experimental property measurements to simulation and thermodynamic predictions, it is also necessary to understand the mechanisms behind this thermophysical property behavior. In this dissertation, internal structure is used to understand macroscopic properties to better understand how occurrences in a fuel system may affect reactor behavior.Activation energy of viscosity in halide molten salts is shown to scale with increasing concentration of strong complex-forming cation species. This amount by which this parameter scales is dependent on the degree of polymerization and coordination... https://escholarship.org/uc/item/9mr3g8r4 Wed, 2 Sep 2026 00:00:00 +0000 Gardner, D. Nathanael https://escholarship.org/uc/item/9mj9q6z4 This dissertation investigates the historical and philosophical formation of the Dzogchen Ground/ground (gzhi) in the Heart Essence (snying thig) tradition of Tibetan Buddhism from the ninth through the fourteenth centuries. The Dzogchen Ground/ground is inspired by Yogācāra and tantric ideas of ālayavijñāna, Buddha-nature, and the universal ground of awakened mind, while synthesizing Tibetan doctrinal innovations. Through constant reinterpretation, the Ground/ground comes to signify a distinctive nondual reality–a basis for both delusion and liberation, as well as the true nature of reality.The study begins with an examination of Dunhuang Yogācāra manuscripts, notably Pelliot tibétain 654 and IOL Tib J 708, which preserve early Tibetan scholastic engagements with the theory of ālayavijñāna and demonstrate the integration of universal ground consciousness discourse into Tibet. It then analyzes the writings of Nupchen Sangye Yeshe, whose works reinterpret the universal ground within... https://escholarship.org/uc/item/9mj9q6z4 Wed, 2 Sep 2026 00:00:00 +0000 Yeshi, Khenpo https://escholarship.org/uc/item/9kc1z6ph The acoustic signal is the carrier of not only linguistic (as in, what the interlocutor intends to communicate) but also socio-cultural meaning (as in, who a speaker-listener is). The field of sociophonetics hinges on the inextricability of these two facets of speech sounds. It has been established that speakers can both intentionally and unintentionally communicate aspects of social identity, including gender and sexuality, in the subtle distinctions present in their production of sounds in a given language system. In turn, listeners may draw upon the biases engendered by previous encounters with speech to make assumptions about the speaker that contextualize how linguistic information is received and understood. However, particularly in the realm of sexual identity, with its close relationship to gender, findings are not yet robust, in either production or perception, in terms of how particular phonetic variables pattern with identity types. The most supported finding has been... https://escholarship.org/uc/item/9kc1z6ph Wed, 2 Sep 2026 00:00:00 +0000 Galvano, Amber C. https://escholarship.org/uc/item/9hz894kv Alternative schools serve students who are underserved, excluded, or marginalized within traditional school settings. In California, close to 169,000 students attended alternative schools during the 2024–25 school year, with nearly 85% identifying as students of color. Despite a growing body of literature on effective alternative school design, insufficient attention has been paid to the leadership practices, challenges, and professional development needs of the principals who lead these schools. This qualitative single-case study examined the leadership practices of 16 current, former, or retired alternative school and systems leaders in the San Francisco/Oakland Bay Area, selected through purposeful sampling. Data were collected through semi-structured interviews, a demographic questionnaire, and the administration of Individual and Collective Hope Scales. Four primary findings emerged. First, alternative school leaders engage in intentional relationship-building with students,... https://escholarship.org/uc/item/9hz894kv Wed, 2 Sep 2026 00:00:00 +0000 Muñoz Daniels, Irma https://escholarship.org/uc/item/9hv1q24q This Dissertation explores the tensions between free speech and regulation in an era of increased marketing and advertising, as well as the proliferation of digital services, including online marketplaces, social media platforms, and search engines, which facilitate the spread of both information and disinformation.Intellectual property rights and the First Amendment frequently exist in tension. Although the First Amendment prohibits the suppression of speech, the enforcement of trademark rights may restrict someone else’s speech. This conflict is also evident in the evaluation and management of user-generated content on social platforms. Because these regulatory regimes can prevent individuals from using particular language, they raise substantial free speech concerns. While scholars have acknowledged the tension among intellectual property, content moderation, and free expression, several critical areas remain insufficiently explored. Likewise, in recent years, there have been... https://escholarship.org/uc/item/9hv1q24q Wed, 2 Sep 2026 00:00:00 +0000 Laketic, Jelena https://escholarship.org/uc/item/9ht9k3cg Software is one of the most leveraged forms of human labor: it underlies nearly every modern system, and writing, modifying, and maintaining it occupies a substantial fraction of skilled work worldwide. AI systems have begun to perform meaningful parts of this work, and the question of how capable they really are is no longer hypothetical. Software is also uniquely well-suited as a substrate for AI agents: actions are cheap to execute, consequences are immediate, and the surrounding tooling (compilers, test runners, type checkers, profilers) produces precise feedback. This feedback loop rests on two abstractions: environments, which let agents execute code in realistic settings, and verifiers, which assess the correctness and quality of their outputs.This thesis finds that environments and verifiers are themselves a primary axis of capability for AI in software engineering, and that scaling them reveals what AI can and cannot do. First, executable environments can be constructed... https://escholarship.org/uc/item/9ht9k3cg Wed, 2 Sep 2026 00:00:00 +0000 Molahalli, Manish Shetty https://escholarship.org/uc/item/9h9087dm Alcohol, specifically ethanol, is estimated to have been consumed by 46% of the world’s population above the age of 15 in 2019. Human societies have been intentionally brewing alcohol for consumption for at least 9,000 years, likely longer. Why are humans so attracted to ethanol consumption? Is it merely the result of a happy biochemical accident? Or could there be an evolutionary explanation? The "drunken monkey" hypothesis points out that the ancestors of humans were frugivorous for tens of millions of years and predicts that the fruits they ate underwent microbial fermentation resulting in the accumulation of ethanol. Olfaction of fruit ethanol by foraging animals may facilitate the localization of fruit crops and indicate the suitability of individual fruits for consumption. The ingestion of fruit ethanol may itself provide beneficial services such as appetite stimulation. Physiological tolerance of increasingly ethanolic fruits by primates and other animals may therefore... https://escholarship.org/uc/item/9h9087dm Wed, 2 Sep 2026 00:00:00 +0000 Maro, Aleksey https://escholarship.org/uc/item/9dc8j5p9 Since the desegregation of schools, Black children have been punished for not assimilating to White supremacist, heteropatriarchy in their traditional U.S. school context. For those Black children who grow up to become teachers, there is a synergy between their teaching experiences and (mis)treatment of Black children. For decades, Black women have mobilized education as a means for social mobility and are grappling with a double bind in their misrepresentation. This qualitative study utilizes semi-structured interview data to explore the beliefs and standpoints of Black women teachers (BWT) as they seek to dignify BIPOC children. Through a frame of anti-Blackness and Black feminist liberation, many Black women conceptualize a loving gaze and politicized praxis as they prioritize racial healing. Some BWT embody radicalism and collective love in their pedagogy, which creates spatial havens for youth empowerment amongst constant threats of pathologization and exclusion. Not only... https://escholarship.org/uc/item/9dc8j5p9 Wed, 2 Sep 2026 00:00:00 +0000 Jones, Makaela https://escholarship.org/uc/item/99t1d2x3 This dissertation tests the premise that, for beginning readers, speech is the ‘object of segmentation’ (Schreuder and van Bon, 1989). I posit that one aspect of speech in particular— coarticulation as proxied by gestural overlap—is likely to influence segmentation performance: reduced coarticulation should enable segmentation, while greater overlap should make it more difficult. Using empirical evidence of gestural overlap from articulatory Phonetics, I make predictions about the segmentability of English and Spanish stimuli, both between and within languages. Participants were 472 TK – Grade 2 children from various educational contexts. Students were asked to segment a spoken word. Their responses received two scores: first sound isolation (0/1) and unitization (0/1), the latter denoting instances where the first two phonemes of a stimulus were identified as a single unit (e.g., /fl/ for flee). For each outcome, 12,388 responses were analyzed within a Generalized Linear... https://escholarship.org/uc/item/99t1d2x3 Wed, 2 Sep 2026 00:00:00 +0000 Inciarte, Himilcon https://escholarship.org/uc/item/99r1g2sm Catabolism is an operation on tableaux originally defined by Lascoux. Though the operation is not difficult to compute, many of its properties remain elusive. The combinatorics of catabolism has deep ties to various generalizations and subspaces of the Type A coinvariant ring and their structures as graded symmetric group representations. This in turn gives connections to the corresponding q-symmetric functions that give the graded Frobenius characters of these spaces. In this dissertation, we explore these connections from both the combinatorial and algebraic perspective. We use catabolizability to construct bases for different families of generalized coinvariant rings. Our constructions give bridges between the combinatorics of catabolism and the representation theoretic structures: in particular, we recover combinatorial formulas for the Schur expansions of their graded Frobenius characters. We also study the combinatorics of catabolism directly in an attempt to identify submodules... https://escholarship.org/uc/item/99r1g2sm Wed, 2 Sep 2026 00:00:00 +0000 Hanada, Mitsuki https://escholarship.org/uc/item/9874x5vt Within economic theory, the study of matching generally focuses on the problem of allocating heterogeneous, indivisible resources to agents, often without the use of prices. In this dissertation I focus on one-sided matching problems, in which a set of agents have preferences over a set of objects which are to be potentially allocated.In Chapter 1, coauthored with Andrew Tai, we study the classic house swapping problem of Shapley and Scarf (1974) but relax the usual assumption that agents have strict preferences over the objects. Top trading cycles with fixed tie-breaking (TTC) has been suggested to deal with indifferences in these kinds of object allocation problems. Unfortunately, under general indifferences, TTC is neither Pareto efficient nor group strategy proof. Furthermore, it may not select an allocation in the core of the market, even when the core is nonempty. However, when indifferences are agreed upon by all agents (“objective indifferences”), TTC maintains Pareto... https://escholarship.org/uc/item/9874x5vt Wed, 2 Sep 2026 00:00:00 +0000 Sandholtz, Will https://escholarship.org/uc/item/97t8b765 One of the largest obstacles in modern drug discovery is that a significant portion (>90%) of the proteome is considered “undruggable,” in that these proteins lack a characterized, functional binding pocket or ligandable hotspot which small molecules can bind to and modulate the protein’s function for therapeutic benefit. To overcome such disease-causing proteins, targeted protein degradation (TPD) strategies have arisen, where the cell’s endogenous degradation machinery is hijacked to ubiquitinate and degrade the classically undruggable protein. Molecular glue degraders serve as a promising modality to achieve TPD. These are monovalent compounds that induce the proximity of a target protein with a component of the ubiquitin proteasome system to degrade the protein of interest. While our research group has previously identified a fumarate-based electrophilic handle that covalently modifies the E3 ligase RNF126 to enable degradation of multiple protein targets, the high intrinsic... https://escholarship.org/uc/item/97t8b765 Wed, 2 Sep 2026 00:00:00 +0000 Modi, Aman Devare https://escholarship.org/uc/item/979490sd Machine learning models for vision applications are inevitably deploying in environments that differ from the condition under which the models were trained and evaluated. Throughout this dissertation we investigate various ways to account for these data shifts and produce models whose performance is improved and better understood in complex settings. We first present a study on automatically evaluating model performance under various kinds of unseen distribution shifts. Establishing synthetic and natural distribution shifts as the distinct and informative types of shift worthy of study we evaluate the predictive ability of prior methods in the synthetic to synthetic shift setting, the synthetic to natural shift setting, and natural to synthetic shift settings. While a variety of prior works perform well in synthetic to synthetic shift settings, none of them outperform our naive average confidence baseline on the more meaningful task of predicting natural distribution shifts. Our... https://escholarship.org/uc/item/979490sd Wed, 2 Sep 2026 00:00:00 +0000 Guillory, Devin L https://escholarship.org/uc/item/96c4x74d Designing complex engineering systems often requires solving an optimization problem: given a target behavior, find the materials, geometry, or operating policy that achieves it. Depending on the domain, faithful design of the underlying system may involve computational fluid dynamics, finite element analysis, stochasticity, or other complex physics, all of which tend to be computationally expensive and produce objective surfaces that are non-smooth, non-convex, and high-dimensional. This combination calls for two solutions: 1) physics-based reduced-order models that make objective function evaluation cheap and 2) derivative-free optimization, such as with a genetic algorithm (GA). In this dissertation we present a unified approach in which reduced-order models from first-principles physics are paired with genetic algorithms to address optimal design questions across three distinct scales of engineering decisions. At the material scale, we pair the Hashin--Shtrikman bounds... https://escholarship.org/uc/item/96c4x74d Wed, 2 Sep 2026 00:00:00 +0000 Becker, Carla Joy https://escholarship.org/uc/item/94z269m7 TnpB is a compact RNA-guided endonuclease and evolutionary ancestor of CRISPR-Cas12 that offers a promising platform for genome engineering. However, the genome-editing activity of TnpBs remains limited and its underlying determinants are poorly understood. Here, we used biochemical and single-molecule assays to examine the DNA-unwinding mechanism of Youngiibacter multivorans TnpB (Ymu1 TnpB). DNA unwinding proceeds through a discrete, long-lived partially unwound intermediate state before reaching a fully unwound open state. The open state forms inefficiently and collapses readily in the absence of negative supercoiling. An optimized variant, Ymu1-WFR, stabilizes formation of both the intermediate and open states, resulting in enhanced DNA cleavage in vitro and increased genome editing in vivo. These findings identify the physical basis for the observed minimal activities of natural TnpBs, revealing how stabilizing specific unwinding states enables efficient DNA targeting. https://escholarship.org/uc/item/94z269m7 Wed, 2 Sep 2026 00:00:00 +0000 Zhou, Zehan https://escholarship.org/uc/item/94v6180j This dissertation reconstructs the formation of a colonized elite and the anti-Chinese commercial nationalist movement they launched in turn of the century Cochinchina. The movement began in 1901 with the newspaper Talking Business Over Tea [Nông cổ mín đàm] and culminated in Gilbert Trần Chánh Chiếu's Enlightened Renewal Movement [Minh Tân] of 1907–1909. The movement united an elite bloc of Confucian gentry based chiefly in the Orchard Heartlands [Miệt Vườn] region of the Mekong Delta with Saigon bureaucrats in a program of commercial and industrial investment aimed at displacing immigrant Asian merchants from the colonial economy. The movement’s program was accommodationist toward French rule, not anti-colonial. It became the dominant form of southern Vietnamese nationalism and the direct ancestor of the interwar Constitutionalist Party. However, the movement proved unable to overcome the factional divisions inhering in its constituencies, and it unraveled when rival elites... https://escholarship.org/uc/item/94v6180j Wed, 2 Sep 2026 00:00:00 +0000 Morreale, Anthony Peter https://escholarship.org/uc/item/93w4r98n Post-treatment events are common in causal inference in both randomized trials and observational studies. They appear in many scientific problems, including noncompliance, truncation by death, mediation, and surrogate endpoint evaluation, and may reveal important information about treatment response, alter the interpretation of the outcome, or make the outcome unobservable. This dissertation studies three related problems involving post-treatment events, with the common goal of defining scientifically meaningful causal estimands and developing rigorous identification, semiparametric efficiency theory, and estimation and inference procedures.The first chapter studies principal stratification with continuous post-treatment variables. Principal stratification is a strategy to address these challenges by adjusting for the potential values of the post-treatment variables, defined as the principal strata. It allows for characterizing treatment effect heterogeneity across principal strata... https://escholarship.org/uc/item/93w4r98n Wed, 2 Sep 2026 00:00:00 +0000 Lu, Sizhu https://escholarship.org/uc/item/93q978gb Modern machine learning and scientific computing are increasingly driven by large-scale models, iterative optimization methods, and randomized numerical algorithms whose empirical performance often exceeds the reach of existing mathematical theory. This dissertation develops theoretical frameworks for three such settings: contrastive pre-training for multimodal generative AI, the implicit bias of gradient descent in non-homogeneous deep networks, and stochastic density functional theory for large-scale electronic structure calculations. Across these topics, the goal is to identify mathematical structure that explains when widely used computational procedures succeed and how their performance scales.The first part studies contrastive pre-training for multimodal learning. We introduce approximate sufficient statistics as a way to quantify the quality of representations learned from paired data, and we show that near-minimizers of the contrastive loss are approximately sufficient... https://escholarship.org/uc/item/93q978gb Wed, 2 Sep 2026 00:00:00 +0000 Cai, Yuhang https://escholarship.org/uc/item/91j8z12b The kinetics and mechanisms of first order phase transitions are determined by tradeoffs between surface tension and chemical potential difference. More than a century of work has refined our understanding of such phase transformations, beginning with the classical theory of nucleation and pushing beyond it toward quantitative accuracy. It is often most common to consider the nucleation and growth of a dense phase into an infinite or semi-infinite dilute phase, but the proliferation of nanomaterials demands consideration of phase transformations that are arrested at the nanoscale. On the nanoscale, not only surface effects but also thermal fluctuations and transitions between intermediate metastable states can play a more significant role in the mechanism of a phase transformation. This thesis will computationally and theoretically analyze the kinetics of nucleation and growth in a set of nano-confined systems and demonstrate several novel growth mechanisms that arise from these... https://escholarship.org/uc/item/91j8z12b Wed, 2 Sep 2026 00:00:00 +0000 Oaks-Leaf, Sam https://escholarship.org/uc/item/8tt6n7h0 Neural network controllers have shown remarkable performance in control tasks, but their deployment in safety-critical applications is hindered by a lack of formal robustness guarantees. We propose methods to synthesize and verify neural network controllers robust to both model uncertainty and exogenous disturbances. We consider robustness specifications through the framework of dissipativity, which naturally accommodates exogenous disturbances, and ensure these dissipation inequalities hold for model perturbations characterized by integral quadratic constraints (IQCs). We first propose linear matrix inequality-based methods for verifying robust dissipation inequalities, and a reinforcement learning framework that alternates learning steps with semidefinite programming to train robust neural network controllers that maximize reward. We then propose a counterexample guided training procedure for neural network controllers, paired with a branch-and-bound-based verification approach... https://escholarship.org/uc/item/8tt6n7h0 Wed, 2 Sep 2026 00:00:00 +0000 Junnarkar, Neelay https://escholarship.org/uc/item/8tt6g2fn Externalities, market failures where activities impose costs or benefits that are not reflected in market prices, are a core area of study in economics. This is also an area where economists can influence policy, because market failures are often best corrected via government intervention. Yet effective externality-correcting policy requires information. This dissertation focuses on two types of information that are central to policymaking: understanding who is harmed by an externality and how severely, and knowing what is regulated by both existing and proposed policies.Policy effectiveness breaks down without sufficient information. Externalities that are difficult to measure may be ignored or addressed with suboptimal policies. When the rules themselves are ambiguous, regulated entities expend resources interpreting their obligations, face litigation risk, and may still comply imperfectly. The result is policy that costs more, delivers less, and is harder to improve than it... https://escholarship.org/uc/item/8tt6g2fn Wed, 2 Sep 2026 00:00:00 +0000 Greenhill, Simon David https://escholarship.org/uc/item/8tj0n2zz Ensuring the positive development of children and adolescents holds great importance, as it enables individuals to thrive and achieve their full potential and to lead meaningful lives as members of society in adulthood. School bullying is a prevalent and serious problem that hinders the positive development of children and adolescents, given its adverse impact on a wide range of outcomes not only for victims but also for those involved as bullies or bully-victims. Bullying refers to unwanted aggression rooted in a power imbalance between individuals (excluding siblings or dating partners), often repeated over time and manifested in various forms (e.g., physical, verbal, relational). Disabled youth may be more likely to be involved in bullying than nondisabled youth, as disability—reflecting a misfit or mismatch between individual characteristics and surrounding environments—may contribute to power imbalances between disabled and nondisabled students and thereby increase the risk... https://escholarship.org/uc/item/8tj0n2zz Wed, 2 Sep 2026 00:00:00 +0000 Chung, Eunkyung https://escholarship.org/uc/item/8st7n15s The subject of this thesis is the initial data sets on R 3 for the vacuum Einstein equation, i.e., pairs (g, k) = (gij , kij ) of symmetric tensor fields on R 3 (where g is positive definite) solving the vacuum constraint equationIndeed, any triple (M3 , g, k) of a 3-manifold, Riemannian metric g and a covariant symmetric 2-tensor k satisfying (0.0.1) is called an initial data set for the vacuum Einstein equationVarious applications of initial data construction include global stability of Minkowski spacetime [CK93], weak cosmic censorship and naked singularity formation [Chr94], disproving the third law of black hole thermodynamics [KU22], and formation of black holes ([LY15] as one example), etc.To understand the nature of the nonlinear PDE system (0.0.1), it is instructive to consider the direct linearization of this system around the simplest solution – i.e., the flat solution (δ, 0) – that takes the form, respectively,This system is comprised of underdetermined PDEs for a... https://escholarship.org/uc/item/8st7n15s Wed, 2 Sep 2026 00:00:00 +0000 Mao, Yuchen https://escholarship.org/uc/item/8s12v525 Angucycline natural products are one of the largest classes of polyketide natural products, and by far the largest class of type II PKS derived aromatic polyketides. They display a wide variety of structural complexities, as well as seemingly ubiquitous antibacterial, antifungal, and cytotoxic biological activities. This dissertation reviews the structures of many of the angucyclines which have been isolated to date, with intentional comparison of similar structures which may have been isolated from different sources or sub-families. Next, the biosynthesis of several angucyclines is covered with special attention given to the non-enzymatic sulfur incorporation proposed for many thioangucyclines to provide important context for the retrosynthetic strategy employed toward grisemycin. Additionally, a broad overview of both traditional and recent chemical synthesis of angucycline natural products is discussed. The chemical syntheses are organized by the key synthetic strategies employed... https://escholarship.org/uc/item/8s12v525 Wed, 2 Sep 2026 00:00:00 +0000 Stevenson, Kincade https://escholarship.org/uc/item/8rh736g8 The discovery of new small-molecule therapeutics remains a notoriously time-consuming and capital-intensive endeavor, constrained by the vastness of synthesizable chemical space and the limitations of traditional screening approaches. Generative artificial intelligence has fundamentally reshaped this landscape by enabling the direct sampling of molecules with desired properties, yet persistent challenges in synthesizability, data quality, and translational validation continue to limit the practical impact of computational methods. This dissertation addresses these challenges across three interconnected pillars: Large Language Model (LLM)-based generative molecular design, rigorous structure-based benchmarking, and prospective validation in competitive drug discovery settings.In Chapters 2 and 3, I develop two LLM-driven generative frameworks that address core bottlenecks in molecular design. SynLlama, described in Chapter 2, tackles the synthesizability problem by fine-tuning... https://escholarship.org/uc/item/8rh736g8 Wed, 2 Sep 2026 00:00:00 +0000 Sun, Kunyang https://escholarship.org/uc/item/8p68g9vq Semiconductor nanocrystals exhibit size-, shape-, and composition-dependent electronic and optical properties governed by quantum confinement. Predictive computational modeling of these effects remains challenging because accurate first-principles excited-state methods such as GW/BSE scale too steeply for nanocrystals containing hundreds to thousands of atoms, while traditional semi-empirical approaches sometimes struggle to achieve the flexibility and transferability needed across diverse semiconductor materials and alloy compositions. At the same time, despite major advances in machine learning for ground-state materials modeling, its application to excited-state properties, electron-phonon interactions, optical response, and charge carrier dynamics of large nanomaterial systems remains limited.This dissertation addresses this challenge through the development of DeepPseudopot, a machine-learned atomistic semi-empirical pseudopotential framework for nanomaterials. By combining... https://escholarship.org/uc/item/8p68g9vq Wed, 2 Sep 2026 00:00:00 +0000 Lin, Kailai https://escholarship.org/uc/item/8nw7j924 In this thesis, I study the geometry of algebraic stacks and some important cohomological invariants on them, motivated by moduli-theoretic problems. The language of stacks provides a natural framework for understanding moduli problems that cannot be captured by schemes alone. The first main chapter is based on "Picard Groups of Stacky Curves," and investigates the geometry and Picard groups of stacky curves and gerbes. I further develop the theory of rigidification and show that every stacky curve can be written as a gerbe over a stacky curve with trivial generic stabilizer. I calculate the Picard groups of tame stacky curves with trivial generic stabilizers and express the Picard groups of gerbes as an extension of two groups, which depend on the Brauer classes of the gerbes and the Picard groups of the base. These results together give a description of the Picard group of any tame stacky curve as an extension of two groups. I apply the theory to many examples, including... https://escholarship.org/uc/item/8nw7j924 Wed, 2 Sep 2026 00:00:00 +0000 Lopez, Rose Eleanor https://escholarship.org/uc/item/8n50g5x4 This dissertation explores an emerging superfamily of bimetallic oxygenases, designated amidohydrolase-related dinuclear oxygenases (AROs), through the discovery of new enzymes, detailed mechanistic characterization, and comprehensive bioinformatic analysis. Chapter 1 introduces the background and significance of oxygen-activating metalloenzymes. Chapter 2 describes the identification of approximately 3,000 candidate metallooxygenases, thereby revealing and naming the ARO superfamily, and presents the biochemical characterization of several newly discovered members. Chapter 3 focuses on one enzyme, SfbO, and investigates its unprecedented ability to activate dioxygen and catalyze C–H bond hydroxylation using a dimanganese cofactor. Another example, a family of diterpenoid monooxygenases (DitZs), has a strict dependence on diiron cofactor, and its reactivity, evolutionary diversification,and catalytic mechanism are examined in Chapter 4. Building on these findings, a broader bioinformatic... https://escholarship.org/uc/item/8n50g5x4 Wed, 2 Sep 2026 00:00:00 +0000 Liu, Chang https://escholarship.org/uc/item/8m96v5n7 This dissertation consists of three papers in the field of public economics. The first paper is co-authored with Arjan Lejour, Simon Rabaté, and Maarten van ’t Riet. It investigates how the measurement of wealth inequality is affected by offshore tax evasion. Using data from a tax amnesty in the Netherlands, we find that this type of tax evasion is concentrated among the wealthy. Wealth hidden offshore is typically excluded from conventional estimates of inequality, implying that these estimates underestimate the true level of wealth inequality. The concentration of offshore wealth appears to be lower in the Netherlands than in other countries. We explore several explanations for this. One is the availability of low-cost evasion opportunities in neighboring countries. Another is that the Netherlands’ wealthiest households face low effective tax rates, implying that the incentive to evade taxes is low as well.I further explore the topic of effective taxation in the second paper... https://escholarship.org/uc/item/8m96v5n7 Wed, 2 Sep 2026 00:00:00 +0000 Leenders, Wouter https://escholarship.org/uc/item/8ks034dk Large language models (LLMs) are increasingly deployed as autonomous agents—systems that observe, interact with, and pursue goals in real-world environments. Yet agents often pursue goals that diverge from their operators' intentions, a phenomenon known as misalignment. As agents are deployed with greater autonomy in higher-stakes settings, understanding and addressing misalignment becomes critical to ensuring their safety and reliability. This dissertation investigates misalignment in AI agents through three complementary approaches: modeling how it arises, measuring its prevalence, and characterizing its presence in model internals. First, in Section 2, we model misalignment by formalizing the feedback loops inherent to agent deployment. We show that such loops can unexpectedly induce optimization, driving harmful side effects even without explicit training signal, a phenomenon we call in-context reward hacking. Second, in Section 3, we measure misalignment by introducing... https://escholarship.org/uc/item/8ks034dk Wed, 2 Sep 2026 00:00:00 +0000 Pan, Alexander https://escholarship.org/uc/item/8hx6r1jw This dissertation studies accessibility to housing along the housing cycle, understanding the inequitable returns to housing by race, and household expectations of the housing market. The first chapter investigates the mortgage channel of monetary policy transmission to home purchasing behaviors of first-time home buyers and incumbent homeowners. Between 2009 and 2019, the first-time home buyer share of home purchases fell from 35% to 22%, a period in which mortgage rates fell from nearly 7% to 3.5%. First, I construct a new mortgage rate-specific monetary policy shock to use as an IV for mortgage rate changes which predicts future mortgage rates better than existing monetary policy shocks. Next, I provide empirical evidence for three new findings: 1) transacted house prices respond to monetary policy-induced mortgage rate changes within a matter of weeks, indicating a rapid housing demand response to mortgage rates; 2) a negative 25 basis point mortgage rate shock lowers... https://escholarship.org/uc/item/8hx6r1jw Wed, 2 Sep 2026 00:00:00 +0000 Drukker, Leonel Diego https://escholarship.org/uc/item/8hw6n3ck Model predictive control (MPC) and control barrier functions (CBFs) have been widely adopted in autonomous systems research for their ability to rigorously enforce safety constraints. In many controls and planning applications, these constraints depend on the optimality of a lower-level optimization problem. This dissertation uses tools from differentiable convex optimization to address challenges in enforcing such implicit safety constraints for two applications: collision avoidance for convex sets and hierarchical MPC. The dissertation is composed of three parts. In Part I, we consider enforcing distance-based safety constraints for convex sets using CBFs. Using duality theory, we smoothly and nonconservatively reformulate distance-based discrete-time CBF constraints for polytopes, enabling robot navigation in tight environments. For continuous-time dynamical systems, we use sensitivity analysis to enforce CBF constraints for state-dependent convex sets and guarantee strong... https://escholarship.org/uc/item/8hw6n3ck Wed, 2 Sep 2026 00:00:00 +0000 Thirugnanam, Akshay https://escholarship.org/uc/item/8gt3t0kd Backward signal coordination, or BSC for short, is a little-known but promising means of managing traffic in congested cities. When queues in one or more travel directions expand and spillover to street links upstream, there is little value in synchronizing traffic signals to what would be the forward motion of vehicles. Upon seeing green phases, drivers in these hyper congested conditions are constrained from moving forward. This is because links downstream are at least partially filled with standing queues. With BSC, in contrast, green times across neighboring intersections are synchronized to the backward-moving kinematic waves that propagate through queues. In this way, drivers receive green times only after downstream queues have begun moving forward. Far less green time is wasted as a result. As an added benefit, BSC reduces congestion’s damaging footprint on a network by filling queued links to their brims, leaving fewer upstream links with standing queues.The dissertation... https://escholarship.org/uc/item/8gt3t0kd Wed, 2 Sep 2026 00:00:00 +0000 Kong, Sili https://escholarship.org/uc/item/8g03z0kb Autonomous driving requires intelligent systems that can understand the current environment, make decisions from raw sensory observations, and anticipate how the world may evolve over time. Traditional autonomous driving systems typically address these capabilities through a hierarchical pipeline, where perception, planning, and world modeling are developed as largely separate components. While such decomposition offers interpretability and engineering flexibility, it often suffers from information mismatch, error accumulation, and limited scalability. Recent advances in learning-based methods have increasingly blurred the boundaries between these modules, motivating a more unified perspective centered on scalable world representations. This dissertation studies autonomous driving through this lens. The central question of the thesis is how to represent the 3D driving environment in a way that supports scene understanding, decision-making, and future world modeling. Rather... https://escholarship.org/uc/item/8g03z0kb Wed, 2 Sep 2026 00:00:00 +0000 Xie, Yichen https://escholarship.org/uc/item/8dj2b0kj The prospects for a utility-scale, fault-tolerant superconducting quantum computer have grown increasingly promising through demonstrations of quantum advantage and quantum error correction. The next breakthrough will likely be driven by processors at scale, with qubit counts exceeding tens of thousands. Achieving this scale presents a formidable challenge for current architectures, including limitations in chip footprint, cryogenic cooling capacity, and control complexity. One approach to alleviating these constraints is to make each qubit better, thereby reducing the redundancy required for error correction. Protected qubits provide a promising path in this direction by encoding quantum information into a decoherence-free subspace. This can be achieved through Hamiltonian engineering in superconducting circuits, with representative examples including the 0–π qubit, the cos(2φ) qubit, and the GKP qubit. In this thesis, we present the superconducting grid states qubit, or... https://escholarship.org/uc/item/8dj2b0kj Wed, 2 Sep 2026 00:00:00 +0000 Kim, Hyunseong https://escholarship.org/uc/item/8bt0m9dm Our beliefs are increasingly shaped by information we encounter online, yet modern information environments like social media pose distinct epistemic challenges. How do our mechanisms of belief formation and learning meet the demands of digital environments? I present a research program identifying three key capacities that guide belief formation and learning in the digital world. First, I demonstrate that even young children adjust their fact-checking behavior according to the prior quality of their digital learning environment. Second, I show that people use the popularity of a belief as a cue to its reliability, and discuss implications for how engagement metrics on social media may impact beliefs. Third, I discuss eye-tracking findings which suggest that young adults are sensitive to the level of stimulation that is most conducive to their learning, which may help them navigate information overload online. This research begins to map a toolkit for digital epistemic vigilance,... https://escholarship.org/uc/item/8bt0m9dm Wed, 2 Sep 2026 00:00:00 +0000 Orticio, Evan Antonio https://escholarship.org/uc/item/8bd5x6f7 The continuing miniaturization of commercial electronics, micro-electromechanical systems (MEMS), and low-mass spacecraft hardware has created an opportunity to reconsider how interplanetary science missions can be designed, manufactured, and deployed. Conventional spacecraft have enabled extraordinary planetary-science returns, but their high cost, long development cycles, and large system masses limit the number of targets that can be explored. This dissertation presents a MEMS-enabled femtoscale solar-sail spacecraft architecture intended to support low-cost asteroidal imaging, small-body reconnaissance, and future cometary sample-capture missions. The central spacecraft concept developed in this work is the Berkeley Low-cost Interplanetary Solar Sail (BLISS), a nearly 10 g solar-sail spacecraft that uses a roughly 1 m$^2$ reflective sail, commercial-off-the-shelf electronics, onboard imaging, optical communication, and MEMS-based mechanical steering. The BLISS architecture... https://escholarship.org/uc/item/8bd5x6f7 Wed, 2 Sep 2026 00:00:00 +0000 Alvara, Alexander Nicholas https://escholarship.org/uc/item/89b7j2m5 This dissertation comprises three essays in behavioral macroeconomics. The first examines a modified New Keynesian model in which a fraction of agents use simple forecasting rules to form their inflation expectations. The second takes a theory-agnostic empirical approach to investigating the determinants of inflation expectations as measured by surveys. The third asks whether trend-cycle decomposition methods require adjustments to deal with asymmetric business cycles. The first essay develops an adaptive model of inflation expectations in which agents dynamically choose between anchoring to the central bank's announced target and extrapolating from recent inflation, based on the relative forecasting performance of each rule. A multi-armed bandit learning algorithm governs the choice, with a small simplicity bias favoring the anchored rule when the two perform comparably. Central bank credibility, measured as the fraction of agents trusting the target, emerges endogenously... https://escholarship.org/uc/item/89b7j2m5 Wed, 2 Sep 2026 00:00:00 +0000 Smaniotto, Bruno https://escholarship.org/uc/item/85q4p8t8 Mechanical forces generated both within and outside the cell, such as cell-substrate stretching and actin contraction, alter nuclear and chromatin structures and influence gene transcription. Heterochromatin Protein 1 (HP1), a chromosomal protein, has attracted increasing attention for its essential roles in maintaining nuclear and chromatin mechanics by crosslinking chromatin fiber. Degradation or mutation of HP1 has been shown to soften the nucleus and alter epigenetic regulation associated with diseases. However, the molecular mechanisms by which HP1 provides mechanical strength to chromatin and the nucleus remain poorly understood. To elucidate the biomechanical basis of HP1 function, we employed full-atomistic molecular dynamics (MD) simulations to examine HP1 under biochemical and mechanical forces. Our results reveal the mechanosensitivity of HP1, identify the molecular grammar to target HP1, and uncover key determinants of HP1 dissociation from chromatin under pulling... https://escholarship.org/uc/item/85q4p8t8 Wed, 2 Sep 2026 00:00:00 +0000 Tsukamoto, Shingo https://escholarship.org/uc/item/85n036hx The rise of urban areas as the predominant residence for the world’s population has positioned cities as leading contributors to global greenhouse gas and air pollutant emissions. In order to mitigate emissions and their effects efficiently, it is crucial to have a thorough understanding of the urban emission landscape, which remains a challenge given the heterogeneity of emission sources within cities. The atmospheric observation network continues to expand as more and more measurements from satellites, mobile monitoring and aircraft campaigns, and sparse stationary monitoring sites are being made. These observation techniques provide useful information at a wide variety of spatial and temporal scales, but gaps in our understanding of urban air quality still persist. Dense sensor networks fill a niche measurement gap in our current observing system by providing long-term, high frequency measurements of air pollutants at high spatial density across target areas. These measurement... https://escholarship.org/uc/item/85n036hx Wed, 2 Sep 2026 00:00:00 +0000 Patel, Milan Yogesh https://escholarship.org/uc/item/8456d2nx Over the last few years, the massive scaling up of resources poured into the pretraining of large language models (LLMs) has led to rapid advances in LLM capabilities. However, performance on the most challenging reasoning-heavy tasks scales far too slowly as a function of pretraining alone to achieve reasonable performance at a practical scale. A different approach to scaling up computation is therefore needed in order to meaningfully improve performance on these tasks.In this thesis we aim to understand when and if scaling up computation at test-time with LLMs can fulfill this gap, making for an effective alternative to scaling up pretraining compute. We begin by demonstrating the necessity of scaling up pretraining resources in order to broadly improve model capabilities. From here we turn towards test-time scaling as a potential substitute. In particular, we carry out one of the first careful empirical studies comparing the efficacy of different strategies for scaling up computation... https://escholarship.org/uc/item/8456d2nx Wed, 2 Sep 2026 00:00:00 +0000 Snell, Charlie Victor https://escholarship.org/uc/item/84428181 It has long been known that humans are exquisitely sensitive to detecting relative motion; motion of one object relative to a fixed frame of reference. This discriminative ability is considered a hyperacuity because detection thresholds are smaller than the spacing between photoreceptors. Yet, when an image moves in a direction that is directly opposite to the direction of eye motion; in a direction consistent with retinal slip, it appears relatively stable despite the presence of world-fixed background content. It is not known how the visual system leverages frames of reference and the magnitude of retinal slip in governing perception of motion when images move in a direction consistent with retinal slip. Further, it remains unknown whether this phenomenon exists during other eye movements, such as smooth pursuit, when a subject tracks a moving target. We measured motion perception during visual fixation and smooth pursuit eye movements. We used an adaptive optics scanning light... https://escholarship.org/uc/item/84428181 Wed, 2 Sep 2026 00:00:00 +0000 D'Angelo, Josephine C. https://escholarship.org/uc/item/83j1s76d This dissertation comprises three essays examining how the distinctive features of developing economies shape the decisions of firms and households. The unifying argument is that economic findings from high-income settings often fail to transfer cleanly to low- and middle-income countries (LMICs), where incomplete markets, weaker infrastructure, limited social protection, and severe budget constraints fundamentally alter the problems that agents face. Ignoring these differences may harm both academic understanding of low-income economies and the effectiveness of economic development policies. The first chapter asks whether risk aversion prevents firms in developing countries from pursuing profitable but uncertain investments. Standard economic theory assumes firms are risk neutral. This assumption may be correct in wealthy economies where owners hold diversified portfolios. In LMICs, however, most firms are owner-operated, and losses threaten household consumption directly.... https://escholarship.org/uc/item/83j1s76d Wed, 2 Sep 2026 00:00:00 +0000 Killeen, Grady Shea https://escholarship.org/uc/item/81s6s341 Glycolysis is a conserved metabolic pathway that produces ATP and biosynthetic precursors. Multiple allosteric regulators control glycolytic enzymes in vitro. For example, phosphofructokinase (PFK) is allosterically regulated by fructose-2,6-bisphosphate (F26BP), ATP, ADP, AMP, citrate, acyl-CoA, and inorganic phosphate. It is not well understood which properties of homeostasis are enabled by each of these regulators, and whether they perform redundant or distinct functions. Using mathematical modeling and experiments with human cells lacking F26BP, we demonstrate that F26BP alters glycolytic rate independent of cellular ATP demand–a unique function not shared by other regulators. We also identified several downstream glycolytic intermediates as novel regulators of F26BP levels. Our findings clarify the role of F26BP as a unique regulator that controls the glycolytic rate independently of the cellular energy state in response to hormone and biosynthetic precursor levels. The F26BP... https://escholarship.org/uc/item/81s6s341 Wed, 2 Sep 2026 00:00:00 +0000 Kober, Megan Marie https://escholarship.org/uc/item/81c5w7g2 Although Latino students constitute the majority in California schools, Latinos remain significantly underrepresented on school boards. In California, Latinos make up about 30 percent of all school board seats across the state's nearly 950 school districts, even though Latino students make up 56 percent of the student population. This representational gap raises questions about how board members prioritize community needs and whether they consider the needs of Latino students and families. While multiple factors contribute to this underrepresentation, scholars have identified at-large elections as an institutional mechanism that dilutes the voting strength of Latinos, thereby limiting their ability to elect their preferred candidates. Research further shows that Latino candidates are more likely to be elected in district elections, though transitioning and implementing these electoral structures requires a complex political and legal process. However, scholars have largely overlooked... https://escholarship.org/uc/item/81c5w7g2 Wed, 2 Sep 2026 00:00:00 +0000 Quintero, Kevin https://escholarship.org/uc/item/80407548 Quantum simulation of lattice gauge theories offers a path to first-principles calculations of real-time dynamics and finite-density phenomena that are inaccessible to classical Euclidean Monte Carlo methods. Realizing this potential requires a pipeline of controlled approximations---lattice discretization, truncation, digitization, state preparation, time evolution, measurement, and continuum extrapolation---each contributing errors and computational costs. General algorithmic asymptotic complexity arguments capture only part of the cost picture: the practical resources required for end-to-end simulation depend on numerical pre-factors, the cost of the subroutines that various algorithms call, and structural features of the lattice theory, all of which the asymptotic scaling leaves unspecified. This thesis develops contributions that aim to make such dependencies explicit and quantitative. We begin with a rigorous analysis of the asymptotic gate complexity of product formulas... https://escholarship.org/uc/item/80407548 Wed, 2 Sep 2026 00:00:00 +0000 Hariprakash, Siddharth https://escholarship.org/uc/item/7v68z4nv Advancements in vehicle technology and transportation electrification come with changing voltage architectures to support higher power demands. Hybrid switched-capacitor converters and other novel converter architectures are attractive solutions in this application space due to their ability to achieve high power densities and efficiencies. However, there are challenges associated with the adoption and implementation of these converters into automotive subsystems. This dissertation focuses on the design optimization of novel power converter topologies while addressing practical challenges vital to transportation. The critical challenges addressed in this thesis encompass electromagnetic interference, dynamic operation, isolation, and output voltage regulation. Through several hardware design case studies, these challenges are addressed to further advance electronic systems in vehicular environments. https://escholarship.org/uc/item/7v68z4nv Wed, 2 Sep 2026 00:00:00 +0000 Krishnan, Sahana https://escholarship.org/uc/item/7sf2441b My dissertation features an overview of 3 studies relating to the Berkeley Undergraduate Research Evaluations Tools (BURET) project. The BURET project was initiated to develop qualitative and quantitative research tools and protocols to evaluate how learning and mentorship are being experienced by undergraduate students who are involved in faculty laboratory research. The BURET project was also tasked with answering the following three questions: 1) What do undergraduate researchers learn about scientific content and research practices? 2) What practices do graduate students and post-docs report using when teaching and mentoring undergraduate researchers? 3) What teaching and mentoring practices do undergraduates report in their research experiences? I have carried out three studies that together focus on research impacts in the K-12 classroom, measurement of student mentoring and teaching in undergraduate research experiences (UREs), and analysis of interventions carried out... https://escholarship.org/uc/item/7sf2441b Wed, 2 Sep 2026 00:00:00 +0000 Garcia, Anthony Oswaldo https://escholarship.org/uc/item/7qn3h5d9 Wildfires emit a large amount of ozone precursors, nitrogen oxides, and particulate matter into the troposphere and sometimes the stratosphere. With the increasing wildfire events in recent years, the western U.S. regions may experience challenges with higher-than-normal air pollution levels during the fire season (June-October). Therefore, determining the ozone enhancement pattern from wildfire smoke is crucial to understanding the influence of wildfire plume transport on a subregional basis. Chapter one of this study investigates the impact of the 2013 Rim Fire on ozone levels in rural Nevada, employing a combination of ground-based monitors, satellite remote sensing, and atmospheric modeling. The research focuses on understanding how wildfire smoke affects ozone concentrations in downwind regions. Findings indicate significant ozone enhancements on smoke days compared to non-smoke days, highlighting the contribution of wildfire emissions to regional air quality deterioration.... https://escholarship.org/uc/item/7qn3h5d9 Wed, 2 Sep 2026 00:00:00 +0000 Ji, Yi https://escholarship.org/uc/item/7p7200js Majorana fermions arise in a broad range of systems in condensed matter physics, not only as a useful abstraction but also as emergent quasiparticles, e.g., at the boundaries of topological insulators. In this dissertation, we consider two distinct settings wherein Majorana fermions play a central role: (i) non-abelian Kitaev spin liquids and (ii) entanglement dynamics in monitored quantum circuitsThe Kitaev spin liquid is a paradigmatic model featuring non-abelian anyons and a chiral Majorana edge mode when time-reversal symmetry is broken. To this end, it presents an alluring platform for realizing topologically protected quantum information. However, experimental efforts to identify this putative phase have been hampered both by the difficulty of probing the charge neutral quasiparticles and by the presence of phonons and other deviations from the idealized toy model. Here we examine how judicious device design may suppress bulk phonon contributions to thermal transport, allowing... https://escholarship.org/uc/item/7p7200js Wed, 2 Sep 2026 00:00:00 +0000 Klocke, Kai Christian https://escholarship.org/uc/item/7n47r13h Despite their increasingly widespread usage, large language models (LLMs) do not meet all needs of all users in practice. How can we bridge this gap to design LLMs that work reliably and fairly for the broad range of real LLM users? I present research tackling this problem at three stages of model design. First, I discuss how to build LLMs that leverage disagreement among user preferences to work for entire populations of users, using that disagreement as signal to improve model training and evaluation. Then, I discuss designs for rigorous evaluations to extricate challenging harms that diverse users face when using LLMs. Finally, I discuss addressing core technical failures of LLMs, such as miscalibrated confidence, to reduce downstream risks when models are deployed to users with different needs. Combined, these interventions facilitate building LLMs that minimize societal harms, and maximize benefits to a wider range of real-world users. https://escholarship.org/uc/item/7n47r13h Wed, 2 Sep 2026 00:00:00 +0000 Fleisig, Eve https://escholarship.org/uc/item/7k90p79r The sustainability of modern cities hinges on the efficient management of land and water resources. In this dissertation I use tools from empirical and spatial economics to study environmental and natural resource management challenges with a particular focus on cities in the developing and middle income world. In Chapter 1, with Carolina Rodríguez-Zamora we study the costs of and the housing market response to subsidence– the sinking of land areas due to groundwater over-extraction– in Mexico City. We propose an equilibrium model of the housing market that features housing re-development in the face of an evolving environmental hazard that has both realized and expected future impacts to home quality. Our model highlights that while realizations of subsidence attract development by lowering the opportunity cost of re-building units, information frictions affecting the capitalization of future risk lead to an over-supply of housing in risky areas. Guided by model-derived... https://escholarship.org/uc/item/7k90p79r Wed, 2 Sep 2026 00:00:00 +0000 Hackett, Lucy https://escholarship.org/uc/item/7jg4v6h9 Catholic hospitals represent a large and growing segment of the U.S. health care system. The number of Catholic hospitals has grown by over 20% in the last two decades and now approximately one in seven U.S. hospitals is Catholic. These hospitals operate under the Ethical and Religious Directives for Catholic Health Care Services (ERDs), a set of 77 directives that emphasize the Church’s role in caring for poor and vulnerable populations, while also prohibiting certain services such as contraception, sterilization, and abortion. As Catholic health systems continue to expand, a fundamental question emerges: what are the consequences for hospital operations, access to care, and patient outcomes?This dissertation addresses that question across three empirical studies, each using rigorous quasi-experimental methods to provide causal evidence on the consequences of Catholic identity in U.S. health care. Together, the three chapters examine how religious directives shape organizational... https://escholarship.org/uc/item/7jg4v6h9 Wed, 2 Sep 2026 00:00:00 +0000 Schulte, Alex https://escholarship.org/uc/item/7jc6v35f Learning-based policies trained across diverse embodiments have improved generalization in robotics domains such as manipulation and navigation, but those gains are typically expressed at the level of semantic task understanding, while low-level actuation is left to existing inner-loop controllers on each platform. This dissertation asks whether cross-embodiment learning can provide a similar benefit when the learned policy is itself the low-level controller, so that generalization must hold across differences in the internal vehicle dynamics rather than only across high-level task specifications. Quadrotors are used as the testbed: they are intrinsically unstable, must be controlled at the motor-command level, and span orders-of-magnitude differences in mass, inertia, and actuator constants across designs, so failures of generalization are safety-critical rather than merely degradations in performance. The dissertation studies whether a single learned controller can adapt across... https://escholarship.org/uc/item/7jc6v35f Wed, 2 Sep 2026 00:00:00 +0000 Zhang, Dingqi https://escholarship.org/uc/item/7f6227c4 As artificial intelligence and autonomous agents become primary actors in modern economic systems, the dynamics of market interactions have shifted from human-led deliberation to high-frequency, data-driven algorithmic competition. This dissertation investigates the theoretical foundations and practical design of robust algorithmic ecosystems, organized around three core pillars: robust mechanism design, reliable decision-making algorithms, and strategic data governance. Part I addresses the vulnerability of classical auctions to strategic manipulation. We introduce the VCG-Posted Price (V-PoP) mechanism, which uses a collusion detection oracle to maintain incentive compatibility and efficiency in the presence of bid manipulation through coordination. We establish welfare and revenue guarantees, demonstrating how side information can be leveraged to improve on classical mechanisms. Part II focuses on the challenge of building computationally efficient and reliable... https://escholarship.org/uc/item/7f6227c4 Wed, 2 Sep 2026 00:00:00 +0000 Kudva, Sukanya https://escholarship.org/uc/item/7dt2x21z Glaucoma is a chronic neurodegenerative disease characterized by progressive retinal ganglion cell loss and irreversible vision decline. Although elevated intraocular pressure is a major risk factor, pressure-lowering therapies do not fully prevent disease progression, indicating that additional mechanisms contribute to retinal injury. Among these, Müller glia are increasingly recognized as central regulators of retinal homeostasis whose responses to mechanical, metabolic, and inflammatory stress can determine whether the retina adapts to injury or progresses toward chronic dysfunction. This dissertation investigates Müller glia as both mediators and therapeutic targets in glaucomatous neuroinflammation, with particular emphasis on the endogenous neuroprotective lipid mediator Lipoxin B4 (LXB4). Chapter 1 provides the biological framework for the dissertation by reviewing Müller glia in health and disease. It highlights their essential roles in structural support, neurotransmitter... https://escholarship.org/uc/item/7dt2x21z Wed, 2 Sep 2026 00:00:00 +0000 Kumar, Matangi Nandha https://escholarship.org/uc/item/7c2463fc The impressive performance of deep learning models in various fields has attracted great attention. Yet a fundamental understanding of deep learning remains limited and is often overshadowed by its rapid practical development. This dissertation studies two statistical questions contribute to such an understanding and are of interest in their own right. The first question is: How can a statistical procedure be constructed within a classical framework and understood through a neural network representation, and what are its statistical properties? In particular, we focus on the problem of nonparametric two-sample testing and propose a new test, called the Radon-Kolmogorov-Smirnov (RKS) test, which is defined by an integral probability metric over a Radon bounded variation function class. The RKS test generalizes the classical Kolmogorov-Smirnov test to higher dimensions and higher smoothness degrees. Its test statistic is attained by a single neuron with ReLU activation, and... https://escholarship.org/uc/item/7c2463fc Wed, 2 Sep 2026 00:00:00 +0000 Paik, Seunghoon https://escholarship.org/uc/item/7bd7n4nw Credit markets and financial intermediation shape the macroeconomy through channels that can be hard to measure or fall outside standard frameworks. This dissertation contributes new evidence and tools in three such areas: informal bank supervisory pressure, the frictions that link credit access to from pricing, and the bilateral mechanics of direct lending. In the frst chapter, Official Arm-Twisting? Measuring the Federal Reserve’s Use of Moral Suasion, I use large language models to construct the first measure of the Federal Reserve’s efforts to encourage or pressure financial institutions outside formal policy. The measure, built from 45 years of financial industry newspaper coverage, peaks in 1980, 2005–08, and 2020. I find that moral suasion is deployed countercyclically, used most often to manage prudential risks or influence lending, and is most effective when backed by supervisory force and aligned with banks’ incentives. In the second chapter, Borrowing Constraints, Markups,... https://escholarship.org/uc/item/7bd7n4nw Wed, 2 Sep 2026 00:00:00 +0000 Martell, Emily Elizabeth https://escholarship.org/uc/item/77b21034 Deep learning has become a central tool in autonomous navigation, enabling robots to perceive, plan, and act in complex environments. However, deploying these deep learning models safely remains a major challenge: they are often overconfident, poorly calibrated under distributional shift, and opaque in their decision-making. In safety-critical applications, such as social navigation in crowded human spaces or exploration in unstructured outdoor environments, these limitations can compromise both reliability and trustworthiness. This thesis argues that safe, robust, and trustworthy deployment of deep learning-enabled robot navigation systems requires explicitly quantifying, explaining, and acting on the confidence of deep learning models used in the navigation stack. Whether these models are trained to perceive the robot's environment or take actions based on the environmental state, it is critical to estimate how trustworthy their predictions are and use these confidence... https://escholarship.org/uc/item/77b21034 Wed, 2 Sep 2026 00:00:00 +0000 Pohland, Sara Michelle https://escholarship.org/uc/item/75m3q6bw This dissertation studies how information, institutions, and social preferences shape collective responses to public problems. Across three empirical settings, the chapters examine why individuals and communities do or do not act on information that is socially valuable: information about environmental harm, public infrastructure quality, and public-health risks. The chapters share a common concern with the political and social conditions under which information changes behavior. Information may fail to travel when local leaders have incentives to suppress it, may be strategically distorted by political actors before reaching voters, or may be filtered through durable cultural norms and individual preferences. Together, the dissertation shows that information is not merely a technical input into decision-making; its effects depend on who controls it, how it is delivered, and the social environment in which people interpret and act on it. Chapter 1 studies whether information... https://escholarship.org/uc/item/75m3q6bw Wed, 2 Sep 2026 00:00:00 +0000 Cheung, Chiman https://escholarship.org/uc/item/7439x9mk Many natural competitors are co-owned by a small number of large institutional investors. While concerns about the anticompetitive potential of common ownership are growing, empirical evidence on its impact remains mixed. This study examines whether common ownership shapes whom firms internally assess as competitors. Internally assessed competitors reflect firms’ attention allocation and reveal their strategic focus in investment and operational decisions. Using scraped data on named competitors disclosed in the Competition sections of 10-K filings from 1995 to 2023, I track how these relationships evolve over time. I first validate that competitor removals capture meaningful shifts by showing that they coincide with real changes in competitive interactions. Building on this validation, I use associational analyses and a difference-in-differences design leveraging exogenous institutional mergers to show that greater ownership overlap increases the likelihood that firms remove... https://escholarship.org/uc/item/7439x9mk Wed, 2 Sep 2026 00:00:00 +0000 Tian, Ziqing https://escholarship.org/uc/item/740489bw Recent advancements in policy learning have enabled impressive results in robotics and cyber-physical systems (CPS), ranging from dexterous manipulation to bipedal locomotion. Despite these successes, learned robot policies still struggle to integrate high-level reasoning and feedback with low-level control, limiting their efficacy in complex, long-horizon tasks. This dissertation introduces methodologies that enable learned policies to leverage the same skills that humans use to reliably achieve complex goals in the real world: devising high-level plans, learning from partial successes, creating cooperative strategies, and dynamically adjusting behavior through trial-and-error.The central insight across the methods contained in this thesis is to use formal, symbolic structure to express both complex tasks and their feedback mechanisms. In doing so, policies can learn high-level behavior from precise, structured plans while maintaining the expressivity of their underlying (deep)... https://escholarship.org/uc/item/740489bw Wed, 2 Sep 2026 00:00:00 +0000 Shah, Ameesh https://escholarship.org/uc/item/72r3s55s The many successful efforts to reduce air pollution in California since the 1990s through regulations from the California Air Resources Board (CARB) initially targeted on-road vehicles; however, air pollution is not equally distributed, and some populations still experience greater exposure burdens than others. Two populations disproportionately exposed to air pollution were examined in Northern California: communities of Stockton and Fresno, who experience historically poor air quality because of trapped air pollution in the Joaquin Valley; and farmworkers who use off-road diesel-powered agricultural equipment, and whose exposure to carcinogenic diesel exhaust has been poorly characterized. We conducted exposure assessments for three different air pollutants: polycyclic aromatic hydrocarbons (PAHs), nitrogen dioxide (NO2), and diesel particulate matter, measured as black carbon (BC). Participants in Stockton and Fresno were recruited within the San Joaquin Valley Pollution and... https://escholarship.org/uc/item/72r3s55s Wed, 2 Sep 2026 00:00:00 +0000 Zalay, Marley https://escholarship.org/uc/item/70n345tk This dissertation presents empirical studies on how both informational utility and visual salience guide attention in learners across species. An individual may attend to something due to its usefulness for learning or due to its ’bright and shiny’ features like bold colors and high contrast. A long history of work in developmental psychology establishes these features as foundational organizing principles of the human attention system—both predict looking time even in infancy. Digital information environments increasingly leverage the psychology of attention to maximize engagement and ad revenue. In this dissertation I investigate these foundational mechanisms of attention in digital information environments. Chapters 2 and 3 demonstrate that engagement does not predict learning in young children, as adult observers strongly assume, with implications for education, policy, and technology design. We find that attentional capture due to visual salience can disrupt learning in young... https://escholarship.org/uc/item/70n345tk Wed, 2 Sep 2026 00:00:00 +0000 Shepherd, Sarah Stolp https://escholarship.org/uc/item/7003g5fd The origin and evolution of our Solar System, our home planet, and our species are mysteries that depend on a rich, complex history of stochastic formation processes. The discovery of planets orbiting other stars beyond the Solar System challenged canonical wisdom on how and when these processes might have unfolded, signaling a paradigm shift in how we see the Earth, life, and ourselves in relation to the Cosmos. The past 30 yr of exoplanet discovery have continued to reveal an overwhelming diversity of planetary properties, host stars, and environments that are completely unlike the Solar System, leading to more questions than answers in how planets and life may emerge. In this dissertation, I present theoretical simulations and multi-technique observations of exoplanets and their host stars to investigate of how planetary systems similar to ours form, evolve, and become habitable. First, I report on the first investigation of potentially habitable exomoons in the HIP 41378... https://escholarship.org/uc/item/7003g5fd Wed, 2 Sep 2026 00:00:00 +0000 Harada, Caleb Kenji https://escholarship.org/uc/item/6s57t67d First-row transition metal intercalated transition metal dichalcogenides (TMDs), TxMCh2,where T is a first-row transition metal and MCh2 is a van der Waals host lattice, exhibit a wide range of magnetic properties spanning ferromagnetism to altermagnetism. The underlying magnetic exchange interactions and resulting ground states of these materials are highly sensitive to several tuning parameters, such as the intercalant concentration (x) and the degree of disorder within both the intercalant and host layers. This combination of emergent magnetic phenomena and multiple tunable parameters makes intercalated TMDs an exciting platform for exploring novel physics with potential technological relevance. However, the roles of different types of disorder, such as vacancies and the coexistence of multiple superlattice orderings within the intercalant layer, and their impact on magnetic behavior remain poorly understood. Establishing a clear and guided structure–property relationship is... https://escholarship.org/uc/item/6s57t67d Wed, 2 Sep 2026 00:00:00 +0000 Gonzalez, Oscar https://escholarship.org/uc/item/6rg7g5rx This dissertation explores how risk is priced across different sectors of the economy. In particular, Chapter 1 focuses on equity markets while Chapters 2 and 3 look at the commercial banking sector.In Chapter 1 I develop an information-geometric framework for selecting priced risk factors from a high-dimensional candidate set. The stochastic discount factor is cast as a point on a statistical manifold of exponential-affine pricing kernels, and the manifold's dual flatness yields an exact Pythagorean decomposition of Kullback-Leibler divergence across nested SDFs. The decomposition drives a greedy forward-selection algorithm that is isomorphic to the portfolio problem of an investor with multiplier preferences. The investor's ranking of candidates by worst-case welfare loss coincides with the statistical ranking by KL divergence, and the Hansen-Jagannathan distance emerges as a leading-order approximation. A Wald test on estimated risk prices and an entropy bound calibrated to... https://escholarship.org/uc/item/6rg7g5rx Wed, 2 Sep 2026 00:00:00 +0000 Perry, Andrew https://escholarship.org/uc/item/6pz4p8rx This dissertation examines why Pakistan's legal institutions fail to protect women from domestic violence despite three decades of legislative reform. The answer lies not in an absence of law but in an interlocking system of institutional resistance that operates across legislatures, religious review bodies, enforcement agencies, courts, and customary forums. This system sustains the privatization of household harm through three reinforcing dynamics. The first is the cultural and legal construction of domestic violence as a family matter beyond state concern. The second is the institutional shaping of religious interpretation, where the outcome depends on who holds interpretive authority. A clear illustration appears in how the Council of Islamic Ideology and the Federal Shariat Court have read the same sources to reach opposing conclusions on domestic violence legislation. The third is institutional absorption, whereby enforcement bodies divert complaints toward reconciliation... https://escholarship.org/uc/item/6pz4p8rx Wed, 2 Sep 2026 00:00:00 +0000 Moazzam, Mahwish https://escholarship.org/uc/item/6mq50911 The interplay between mathematics and physics has a long and rich history. Recently, ideas from algebraic geometry have played an increasingly important role in the study of physical phenomena, including those arising in particle physics, quantum mechanics, and cosmology. This exchange has led to major advances in both fields and continues to open new directions. In this thesis, I establish a novel connection between algebraic geometry and quantum chemistry. Through methods from nonlinear algebra, with particular emphasis on combinatorics, and representation theory, I develop geometric formulations of coupled cluster theory that lead to new structural, enumerative, and computational results.First, we develop an algebraic-geometric framework for coupled cluster (CC) theory. At the heart of quantum chemistry is the problem of solving the electronic Schrödinger equation, which can be formulated as a finite but high-dimensional eigenvalue problem. To study this problem, we introduce... https://escholarship.org/uc/item/6mq50911 Wed, 2 Sep 2026 00:00:00 +0000 Sverrisdottir, Svala https://escholarship.org/uc/item/6m1230zw Density functional theory (DFT) guarantees a map between a system's electron density and energy. Although the exact density functional is still unknown to us today, with the development and benchmark of modern density functional approximations (DFA), DFT remains the most prevalent method for electronic structure calculations due to its good balance between computational cost and accuracy. However, besides the accurate energies, another main concern to chemists is how to interpret the DFT calculation results for understanding the behavior of interesting chemical systems. This dissertation tries to bridge the gap between accurate DFT calculations and insightful chemical interpretations through the analysis of the key component of DFT: the electron density. The work begins with the comparison between two popular energy decomposition analysis (EDA) schemes for intermolecular interactions, namely the extended transition-state method with natural orbitals for chemical valence... https://escholarship.org/uc/item/6m1230zw Wed, 2 Sep 2026 00:00:00 +0000 Shen, Hengyuan