150813 research outputs found
Sort by
The Death of Quasiparticles: Strongly Interacting Gapless Phases with Fermi Surfaces and Fractional Statistics
No full textThe emergence of quasiparticles at low temperature provides a powerful organizing principle for many quantum phases of matter, ranging from conventional magnets and superconductors to exotic insulators with topological order. In this thesis, I describe my research in gapless quantum phases in which the framework of quasiparticles breaks down. The main characters are two categories of gapless phases that feature the interplay between strong interactions and two additional ingredients – Fermi surfaces and fractional statistics. Chapter 2 through Chapter 5 focus on strongly interacting metals with Fermi surfaces. The most salient examples are a class of Hertz-Millis models describing the onset of spontaneous symmetry breaking in a metallic environment. At the quantum critical point, gapless order parameter fluctuations destroy quasiparticles living on the Fermi surface, giving rise to a strongly coupled non-Fermi liquid metal. A key result of these chapters is the identification of an infinite-dimensional symmetry that survives in these non-Fermi liquid metals despite the death of quasiparticles. This infinite-dimensional symmetry and its quantum anomaly lead to a series of non-perturbative results on thermodynamics and transport, which are confirmed by perturbative diagrammatic calculations in special examples. Chapter 6 through Chapter 8 explore quantum phases in which anyonic quasiparticles with fractional statistics play an essential role. When parameters in the system are tuned to close the anyon energy gaps, the original anyons lose their coherence and a variety of novel phases emerge. A highlight in this direction is a new mechanism for topological superconductivity in itinerant abelian and non-abelian anyon fluids, which could make contact with experiments on doped fractional quantum anomalous Hall states in the near future.Ph.D
LLMs in Citation Intent Classification: Progress, Precision, and Reproducibility Challenges
No full textACM REP ’25, Vancouver, BC, CanadaUnderstanding the intent behind scientific citations is critical for
advancing scholarly search and knowledge mapping. This paper
reflects on the methodological use of large language models (LLMs)
for multi-class citation intent classification. Our experiments evaluating a diverse range of models and approaches reveal striking
disagreement among state-of-the-art (SotA) systems. This inconsistency suggests that citation intent classification remains a challenging task for LLMs raising questions about the robustness, reliability
and replicability of current methods. Moreover, our findings highlight a concerning dependency on proprietary LLMs that, even
with access to compute resources, were necessary to achieve sufficient accuracy. This introduces new challenges, as silent updates,
lack of versioning, and opaque training pipelines pose threats to
methodological transparency and long-term reproducibility in LLMenabled research
Parameter Estimation for Anonymous Hawkes Processes
No full textHawkes Processes are self-exciting point processes used to model many real-life networks in which an event from one agent causes the rate at which events occur from related agents to increase, such as in earthquake networks or social media. This project investigates the question of finding the underlying structure of the Hawkes Processes given a history of when events occurred. This problem has been studied extensively in the regime where the event labels are known, and the bulk of the literature involves parameterizing the model and passing it through statistical learning tools. Our proposed work focuses on the the same question in “anonymous" case where labels are not given. In this regime, the lack of information makes many previous approaches intractable and we develop novel non-parametric approaches for solving cases of the structure learning problem in algorithmic and information theoretic settings. Our results show the ability to learn the entire model under mild assumptions in the information theoretic regime, where we have access to an arbitrarily long Anonymous Hawkes Process transcript, whereas when we’re confined to a polynomially lengthed transcript, the situation is considerably more difficult.M.Eng
Meeting at Crossroads: An exploration of playful listening through a co-creative AI game
No full textCHI EA ’25, Yokohama, JapanActive listening is a well-established cornerstone of empathetic communication and a hallmark of “civic competence”, but is a challenging and energy consuming skill. Games offer a provocative lens to consider how active listening could be explored playfully. In this paper, we present Crossroads, an interactive social game which makes active listening fun by inviting players to co-create images about one another’s personal experiences. Deployed through a tablet-mobile web app, players take turns acting in ‘listener roles’ to generate AI images, and eventually uncover a collective picture along a “crossroad” shaped map. An initial mixed-method evaluation with 36 users demonstrates that players find the experience highly engaging and feel especially heard during in-game conversations. This work contributes a novel game which uses AI to mediate empathetic dialogue, and surfaces questions about the trade-offs of gamifying listening
The Boundawall: A Proposal on the Nature of Black Holes
No full textThis research suggests a new interpretation of black holes in which the event horizon represents the termination of physical reality. In this view, when curvature approaches a critical threshold, the three-dimensional spatial geometry may undergo a dimensional compression into a two-dimensional manifold—the boundawall—that preserves gravitational continuity while preventing further causal evolution. Inside this surface, spacetime would cease to exist. All mass-energy and information would then be confined to the boundawall, forming a structure consistent with the external Schwarzschild geometry and the Bekenstein-Hawking entropy law. We outline a possible Dimensional Conversion Law that could govern this phenomenon, and discuss the conservation, causal, and thermodynamic implications of the boundawall. Finally, we comment on potential observational consistency and on limited predictions such as surface-mode signatures. In this theory, the event horizon is viewed not merely as a limit of observation, but as a potential boundary/wall of existence itself
Engineering Principles for Scalable Connectomics
No full textBrain tissue sectioning presents a significant challenge in connectomics, particularly when scaling to larger volumes. In the MICrONS 1 mm³ mouse visual cortex dataset, 25.1% of scanned images—representing over a month of imaging work—were discarded due to sectioning defects. Current methods result in material loss during cutting and face limitations in tool wear and process efficiency. This thesis examines tissue sectioning through an engineering lens. Drawing from established machining practices and parallel industries, we propose and evaluate potential improvements to sectioning methods. The work aims to contribute to ongoing efforts in mapping larger connectomes, making it more practical and less error-prone.S.B
The Grice Is Right: Grice's Non‐Cooperation Problem and the Structure of Conversation
No full textH. P. Grice seemed to rest his theory of conversational implicature on the assumption that speakers aim to cooperatively exchangeinformation with each other. In the real world, speakers often don’t. Does one of the most influential theories in 20th-centuryphilosophy of language rest on a mistake? Yes—but not in the way that philosophers have thought. I argue that Grice shouldhave rested his theory on a different assumption: that speakers aim to appear to aim to cooperatively exchange informationwith each other. This proposal dissolves Grice’s Non-Cooperation Problem but preserves Grice’s central insights about the natureof conversational implicatures. More generally, it enables the Gricean to illuminate the structure of many non-cooperative orotherwise “non-ideal” conversations
ZeoSyn: A Comprehensive Zeolite Synthesis Dataset Enabling Machine-Learning Rationalization of Hydrothermal Parameters
No full textZeolites, nanoporous aluminosilicates with well-defined porous structures, are versatile materials with applications in catalysis, gas separation, and ion exchange. Hydrothermal synthesis is widely used for zeolite production, offering control over composition, crystallinity, and pore size. However, the intricate interplay of synthesis parameters necessitates a comprehensive understanding of synthesis-structure relationships to optimize the synthesis process. Hitherto, public zeolite synthesis databases only contain a subset of parameters and are small in scale, comprising up to a few thousand synthesis routes. We present ZeoSyn, a dataset of 23,961 zeolite hydrothermal synthesis routes, encompassing 233 zeolite topologies and 921 organic structure-directing agents (OSDAs). Each synthesis route comprises comprehensive synthesis parameters: 1) gel composition, 2) reaction conditions, 3) OSDAs, and 4) zeolite products. Using ZeoSyn, we develop a machine learning classifier to predict the resultant zeolite given a synthesis route with >70% accuracy. We employ SHapley Additive exPlanations (SHAP) to uncover key synthesis parameters for >200 zeolite frameworks. We introduce an aggregation approach to extend SHAP to all building units. We demonstrate applications of this approach to phase-selective and intergrowth synthesis. This comprehensive analysis illuminates the synthesis parameters pivotal in driving zeolite crystallization, offering the potential to guide the synthesis of desired zeolites. The dataset is available at https://github.com/eltonpan/zeosyn_dataset
Computing Skinning Weights via Convex Duality
No full textWe study the problem of optimising for skinning weights through the lens of convex duality. In particular, we show that the popular bounded biharmonic weight (BBW) model for skinning is dual to a non-negative least-squares problem, which is amenable to efficient solution via iterative algorithms; the final weights are then recoverable via a closed-form expression. Our formulation maintains convexity and is provably equivalent to the original problem. We also provide theoretical discussion giving intuition for the dual problem in the smooth case. Our final algorithm, which can be implemented in a few lines of code, achieves efficient convergence times relative to generic quadratic programming tools applied to the primal problem, without nonconvex formulations, relaxations or specialised optimisation techniques
Pressurized plankton observatory offers a new window into deep‐sea larval behavior
No full textThe High-Pressure Plankton Observatory (HiPPO) is designed to quantify motions of zooplankton for behavioral study, including swimming and metabolic responses to environmental perturbations. It builds on prior chamber designs while filling gaps in capability for resolving orientation of small (< 1 mm) plankton, tracking their movements over ecologically relevant spatial scales, and recording in flow-through conditions on a vessel at sea. The HiPPO chamber has a direct light path for silhouette imaging of zooplankton as they move vertically and horizontally across a 3.56 cm diameter viewing area. Seawater forced by a high-performance liquid chromatography pump is exchanged continuously through the chamber, but flushing of zooplankton is prevented by fine mesh at the ports. A high-resolution camera/computer setup enables sustained imaging of plankton motions for quantitative analysis. Application of HiPPO to an investigation of larval behavior of deep-sea hydrothermal vent species revealed swimming behaviors similar to those of shallow-water species, including upward and downward helices, meandering, and short hovers. In conditions with microbial biofilm (a potential settlement cue) on a 2024 expedition, vent larvae unexpectedly swam rapidly upward in tight helices at velocities (0.15 cm s−1) higher than those observed in prior experiments with no biofilm (0.03 cm s−1). Many factors varied between the 2024 and earlier trials, so the difference cannot be attributed with certainty to a cue response. This study describes key new features of HiPPO and demonstrates the system's ability to document novel zooplankton behavior