SearcharxivSearch

arXiv subjects

Jiwoong Park

Publications and source records attributed to Jiwoong Park.

At least 19 recordsLinked to original sources

GateANN: I/O-Efficient Filtered Vector Search on SSDs

We present GateANN, an I/O-efficient SSD-based graph ANNS system that supports filtered vector search on an unmodified graph index. Existing SSD-based systems either waste I/O by post-filtering, or require expensive filter-aware index rebuilds. GateANN avoids both by decoupling graph traversal from vector retrieval. Our key insight is that traversing a node requires only its neighbor list and an approximate distance, neither of which needs the full-precision vector on SSD. Based on this, GateANN introduces graph tunneling. It checks each node's filter predicate in memory before issuing I/O and routes through non-matching nodes entirely in memory, preserving graph connectivity without any SSD read for non-matching nodes. Our experimental results show that it reduces SSD reads by up to 10x and improves throughput by up to 7.6x.

cs.OS

Mind the Gap -- Imaging Buried Interfaces in Twisted Oxide Moirés

The ability to tune electronic structure in twisted stacks of two-dimensional (2D) materials has motivated the exploration of similar moiré physics with twisted oxide membranes. Due to the intrinsic three-dimensional nature of bonding in many oxides, achieving atomic-level coupling is significantly more challenging than with van der Waals materials. Although clean interfaces with atomic-level proximity have been demonstrated in ceramic bicrystals using high-temperature and high-pressure processing to facilitate atomic diffusion that flattens rough interfaces, such conditions are not readily accessible when bonding oxide membranes. This study shows how topographic mismatch due to surface roughness of the membranes can restrict atomic-scale proximity at the interface to isolated patches even after contaminants and amorphous interlayers are eliminated. In interfaces between 2D materials and oxide membranes the reduced ability of the 2D material to conform to the membrane's step-terrace topography also limits atomic-scale contact. When imaging stacked membranes in projection, we find conventional through-focal imaging to be relatively insensitive to the buried interface, whereas electron ptychography detects structural variations on the order of a nanometer. These findings highlight interface roughness as a key challenge for the field of oxide twistronics and emphasize the need for reliable characterization methods, both in cross-section and projection.

cond-mat.mtrl-sci

LG Uplus System with Multi-Speaker IDs and Discriminator-based Sub-Judges for the WildSpoof Challenge

This paper describes our submission to the WildSpoof Challenge Track 2, which focuses on spoof-aware speaker verification (SASV) in the presence of high-quality text-to-speech (TTS) attacks. We adopt a ResNet-221 back-bone and study two speaker-labeling strategies, namelyDual-Speaker IDs and Multi-Speaker IDs, to explicitly enlarge the margin between bona fide and generated speech in the embedding space. In addition, we propose discriminator-based sub-judge systems that reuse internal features from HiFi-GAN and BigVGAN discriminators, aggregated via multi-query multi-head attentive statistics pooling(MQMHA). Experimental results on the SpoofCeleb corpus show that our system design is effective in improving agnostic detection cost function (a-DCF).

eess.AS

Enhancing Generalization in Data-free Quantization via Mixup-class Prompting

Post-training quantization (PTQ) improves efficiency but struggles with limited calibration data, especially under privacy constraints. Data-free quantization (DFQ) mitigates this by generating synthetic images using generative models such as generative adversarial networks (GANs) and text-conditioned latent diffusion models (LDMs), while applying existing PTQ algorithms. However, the relationship between generated synthetic images and the generalizability of the quantized model during PTQ remains underexplored. Without investigating this relationship, synthetic images generated by previous prompt engineering methods based on single-class prompts suffer from issues such as polysemy, leading to performance degradation. We propose \textbf{mixup-class prompt}, a mixup-based text prompting strategy that fuses multiple class labels at the text prompt level to generate diverse, robust synthetic data. This approach enhances generalization, and improves optimization stability in PTQ. We provide quantitative insights through gradient norm and generalization error analysis. Experiments on convolutional neural networks (CNNs) and vision transformers (ViTs) show that our method consistently outperforms state-of-the-art DFQ methods like GenQ. Furthermore, it pushes the performance boundary in extremely low-bit scenarios, achieving new state-of-the-art accuracy in challenging 2-bit weight, 4-bit activation (W2A4) quantization.

cs.CV

AMXFP4: Taming Activation Outliers with Asymmetric Microscaling Floating-Point for 4-bit LLM Inference

As large language models (LLMs) grow in parameter size and context length, computation precision has been reduced from 16-bit to 4-bit to improve inference efficiency. However, this reduction causes accuracy degradation due to activation outliers. Rotation-based INT4 methods address this via matrix calibration, but they introduce multi-hour overheads and leave key computations in full precision. Microscaling (MX) floating-point (FP) formats offer fine-grained representation with a shared scale, enabling fully quantized matrix multiplications through direct casting without calibration. However, existing research shows unsatisfactory empirical results for MXFP4 inference, and the robustness of MX formats remains largely unexplored. In this work, we uncover the fundamental tradeoffs of the MX format: while it effectively suppresses activation outliers, it does so at the cost of increased group-wise asymmetry. To address this, we propose AMXFP4, a 4-bit asymmetric FP format that handles both issues using asymmetric shared scales, without requiring calibration. Our custom MAC engine adds negligible hardware cost while improving accuracy: AMXFP4 outperforms MXFP4 by 3% on VQA and exceeds rotation-based methods by 1.6% on CSQA. It also surpasses recently deployed commercial MXFP4 variants. Code: https://github.com/aiha-lab/MX-QLLM

cs.AI

Bright hybrid excitons in molecularly tunable bilayer crystals

Bilayer crystals, built by stacking crystalline monolayers, generate interlayer potentials that govern excitonic phenomena but are constrained by fixed covalent lattices and orientations. Replacing one layer with an atomically thin molecular crystal overcomes this limitation, as diverse functional groups enable tunable molecular lattices and interlayer potentials, tailoring a wide range of excitonic properties. Here, we report hybrid excitons in four-atom-thick hybrid bilayer crystals (HBCs), directly synthesized with single-crystalline perylene diimide (PDI) molecular crystal atop WS2 monolayers. These excitons arise from a hybridized bilayer band structure, revealed by lattice-scale first-principles calculations, inheriting properties from both monolayers. They exhibit bright photoluminescence with near-unity polarization above and below the WS2 bandgap, along with spectral signatures of exciton delocalization, supported by theory, while their energies and intensities are tuned by modifying the HBC composition by synthesis. Our work introduces a molecule-based 2D quantum materials platform for bottom-up design and control of optoelectronic properties.

cond-mat.mtrl-sci

Quasi-phase-matched up- and down-conversion in periodically poled layered semiconductors

Nonlinear optics lies at the heart of classical and quantum light generation. The invention of periodic poling revolutionized nonlinear optics and its commercial applications by enabling robust quasi-phase-matching in crystals such as lithium niobate. However, reaching useful frequency conversion efficiencies requires macroscopic dimensions, limiting further technology development and integration. Here we realize a periodically poled van der Waals semiconductor (3R-MoS$_2$). Due to its exceptional nonlinearity, we achieve macroscopic frequency conversion efficiency of 0.03% at the relevant telecom wavelength over a microscopic thickness of 3.4$μ$m (that is, 3 poling periods), $10-100\times$ thinner than current systems with similar performances. Due to unique intrinsic cavity effects, the thickness-dependent quasi-phase-matched second harmonic signal surpasses the usual quadratic enhancement by $50\%$. Further, we report the broadband generation of photon pairs at telecom wavelengths via quasi-phase-matched spontaneous parametric down-conversion, showing a maximum coincidence-to-accidental-ratio of $638 \pm 75$. This work opens the new and unexplored field of phase-matched nonlinear optics with microscopic van der Waals crystals, unlocking applications that require simple, ultra-compact technologies such as on-chip entangled photon-pair sources for integrated quantum circuitry and sensing.

physics.optics

Molecular tuning of excitons in four-atom-thick hybrid bilayer crystals

Bilayer crystals, formed by stacking monolayers of two-dimensional (2D) crystals, create interlayer potentials that govern excitonic phenomena but are constrained by their fixed covalent lattices. Replacing one layer with an atomically thin molecular crystal overcomes this limitation, as precise control of functional groups enables tunable 2D molecular lattices and, consequently, electronic structures. Here, we report molecular tuning of lattices and excitons in four-atom-thick hybrid bilayer crystals (HBCs), synthesized as monolayers of perylene-based molecular and transition metal dichalcogenide (TMD) single crystals. In HBCs, we observe an anisotropic photoluminescence signal exhibiting characteristics of both molecular and TMD excitons, directly tuned by molecular geometry and HBC composition. Ab initio calculations reveal that this anisotropic emission arises from hybrid excitons, which inherit properties from both layers through a hybridized bilayer band structure. Our work establishes a synthetically derived, molecule-based 2D quantum materials platform with the potential for engineering interlayer potentials.

cond-mat.mtrl-sci

Spectra-orthogonal optical anisotropy in wafer-scale molecular crystal monolayers

Controlling the spectral and polarization responses of two-dimensional (2D) crystals is vital for developing ultra-thin platforms for compact optoelectronic devices. However, independently tuning optical anisotropy and spectral response remains challenging in conventional semiconductors due to the intertwined nature of their lattice and electronic structures. Here, we report spectra-orthogonal optical anisotropy, where polarization anisotropy is tuned independently of spectral response, in wafer-scale, one-atom-thick 2D molecular crystal (2DMC) monolayers synthesized on monolayer transition metal dichalcogenide (TMD) crystals. Utilizing the concomitant spectral consistency and structural tunability of perylene derivatives, we demonstrate tunable optical polarization anisotropy in 2DMCs with similar spectral profiles, as confirmed by room-temperature scanning tunneling microscopy and cross-polarized reflectance microscopy. Additional angle-dependent analysis of the single- and polycrystalline molecular domains reveals an epitaxial relationship between the 2DMC and the TMD. Our results establish a scalable, molecule-based 2D crystalline platform for unique and tunable functionalities unattainable in covalent 2D solids.

cond-mat.mes-hall

Equivariant Blurring Diffusion for Hierarchical Molecular Conformer Generation

How can diffusion models process 3D geometries in a coarse-to-fine manner, akin to our multiscale view of the world? In this paper, we address the question by focusing on a fundamental biochemical problem of generating 3D molecular conformers conditioned on molecular graphs in a multiscale manner. Our approach consists of two hierarchical stages: i) generation of coarse-grained fragment-level 3D structure from the molecular graph, and ii) generation of fine atomic details from the coarse-grained approximated structure while allowing the latter to be adjusted simultaneously. For the challenging second stage, which demands preserving coarse-grained information while ensuring SE(3) equivariance, we introduce a novel generative model termed Equivariant Blurring Diffusion (EBD), which defines a forward process that moves towards the fragment-level coarse-grained structure by blurring the fine atomic details of conformers, and a reverse process that performs the opposite operation using equivariant networks. We demonstrate the effectiveness of EBD by geometric and chemical comparison to state-of-the-art denoising diffusion models on a benchmark of drug-like molecules. Ablation studies draw insights on the design of EBD by thoroughly analyzing its architecture, which includes the design of the loss function and the data corruption process. Codes are released at https://github.com/Shen-Lab/EBD .

cs.LG

Universal wrinkling of freestanding atomically thin films

Atomically thin films, like transition metal dichalcogenides, can now be synthesized at wafer scale, achieving the same extreme aspect ratio (~10^8) that a sheet of paper would have if it covered an entire city. Yet, the intrinsic (i.e. unconfined) three-dimensional shape of these extreme membranes remains a mystery because of the very fundamentals of mechanical measurements: to measure such an ultra-thin film, one first needs to simultaneously free it and stabilize it without introducing confining boundaries. Here, we introduce a counter-intuitive solution to this problem: place atomically thin films on water. Using atomic force microscopy (AFM) and Raman spectroscopy adapted to water's surface, we reveal that large-scale freestanding membranes spontaneously self-wrinkle into a universal mechanical state with long emergent length scales that follow robust scaling trends. Our analytical and numerical models suggest that these universal trends are controlled by mesoscopic parameters of the polycrystalline domains instead of atomistic details. Moreover, we demonstrate experimentally that the wrinkles result in a large and tunable reduction of elastic stiffness by up to 2 orders of magnitude. The present work illuminates the physical properties of the world's thinnest materials at length scales never probed before and highlights their potential for tunable strain-controlled nanomechanical devices.

cond-mat.mes-hall

Confidence-Based Feature Imputation for Graphs with Partially Known Features

This paper investigates a missing feature imputation problem for graph learning tasks. Several methods have previously addressed learning tasks on graphs with missing features. However, in cases of high rates of missing features, they were unable to avoid significant performance degradation. To overcome this limitation, we introduce a novel concept of channel-wise confidence in a node feature, which is assigned to each imputed channel feature of a node for reflecting certainty of the imputation. We then design pseudo-confidence using the channel-wise shortest path distance between a missing-feature node and its nearest known-feature node to replace unavailable true confidence in an actual learning process. Based on the pseudo-confidence, we propose a novel feature imputation scheme that performs channel-wise inter-node diffusion and node-wise inter-channel propagation. The scheme can endure even at an exceedingly high missing rate (e.g., 99.5\%) and it achieves state-of-the-art accuracy for both semi-supervised node classification and link prediction on various datasets containing a high rate of missing features. Codes are available at https://github.com/daehoum1/pcfi.

cs.LG

The Role of Self-Torques in Transition Metal Dichalcogenide/Ferromagnet Bilayers

Recently, transition metal dichalcogenides (TMDs) have been extensively studied for their efficient spin-orbit torque generation in TMD/ferromagnetic bilayers, owing to their large spin-orbit coupling, variety in crystal symmetries, and pristine interfaces. Although the TMD layer was considered essential for the generation of the observed SOTs, recent reports show the presence of a self-torque in single-layer ferromagnetic devices with magnitudes comparable to TMD/ferromagnetic devices. Here, we perform second-harmonic Hall SOT measurements on metal-organic chemical vapor deposition (MOCVD) grown MoS$_{2}$/permalloy/Al$_{2}$O$_{3}$ devices and compare them to a single-layer permalloy/Al$_{2}$O$_{3}$ device to accurately disentangle the role of self-torques from contributions from the TMD layer. We report a damping-like self-torque conductivity of opposite sign in our single-layer permalloy/Al$_{2}$O$_{3}$ device compared to one MoS$_{2}$/permalloy/Al$_{2}$O$_{3}$ device, and find no significant one for all other MoS$_{2}$/permalloy/Al$_{2}$O$_{3}$ devices. This indicates a competition between the self-torque and the torque arising from the TMD layer, which would reduce the observed torque in these bilayers. In addition, we find a field-like spin-torque conductivity of comparable magnitude to control MoS$_{2}$/permalloy/Al$_{2}$O$_{3}$ devices, indicating only a minor role of the MoS$_{2}$ layer. Finally, we find a linear dependence of the SOT conductivity on the Hall bar leg/channel width ratio of our devices, indicating that the Hall bar dimensions are of significant importance for the reported SOT strength. Our results accentuate the importance of delicate details, like device asymmetry, Hall bar dimensions, and self-torque generation, for the correct disentanglement of the microscopic origins underlying the SOTs, essential for future energy-efficient spintronic applications.

cond-mat.mes-hall

Torsional Periodic Lattice Distortions and Diffraction of Twisted 2D Materials

Twisted 2D materials form complex moiré structures that spontaneously reduce symmetry through picoscale deformation within a mesoscale lattice. We show twisted 2D materials contain a torsional displacement field comprised of three transverse periodic lattice distortions (PLD). The torsional PLD amplitude provides a single order parameter that concisely describes the structural complexity of twisted bilayer moirés. Moreover, the structure and amplitude of a torsional periodic lattice distortion is quantifiable using rudimentary electron diffraction methods sensitive to reciprocal space. In twisted bilayer graphene, the torsional PLD begins to form at angles below 3.89° and the amplitude reaches 8 pm around the magic angle of 1.1°. At extremely low twist angles (e.g. below 0.25°) the amplitude increases and additional PLD harmonics arise to expand Bernal stacked domains separated by well defined solitonic boundaries. The torsional distortion field in twisted bilayer graphene is analytically described and has an upper bound of 22.6 pm. Similar torsional distortions are observed in twisted WS$_2$, CrI$_3$, and WSe$_2$ / MoSe$_2$.

cond-mat.mtrl-sci

Meta-node: A Concise Approach to Effectively Learn Complex Relationships in Heterogeneous Graphs

Existing message passing neural networks for heterogeneous graphs rely on the concepts of meta-paths or meta-graphs due to the intrinsic nature of heterogeneous graphs. However, the meta-paths and meta-graphs need to be pre-configured before learning and are highly dependent on expert knowledge to construct them. To tackle this challenge, we propose a novel concept of meta-node for message passing that can learn enriched relational knowledge from complex heterogeneous graphs without any meta-paths and meta-graphs by explicitly modeling the relations among the same type of nodes. Unlike meta-paths and meta-graphs, meta-nodes do not require any pre-processing steps that require expert knowledge. Going one step further, we propose a meta-node message passing scheme and apply our method to a contrastive learning model. In the experiments on node clustering and classification tasks, the proposed meta-node message passing method outperforms state-of-the-arts that depend on meta-paths. Our results demonstrate that effective heterogeneous graph learning is possible without the need for meta-paths that are frequently used in this field.

cs.LG

Unsupervised Hyperbolic Representation Learning via Message Passing Auto-Encoders

Most of the existing literature regarding hyperbolic embedding concentrate upon supervised learning, whereas the use of unsupervised hyperbolic embedding is less well explored. In this paper, we analyze how unsupervised tasks can benefit from learned representations in hyperbolic space. To explore how well the hierarchical structure of unlabeled data can be represented in hyperbolic spaces, we design a novel hyperbolic message passing auto-encoder whose overall auto-encoding is performed in hyperbolic space. The proposed model conducts auto-encoding the networks via fully utilizing hyperbolic geometry in message passing. Through extensive quantitative and qualitative analyses, we validate the properties and benefits of the unsupervised hyperbolic representations. Codes are available at https://github.com/junhocho/HGCAE.

cs.LG

Imaging Polarity in Two Dimensional Materials by Breaking Friedel's Law

Friedel's law guarantees an inversion-symmetric diffraction pattern for thin, light materials where a kinematic approximation or a single-scattering model holds. Typically, breaking Friedel symmetry is ascribed to multiple scattering events within thick, non-centrosymmetric crystals. However, two-dimensional (2D) materials such as a single monolayer of MoS$_2$ can also violate Friedel's law, with unexpected contrast between conjugate Bragg peaks. We show analytically that retaining higher order terms in the power series expansion of the scattered wavefunction can describe the anomalous contrast between $hkl$ and $\overline{hkl}$ peaks that occurs in 2D crystals with broken in-plane inversion symmetry. These higher-order terms describe multiple scattering paths starting from the same atom in an atomically thin material. Furthermore, 2D materials containing heavy elements, such as WS$_2$, always act as strong phase objects, violating Friedel's law no matter how high the energy of the incident electron beam. Experimentally, this understanding can enhance diffraction-based techniques to provide rapid imaging of polarity, twin domains, in-plane rotations, or other polar textures in 2D materials.

cond-mat.mtrl-sci

Symmetric Graph Convolutional Autoencoder for Unsupervised Graph Representation Learning

We propose a symmetric graph convolutional autoencoder which produces a low-dimensional latent representation from a graph. In contrast to the existing graph autoencoders with asymmetric decoder parts, the proposed autoencoder has a newly designed decoder which builds a completely symmetric autoencoder form. For the reconstruction of node features, the decoder is designed based on Laplacian sharpening as the counterpart of Laplacian smoothing of the encoder, which allows utilizing the graph structure in the whole processes of the proposed autoencoder architecture. In order to prevent the numerical instability of the network caused by the Laplacian sharpening introduction, we further propose a new numerically stable form of the Laplacian sharpening by incorporating the signed graphs. In addition, a new cost function which finds a latent representation and a latent affinity matrix simultaneously is devised to boost the performance of image clustering tasks. The experimental results on clustering, link prediction and visualization tasks strongly support that the proposed model is stable and outperforms various state-of-the-art algorithms.

cs.LG