arXiv subjects

Suyoung Kim

Publications and source records attributed to Suyoung Kim.

17 recordsLinked to original sources

arXiv2026

Magnetic properties of a quasi-two-dimensional spin-1/2 antiferromagnet Y2CuGe4O12

Competing magnetic interactions and frustration-induced quantum fluctuations in spatially anisotropic low-dimensional magnets often give rise to exotic magnetic phenomena, including field-induced phases. Here, we present crystal structure, magnetic susceptibility, specific heat, and electron spin resonance (ESR) measurements on polycrystalline Y$2$CuGe$4$O${12}$, supported by density functional theory (DFT) calculations. In this compound, the Cu$^{2+}$ ions form a distorted triangular lattice with competing intraplanar ferromagnetic ($J_1 \approx 0.138$ K and $J_2 \approx 0.01$ K) and antiferromagnetic ($J_3 \approx -3.22$ K) exchange interactions, together with a weaker interplanar antiferromagnetic coupling ($J_4 \approx -1.56$ K). These interactions account for the small Curie--Weiss temperature, $θ{\rm CW}=-1.8$ K. Despite the dominant antiferromagnetic interactions, no signature of long-range magnetic ordering is observed down to 0.4 K. Instead, broad maxima in both the magnetic susceptibility and magnetic specific heat reveal the development of short-range spin correlations, further supported by the critical ESR linewidth broadening characteristic of low-dimensional frustrated magnets. Application of an external magnetic field progressively suppresses the broad maximum in the magnetic specific heat, reflecting competition between the Zeeman and exchange energy scales, and drives the system into a field-polarized state above the saturation field, $μ_0H_{\rm s}=2.6$ T. In this regime, the magnetic specific heat exhibits an exponential temperature dependence, consistent with gapped magnon excitations. These results establish Y$_2$CuGe$4$O${12}$ as a rare distorted triangular-lattice magnet in which further-neighbor exchange interactions dominate the magnetic behavior, providing a promising platform for exploring frustration-driven quantum phenomena.

cond-mat.str-el↗

arXiv2026

Zero-Shot Quantization for Object Detectors using Off-the-Shelf Generative Models

With an increasing number of Object Detection (OD) models being deployed on edge devices, Zero-Shot Quantization for OD (ZSQ-OD) aims to quantize these models when access to the original training data is prohibited. Existing research on Zero-Shot Quantization-Aware Training (QAT) for OD synthesizes training sets through noise optimization. However, this approach struggles to maintain performance in low-bit regions. In this paper, we introduce GoodQ (Generative off-the-shelf models for object detector Quantization), a QAT pipeline that utilizes off-the-shelf generative models to construct a training set. We first identify three challenges that arise when introducing a generative model to the ZSQ-OD task: 1) each image contains dense information with multiple instances, 2) the class-wise distribution in the original dataset is imbalanced, and 3) the pseudo-labels assigned to the generated images can potentially act as noisy signals during QAT. GoodQ addresses these challenges by 1) introducing an Information-Dense Prompting strategy to generate multi-instance images, 2) applying Intrinsic Distribution-Aware Selection to match the pretrained class distribution, and 3) employing Teacher-guided Adaptive Noise Reduction to mitigate noise arising from the QAT process. Our framework achieves state-of-the-art performance in low-bit ZSQ (W4A4) and extends quantization to extreme bit-widths (W3A3). Furthermore, we conduct an extensive analysis to uncover the underlying factors contributing to the efficacy of GoodQ.

cs.LG↗

arXiv2026

ReSpinQuant: Efficient Layer-Wise LLM Quantization via Subspace Residual Rotation Approximation

Rotation-based Post-Training Quantization (PTQ) has emerged as a promising solution for mitigating activation outliers in the quantization of Large Language Models (LLMs). Global rotation methods achieve inference efficiency by fusing activation rotations into attention and FFN blocks, but suffer from limited expressivity as they are constrained to use a single learnable rotation matrix across all layers. To tackle this, layer-wise transformation methods emerged, achieving superior accuracy through localized adaptation. However, layer-wise methods cannot fuse activation rotation matrices into weights, requiring online computations and causing significant overhead. In this paper, we propose ReSpinQuant, a quantization framework that resolves such overhead by leveraging offline activation rotation fusion and matching basis using efficient residual subspace rotation. This design reconciles the high expressivity of layer-wise adaptation with only negligible inference overhead. Extensive experiments on W4A4 and W3A3 quantization demonstrate that ReSpinQuant achieves state-of-the-art performance, outperforming global rotation methods and matching the accuracy of computationally expensive layer-wise methods with minimal overhead.

cs.CV↗

arXiv2026

Safety-Preserving PTQ via Contrastive Alignment Loss

Post-Training Quantization (PTQ) has become the de-facto standard for efficient LLM deployment, yet its optimization objective remains fundamentally incomplete. Standard PTQ methods minimize reconstruction error (e.g., MSE or KL divergence) without accounting for behavioral alignment--a critical property instilled through safety fine-tuning. We demonstrate that this objective mismatch introduces a systematic vulnerability: models can maintain low perplexity yet exhibit significant degradation in safety alignment, revealing that perplexity alone is an insufficient and often misleading proxy for deployment readiness. To address this, we propose Contrastive Alignment Quantization (CAQ), which extends the PTQ objective design space by integrating a Contrastive Alignment Loss (CAL). CAL introduces a principled push-pull mechanism that jointly optimizes distributional fidelity and behavioral alignment: it steers the quantized model toward its safe, instruction-tuned counterpart while diverging from the unaligned, pre-trained reference. CAQ requires no specialized safety datasets, relying solely on standard calibration data, and introduces negligible computational overhead over existing transformation-based PTQ pipelines. We show that CAQ enables robust 4-bit (W4A4) quantization across diverse model families--including LLaMA, Qwen, and Mistral--achieving superior safety alignment where state-of-the-art PTQ methods fail, without sacrificing general capabilities. Anonymized code is available in the supplementary material.

cs.AI↗

arXiv2026

Magnetic ground state of a Jeff = 1/2 based frustrated triangular lattice antiferromagnet

The subtle interplay between competing degrees of freedom, crystal electric fields, and spin correlations can lead to exotic quantum states in 4f ion-based frustrated triangular lattice antiferromagnets. We present the crystal structure, thermodynamic and muon spin relaxation (μSR) studies of the 4f ion-based frustrated magnet Ba4YbReWO12, wherein Yb3+ ions constitute a triangular lattice. The magnetic susceptibility does not show any signature of spin freezing down to 1.9 K or long-range magnetic ordering down to 0.4 K. The low-temperature Curie-Weiss fit to the inverse magnetic susceptibility data reveals a weak antiferromagnetic exchange interaction, which is corroborated by the fit of magnetic specific heat data following the J1-J2 model with the nearest neighbor exchange interaction of J1 = -0.197 K between the Jeff = 1/2 states of the Yb3+ moments in the lowest Kramers doublet. The lowest Kramers ground state doublet is well separated from the first excited state with a gap of 278 K, as evidenced by our μSR experiments that support the realization of Jeff = 1/2 at low temperatures. The specific heat experiments do not detect a phase transition down to 56 mK. The magnetic specific heat shows a broad maximum 90 mK suggesting a disordered ground state with short range spin correlations. The associated magnetic entropy release at low temperatures is consistent with that expected for the Jeff = 1/2 state. The zero-field μSR measurements show neither the signature of spin freezing nor a phase transition, at least down to 43 mK. Our results suggest a dynamic, disordered ground state in this Jeff = 1/2 frustrated triangular lattice antiferromagnet. Ba4RReWO12 (R=rare earth) offers a viable platform to realize intriguing quantum states borne out of spin-orbit coupling and frustration

cond-mat.str-el↗

arXiv2025

The Role of Teacher Calibration in Knowledge Distillation

Knowledge Distillation (KD) has emerged as an effective model compression technique in deep learning, enabling the transfer of knowledge from a large teacher model to a compact student model. While KD has demonstrated significant success, it is not yet fully understood which factors contribute to improving the student's performance. In this paper, we reveal a strong correlation between the teacher's calibration error and the student's accuracy. Therefore, we claim that the calibration of the teacher model is an important factor for effective KD. Furthermore, we demonstrate that the performance of KD can be improved by simply employing a calibration method that reduces the teacher's calibration error. Our algorithm is versatile, demonstrating effectiveness across various tasks from classification to detection. Moreover, it can be easily integrated with existing state-of-the-art methods, consistently achieving superior performance.

cs.LG↗

arXiv2025

Coexistence of anomalous spin dynamics and weak magnetic order in a chiral trillium lattice K2FeSn(PO4)3

Trillium lattices, where magnetic ions form a three-dimensional chiral network of corner-sharing equilateral triangular motifs, offer a prominent platform to explore exotic quantum states. In this work, we report ground-state properties of the $S$ = 5/2 trillium lattice compound K$_{2}$FeSn(PO$_{4}$)$_{3}$ through thermodynamic, electron spin resonance (ESR), and muon spin relaxation ($μ$SR) experiments. Thermodynamic and ESR measurements reveal the two-step evolution of magnetic correlations across $T^{*}$ = 11 K, which results from an interplay between dominant antiferromagnetic Heisenberg interactions and subleading interactions. Below $T^{*}$, \textit{dc} and \textit{ac} magnetic susceptibilities indicate weak \textcolor{black}{magnetic ordering} at $T_{\rm N} \approx 2$ K under low fields, which is suppressed for $μ_{0}H \geq 2$ T, consistent with a power-law dependence of magnetic specific heat at low temperatures. $μ$SR experiments confirm the dominance of persistent spin dynamics and the absence of conventional spin freezing, supporting the subtle nature of weak magnetic ordering coexisting with spin-liquid-like fluctuations. These findings underscore the potential for realizing a classical spin-liquid ground state with exotic excitations in high-spin trillium lattice systems.

cond-mat.str-el↗

arXiv2025

Spin dynamics and 1/3 magnetization plateau in a coupled distorted diamond chain compound K2Cu3(MoO4)4

We investigate magnetic properties of the $s$ = 1/2 compound K$_{2}$Cu$_{3}$(MoO$_{4}$)$_{4}$ by combining magnetic susceptibility, magnetization, specific heat, and electron spin resonance (ESR) with density functional calculations. Its monoclinic structure features alternating Cu$^{2+}$ ($s$ = 1/2) monomers and edge-shared dimers linked by MoO$_{4}$ units, forming a distorted diamond chain along the $a$-axis. Antiferromagnetic order occurs at $T_{\rm N}$ = 2.3 K, as evident from a $λ$-type anomaly in specific heat and magnetic susceptibility derivatives. Inverse magnetic susceptibility reveals coexisting ferro- and antiferromagnetic interactions. Specific heat and ESR data show two characteristic temperatures: one at 20 K, associated with spin-singlet formation in Cu$_{2}$O$_{9}$ dimers, and another at 3.68 K, indicating short-range correlations between dimers and monomers. Magnetization measurements reveal a metamagnetic transition at 2.6 T and a critical magnetic field $μ_{0}H_{c}$ = 3.4 T, where a 1/3 magnetization plateau emerges with saturation near 0.35 $μ_{\rm B}$. Low-temperature specific heat and magnetization data reveal the suppression of long-range order at $μ_{0}H_{c}$, enabling the construction of a temperature-magnetic field phase diagram showing multiple magnetic phases near the $μ_{0}H_{c}$. Density functional theory confirms a distorted diamond chain with $J_{1}$ dimers and competing $J_2$, $J_4$, $J_3$, and $J_5$ interactions with monomer spins as an effective low-temperature spin model.

cond-mat.str-el↗

arXiv2025

A Revisit to the Decoder for Camouflaged Object Detection

Camouflaged object detection (COD) aims to generate a fine-grained segmentation map of camouflaged objects hidden in their background. Due to the hidden nature of camouflaged objects, it is essential for the decoder to be tailored to effectively extract proper features of camouflaged objects and extra-carefully generate their complex boundaries. In this paper, we propose a novel architecture that augments the prevalent decoding strategy in COD with Enrich Decoder and Retouch Decoder, which help to generate a fine-grained segmentation map. Specifically, the Enrich Decoder amplifies the channels of features that are important for COD using channel-wise attention. Retouch Decoder further refines the segmentation maps by spatially attending to important pixels, such as the boundary regions. With extensive experiments, we demonstrate that ENTO shows superior performance using various encoders, with the two novel components playing their unique roles that are mutually complementary.

cs.CV↗

arXiv2024

Do not think about pink elephant!

Large Models (LMs) have heightened expectations for the potential of general AI as they are akin to human intelligence. This paper shows that recent large models such as Stable Diffusion and DALL-E3 also share the vulnerability of human intelligence, namely the "white bear phenomenon". We investigate the causes of the white bear phenomenon by analyzing their representation space. Based on this analysis, we propose a simple prompt-based attack method, which generates figures prohibited by the LM provider's policy. To counter these attacks, we introduce prompt-based defense strategies inspired by cognitive therapy techniques, successfully mitigating attacks by up to 48.22\%.

cs.CL↗

arXiv2024

Electronic and magnetic properties of LaCu$_{x}$Sb$_{2}$ tuned by Cu occupancy

We report thermodynamic and transport properties of LaCu$_{x}$Sb$_{2}$ ($0.92 \leq x \leq 1.12$), synthesized by controlling the initial loading composition and investigated by magnetization, electrical resistivity, and specific heat measurements. The physical properties of this system are highly dependent on Cu-site occupancy $x$, where residual resistivity ratio (RRR), magnetoresistance (MR), superconducting transition temperature ($T_{c}$), and electronic specific heat coefficient ($γ$) indicate a systematic variation as a function of $x$. The Shubnikov-de Haas quantum oscillations are observed in magnetoresistance measurements for samples close to the Cu stoichiometry $x \sim 1$, while the de Haas-van Alphen oscillations are detected in a wide range of $x$ ($0.92 \leq x \le 1.12$). For $H \parallel c$, the oscillation frequency indicates a clear $x$-dependence, implying a systematic change of Fermi surface. DFT calculations for the sample closest to ideal Cu stoichiometry reveal electronic structures with a common feature of the square-net-based semimetals, which is in good agreement with the experimental observations. The magnetic response of LaCu$_{x}$Sb$_{2}$ to magnetic fields is anisotropic owing to the Fermi surface anisotropy. Our results show how the physical properties are influenced by the Cu-site occupancy $x$, linked to the electronic bands arising from the Sb square net.

cond-mat.str-el↗

arXiv2024

Contrastive and Consistency Learning for Neural Noisy-Channel Model in Spoken Language Understanding

Recently, deep end-to-end learning has been studied for intent classification in Spoken Language Understanding (SLU). However, end-to-end models require a large amount of speech data with intent labels, and highly optimized models are generally sensitive to the inconsistency between the training and evaluation conditions. Therefore, a natural language understanding approach based on Automatic Speech Recognition (ASR) remains attractive because it can utilize a pre-trained general language model and adapt to the mismatch of the speech input environment. Using this module-based approach, we improve a noisy-channel model to handle transcription inconsistencies caused by ASR errors. We propose a two-stage method, Contrastive and Consistency Learning (CCL), that correlates error patterns between clean and noisy ASR transcripts and emphasizes the consistency of the latent features of the two transcripts. Experiments on four benchmark datasets show that CCL outperforms existing methods and improves the ASR robustness in various noisy environments. Code is available at https://github.com/syoung7388/CCL.

cs.CL↗

arXiv2024

Coreset Selection for Object Detection

Coreset selection is a method for selecting a small, representative subset of an entire dataset. It has been primarily researched in image classification, assuming there is only one object per image. However, coreset selection for object detection is more challenging as an image can contain multiple objects. As a result, much research has yet to be done on this topic. Therefore, we introduce a new approach, Coreset Selection for Object Detection (CSOD). CSOD generates imagewise and classwise representative feature vectors for multiple objects of the same class within each image. Subsequently, we adopt submodular optimization for considering both representativeness and diversity and utilize the representative vectors in the submodular optimization process to select a subset. When we evaluated CSOD on the Pascal VOC dataset, CSOD outperformed random selection by +6.4%p in AP$_{50}$ when selecting 200 images.

cs.CV↗

arXiv2023

Triple-sinusoid hedgehog lattice in a centrosymmetric Kondo metal

Superposed symmetry-equivalent magnetic ordering wave vectors can lead to topologically non-trivial spin textures, such as magnetic skyrmions and hedgehogs, and give rise to novel quantum phenomena due to fictitious magnetic fields associated with a non-zero Berry curvature of these spin textures. To date, all known spin textures are constructed through the superposition of multiple spiral orders, where spins vary in directions with constant amplitude. Recent theoretical studies have suggested that multiple sinusoidal orders, where collinear spins vary in amplitude, can construct distinct topological spin textures regarding chirality properties. However, such textures have yet to be experimentally realised. In this work, we report the observation of a zero-field magnetic hedgehog lattice from a superposition of triple sinusoidal wave vectors in the magnetically frustrated Kondo lattice CePtAl4Ge2. Notably, we also observe the emergence of anomalous electrical and thermodynamic behaviours near the field-induced transition from the zero-field topological hedgehog lattice to a non-topological sinusoidal state. These observations highlight the role of Kondo coupling in stabilising the zero-field hedgehog state in the Kondo lattice and warrant an expedited search for other topological magnetic structures coupled with Kondo coupling.

cond-mat.str-el↗

arXiv2023

Antiferromagnetism of CeCd$_{0.67}$As$_{2}$ existing deep inside the narrow gap semiconducting state

Single crystals of $R$Cd$_{0.67}$As$_2$ ($R$ = La and Ce) have been synthesized by high temperature ternary melt and their physical properties have been explored by means of magnetization, specific heat, electrical resistivity, Hall coefficient, and thermoelectric power measurements. $R$Cd$_{0.67}$As$_2$ compounds indicate a (structural) phase transition at high temperatures, accompanied by a remarkable increase of the electrical resistivity with an extremely low carrier concentration. CeCd$_{0.67}$As$_2$ exhibits a large magnetic anisotropy and an antiferromagnetic (AFM) order below $T_{N} = 4$~K. Magnetic susceptibility curves, together with magnetization isotherms and specific heat, are analyzed by the point charge model of crystalline electric field (CEF). In the paramagnetic state, the observed magnetic properties can be well explained by the CEF effects, implying that the 4$f$ moments remain localized. Electrical resistivity measurements, together with Hall resistivity and thermoelectric power, also suggest highly localized 4$f$ electrons, where Kondo contributions are negligible. The low temperature physical properties manifest strong magnetic field dependencies. For $H \perp c$, $T_{N}$ shifts to lower temperature as magnetic field increases, and eventually disappears at $H_{c} \sim 60$~ kOe. Inside the AFM state, three metamagnetic transitions are clearly evidenced from the magnetization isotherms. The RKKY interaction may be responsible for the AFM ordering in CeCd$_{0.67}$As$_2$, however it would have to be mediated by extremely low charge carriers. Although the AFM ordering temperature in CeCd$_{0.67}$As$_2$ can be continuously suppressed to zero, no AFM quantum phase transition is expected due to the lack of conduction electron clouds to screen the 4$f$ moments.

cond-mat.str-el↗

arXiv2023

The influence of crystalline electric field on the magnetic properties of CeCd3X3 (X = P and As)

CeCd$_3$P$_3$ and CeCd$_3$As$_3$ compounds adopt the hexagonal ScAl$_3$C$_3$-type structure, where magnetic Ce ions on a triangular lattice order antiferromagnetically below $T_\text{N} \sim$0.42~K. Their crystalline electric field (CEF) level scheme has been determined by fitting magnetic susceptibility curves, magnetization isotherms, and Schottky anomalies in specific heat. The calculated results, incorporating the CEF excitation, Zeeman splitting, and molecular field, are in good agreement with the experimental data. The CEF model, with Ce$^{3+}$ ions in a trigonal symmetry, explains the strong easy-plane magnetic anisotropy that has been observed in this family of materials. A detailed examination of the CEF parameters suggests that the fourth order CEF parameter $B_{4}^{3}$ is responsible for the strong CEF induced magnetocrystalline anisotropy, with a large $ab$-plane moment and a small $c$-axis moment. The reliability of our CEF analysis is assessed by comparing the current study with earlier reports of CeCd$_{3}$As$_{3}$. For both CeCd$_{3}X_{3}$ ($X$ = P and As) compounds, less than 40 \% of $R\ln(2)$ magnetic entropy is recovered by $T_\text{N}$ and full $R\ln(2)$ entropy is achieved at the Weiss temperature $θ_{p}$. Although the observed magnetic entropy is reminiscent of delocalized 4$f$-electron magnetism with significant Kondo screening, the electrical resistivity of these compounds follows a typical metallic behavior. Measurements of thermoelectric power further validate the absence of Kondo contribution in CeCd$_{3}X_{3}$.

cond-mat.str-el↗

arXiv2023

Spectroscopic Evidence for the Superconductivity of Elemental Metal Y under Pressure

Very high applied pressure induces superconductivity with the transition temperature ($T_c$) exceeding 19 K in elemental yttrium, but relatively little is known about the nature of that superconductivity. From point-contact spectroscopy (PCS) measurements in a diamond anvil cell (DAC), a strong enhancement in the differential conductance is revealed near the zero-biased voltage owing to Andreev reflection, a hallmark of the superconducting (SC) phase. Analysis of the PCS spectra based on the extended Blonder-Tinkham-Klapwijk (BTK) model indicates two SC gaps at 48.6 GPa, where the large gap $Δ_L$ is 3.63 meV and the small gap $Δ_S$ is 0.46 meV. When scaled against a reduced temperature, both small and large SC gaps collapse on a single curve that follows the prediction from BCS theory. The SC gap-to-$T_c$ ratio is 8.2 for the larger gap, and the initial slope of the upper critical field is -1.9 T/K, indicating that Y belongs to a family of strongly coupled BCS superconductors. The successful application of PCS to Y in DAC environments demonstrates its utility for future research on other pressure-induced high-$T_c$ superconductors.

cond-mat.supr-con↗