arXiv subjects

Taegyun Kim

Publications and source records attributed to Taegyun Kim.

14 recordsLinked to original sources

arXiv2026

PetQA: Benchmarking Veterinary Knowledge and Clinical Reasoning

We introduce PetQA, a Korean long-form question-answering (QA) benchmark for evaluating veterinary knowledge and clinical reasoning in large language models (LLMs) and large vision-language models (LVLMs). PetQA contains 10,076 text-only and 8,751 multimodal QA pairs derived from real-world questions about dogs and cats, paired with answers from expert veterinarians. Its test split, PetQA-Bench, further includes annotations for question types and clinical conditions. We evaluate eighteen models using ROUGE, BERTScore, and LLM-as-a-judge metrics for factuality and helpfulness under three settings: zero-shot inference, retrieval-augmented generation (RAG), and supervised fine-tuning (SFT). The benchmarking results provide an overview of the strengths and limitations of current models in addressing veterinary clinical queries and highlight the need for more effective adaptation methods to develop clinically reliable AI systems for veterinary care. To facilitate broader use, we additionally provide translated versions of PetQA-Bench in five languages.

cs.CL↗

arXiv2026

Non-Shattering at and Above the Dynamical Temperature in the Spherical Pure p-Spin Model

We consider the notion of shattering introduced by Ben Arous and Jagannath for spherical pure $p$-spin glasses with overlap $q$. For every $p\geq 3$ and $0<β\leqβ_{\mathrm{sh}}(p)$, we rule out shattering whenever $q\leq2^{-1/2}$ or $q>\sqrt{(p-2)/(p-1)}$. The proof combines a deterministic $N+1$ bound for disjoint bands in the first range with a general-$p$ sign law showing that their total marked weight has subdominant free energy in the second. A spherical-code bound and Hölder's inequality give an additional $q$-dependent obstruction; in particular, they rule out every fixed overlap for $0<β\leq\sqrt{\log2}$. For $p=3$, the first two ranges already exhaust every fixed $q\in(0,1)$, so the landscape is not shattered at any $T\geq T_{\mathrm{sh}}$. For $p\geq4$, the cases not covered by our criteria are confined to $2^{-1/2}<q\leq\sqrt{(p-2)/(p-1)}$ and $\sqrt{\log2}<β\leqβ_{\mathrm{sh}}(p)$. In particular, this paper partially resolves Conjecture 1 of the paper above and also suggests new methods to show non-shattering.

math.PR↗

arXiv2026

Regularized Bulk Universality versus Bounded-Disorder Nonuniversality for Annealed Complexity of Spherical $p$-Spin Landscapes

Fix $p\ge 3$. We establish a separation between regularized bulk universality and unrestricted annealed complexity for the pure spherical $p$-spin Hamiltonian with independent non-Gaussian tensor coordinates. There is a symmetric, compactly supported disorder law with a smooth density, matching the first $2p$ Gaussian moments, for which the expectation of a positive, frame-averaged regularization of the one-point Kac-Rice functional is asymptotic to its Gaussian counterpart, while the unrestricted critical-point count in a compact energy window has a lower exponential rate strictly above the Gaussian limit. The obstruction is a coherent block involving on the order of $N^{1/p}$ coordinates, so neither finite moment matching nor a uniform entry bound restores unregularized annealed universality. For uniformly subexponential disorder matching the first $m$ Gaussian moments, the regularized functional on incoherent frames satisfies $\left|\log\left(\mathbb{E}\mathcal{Z}_{N,L}^J/\mathbb{E}\mathcal{Z}_{N,L}^G\right)\right|\le C N q_N^{m-1}$, where $q_N=L^{p-1}N^{-(p-2)/2}(\log N)^{(p-1)/2}$. Thus mean and variance matching imply regularized pressure universality for every $p\ge 3$, and the expectation ratio tends to one when $(m-1)(p-2)>2$. We identify the Gaussian variational limit and prove the Gaussian quadratic energy-excursion upper bound uniformly over profiles asymptotically supported on $o(N)$ coordinates under a Gaussian-rate profile-tail condition. Finally, we give a conditional reduction to exact universality: once two one-sided de-regularization defects vanish, an exact max formula makes control of the localized complement necessary and sufficient. The defects vanish in the Gaussian model.

math.PR↗

arXiv2026

Impostor Among $ν$s: Dark Radiation Masquerading as Self-Interacting Neutrinos

Multiple cosmological observations hint at neutrino self-interactions beyond the Standard Model, yet such interactions face severe constraints from terrestrial experiments. We resolve this tension by introducing a model where active neutrinos resonantly convert to self-interacting dark radiation after BBN but before CMB epoch. This exploits the fact that cosmological observables cannot distinguish between neutrinos and dark radiation with the same abundance and free-streaming properties. Our mechanism, based on a simple type-I seesaw framework along with a keV-scale scalar mediator, achieves two objectives: (i) it produces strongly self-interacting dark radiation that imitates neutrino self-interactions favored by cosmological data, and (ii) it depletes the active neutrino energy density, relaxing cosmological neutrino mass bounds and easing the tension with neutrino oscillation data. The model naturally evades laboratory constraints through suppression of the neutrino-mediator coupling by the squared mass ratio of active and sterile neutrinos. We show that this scenario is favored over $Λ$CDM by the combined Planck and DESI data, while being consistent with all other constraints. Our mechanism is testable in future laboratory probes of absolute neutrino mass and searches for sterile neutrinos.

hep-ph↗

arXiv2026

A Sharp Universality Dichotomy for the Free Energy of Spherical Spin Glasses

We study the free energy for pure and mixed spherical $p$-spin models with i.i.d.\ disorder. In the mixed case, each $p$-interaction layer is assumed either to have regularly varying tails with exponent $α_p$ or to satisfy a finite $2p$-th moment condition. For the pure spherical $p$-spin model with regularly varying disorder of tail index $α$, we introduce a tail-adapted normalization that interpolates between the classical Gaussian scaling and the extreme-value scale, and we prove a sharp universality dichotomy for the quenched free energy. In the subcritical regime $α<2p$, the thermodynamics is driven by finitely many extremal couplings and the free energy converges to a non-degenerate random limit described by the NIM (non-intersecting monomial) model, depending only on extreme-order statistics. At the critical exponent $α=2p$, we obtain a random one-dimensional TAP-type variational formula capturing the coexistence of an extremal spike and a universal Gaussian bulk on spherical slices. In the supercritical regime $α>2p$ (more generally, under a finite $2p$-th moment assumption), the free energy is universal and agrees with the deterministic Crisanti--Sommers/Parisi value of the corresponding Gaussian model, as established in [Sawhney-Sellke'24]. We then extend the subcritical and critical results to mixed spherical models in which each $p$-layer is either heavy-tailed with $α_p\le 2p$ or has finite $2p$-th moment. In particular, we derive a TAP-type variational representation for the mixed model, yielding a unified universality classification of the quenched free energy across tail exponents and mixtures.

math.PR↗

arXiv2025

Leveraging intermediate resonances to probe CP violation at colliders

We explore the phenomenological impact of interference in tree-level contributions to three-body final states in $2\to 3$ scattering processes. This work introduces a novel search strategy leveraging asymmetries to enable sensitivity to CP-violating effects in less well-explored regions of phase space. Analytically, we demonstrate the effectiveness of this observable in probing interference between Standard Model charged-current decays and effective left-handed vector interactions, illustrated in a toy model featuring a scalar leptoquark, $S_1 \sim (3, 1, -1/3)$. Numerically, we apply this framework to studying the process $pp\to b τν$; unlike traditional high-$p_T$ searches or ``bump hunts", this approach utilizes an intermediate energy regime -- where new physics is neither light enough to be produced on shell or heavy enough to justify an effective field theory treatment. A proof-of-principle analysis at parton level demonstrates a percent-level asymmetry, with sensitivity also to BSM weak-CP phase. While the specific phase sensitivity is diminished at particle level due to showering and detector effects, a machine learning classifier can recover sensitively to the presence of SM-BSM interference, significantly outperforming standard analysis methods. Notably discrimination between BSM signal and SM background could be achieved at the 2$σ$ level for the current LHC dataset and 8$σ$ at the High-Luminosity LHC. Moreover, this asymmetry observable as defined can also be more broadly applied to other searches for CP-violation in $2\to 3$ processes in present and future collider environments.

hep-ph↗

arXiv2025

High-Frequency Gravitational Waves on BREAD

The use of light axion dark matter experiments as high-frequency gravitational wave (HFGW) detectors has garnered increasing attention in recent years. We explore the capabilities of the Broadband Reflector Experiment for Axion Detection (BREAD) in probing the GW parameter space and study the directional dependence of its coverage. This detector can investigate frequencies ranging from 0.05 to 200 THz. We find that employing single photon detectors BREAD is sensitive to GWs with characteristic strains as low as $10^{-21}$ at 0.1 THz and $10^{-25}$ at 200 THz with a year exposure time, making it competitive with other proposals operating at similar frequencies.

hep-ph↗

arXiv2025

Heavy-Tailed Mixed p-Spin Spherical Model: Breakdown of Ultrametricity and Failure of the Parisi Formula

We prove that the two cornerstones of mean-field spin glass theory -- the Parisi variational formula and the ultrametric organization of pure states -- break down under heavy-tailed disorder. For the mixed spherical $p$-spin model whose couplings have tail exponent $α<2$, we attach to each $p$ an explicit threshold $H_p^{*}$. If any coupling exceeds its threshold, a single dominant monomial governs both the limiting free energy and the entire Gibbs measure; the resulting energy landscape is intrinsically probabilistic, with a sharp failure of ultrametricity for $p\ge4$ and persistence of only a degenerate 1-RSB structure for $p\le3$. When all couplings remain below their thresholds, the free energy is $O(n^{-1})$ and the overlap is near zero, resulting in a trivial Gibbs geometry. For $α<1$ we further obtain exact fluctuations of order $n^{1-p}$. Our proof introduces Non-Intersecting Monomial Reduction (NIMR), an algebraic-combinatorial technique that blends convexity analysis, extremal combinatorics and concentration on the sphere, providing the first rigorous description of both regimes for heavy-tailed spin glasses with $p\ge3$.

math.PR↗

arXiv2024

Fluctuations of the free energy of the spherical Sherrington-Kirkpatrick model with heavy-tailed interaction

We consider the 2-spin spherical Sherrington--Kirkpatrick model without external magnetic field where the interactions between the spins are given as random variables with heavy-tailed distribution. We show that the free energy exhibits a sharp phase transition depending on the location of the largest eigenvalue of the interaction matrix. We also prove the order of the limiting free energy and the limiting distribution of the fluctuation of the free energy for both regimes.

math.PR↗

arXiv2023

Probing Cosmological Particle Production and Pairwise Hotspots with Deep Neural Networks

Particles with masses much larger than the inflationary Hubble scale, $H_I$, can be pair-produced non-adiabatically during inflation. Due to their large masses, the produced particles modify the curvature perturbation around their locations. These localized perturbations eventually give rise to localized signatures on the Cosmic Microwave Background (CMB), in particular, pairwise hotspots (PHS). In this work, we show that Convolutional Neural Networks (CNN) provide a powerful tool for identifying PHS on the CMB. While for a given hotspot profile a traditional Matched Filter Analysis is known to be optimal, a Neural Network learns to effectively detect the large variety of shapes that can arise in realistic models of particle production. Considering an idealized situation where the dominant background to the PHS signal comes from the standard CMB fluctuations, we show that a CNN can isolate the PHS with $\mathcal{O}(10)\%$ efficiency even if the hotspot temperature is $\mathcal{O}(10)$ times smaller than the average CMB fluctuations. Overall, the CNN search is sensitive to heavy particle masses $M_0/H_I=\mathcal{O}(200)$, and constitutes one of the unique probes of very high energy particle physics.

hep-ph↗

arXiv2022

Monolepton production in SMEFT to $\mathcal O(1/Λ^4)$ and beyond

We calculate $pp \to \ell^{+}ν, \ell^-\bar ν$ to ${\cal{O}}(1/Λ^4)$ within the Standard Model Effective Field Theory (SMEFT) framework. In particular, we calculate the four-fermion contribution from dimension six and eight operators, which dominates at large center of mass energy. We explore the relative size of the $\mathcal O(1/Λ^4)$ and $\mathcal O(1/Λ^2)$ results for various kinematic regimes and assumptions about the Wilson coefficients. Results for Drell-Yan production $pp \to \ell^+\ell^-$ at ${\cal{O}}(1/Λ^4)$ are also provided. Additionally, we develop the form for four fermion contact term contributions to $pp \to \ell^{+}ν, \ell^-\bar ν, pp \to \ell^+\ell^-$ of arbitrary mass dimension. This allows us to estimate the effects from even higher dimensional (dimension $> 8$) terms in the SMEFT framework.

hep-ph↗

arXiv2021

A $W^\pm$ polarization analyzer from Deep Neural Networks

In this paper, we train a Convolutional Neural Network to classify longitudinally and transversely polarized hadronic $W^\pm$ using the images of boosted $W^{\pm}$ jets as input. The images capture angular and energy information from the jet constituents that is faithful to properties of the original quark/anti-quark $W^{\pm}$ decay products without the need for invasive substructure cuts. We find that the difference between the polarizations is too subtle for the network to be used as an event-by-event tagger. However, given an ensemble of $W^{\pm}$ events with unknown polarization, the average network output from that ensemble can be used to extract the longitudinal fraction $f_L$. We test the network on Standard Model $pp \to W^{\pm}Z$ events and on $pp \to W^{\pm}Z$ in the presence of dimension-6 operators that perturb the polarization composition.

hep-ph↗

arXiv2020

Discover Your Social Identity from What You Tweet: a Content Based Approach

An identity denotes the role an individual or a group plays in highly differentiated contemporary societies. In this paper, our goal is to classify Twitter users based on their role identities. We first collect a coarse-grained public figure dataset automatically, then manually label a more fine-grained identity dataset. We propose a hierarchical self-attention neural network for Twitter user role identity classification. Our experiments demonstrate that the proposed model significantly outperforms multiple baselines. We further propose a transfer learning scheme that improves our model's performance by a large margin. Such transfer learning also greatly reduces the need for a large amount of human labeled data.

cs.SI↗

arXiv2017

An upper bound on tricolored ordered sum-free sets

We present a strengthening of the lemma on the lower bound of the slice rank by Tao (2016) motivated by the Croot-Lev-Pach-Ellenberg-Gijswijt bound on cap sets (2017, 2017). The Croot-Lev-Pach-Ellenberg-Gijswijt method and the lemma of Tao are based on the fact that the rank of a diagonal matrix is equal to the number of non-zero diagonal entries. Our lemma is based on the rank of upper-triangular matrices. This stronger lemma allows us to prove the following extension of the Ellenberg-Gijswijt result (2017). A tricolored ordered sum-free set in $\mathbb F_p^n$ is a collection $\{(a_i,b_i,c_i):i=1,2,\ldots,m\}$ of ordered triples in $(\mathbb F_p^n )^3$ such that $a_i+b_i+c_i=0$ and if $a_i+b_j+c_k=0$, then $i\le j\le k$. By using the new lemma, we present an upper bound on the size of a tricolored ordered sum-free set in $\mathbb F_p^n$.

math.CO↗