arXiv subjects

Sungjae Cho

Publications and source records attributed to Sungjae Cho.

18 recordsLinked to original sources

arXiv2025

MedBLINK: Probing Basic Perception in Multimodal Language Models for Medicine

Multimodal language models (MLMs) show promise for clinical decision support and diagnostic reasoning, raising the prospect of end-to-end automated medical image interpretation. However, clinicians are highly selective in adopting AI tools; a model that makes errors on seemingly simple perception tasks such as determining image orientation or identifying whether a CT scan is contrast-enhance are unlikely to be adopted for clinical tasks. We introduce Medblink, a benchmark designed to probe these models for such perceptual abilities. Medblink spans eight clinically meaningful tasks across multiple imaging modalities and anatomical regions, totaling 1,429 multiple-choice questions over 1,605 images. We evaluate 19 state-of-the-art MLMs, including general purpose (GPT4o, Claude 3.5 Sonnet) and domain specific (Med Flamingo, LLaVA Med, RadFM) models. While human annotators achieve 96.4% accuracy, the best-performing model reaches only 65%. These results show that current MLMs frequently fail at routine perceptual checks, suggesting the need to strengthen their visual grounding to support clinical adoption. Data is available on our project page.

cs.AI↗

arXiv2025

Lee and Seung (2000)'s Algorithms for Non-negative Matrix Factorization: A Supplementary Proof Guide

Lee and Seung (2000) introduced numerical solutions for non-negative matrix factorization (NMF) using iterative multiplicative update algorithms. These algorithms have been actively utilized as dimensionality reduction tools for high-dimensional non-negative data and learning algorithms for artificial neural networks. Despite a considerable amount of literature on the applications of the NMF algorithms, detailed explanations about their formulation and derivation are lacking. This report provides supplementary details to help understand the formulation and derivation of the proofs as used in the original paper.

math.NA↗

arXiv2022

Charge-Density-Wave Proximity Effects in Graphene

Certain layered transition metal dichalcogenides (TMDCs), such as 1T-TaS2, show a rich collection of charge density wave (CDW) phases at different temperatures, and their atomic structures and electron conductions have been widely studied. However, the properties of CDW systems that are integrated with other electronic materials have not yet been investigated. Here, we incorporate the CDW properties of TMDCs into the electronic transport of graphene for the first time. During CDW phase transitions, anomalous transport behaviors that are closely related to the formation of correlated disorder in TMDCs were observed in the graphene sample used in this study. In particular, the commensurate CDW phase forms a periodic charge distribution with potential fluctuations, and thus constitutes correlated charged impurities, which decreases resistivity and enhances carrier mobility in graphene. The CDW-graphene heterostructure system demonstrated here paves the way to controlling the temperature-dependent carrier mobility and resistivity of graphene and to developing novel functional electronic devices such as graphene-based sensors and memory devices.

cond-mat.mes-hall↗

arXiv2021

Multi-speaker Emotional Text-to-speech Synthesizer

We present a methodology to train our multi-speaker emotional text-to-speech synthesizer that can express speech for 10 speakers' 7 different emotions. All silences from audio samples are removed prior to learning. This results in fast learning by our model. Curriculum learning is applied to train our model efficiently. Our model is first trained with a large single-speaker neutral dataset, and then trained with neutral speech from all speakers. Finally, our model is trained using datasets of emotional speech from all speakers. In each stage, training samples of each speaker-emotion pair have equal probability to appear in mini-batches. Through this procedure, our model can synthesize speech for all targeted speakers and emotions. Our synthesized audio sets are available on our web page.

cs.CL↗

arXiv2021

Dirac-Source Diode with Sub-unity Ideality Factor

An increase in power consumption necessitates a low-power circuit technology to extend Moore's law. Low-power transistors, such as tunnel field-effect transistors (TFETs), negative-capacitance field-effect transistors (NC-FETs), and Dirac-source field-effect transistors (DS-FETs), have been realised to break the thermionic limit of the subthreshold swing (SS). However, a low-power diode rectifier, which breaks the thermionic limit of an ideality factor (n) of 1 at room temperature, has not been proposed yet. In this study, we have realised a DS Schottky diode, which exhibits a steep-slope characteristic curve, by utilising the linear density of states (DOSs) of graphene. For the developed DS Schottky diode, n<1 for more than two decades of drain current with a minimum value of 0.8, and the rectifying ratio is large (100000). The realisation of a DS Schottky diode paves the way for the development of low-power electronic circuits.

cond-mat.mes-hall↗

arXiv2020

Gate-Tunable Reversible Rashba-Edelstein Effect in a Few-Layer Graphene/2H-TaS2 Heterostructure at Room Temperature

We report the observation of current-induced spin polarization, the Rashba-Edelstein effect (REE), and its Onsager reciprocal phenomenon, the spin galvanic effect (SGE), in a few-layer graphene/2H-TaS2 heterostructure at room temperature. Spin-sensitive electrical measurements unveil full spin-polarization reversal by an applied gate voltage. The observed gate-tunable charge-to-spin conversion is explained by the ideal work function mismatch between 2H-TaS2 and graphene, which allows strong interface-induced Bychkov-Rashba interaction with a spin-gap reaching 70 meV, while keeping the Dirac nature of the spectrum intact across electron and hole sectors. The reversible electrical generation and control of the nonequilibrium spin polarization vector, not previously observed in a nonmagnetic material, are elegant manifestations of emergent 2D Dirac fermions with robust spin-helical structure. Our experimental findings, supported by first-principles relativistic electronic structure and transport calculations, demonstrate a route to design low-power spin-logic circuits from layered materials.

cond-mat.mes-hall↗

arXiv2020

Thickness-Controlled Black Phosphorus Tunnel Field-Effect Transistor for Low Power Switches

The continuous down-scaling of transistors has been the key to the successful development of current information technology. However, with Moore's law reaching its limits, the development of alternative transistor architectures is urgently needed. Transistors require a switching voltage of at least 60 mV for each tenfold increase in current, that is, a subthreshold swing (SS) of 60 mV per decade (dec). Alternative tunnel field-effect transistors (TFETs) are widely studied to achieve a sub-thermionic SS and high I60 (the current where SS becomes 60 mV/dec). Heterojunction (HJ) TFETs show promise for delivering a high I60, but experimental results do not meet theoretical expectations due to interface problems in the HJs constructed from different materials. Here, we report a natural HJ-TFET with spatially varying layer thickness in black phosphorus without interface problems. We achieved record-low average SS values over 4-5 dec of current (SSave_4dec = ~22.9 mV/dec and SSave_5dec = ~26.0 mV/dec) with record-high I60 (I60 = 0.65-1 uA/um), paving the way for application in low-power switches.

physics.app-ph↗

arXiv2019

Emotional Voice Conversion using Multitask Learning with Text-to-speech

Voice conversion (VC) is a task to transform a person's voice to different style while conserving linguistic contents. Previous state-of-the-art on VC is based on sequence-to-sequence (seq2seq) model, which could mislead linguistic information. There was an attempt to overcome it by using textual supervision, it requires explicit alignment which loses the benefit of using seq2seq model. In this paper, a voice converter using multitask learning with text-to-speech (TTS) is presented. The embedding space of seq2seq-based TTS has abundant information on the text. The role of the decoder of TTS is to convert embedding space to speech, which is same to VC. In the proposed model, the whole network is trained to minimize loss of VC and TTS. VC is expected to capture more linguistic information and to preserve training stability by multitask learning. Experiments of VC were performed on a male Korean emotional text-speech dataset, and it is shown that multitask learning is helpful to keep linguistic contents in VC.

eess.AS↗

arXiv2019

Simulating Problem Difficulty in Arithmetic Cognition Through Dynamic Connectionist Models

The present study aims to investigate similarities between how humans and connectionist models experience difficulty in arithmetic problems. Problem difficulty was operationalized by the number of carries involved in solving a given problem. Problem difficulty was measured in humans by response time, and in models by computational steps. The present study found that both humans and connectionist models experience difficulty similarly when solving binary addition and subtraction. Specifically, both agents found difficulty to be strictly increasing with respect to the number of carries. Another notable similarity is that problem difficulty increases more steeply in subtraction than in addition, for both humans and connectionist models. Further investigation on two model hyperparameters --- confidence threshold and hidden dimension --- shows higher confidence thresholds cause the model to take more computational steps to arrive at the correct answer. Likewise, larger hidden dimensions cause the model to take more computational steps to correctly answer arithmetic problems; however, this effect by hidden dimensions is negligible.

cs.NE↗

arXiv2015

Kondo-like zero-bias conductance anomaly in a three-dimensional topological insulator nanowire

Zero bias anomalies in topological nanowires have recently captured significant attention, as they are possible signatures of Majorana modes. Yet there are many other possible origins of zero bias peaks in nanowires, for example, weak localization, Andreev bound states, or the Kondo effect. Here, we discuss observations of differential conductance peaks at zero bias voltage in nonsuperconducting electronic transport through a 3D topological insulator nanowire. The zero bias conductance peaks show logarithmic temperature dependence and often linear splitting with magnetic fields, both of which are signatures of the Kondo effect in quantum dots. Here, we characterize the zero bias peaks and discuss their origin.

cond-mat.mes-hall↗

arXiv2014

Disorder induced magnetoresistance in a two dimensional electron system

We predict and demonstrate that a disorder induced carrier density inhomogeneity causes magnetoresistance (MR) in a two-dimensional electron system. Our experiments on graphene show a quadratic MR persisting far from the charge neutrality point. Effective medium calculations show that for charged impurity disorder, the low-field MR is a universal function of the ratio of carrier density to fluctuations in carrier density, a power-law when this ratio is large, in excellent agreement with experiment. The MR is generic and should occur in other materials with large carrier density inhomogeneity.

cond-mat.mes-hall↗

arXiv2012

Symmetry Protected Josephson Supercurrents in Three-Dimensional Topological Insulators

Coupling the surface state of a topological insulator (TI) to an s-wave superconductor is predicted to produce the long-sought Majorana quasiparticle excitations. However, superconductivity has not been measured in surface states when the bulk charge carriers are fully depleted, i.e., in the true topological regime that is relevant for investigating Majorana modes. Here, we report measurements of DC Josephson effects in TI-superconductor junctions as the chemical potential is moved from the bulk bands into the band gap, or through the true topological regime characterized by the presence of only surface currents. We examine the relative behavior of the system at different bulk/surface ratios, determining the effects of strong bulk/surface mixing, disorder, and magnetic field. We compare our results to 3D quantum transport simulations to conclude that the supercurrent is largely carried by surface states, due to the inherent topology of the bands, and that it is robust against disorder.

cond-mat.mes-hall↗

arXiv2012

Surface conduction of topological Dirac electrons in bulk insulating Bi2Se3

The newly-discovered three-dimensional strong topological insulators (STIs) exhibit topologically-protected Dirac surface states. While the STI surface state has been studied spectroscopically by e.g. photoemission and scanned probes, transport experiments have failed to demonstrate the most fundamental signature of the STI: ambipolar metallic electronic transport in the topological surface of an insulating bulk. Here we show that the surfaces of thin (<10 nm), low-doped Bi2Se3 (\approx10^17/cm3) crystals are strongly electrostatically coupled, and a gate electrode can completely remove bulk charge carriers and bring both surfaces through the Dirac point simultaneously. We observe clear surface band conduction with linear Hall resistivity and well-defined ambipolar field effect, as well as a charge-inhomogeneous minimum conductivity region. A theory of charge disorder in a Dirac band explains well both the magnitude and the variation with disorder strength of the minimum conductivity (2 to 5 e^2/h per surface) and the residual (puddle) carrier density (0.4 x 10^12 cm^-2 to 4 x 10^12 cm^-2). From the measured carrier mobilities 320 cm^2/Vs to 1,500 cm^2/Vs, the charged impurity densities 0.5 x 10^13 cm^-2 to 2.3 x 10^13 cm^-2 are inferred. They are of a similar magnitude to the measured doping levels at zero gate voltage (1 x 10^13 cm^-2 to 3 x 10^13 cm^-2), identifying dopants as the charged impurities.

cond-mat.mes-hall↗

arXiv2012

Insulating behavior in ultra-thin bismuth selenide field effect transistors

Ultrathin (~3 quintuple layer) field-effect transistors (FETs) of topological insulator Bi2Se3 are prepared by mechanical exfoliation on 300nm SiO2/Si susbtrates. Temperature- and gate-voltage dependent conductance measurements show that ultrathin Bi2Se3 FETs are n-type, and have a clear OFF state at negative gate voltage, with activated temperature-dependent conductance and energy barriers up to 250 meV.

cond-mat.mes-hall↗

arXiv2012

Topological insulator quantum dot with tunable barriers

Thin (6-7 quintuple layer) topological insulator Bi2Se3 quantum dot devices are demonstrated using ultrathin (2~4 quintuple layer) Bi2Se3 regions to realize semiconducting barriers which may be tuned from Ohmic to tunneling conduction via gate voltage. Transport spectroscopy shows Coulomb blockade with large charging energy >5 meV, with additional features implying excited states.

cond-mat.mes-hall↗

arXiv2009

Massless and massive particle-in-a-box states in single-and bi-layer graphene

Electron transport through short, phase-coherent metal-graphene-metal devices occurs via resonant transmission through particle-in-a-box-like states defined by the atomically-sharp metal leads. we study the spectrum of particle-in-a-box states for single- and bi-layer graphene, corresponding to massless and massive two-dimensional (2d) Fermions.The density of states D as a function of particle number n shows the expected relations D(n) ~ n1/2 for massless 2d Fermions (electrons in single-layer graphene) and D(n) ~ constant for massive 2d Fermions (electrons in bi-layer graphene). The single parameters of the massless and massive dispersion relations are found, Fermi velocity vF = 1.1 x 106m/s and effective mass m* = 0.032 me, where me is the electron mass, in excellent agreement with theoretical expectations.

cond-mat.mes-hall↗

arXiv2007

Graphene Spin Transistor

Graphitic nanostructures, e.g. carbon nanotubes (CNT) and graphene, have been proposed as ideal materials for spin conduction[1-7]; they have long electronic mean free paths[8] and small spin-orbit coupling[9], hence are expected to have very long spin-scattering times. In addition, spin injection and detection in graphene opens new opportunities to study exotic electronic states such as the quantum Hall[10,11] and quantum spin Hall[9] states, and spin-polarized edge states[12] in graphene ribbons. Here we perform the first non-local four-probe experiments[13] on graphene contacted by ferromagnetic Permalloy electrodes. We observe sharp switching and often sign-reversal of the non-local resistance at the coercive field of the electrodes, indicating definitively the presence of a spin current between injector and detector. The non-local resistance changes magnitude and sign quasi-periodically with back-gate voltage, and Fabry-Perot-like oscillations[6,14,15] are observed, consistent with quantum-coherent transport. The non-local resistance signal can be observed up to at least T = 300 K.

cond-mat.mes-hall↗

arXiv2007

Charge Transport and Inhomogeneity near the Charge Neutrality Point in Graphene

The magnetic field-dependent longitudinal and Hall components of the resistivity rho_xx(H) and rho_xy(H) are measured in graphene on silicon dioxide substrates at temperatures from 1.6 K to room temperature. At charge densities near the charge-neutrality point rho_xx(H) is strongly enhanced and rho_xy(H) is suppressed, indicating nearly equal electron and hole contributions to the transport current. The data are inconsistent with uniformly distributed electron and hole concentrations (two-fluid model) but in excellent agreement with the recent theoretical prediction for inhomogeneously distributed electron and hole regions of equal mobility. At low temperatures and high magnetic fields rho_xx(H) saturates to a value ~h/e^2, with Hall conductivity << e^2/h, which may indicate a regime of localized v = 2 and v = -2 quantum Hall puddles.

cond-mat.mes-hall↗