arXiv subjects

Seungjun Lee

Publications and source records attributed to Seungjun Lee.

At least 37 records · Page 2Linked to original sources

arXiv2026

AutoGNN: End-to-End Hardware-Driven Graph Preprocessing for Enhanced GNN Performance

Graph neural network (GNN) inference faces significant bottlenecks in preprocessing, which often dominate overall inference latency. We introduce AutoGNN, an FPGA-based accelerator designed to address these challenges by leveraging FPGA's reconfigurability and specialized components. AutoGNN adapts to diverse graph inputs, efficiently performing computationally intensive tasks such as graph conversion and sampling. By utilizing components like adder trees, AutoGNN executes reduction operations in constant time, overcoming the limitations of serialization and synchronization on GPUs. AutoGNN integrates unified processing elements (UPEs) and single-cycle reducers (SCRs) to streamline GNN preprocessing. UPEs enable scalable parallel processing for edge sorting and unique vertex selection, while SCRs efficiently handle sequential tasks such as pointer array construction and subgraph reindexing. A user-level software framework dynamically profiles graph inputs, determines optimal configurations, and reprograms AutoGNN to handle varying workloads. Implemented on a 7$n$m enterprise FPGA, AutoGNN achieves up to 9.0$\times$ and 2.1$\times$ speedup compared to conventional and GPU-accelerated preprocessing systems, respectively, enabling high-performance GNN preprocessing across diverse datasets.

cs.AR↗

arXiv2026

Segment Any Events with Language

Scene understanding with free-form language has been widely explored within diverse modalities such as images, point clouds, and LiDAR. However, related studies on event sensors are scarce or narrowly centered on semantic-level understanding. We introduce SEAL, the first Semantic-aware Segment Any Events framework that addresses Open-Vocabulary Event Instance Segmentation (OV-EIS). Given the visual prompt, our model presents a unified framework to support both event segmentation and open-vocabulary mask classification at multiple levels of granularity, including instance-level and part-level. To enable thorough evaluation on OV-EIS, we curate four benchmarks that cover label granularity from coarse to fine class configurations and semantic granularity from instance-level to part-level understanding. Extensive experiments show that our SEAL largely outperforms proposed baselines in terms of performance and inference speed with a parameter-efficient architecture. In the Appendix, we further present a simple variant of our SEAL achieving generic spatiotemporal OV-EIS that does not require any visual prompts from users in the inference. Check out our project page in https://0nandon.github.io/SEAL

cs.CV↗

arXiv2026

Emergence of unconventional magnetic order in strain-engineered RuO2/TiO2 superlattices

The spin ordering in RuO2 remains a highly debated topic, owing to its elusive nature, with reports ranging from a nonmagnetic ground state to signatures of unconventional magnetic order. Here we provide the first unambiguous, and direct evidence of unconventional magnetism in epitaxial, fully strained RuO2/TiO2 superlattices on TiO2 (110) substrate grown by hybrid molecular beam epitaxy. Polarized neutron reflectometry reveals a finite magnetic moment localized within the compressively strained RuO2 layers, consistent with predictions obtained from first-principles calculations. Complementary density functional theory and X-ray photoemission spectroscopy show that epitaxial strain drives the Ru 4d states toward the Fermi level, triggering a Stoner-type instability that stabilizes non-compensated magnetic order. These unique results reveal that RuO2 exhibits unconventional magnetic states under epitaxial strain, which are not accessible in bulk and establish strain engineering as a powerful route to uncover and control magnetic phases in RuO2 and related oxides.

cond-mat.mtrl-sci↗

arXiv2026

EXAONE 3.0 7.8B Instruction Tuned Language Model

We introduce EXAONE 3.0 instruction-tuned language model, the first open model in the family of Large Language Models (LLMs) developed by LG AI Research. Among different model sizes, we publicly release the 7.8B instruction-tuned model to promote open research and innovations. Through extensive evaluations across a wide range of public and in-house benchmarks, EXAONE 3.0 demonstrates highly competitive real-world performance with instruction-following capability against other state-of-the-art open models of similar size. Our comparative analysis shows that EXAONE 3.0 excels particularly in Korean, while achieving compelling performance across general tasks and complex reasoning. With its strong real-world effectiveness and bilingual proficiency, we hope that EXAONE keeps contributing to advancements in Expert AI. Our EXAONE 3.0 instruction-tuned model is available at https://huggingface.co/LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct.

cs.CL↗

arXiv2025

D3D-VLP: Dynamic 3D Vision-Language-Planning Model for Embodied Grounding and Navigation

Embodied agents face a critical dilemma that end-to-end models lack interpretability and explicit 3D reasoning, while modular systems ignore cross-component interdependencies and synergies. To bridge this gap, we propose the Dynamic 3D Vision-Language-Planning Model (D3D-VLP). Our model introduces two key innovations: 1) A Dynamic 3D Chain-of-Thought (3D CoT) that unifies planning, grounding, navigation, and question answering within a single 3D-VLM and CoT pipeline; 2) A Synergistic Learning from Fragmented Supervision (SLFS) strategy, which uses a masked autoregressive loss to learn from massive and partially-annotated hybrid data. This allows different CoT components to mutually reinforce and implicitly supervise each other. To this end, we construct a large-scale dataset with 10M hybrid samples from 5K real scans and 20K synthetic scenes that are compatible with online learning methods such as RL and DAgger. Our D3D-VLP achieves state-of-the-art results on multiple benchmarks, including Vision-and-Language Navigation (R2R-CE, REVERIE-CE, NavRAG-CE), Object-goal Navigation (HM3D-OVON), and Task-oriented Sequential Grounding and Navigation (SG3D). Real-world mobile manipulation experiments further validate the effectiveness.

cs.CV↗

arXiv2025

Improved Decoupled Control of Modular Multilevel Converter under Constaint of Nearest Level Modulation via Disturbance Observer Design

Nearest level modulation (NLM) is an attractive modulation method for its implementation simplicity in modular multilevel converter (MMC). However, it introduces significant voltage and current distortion when the number of submodules (SMs) per arm is small, as in medium-voltage applications. While indirect modulation offers fully decoupled control of ac-side current, dc-side current and SM capacitor energy, its performance is fundamentally reliant on accurate arm voltage synthesis, making it incompatible with the large quantization error inherent in NLM. To resolve this conflict, this paper proposes a new control strategy based on a disturbance observer (DOB). The key idea is to estimate and actively compensate for the inevitable arm voltage synthesis error induced by NLM, thereby enabling fully decoupled control of indirect-modulated MMC even under NLM operation with a small number of SMs. A key advantage is its ease of implementation, as it requires no modifications to the conventional NLM and decoupled control structure. The validity and effectiveness of the proposed method in improving current quality and decoupled SM energy control are verified through both simulation and experimental results.

eess.SY↗

arXiv2025

Computational search for materials having a giant anomalous Hall effect in the pyrochlore and spinel crystal structures

Ferromagnetic pyrochlore and spinel materials with topological flat bands are of interest for their potential to exhibit a giant anomalous Hall effect (AHE). In this work, we present computational predictions of stability and electronic structure for 448 compositions within the pyrochlore (A2B2O7) and spinel (AB2O4) frameworks. Of these, 92 are predicted to be thermodynamically stable or close (< 100 meV/atom) to the convex hull, with trends deviating from expectations based on ionic radius-ratio rules. Thirteen are predicted to adopt a ferromagnetic ground state among the collinear configurations considered. Two additional materials meeting these criteria were also identified from open materials databases. Calculations of anomalous Hall angles (AHA) and conductivities reveal that 11 of the screened materials are promising candidates for spintronic applications requiring high electronic conductivity and a giant AHE. Our results suggest that the AHA can be further enhanced by tuning the Fermi level, for example through chemical doping. Using this approach, we identify five materials whose AHA exceed 0.2 under the approximation of collinear magnetism. Notably, Ag2Pt2O7 exhibits a high AHA of 0.405 when its Fermi level is optimized. These findings provide a roadmap for the targeted synthesis of new pyrochlore and spinel compounds with enhanced AHE properties. They also broaden the compositional design space for these structures and support the discovery of high-performance materials for next-generation spintronic applications.

cond-mat.mtrl-sci↗

arXiv2025

Momentum-Resolved Relaxation-Time Approach for Size-Dependent Conductivity in Anisotropic Metallic Films

Shrinking CMOS interconnect dimensions to the nanometer scale intensifies electron scattering at surfaces, interfaces, and grain boundaries, causing severe conductivity loss and challenging copper-based designs. Here we present a momentum-resolved relaxation time framework that integrates density functional theory with the semiclassical Boltzmann transport equation to predict size-dependent resistivity in metallic thin films. Electron phonon interactions are computed from first principles, and anisotropic surface and grain boundary scattering is captured through a momentum dependent mean free path, allowing relaxation times to vary spatially and directionally without empirical fitting. Applied to isotropic (Cu, Ag, Au) and anisotropic (W, Ti$_2$GeC) metals, the model achieves excellent agreement with experiments and uncovers the critical role of crystallographic anisotropy in transport. We further identify layered MAX phase compounds as promising ultrathin interconnects. This work provides a predictive, physically rigorous, and computationally efficient route to designing high-performance conductors for next generation nanoelectronics.

cond-mat.mtrl-sci↗

arXiv2025

Strain-Stabilized Interfacial Polarization Tunes Work Function Over 1 eV in RuO2/TiO2 Heterostructures

Interfacial polarization-charge accumulation at the heterointerface-is a well-established tool in semiconductors, but its influence in metals remains unexplored. Here, we demonstrate that interfacial polarization can robustly modulate surface work function in metallic rutile RuO2 layers in epitaxial RuO2/TiO2 heterostructures grown by hybrid molecular beam epitaxy. Using multislice electron ptychography, we directly visualize polar displacements of transition metal ions relative to oxygen octahedra near the interface, despite the conductive nature of RuO2. This interfacial polarization enables over 1 eV modulation of the RuO2 work function, controlled by small thickness variation (2-4 nm) as measured by Kelvin probe probe microscopy, with a critical thickness of 4 nm - corresponding to the transition from fully strained to relaxed film. These results establish interfacial polarization as a powerful route to control electronic properties in metals and have implications for designing tunable electronic, catalytic, and quantum devices through interfacial control in polar metallic systems.

cond-mat.mtrl-sci↗

arXiv2025

Moiré Collapse and Luttinger Liquids In Twisted Anisotropic Homobilayers

We introduce twisted anisotropic homobilayers as a distinct class of moiré systems, characterized by a distinctive ``magic angle", $θ_M$, where both the moiré unit cell and Brillouin zone collapse. Unlike conventional studies of moiré materials, which primarily focus on small lattice misalignments, we demonstrate that this moiré collapse occurs at large twist angles in generic twisted anisotropic homobilayers. The collapse angle, $θ_M$, is likely to give rise quasi-crystal behavior as well as to the formation of strongly correlated states, that arise not from flat bands, but from the presence of ultra-anisotropic electronic states, where non-Fermi liquid phases can be stabilized. In this work, we develop a continuum model for electrons based on extensive \textit{ab initio} calculations for twisted bilayer black phosphorus, enabling a detailed study of the emerging moiré collapse features in this archetypal system. We show that the (temperature) stability criterion for the emergence of (sliding) Luttinger liquids is generally met as the twist angle approaches $θ_M$. Furthermore, we explicitly formulate the collapsed single-particle one-dimensional (1D) continuum Hamiltonian and provide the \textit{fully interacting}, bosonized Hamiltonian applicable at low doping levels. Our analysis reveals a rich landscape of multichannel Luttinger liquids, potentially enhanced by valley degrees of freedom at large twist angles.

cond-mat.str-el↗

arXiv2025

Generalized energy band alignment model for van der Waals heterostructures with a charge spillage dipole

The energy band alignment at the interface of van der Waals heterostructures (vdWHs) is a key design parameter for next-generation electronic and optoelectronic devices. Although the Anderson and midgap models have been widely adopted for bulk semiconductor heterostructures, they exhibit severe limitations when applied to vdWHs, particularly for type-III systems. Based on first-principles calculations for approximately $10^3$ vdWHs, we demonstrate these traditional models miss a critical dipole arising from interlayer charge spillage. We introduce a generalized linear response (gLR) model that includes this dipole through a quantum capacitance term while remaining analytically compact. With only two readily computed inputs, the charge neutrality level offset and the sum of the isolated-layer bandgaps, the gLR reproduces DFT band line-ups with $r^2\sim$0.9 across type-I, II, and III stacks. Machine-learning feature analysis confirms that these two descriptors dominate the underlying physics, indicating the model is near-minimal and broadly transferable. The gLR framework therefore provides both mechanistic insight and a fast, accurate surrogate for high-throughput screening of the vast vdW heterostructure design space.

cond-mat.mtrl-sci↗

arXiv2025

CXL-GPU: Pushing GPU Memory Boundaries with the Integration of CXL Technologies

This work introduces a GPU storage expansion solution utilizing CXL, featuring a novel GPU system design with multiple CXL root ports for integrating diverse storage media (DRAMs and/or SSDs). We developed and siliconized a custom CXL controller integrated at the hardware RTL level, achieving two-digit nanosecond roundtrip latency, the first in the field. This study also includes speculative read and deterministic store mechanisms to efficiently manage read and write operations to hide the endpoint's backend media latency variation. Performance evaluations reveal our approach significantly outperforms existing methods, marking a substantial advancement in GPU storage technology.

cs.AR↗

arXiv2025

From Block to Byte: Transforming PCIe SSDs with CXL Memory Protocol and Instruction Annotation

This paper explores how Compute Express Link (CXL) can transform PCIe-based block storage into a scalable, byte-addressable working memory. We address the challenges of adapting block storage to CXL's memory-centric model by emphasizing cacheability as a key enabler and advocating for Type 3 endpoint devices, referred to as CXL-SSDs. To validate our approach, we prototype a CXL-SSD on a custom FPGA platform and propose annotation mechanisms, Determinism and Bufferability, to enhance performance while preserving data persistency. Our simulation-based evaluation demonstrates that CXL-SSD achieves 10.9x better performance than PCIe-based memory expanders and further reduces latency by 5.4x with annotation enhancements. In workloads with high locality, CXL-SSD approaches DRAM-like performance due to efficient on-chip caching. This work highlights the feasibility of integrating block storage into CXL's ecosystem and provides a foundation for future memory-storage convergence.

cs.AR↗

arXiv2025

PDEfuncta: Spectrally-Aware Neural Representation for PDE Solution Modeling

Scientific machine learning often involves representing complex solution fields that exhibit high-frequency features such as sharp transitions, fine-scale oscillations, and localized structures. While implicit neural representations (INRs) have shown promise for continuous function modeling, capturing such high-frequency behavior remains a challenge-especially when modeling multiple solution fields with a shared network. Prior work addressing spectral bias in INRs has primarily focused on single-instance settings, limiting scalability and generalization. In this work, we propose Global Fourier Modulation (GFM), a novel modulation technique that injects high-frequency information at each layer of the INR through Fourier-based reparameterization. This enables compact and accurate representation of multiple solution fields using low-dimensional latent vectors. Building upon GFM, we introduce PDEfuncta, a meta-learning framework designed to learn multi-modal solution fields and support generalization to new tasks. Through empirical studies on diverse scientific problems, we demonstrate that our method not only improves representational quality but also shows potential for forward and inverse inference tasks without the need for retraining.

cs.LG↗

arXiv2025

Dynam3D: Dynamic Layered 3D Tokens Empower VLM for Vision-and-Language Navigation

Vision-and-Language Navigation (VLN) is a core task where embodied agents leverage their spatial mobility to navigate in 3D environments toward designated destinations based on natural language instructions. Recently, video-language large models (Video-VLMs) with strong generalization capabilities and rich commonsense knowledge have shown remarkable performance when applied to VLN tasks. However, these models still encounter the following challenges when applied to real-world 3D navigation: 1) Insufficient understanding of 3D geometry and spatial semantics; 2) Limited capacity for large-scale exploration and long-term environmental memory; 3) Poor adaptability to dynamic and changing environments.To address these limitations, we propose Dynam3D, a dynamic layered 3D representation model that leverages language-aligned, generalizable, and hierarchical 3D representations as visual input to train 3D-VLM in navigation action prediction. Given posed RGB-D images, our Dynam3D projects 2D CLIP features into 3D space and constructs multi-level 3D patch-instance-zone representations for 3D geometric and semantic understanding with a dynamic and layer-wise update strategy. Our Dynam3D is capable of online encoding and localization of 3D instances, and dynamically updates them in changing environments to provide large-scale exploration and long-term memory capabilities for navigation. By leveraging large-scale 3D-language pretraining and task-specific adaptation, our Dynam3D sets new state-of-the-art performance on VLN benchmarks including R2R-CE, REVERIE-CE and NavRAG-CE under monocular settings. Furthermore, experiments for pre-exploration, lifelong memory, and real-world robot validate the effectiveness of practical deployment.

cs.CV↗

arXiv2025

MolMole: Molecule Mining from Scientific Literature

The extraction of molecular structures and reaction data from scientific documents is challenging due to their varied, unstructured chemical formats and complex document layouts. To address this, we introduce MolMole, a vision-based deep learning framework that unifies molecule detection, reaction diagram parsing, and optical chemical structure recognition (OCSR) into a single pipeline for automating the extraction of chemical data directly from page-level documents. Recognizing the lack of a standard page-level benchmark and evaluation metric, we also present a testset of 550 pages annotated with molecule bounding boxes, reaction labels, and MOLfiles, along with a novel evaluation metric. Experimental results demonstrate that MolMole outperforms existing toolkits on both our benchmark and public datasets. The benchmark testset will be publicly available, and the MolMole toolkit will be accessible soon through an interactive demo on the LG AI Research website. For commercial inquiries, please contact us at \href{mailto:contact_ddu@lgresearch.ai}{contact\_ddu@lgresearch.ai}.

cs.LG↗

arXiv2025

SCENT: Robust Spatiotemporal Learning for Continuous Scientific Data via Scalable Conditioned Neural Fields

Spatiotemporal learning is challenging due to the intricate interplay between spatial and temporal dependencies, the high dimensionality of the data, and scalability constraints. These challenges are further amplified in scientific domains, where data is often irregularly distributed (e.g., missing values from sensor failures) and high-volume (e.g., high-fidelity simulations), posing additional computational and modeling difficulties. In this paper, we present SCENT, a novel framework for scalable and continuity-informed spatiotemporal representation learning. SCENT unifies interpolation, reconstruction, and forecasting within a single architecture. Built on a transformer-based encoder-processor-decoder backbone, SCENT introduces learnable queries to enhance generalization and a query-wise cross-attention mechanism to effectively capture multi-scale dependencies. To ensure scalability in both data size and model complexity, we incorporate a sparse attention mechanism, enabling flexible output representations and efficient evaluation at arbitrary resolutions. We validate SCENT through extensive simulations and real-world experiments, demonstrating state-of-the-art performance across multiple challenging tasks while achieving superior scalability.

cs.LG↗

arXiv2025

DiET-GS: Diffusion Prior and Event Stream-Assisted Motion Deblurring 3D Gaussian Splatting

Reconstructing sharp 3D representations from blurry multi-view images are long-standing problem in computer vision. Recent works attempt to enhance high-quality novel view synthesis from the motion blur by leveraging event-based cameras, benefiting from high dynamic range and microsecond temporal resolution. However, they often reach sub-optimal visual quality in either restoring inaccurate color or losing fine-grained details. In this paper, we present DiET-GS, a diffusion prior and event stream-assisted motion deblurring 3DGS. Our framework effectively leverages both blur-free event streams and diffusion prior in a two-stage training strategy. Specifically, we introduce the novel framework to constraint 3DGS with event double integral, achieving both accurate color and well-defined details. Additionally, we propose a simple technique to leverage diffusion prior to further enhance the edge details. Qualitative and quantitative results on both synthetic and real-world data demonstrate that our DiET-GS is capable of producing significantly better quality of novel views compared to the existing baselines. Our project page is https://diet-gs.github.io

cs.CV↗