arXiv subjects

K. Wong

Publications and source records attributed to K. Wong.

6 recordsLinked to original sources

arXiv2025

Rethinking the Potential of Multimodality in Collaborative Problem Solving Diagnosis with Large Language Models

Detecting collaborative and problem-solving behaviours from digital traces to interpret students' collaborative problem solving (CPS) competency is a long-term goal in the Artificial Intelligence in Education (AIEd) field. Although multimodal data and advanced models are argued to have the potential to detect complex CPS behaviours, empirical evidence on their value remains limited with some contrasting evidence. In this study, we investigated the potential of multimodal data to improve model performance in diagnosing 78 secondary school students' CPS subskills and indicators in authentic educational settings. In particular, text embeddings from verbal data and acoustic embeddings from audio data were used in a multimodal classification model for CPS diagnosis. Both unimodal and multimodal transformer-based models outperformed traditional models in detecting CPS classes. Although the inclusion of multimodality did not improve the performance of traditional unimodal models, its integration into transformer-based models demonstrated improved performance for diagnosing social-cognitive CPS classes compared to unimodal transformer-based models. Based on the results, the paper argues that multimodality and the selection of a particular modelling technique should not be taken for granted to achieve the best performance in the automated detection of every CPS subskill and indicator. Rather, their value is limited to certain types of CPS indicators, affected by the complexity of the labels, and dependent on the composition of indicators in the dataset. We conclude the paper by discussing the required nuance when considering the value of LLMs and multimodality in automated CPS diagnosis, highlighting the need for human-AI complementarity, and proposing the exploration of relevant model architectures and techniques to improve CPS diagnosis in authentic educational contexts.

cs.CL↗

arXiv2017

The mystery of the 'Kite' radio source in Abell 2626: insights from new Chandra observations

We present the results of a new Chandra study of the galaxy cluster A2626. The radio emission of the cluster shows a complex system of four symmetric arcs without known correlations with the X-ray emission. The mirror symmetry of the radio arcs toward the center and the presence of two optical cores in the central galaxy suggested that they may be created by pairs of precessing radio jets powered by dual AGNs inside the cD galaxy. However, previous observations failed to observe the second jetted AGN and the spectral trend due to radiative age along the radio arcs, thus challenging this interpretation. The new Chandra observation had several scientific objectives, including the search for the second AGN that would support the jet precession model. We focus here on the detailed study of the local properties of the thermal and non-thermal emission in the proximity of the radio arcs, in order to get more insights into their origin. We performed a standard data reduction of the Chandra dataset deriving the radial profiles of temperature, density, pressure and cooling time of the intra-cluster medium. We further analyzed the 2D distribution of the gas temperature, discovering that the south-western junction of the radio arcs surrounds the cool core of the cluster. We studied the X-ray SB and spectral profiles across the junction, finding a cold front spatially coincident with the radio arcs. This may suggest a connection between the sloshing of the thermal gas and the nature of the radio filaments, raising new scenarios for their origin. A possibility is that the radio arcs trace the projection of a complex surface connecting the sites where electrons are most efficiently reaccelerated by the turbulence that is generated by the gas sloshing. In this case, diffuse emission embedded by the arcs and with extremely steep spectrum should be most visible at very low radio frequencies.

astro-ph.GA↗

arXiv2017

Delayed Slater determinant update algorithms for high efficiency quantum Monte Carlo

Within ab initio Quantum Monte Carlo simulations, the leading numerical cost for large systems is the computation of the values of the Slater determinants in the trial wavefunction. Each Monte Carlo step requires finding the determinant of a dense matrix. This is most commonly iteratively evaluated using a rank-1 Sherman-Morrison updating scheme to avoid repeated explicit calculation of the inverse. The overall computational cost is therefore formally cubic in the number of electrons or matrix size. To improve the numerical efficiency of this procedure, we propose a novel multiple rank delayed update scheme. This strategy enables probability evaluation with application of accepted moves to the matrices delayed until after a predetermined number of moves, K. The accepted events are then applied to the matrices en bloc with enhanced arithmetic intensity and computational efficiency via matrix-matrix operations instead of matrix-vector operations. This procedure does not change the underlying Monte Carlo sampling or its statistical efficiency. For calculations on large systems and algorithms such as diffusion Monte Carlo where the acceptance ratio is high, order of magnitude improvements in the update time can be obtained on both multi-core CPUs and GPUs.

physics.comp-ph↗

arXiv2013

An Analysis of the Rapidly Rotating Bp Star HD 133880

HD 133880 is a rapidly rotating Bp star and host to one of the strongest magnetic fields of any Bp star. A member of the Upper Centaurus Lupus association, it is a star with a well-determined age of 16 Myr. Twelve new spectra obtained from the FEROS, ESPaDOnS, and HARPS instruments, provide sufficient material from which to re-evaluate the magnetic field and obtain a first approximation to the atmospheric abundance distributions of various elements. Using the new magnetic field measurements and optical photometry together with previously published data, we refine the period of HD 133880 to P = 0.877476 \pm 0.000009 days. The magnetic field structure was characterised by a colinear multipole expansion from the observed variations of the longitudinal and surface fields with rotational phase. This simple axisymmetric magnetic field model is based on a predominantly quadrupolar component that roughly describes the field variations. Using spectrum synthesis, we derived mean abundances for O, Mg, Si, Ti, Cr, Fe, Nd and Pr. All elements, except Mg (which has a uniform distribution), are overabundant compared to the Sun and are more abundant in the negative than in the positive magnetic hemisphere. In contrast to most Bp stars the abundance of O in HD 133880 is overabundant compared to the solar abundance ratio. In studying the Halpha and Paschen lines in the optical spectra we could not unambiguously detect information about the magnetosphere of HD 133880. However, radio emission data at both 3 and 6 cm suggests that the magnetospheric plasma is held in rigid rotation with the star by the magnetic field and further supported against collapse by the rapid rotation. Subtle differences in the shapes of the optically thick radio light curves at 3 and 6 cm suggest that the large-scale magnetic field is not fully axisymmetric at large distances from the star.

astro-ph.SR↗

arXiv2006

Charm physics with highly improved staggered quarks

We use a relativistic highly improved staggered quark action to discretize charm quarks on the lattice. We calculate the masses and the dispersion relation for heavy-heavy and heavy-light meson states, and show that for lattice spacings below .1 fm, the discretization errors are at the few percent level. We also discuss the prospects for accurate calculations at the few percent level of f_D_s, f_D, and the leptonic width of the psi and phi.

hep-lat↗

arXiv2006

Highly Improved Staggered Quarks on the Lattice, with Applications to Charm Physics

We use perturbative Symanzik improvement to create a new staggered-quark action (HISQ) that has greatly reduced one-loop taste-exchange errors, no tree-level order a^2 errors, and no tree-level order (am)^4 errors to leading order in the quark's velocity v/c. We demonstrate with simulations that the resulting action has taste-exchange interactions that are at least 3--4 times smaller than the widely used ASQTAD action. We show how to estimate errors due to taste exchange by comparing ASQTAD and HISQ simulations, and demonstrate with simulations that such errors are no more than 1% when HISQ is used for light quarks at lattice spacings of 1/10 fm or less. The suppression of (am)^4 errors also makes HISQ the most accurate discretization currently available for simulating c quarks. We demonstrate this in a new analysis of the psi-eta_c mass splitting using the HISQ action on lattices where a m_c=0.43 and 0.66, with full-QCD gluon configurations (from MILC). We obtain a result of~111(5) MeV which compares well with experiment. We discuss applications of this formalism to D physics and present our first high-precision results for D_s mesons.

hep-lat↗