arXiv subjects

Christopher Chiu

Publications and source records attributed to Christopher Chiu.

14 recordsLinked to original sources

arXiv2026

Automatic Construction of Clinical Scoring Systems with LLM Agents

Modern clinical practice relies on evidence-based guidelines implemented as compact scoring systems composed of a small number of interpretable decision rules. While machine-learning models achieve strong performance, many fail to translate into routine clinical use due to misalignment with workflow constraints such as memorability, auditability, and bedside execution. We argue that this gap arises not from insufficient predictive power, but from optimizing over model classes that are incompatible with guideline deployment. Deployable guidelines often take the form of unit-weighted clinical checklists, formed by thresholding the sum of binary rules, but learning such scores requires searching an exponentially large discrete space of possible rule sets. We introduce AgentScore, which performs semantically guided optimization in this space by using LLMs to propose candidate rules and a deterministic, data-grounded verification-and-selection loop to enforce statistical validity and deployability constraints. Across eight clinical prediction tasks, AgentScore outperforms existing score-generation methods and achieves AUROC comparable to more flexible interpretable models despite operating under stronger structural constraints. On two additional externally validated tasks, AgentScore achieves higher discrimination than established guideline-based scores.

cs.LG↗

arXiv2025

When AI Agents Compete for Jobs: Strategic Capabilities and Economic Dynamics of AI Labour Markets

Emerging agentic marketplaces provide the economic infrastructure for matching and coordinating the large amounts of AI agents used in agentic swarms. Unlike human workers, AI agents can operate on multiple jobs simultaneously, acquire skills rapidly, and labor without wage floors. These differences introduce a new segment of $\textbf{AI labor markets}$, where AI agents interact with each other at a much higher frequency than human markets. Yet we lack frameworks to understand how such markets behave in light of economic forces that shape labor markets, such as adverse selection and reputation dynamics. To explore this, we introduce $\texttt{AI-Work}$, a tractable, simulated gig economy where Large Language Model (LLM) agents compete for jobs, develop skills, and adapt their strategies under uncertainty and competitive pressure. Our experiments examine three domains of capabilities that successful agents possess: $\textbf{metacognition}$ (accurate self-assessment of skills), $\textbf{competitive awareness}$ (modeling rivals and market dynamics), and $\textbf{long-horizon strategic planning}$. Agents with these capabilities consistently achieve higher profits, market share, and stronger adaptation than competing agents. Through $\texttt{AI-Work}$, we hope to provide a foundation to explore the microeconomic properties of AI-only labor markets, and a conceptual framework to study the strategic reasoning capabilities of participating AI agents.

cs.MA↗

arXiv2025

Simulating Viva Voce Examinations to Evaluate Clinical Reasoning in Large Language Models

Clinical reasoning in medicine is a hypothesis-driven process where physicians refine diagnoses from limited information through targeted history, physical examination, and diagnostic investigations. In contrast, current medical benchmarks for large language models (LLMs) primarily assess knowledge recall through single-turn questions, where complete clinical information is provided upfront. To address this gap, we introduce VivaBench, a multi-turn benchmark that evaluates sequential clinical reasoning in LLM agents. Our dataset consists of 1762 physician-curated clinical vignettes structured as interactive scenarios that simulate a (oral) examination in medical training, requiring agents to actively probe for relevant findings, select appropriate investigations, and synthesize information across multiple steps to reach a diagnosis. While current LLMs demonstrate competence in diagnosing conditions from well-described clinical presentations, their performance degrades significantly when required to navigate iterative diagnostic reasoning under uncertainty in our evaluation. Our analysis identified several failure modes that mirror common cognitive errors in clinical practice, including: (1) fixation on initial hypotheses, (2) inappropriate investigation ordering, (3) premature diagnostic closure, and (4) failing to screen for critical conditions. These patterns reveal fundamental limitations in how current LLMs reason and make decisions under uncertainty. Through VivaBench, we provide a standardized benchmark for evaluating conversational medical AI systems for real-world clinical decision support. Beyond medical applications, we contribute to the larger corpus of research on agentic AI by demonstrating how sequential reasoning trajectories can diverge in complex decision-making environments.

cs.LG↗

arXiv2025

Singular loci in varieties of tensors

A Vec-variety is a suitable functor from finite-dimensional vector spaces to finite-dimensional varieties. Most varieties in the geometry of tensors, e.g. the variety of d-way tensors of slice rank at most r, are of this form. We prove that the singular locus of a Vec-variety is a proper closed Vec-subvariety, analogously to the situation for ordinary finite-dimensional varieties. Via earlier work of the third author, this implies that these singular loci admit a description by finitely many polynomial equations. A natural follow-up question to our main result is whether a Vec-variety also admits a suitably functorial resolution of singularities. We establish some preliminary results in this direction in the regime where the dimension of evaluations of a Vec-variety grows linearly with that of the input vector space.

math.AG↗

arXiv2024

Fine-Grained Classification for Poisonous Fungi Identification with Transfer Learning

FungiCLEF 2024 addresses the fine-grained visual categorization (FGVC) of fungi species, with a focus on identifying poisonous species. This task is challenging due to the size and class imbalance of the dataset, subtle inter-class variations, and significant intra-class variability amongst samples. In this paper, we document our approach in tackling this challenge through the use of ensemble classifier heads on pre-computed image embeddings. Our team (DS@GT) demonstrate that state-of-the-art self-supervised vision models can be utilized as robust feature extractors for downstream application of computer vision tasks without the need for task-specific fine-tuning on the vision backbone. Our approach achieved the best Track 3 score (0.345), accuracy (78.4%) and macro-F1 (0.577) on the private test set in post competition evaluation. Our code is available at https://github.com/dsgt-kaggle-clef/fungiclef-2024.

cs.CV↗

arXiv2024

Topological Noetherianity of the infinite half-spin representations

We prove that the infinite half-spin representations are topologically Noetherian with respect to the infinite spin group. As a consequence we obtain that half-spin varieties, which we introduce, are defined by the pullback of equations at a finite level. The main example for such varieties is the infinite isotropic Grassmannian in its spinor embedding, for which we explicitly determine its defining equations.

math.AG↗

arXiv2023

On the formal neighborhood of a degenerate arc

We prove a result describing the structure of the formal neighborhoods of certain arcs in the arc space of an algebraic variety which are completely contained in the singular locus. In particular, we provide a precise formulation of the intuitive statement that constant arcs centered in the singular locus are the most singular points of the arc space.

math.AG↗

arXiv2022

On arc fibers of morphisms of schemes

Given a morphism $f \colon X \to Y$ of schemes over a field, we prove several finiteness results about the fibers of the induced map on arc spaces $f_\infty \colon X_\infty \to Y_\infty$. Assuming that $f$ is quasi-finite and $X$ is separated and quasi-compact, our theorem states that $f_\infty$ has topologically finite fibers of bounded cardinality and its restriction to $X_\infty \setminus R_\infty$, where $R$ is the ramification locus of $f$, has scheme-theoretically finite reduced fibers. We also provide an effective bound on the cardinality of the fibers of $f_\infty$ when $f$ is a finite morphism of varieties over an algebraically closed field, describe the ramification locus of $f_\infty$, and prove a general criterion for $f_\infty$ to be a morphism of finite type. We apply these results to further explore the local structure of arc spaces. One application is that the local ring at a stable point of the arc space of a variety has finitely generated maximal ideal and topologically Noetherian spectrum, something that should be contrasted with the fact that these rings are not Noetherian in general; a lower-bound to the dimension of these rings is also obtained. Another application gives a semicontinuity property for the embedding dimension and embedding codimension of arc spaces which extends to this setting a theorem of Lech on Noetherian local rings and translates into a semicontinuity property for Mather log discrepancies. Other applications are discussed in the paper.

math.AG↗

arXiv2022

On Real-time Image Reconstruction with Neural Networks for MRI-guided Radiotherapy

MRI-guidance techniques that dynamically adapt radiation beams to follow tumor motion in real-time will lead to more accurate cancer treatments and reduced collateral healthy tissue damage. The gold-standard for reconstruction of undersampled MR data is compressed sensing (CS) which is computationally slow and limits the rate that images can be available for real-time adaptation. Here, we demonstrate the use of automated transform by manifold approximation (AUTOMAP), a generalized framework that maps raw MR signal to the target image domain, to rapidly reconstruct images from undersampled radial k-space data. The AUTOMAP neural network was trained to reconstruct images from a golden-angle radial acquisition, a benchmark for motion-sensitive imaging, on lung cancer patient data and generic images from ImageNet. Model training was subsequently augmented with motion-encoded k-space data derived from videos in the YouTube-8M dataset to encourage motion robust reconstruction. We find that AUTOMAP-reconstructed radial k-space has equivalent accuracy to CS but with much shorter processing times after initial fine-tuning on retrospectively acquired lung cancer patient data. Validation of motion-trained models with a virtual dynamic lung tumor phantom showed that the generalized motion properties learned from YouTube lead to improved target tracking accuracy. Our work shows that AUTOMAP can achieve real-time, accurate reconstruction of radial data. These findings imply that neural-network-based reconstruction is potentially superior to existing approaches for real-time image guidance applications.

physics.med-ph↗

arXiv2021

Isosingular loci of algebraic varieties

We define the notion of isosingular loci of algebraic varieties, following the analytic case first studied by Ephraim. In particular, we give a partial extension of his main result in arbitrary characteristic and a full extension assuming characteristic $0$. One of the main obstructions in the positive characteristic case is the non-separability of the orbit map associated to the contact group, as first observed by Greuel and Pham for isolated singularities.

math.AG↗

arXiv2020

Adaptive multi-channel event segmentation and feature extraction for monitoring health outcomes

$\textbf{Objective}$: To develop a multi-channel device event segmentation and feature extraction algorithm that is robust to changes in data distribution. $\textbf{Methods}$: We introduce an adaptive transfer learning algorithm to classify and segment events from non-stationary multi-channel temporal data. Using a multivariate hidden Markov model (HMM) and Fisher's linear discriminant analysis (FLDA) the algorithm adaptively adjusts to shifts in distribution over time. The proposed algorithm is unsupervised and learns to label events without requiring $\textit{a priori}$ information about true event states. The procedure is illustrated on experimental data collected from a cohort in a human viral challenge (HVC) study, where certain subjects have disrupted wake and sleep patterns after exposure to a H1N1 influenza pathogen. $\textbf{Results}$: Simulations establish that the proposed adaptive algorithm significantly outperforms other event classification methods. When applied to early time points in the HVC data the algorithm extracts sleep/wake features that are predictive of both infection and infection onset time. $\textbf{Conclusion}$: The proposed transfer learning event segmentation method is robust to temporal shifts in data distribution and can be used to produce highly discriminative event-labeled features for health monitoring. $\textbf{Significance}$: Our integrated multisensor signal processing and transfer learning method is applicable to many ambulatory monitoring applications.

eess.SP↗

arXiv2020

Higher derivations of modules and the Hasse-Schmidt module

In this paper we revisit Ribenboim's notion of higher derivations of modules and relate it to the recent work of De Fernex and Docampo on the sheaf of differentials of the arc space. In particular, we derive their formula for the K\"ahler differentials of the Hasse-Schmidt algebra as a consequence of the fact that the Hasse-Schmidt algebra functors commute.

math.AC↗

arXiv2020

Embedding codimension of the space of arcs

We introduce a notion of embedding codimension of an arbitrary local ring, establish some general properties, and study in detail the case of arc spaces of schemes of finite type over a field. Viewing the embedding codimension as a measure of singularities, our main result can be interpreted as saying that the singularities of the arc space are maximal at the arcs that are fully embedded in the singular locus of the underlying scheme, and progressively improve as we move away from said locus. As an application, we complement a theorem of Drinfeld, Grinberg, and Kazhdan on formal neighborhoods in arc spaces by providing a converse to their theorem, an optimal bound for the embedding codimension of the formal model appearing in the statement, a precise formula for the embedding dimension of the model constructed in Drinfeld's proof, and a geometric meaningful way of realizing the decomposition stated in the theorem.

math.AG↗

arXiv2013

ExoMol line lists II: The ro-vibrational spectrum of SiO

Accurate rotation-vibration line lists are calculated for silicon monoxide. Line lists are presented for the main isotopologue, $^{28}$Si$^{16}$O, and for four monosubsituted isotopologues ($^{29}$Si$^{16}$O, $^{30}$Si$^{16}$O, $^{28}$Si$^{18}$O and $^{28}$Si$^{17}$O), in their ground electronic states. These line lists are suitable for high temperatures (up to 9000 K) including those relevant to exoplanetary atmospheres and cool stars. A combination of empirical and \textit{ab initio} methods is used: the potential energy curves are determined to high accuracy by fitting to extensive data from the analysis of both laboratory and sunspot spectra; a high quality {\it ab initio} dipole moment curve is calculated at the large basis set, multi-reference configuration interaction level. A partition function plus full line lists of rotation-vibration transitions are made available in an electronic form as supplementary data to this article and at \url{www.exomol.com}.

astro-ph.SR↗