SearcharxivSearch

arXiv subjects

Sunyoung Shin

Publications and source records attributed to Sunyoung Shin.

At least 19 recordsLinked to original sources

Are Video Models Zero-Shot Learners and Reasoners in Education? EduVideoBench, A Knowledge-Skills-Attitude Benchmark for Educational Video Generation

Video generation models (VGMs) are rapidly entering classrooms, yet existing benchmarks evaluate only perceptual quality, intrinsic faithfulness, generic safety, or video as a reasoning medium, and none assesses whether the outputs are educationally valid. In this work, we present EduVideoBench, the first balanced benchmark in the education domain, grounded in the Knowledge-Skills-Attitude (KSA) framework so that pedagogical adequacy and educational safety are evaluated jointly rather than as ad-hoc quality dimensions. Across five frontier VGMs, our results show substantial room for improvement across knowledge, skills, and attitude before they are classroom-ready. We complement this with a qualitative analysis of expert comments, finding that educational validity is multi-component, where a single misaligned element such as pacing, legibility, or notation can invalidate an otherwise correct video. We hope EduVideoBench will guide the development of VGMs that are pedagogically grounded and safe for the classroom.

cs.CL

Soohak: A Mathematician-Curated Benchmark for Evaluating Research-level Math Capabilities of LLMs

Following the recent achievement of gold-medal performance on the IMO by frontier LLMs, the community is searching for the next meaningful and challenging target for measuring LLM reasoning. Whereas olympiad-style problems measure step-by-step reasoning alone, research-level problems use such reasoning to advance the frontier of mathematical knowledge itself, emerging as a compelling alternative. Yet research-level math benchmarks remain scarce because such problems are difficult to source (e.g., Riemann Bench and FrontierMath-Tier 4 contain 25 and 50 problems, respectively). To support reliable evaluation of next-generation frontier models, we introduce Soohak, a 439-problem benchmark newly authored from scratch by 64 mathematicians. Soohak comprises two subsets. On the Challenge subset, frontier models including Gemini-3-Pro, GPT-5, and Claude-Opus-4.5 reach 30.4%, 26.4%, and 10.4% respectively, leaving substantial headroom, while leading open-weight models such as Qwen3-235B, GPT-OSS-120B, and Kimi-2.5 remain below 15%. Notably, beyond standard problem solving, Soohak introduces a refusal subset that probes a capability intrinsic to research mathematics: recognizing ill-posed problems and pausing rather than producing confident but unjustified answers. On this subset, no model exceeds 50%, identifying refusal as a new optimization target that current models do not directly address. To prevent contamination, the dataset will be publicly released in late 2026, with model evaluations available upon request in the interim.

cs.CL

Pedagogical Alignment for Vision-Language-Action Models: A Comprehensive Framework for Data, Architecture, and Evaluation in Education

Science demonstrations are important for effective STEM education, yet teachers face challenges in conducting them safely and consistently across multiple occasions, where robotics can be helpful. However, current Vision-Language-Action (VLA) models require substantial computational resources and sacrifice language generation capabilities to maximize efficiency, making them unsuitable for resource-constrained educational settings that require interpretable, explanation-generating systems. We present \textit{Pedagogical VLA Framework}, a framework that applies pedagogical alignment to lightweight VLA models through four components: text healing to restore language generation capabilities, large language model (LLM) distillation to transfer pedagogical knowledge, safety training for educational environments, and pedagogical evaluation adjusted to science education contexts. We evaluate Pedagogical VLA Framework across five science demonstrations spanning physics, chemistry, biology, and earth science, using an evaluation framework developed in collaboration with science education experts. Our evaluation assesses both task performance (success rate, protocol compliance, efficiency, safety) and pedagogical quality through teacher surveys and LLM-as-Judge assessment. We additionally provide qualitative analysis of generated texts. Experimental results demonstrate that Pedagogical VLA Framework achieves comparable task performance to baseline models while producing contextually appropriate educational explanations.

cs.CL

Junctions of mass-deformed nonlinear sigma models on $SO(2N)/U(N)$ and $Sp(N)/U(N)$ II

We study vacua, walls and three-pronged junctions of mass-deformed nonlinear sigma models on $SO(2N)/U(N)$ and $Sp(N)/U(N)$ for generic $N$. We review and discuss the on-shell component Lagrangians of the ${\mathcal{N}}=2$ nonlinear sigma model on the Grassmann manifold, which are obtained in the ${\mathcal{N}}=1$ superspace formalism and in the harmonic superspace formalism. We also show that the K\"{a}hler potential of the ${\mathcal{N}}=2$ nonlinear sigma model on the complex projective space, which is obtained in the projective superspace formalism, is equivalent to the K\"{a}hler potential of the ${\mathcal{N}}=2$ nonlinear sigma model with the Fayet-Iliopoulos parameters $c^a=(0,0,c=1)$ on the complex projective space, which is obtained in the ${\mathcal{N}}=1$ superspace formalism.

hep-th

Junctions of mass-deformed nonlinear sigma models on $SO(2N)/U(N)$ and $Sp(N)/U(N)$ I

We construct on-shell ${\mathcal{N}}=2$ nonlinear sigma models on $SO(2N)/U(N)$ and $Sp(N)/U(N)$ by holomorphically embedding the models in the hyper-K\"{a}hler nonlinear sigma model on the cotangent bundle of the Grassmann manifold $T^\ast G_{2N,N}$ in the ${\mathcal{N}}=1$ superspace formalism. We apply the moduli matrix formalism to the mass-deformed nonlinear sigma models on the quadrics to study three-pronged junctions by using a recently proposed diagram method.

hep-th

Vacua, walls and junctions in $G_{N_F,N_C}$

We discuss vacua, walls and three-pronged junctions of the mass-deformed nonlinear sigma models on the Grassmann manifold $G_{N_F,N_C}=\frac{SU(N_F)}{SU(N_C)\times SU(N_F-N_C)\times U(1)}$, which are non-Abelian gauge theories for $N_C\geq 2$. Polyhedra are proposed in \cite{Eto:2005cp} to describe Bogomol'nyi-Prasad-Sommerfield objects of the mass-deformed nonlinear sigma models on the complex projective space, which are Abelian gauge theories. We show that we can produce similar polyhedra for the mass-deformed nonlinear sigma models on the Grassmann manifold by applying the moduli matrix formalism \cite{Isozumi:2004jc} and the pictorial representation \cite{Lee:2017kaj}. Non-Abelian junctions can be analysed by making use of the polyhedra instead of the Pl\"{u}cker embedding. We present diagrams for vacua, walls and three-pronged junctions, and compute three-pronged junction positions of the mass-deformed nonlinear sigma models on the Grassmann manifold. We show that the results are consistent with the known results of \cite{Eto:2005fm}, which are worked out by using the Pl\"{u}cker embedding.

hep-th

On Gauge Invariant Cosmological Perturbations in UV-modified Horava Gravity

We consider gauge invariant cosmological perturbations in UV-modified, z=3 Horava gravity with one scalar matter field, which has been proposed as a renormalizable gravity theory without the ghost problem in four dimensions. In order to exhibit its dynamical degrees of freedom, we consider the Hamiltonian reduction method and find that, by solving "all" the constraint equations, the degrees of freedom are the same as those of Einstein gravity: One scalar and two tensor (graviton) modes when a scalar matter field presents. However, we confirm that there is no extra graviton modes and general relativity is recovered in IR, which achieves the consistency of the model. From the UV-modification terms which break the detailed balance condition in UV, we obtain scale-invariant power spectrums for "non"-inflationary backgrounds, like the power-law expansions, without knowing the details of early expansion history of Universe. This could provide a new framework for the Big Bang cosmology. Moreover, we find that "tensor and scalar fluctuations travel differently in UV, generally". We present also some clarifying remarks about confusing points in the literatures.

hep-th

Generalized complex marginal deformation of pp-waves and giant gravitons

We present the Penrose limits of a complex marginal deformation of $AdS_5\times S^5$, which incorporates the $SL(2,\mathbb{R})$ symmetry of type IIB theory, along the $(J,0,0)$ geodesic and along the $(J,J,J)$ geodesic. We discuss giant gravitons on the deformed $(J,0,0)$ pp-wave background.

hep-th

Soft Wall Model in the Hadronic Medium

We study the holographic QCD in the hadronic medium by using the soft wall model. We discuss the Hawking-Page transition between Reissner-Nordström AdS black hole and thermal charged AdS of which the geometries correspond to deconfinement and confinement phases respectively. We also present the numerical result of the vector and axial vector meson spectra depending on the quark density.

hep-th

$β$-deformation on a slanted torus and deformed pp-wave

We discuss the $β$-deformation of $AdS_5\times S^5$ which incorporates the $SL(2,\mathbb{R})$ symmetry of the type IIB theory. The axion-dilaton is identified with a two-torus from an eleven dimensional viewpoint. We consider the null geodesic with equal component angular momenta to take the Penrose limit of the deformed $AdS_5\times S^5$. We study the bosonic part of the string sigma model and the spectrum of the string in the pp-wave background.

hep-th

$T^3$ deformations and $β$-deformed geometries

We discuss $β$-deformed geometries on two types of $T^3$'s where the direction along the third coordinate is not orthogonal to the direction along the second coordinate or the direction along the first coordinate. We show that the intersection angle between the direction along the third coordinate and the direction along the second coordinate corresponds to the parameter of the S-duality of the $β$-deformation while the intersection angle between the direction along the third coordinate and the direction along the first coordinate generalizes the $β$-deformed geometry.

hep-th

Holographic Meson Spectra in the Dense Medium with Chiral Condensate

We study two $1/N_c$ effects on the meson spectra by using the AdS/CFT correspondence where the $1/N_c$ corrections from the chiral condensate and the quark density are controlled by the gravitational backreaction of the massive scalar field and U(1) gauge field respectively. The dual geometries with zero and nonzero current quark masses are obtained numerically. We discuss meson spectra and binding energy of heavy quarkonium with the subleading corrections in the hard wall model.

hep-th

Self-trapping nature of Tl nanoclusters on Si(111)-7$\times$7 surface

We have investigated electronic and structural properties of thallium (Tl) nanoclusters formed on the Si(111)-7$\times$7 surface at room temperature (RT) by utilizing photoemission spectroscopy (PES) and high-resolution electron-energy-loss spectroscopy (HREELS) combined with first principles calculations. Our PES data show that the state S2 stemming from Si restatoms remains quite inert with Tl coverage $θ$ while S1 from Si adatoms gradually changes, in sharp contrast with the rapidly decaying states of Na or Li nanoclusters. No Tl-induced surface state is observed until $θ$=0.21 ML where Tl nanoclusters completely cover the faulted half unit cells (FHUCs) of the surface. These spectral behaviors of surface states and a unique loss peak L$_2$ associated with Tl in HREELS spectra indicate no strong Si-Tl bonding and are well understood in terms of gradual filling of Si dangling bonds with increasing $θ$. Our calculational results further reveal that there are several metastable atomic structures for Tl nanoclusters at RT transforming from each other faster than 10$^{10}$ flippings per second. We thus conclude that the highly mobile Tl atoms form self-trapped nanoclusters within FHUC at RT with several metastable phases. The mobile and multi-phased nature of Tl nanoclusters not only account for all the existing experimental observations including the fuzzy scanning tunneling microscope images and a dynamical model proposed by recent x-ray study but also provides an example of self-trapping of atoms in a nanometer-scale region.

cond-mat.mes-hall

The stability of graphene band structures against an external periodic perturbation; Na on Graphene

We report that the $π$ band of graphene sensitively changes as a function of an external potential induced by Na especially when the potential becomes periodic at low temperature. We have measured the band structures from the graphene layers formed on the 6H-SiC(0001) substrate using angle-resolved photoemission spectroscopy with synchrotron photons. With increasing Na dose, the $π$ band appears to be quickly diffused into background at 85 K whereas it becomes significantly enhanced its spectral intensity at room temperature (RT). A new parabolic band centered at $k\sim$1.15 Å$^{-1}$ also forms near Fermi energy with Na at 85 K while no such a band observed at RT. Such changes in the band structure are found to be reversible with temperature. Analysis based on our first principles calculations suggests that the changes of the $π$ band of graphene be mainly driven by the Na-induced potential especially at low temperature where the potential becomes periodic due to the crystallized Na overlayer. The new parabolic band turns to be the $π$ band of the underlying buffer layer partially filled by the charge transfer from Na adatoms. The five orders of magnitude increased hopping rate of Na adatoms at RT preventing such a charge transfer explains the absence of the new band at RT.

cond-mat.mes-hall

Walls of massive Kähler sigma models on SO(2N)/U(N) and Sp(N)/U(N)

We study the Bogomol'nyi-Prasad-Sommerfield wall solutions in massive Kähler nonlinear sigma models on SO(2N)/U(N) and Sp(N)/U(N) in three-dimensional spacetime. We show that SO(2N)/U(N) and Sp(N)/U(N) models have 2^{N-1} and 2^N discrete vacua, respectively. We explicitly construct the exact BPS multiwall solutions for N\le 3.

hep-th