arXiv subjects

Dongwoo Lee

Publications and source records attributed to Dongwoo Lee.

At least 19 recordsLinked to original sources

arXiv2026

Understanding LoRA as Knowledge Memory: An Empirical Analysis

Continuous knowledge updating for pre-trained large language models (LLMs) is increasingly necessary yet remains challenging. Although inference-time methods like In-Context Learning (ICL) and Retrieval-Augmented Generation (RAG) are popular, they face constraints in context budgets, costs, and retrieval fragmentation. Departing from these context-dependent paradigms, this work investigates a parametric approach using Low-Rank Adaptation (LoRA) as a modular knowledge memory. Although few recent works examine this concept, the fundamental mechanics governing its capacity and composability remain largely unexplored. We bridge this gap through the first systematic empirical study mapping the design space of LoRA-based memory, ranging from characterizing storage capacity and optimizing internalization to scaling multi-module systems and evaluating long-context reasoning. Rather than proposing a single architecture, we provide practical guidance on the operational boundaries of LoRA memory. Overall, our findings position LoRA as the complementary axis of memory alongside RAG and ICL, offering distinct advantages.

cs.LG↗

arXiv2026

A Dataset and Benchmark for Robotic Cloth Unfolding Grasp Selection: The ICRA 2024 Cloth Competition

Robotic cloth manipulation suffers from a lack of standardized benchmarks and shared datasets for evaluating and comparing different approaches. To address this, we created a benchmark and organized the ICRA 2024 Cloth Competition, a unique head-to-head evaluation focused on grasp pose selection for in-air robotic cloth unfolding. Eleven diverse teams participated in the competition, utilizing our publicly released dataset of real-world robotic cloth unfolding attempts and a variety of methods to design their unfolding approaches. Afterwards, we also expanded our dataset with 176 competition evaluation trials, resulting in a dataset of 679 unfolding demonstrations across 34 garments. Analysis of the competition results revealed insights about the trade-off between grasp success and coverage, the surprisingly strong achievements of hand-engineered methods and a significant discrepancy between competition performance and prior work, underscoring the importance of independent, out-of-the-lab evaluation in robotic cloth manipulation. The associated dataset is a valuable resource for developing and evaluating grasp selection methods, particularly for learning-based approaches. We hope that our benchmark, dataset and competition results can serve as a foundation for future benchmarks and drive further progress in data-driven robotic cloth manipulation. The dataset and benchmarking code are available at https://airo.ugent.be/cloth_competition.

cs.RO↗

arXiv2026

NavFormer: IGRF Forecasting in Moving Coordinate Frames

Triad magnetometer components change with sensor attitude even when the IGRF total intensity target stays invariant. NavFormer forecasts this invariant target with rotation invariant scalar features and a Canonical SPD module that stabilizes the spectrum of window level second moments of the triads without sign discontinuities. The module builds a canonical frame from a Gram matrix per window and applies state dependent spectral scaling in the original coordinates. Experiments across five flights show lower error than strong baselines in standard training, few shot training, and zero shot transfer. The code is available at: https://anonymous.4open.science/r/NavFormer-Robust-IGRF-Forecasting-for-Autonomous-Navigators-0765

cs.LG↗

arXiv2025

Adaptive Inference-Time Scaling via Cyclic Diffusion Search

Diffusion models have demonstrated strong generative capabilities across domains ranging from image synthesis to complex reasoning tasks. However, most inference-time scaling methods rely on fixed denoising schedules, limiting their ability to allocate computation based on instance difficulty or task-specific demands adaptively. We introduce the challenge of adaptive inference-time scaling-dynamically adjusting computational effort during inference-and propose Adaptive Bi-directional Cyclic Diffusion (ABCD), a flexible, search-based inference framework. ABCD refines outputs through bi-directional diffusion cycles while adaptively controlling exploration depth and termination. It comprises three components: Cyclic Diffusion Search, Automatic Exploration-Exploitation Balancing, and Adaptive Thinking Time. Experiments show that ABCD improves performance across diverse tasks while maintaining computational efficiency.

cs.LG↗

arXiv2025

The Emergence of Strategic Reasoning of Large Language Models

As large language models (LLMs) have demonstrated strong reasoning abilities in structured tasks (e.g., coding and mathematics), we explore whether these abilities extend to strategic multi-agent environments. We investigate strategic reasoning capabilities -- the process of choosing an optimal course of action by predicting and adapting to others' actions -- of LLMs by analyzing their performance in three classical games from behavioral economics. Using hierarchical models of bounded rationality, we evaluate three standard LLMs (ChatGPT-4, Claude-3.5-Sonnet, Gemini 1.5) and three reasoning LLMs (OpenAI-o1, Claude-4-Sonnet-Thinking, Gemini Flash Thinking 2.0). Our results show that reasoning LLMs exhibit superior strategic reasoning compared to standard LLMs (which do not demonstrate substantial capabilities) and often match or exceed human performance; this represents the first and thus most fundamental transition in strategic reasoning capabilities documented in LLMs. Since strategic reasoning is fundamental to future AI systems (including Agentic AI), our findings demonstrate the importance of dedicated reasoning capabilities in achieving effective strategic reasoning.

econ.GN↗

arXiv2025

Bayesian Neural Scaling Law Extrapolation with Prior-Data Fitted Networks

Scaling has been a major driver of recent advancements in deep learning. Numerous empirical studies have found that scaling laws often follow the power-law and proposed several variants of power-law functions to predict the scaling behavior at larger scales. However, existing methods mostly rely on point estimation and do not quantify uncertainty, which is crucial for real-world applications involving decision-making problems such as determining the expected performance improvements achievable by investing additional computational resources. In this work, we explore a Bayesian framework based on Prior-data Fitted Networks (PFNs) for neural scaling law extrapolation. Specifically, we design a prior distribution that enables the sampling of infinitely many synthetic functions resembling real-world neural scaling laws, allowing our PFN to meta-learn the extrapolation. We validate the effectiveness of our approach on real-world neural scaling laws, comparing it against both the existing point estimation methods and Bayesian approaches. Our method demonstrates superior performance, particularly in data-limited scenarios such as Bayesian active learning, underscoring its potential for reliable, uncertainty-aware extrapolation in practical applications.

cs.LG↗

arXiv2025

CytoFM: The first cytology foundation model

Cytology is essential for cancer diagnostics and screening due to its minimally invasive nature. However, the development of robust deep learning models for digital cytology is challenging due to the heterogeneity in staining and preparation methods of samples, differences across organs, and the limited availability of large, diverse, annotated datasets. Developing a task-specific model for every cytology application is impractical and non-cytology-specific foundation models struggle to generalize to tasks in this domain where the emphasis is on cell morphology. To address these challenges, we introduce CytoFM, the first cytology self-supervised foundation model. Using iBOT, a self-supervised Vision Transformer (ViT) training framework incorporating masked image modeling and self-distillation, we pretrain CytoFM on a diverse collection of cytology datasets to learn robust, transferable representations. We evaluate CytoFM on multiple downstream cytology tasks, including breast cancer classification and cell type identification, using an attention-based multiple instance learning framework. Our results demonstrate that CytoFM performs better on two out of three downstream tasks than existing foundation models pretrained on histopathology (UNI) or natural images (iBOT-Imagenet). Visualizations of learned representations demonstrate our model is able to attend to cytologically relevant features. Despite a small pre-training dataset, CytoFM's promising results highlight the ability of task-agnostic pre-training approaches to learn robust and generalizable features from cytology data.

cs.CV↗

arXiv2025

Underwater Willis lens for broadband low-frequency focusing

Broadband underwater sound focusing in the low-frequency range is essential for various applications such as battery-free environmental monitoring and sensing. However, achieving low-frequency underwater focusing typically necessitates bulky, heavy structures that hinder practical deployment. Here, we introduce a three-dimensional underwater lens comprising cavity-based locally resonant asymmetric structures, enabling the efficient manipulation of low-frequency waterborne sound through a densely packed lattice configuration. We experimentally validated its broadband focusing performance over a range of 20-35 kHz. In addition, we observed that our lens exhibits asymmetric backscattering-a distinctive effect arising from its bianisotropic nature-which we term the Willis lens. Unlike conventional underwater lenses that rely on fully filled structures, our design employs cavity-based scatterers, achieving a lighter yet robust focusing performance. With its lightweight, efficient, and reliable design, the Willis lens provides a promising platform for underwater sensor networks and future advancements in on-demand waterborne sound focusing.

cond-mat.mes-hall↗

arXiv2025

Sabotage and Free Riding in Contests with a Group-Specific Public-Good/Bad Prize

We study contests in which two groups compete to win (or not to win) a group-specific public-good/bad prize. Each player in the groups can exert two types of effort: one to help her own group win the prize, and one to sabotage her own group's chances of winning it. The players in the groups choose their effort levels simultaneously and independently. We introduce a specific form of contest success function that determines each group's probability of winning the prize, taking into account players' sabotage activities. We show that two types of purestrategy Nash equilibrium occur, depending on parameter values: one without sabotage activities and one with sabotage activities. In the first type, only the highest-valuation player in each group expends positive effort, whereas, in the second type, only the lowest-valuation player in each group expends positive effort.

econ.TH↗

arXiv2024

GS-Blur: A 3D Scene-Based Dataset for Realistic Image Deblurring

To train a deblurring network, an appropriate dataset with paired blurry and sharp images is essential. Existing datasets collect blurry images either synthetically by aggregating consecutive sharp frames or using sophisticated camera systems to capture real blur. However, these methods offer limited diversity in blur types (blur trajectories) or require extensive human effort to reconstruct large-scale datasets, failing to fully reflect real-world blur scenarios. To address this, we propose GS-Blur, a dataset of synthesized realistic blurry images created using a novel approach. To this end, we first reconstruct 3D scenes from multi-view images using 3D Gaussian Splatting (3DGS), then render blurry images by moving the camera view along the randomly generated motion trajectories. By adopting various camera trajectories in reconstructing our GS-Blur, our dataset contains realistic and diverse types of blur, offering a large-scale dataset that generalizes well to real-world blur. Using GS-Blur with various deblurring methods, we demonstrate its ability to generalize effectively compared to previous synthetic or real blur datasets, showing significant improvements in deblurring performance.

cs.CV↗

arXiv2024

Bound States-to-Bands in the Continuum in Cylindrical Granular Crystals

We theoretically investigate and experimentally demonstrate that genuine bound states in the continuum (BICs) -- polarization-protected BICs -- can be completely localized within finite-size solid resonators. This bound mode is realized in a highly tunable mechanical system made of cylindrical granular crystals, where tunning the contact boundaries enables the in situ transition from the BICs to quasi-BICs in a controllable manner. Since a single-particle resonator can support BICs itself, these bound states can extend to form bound bands within periodic structures composed of such resonators. We experimentally demonstrate the emergence of a quasi-bound (flat) band in a finite chain with broken resonator symmetry, using a laser Doppler vibrometer. Remarkably, we show that all cylindrical resonators within the entire chain exhibit high-Q and dispersionless resonance.

physics.app-ph↗

arXiv2024

Machine Learning Prediction Models for Solid Electrolytes based on Lattice Dynamics Properties

Recently, machine-learning approaches have accelerated computational materials design and the search for advanced solid electrolytes. However, the predictors are currently limited to static structural parameters, which may not fully account for the dynamic nature of ionic transport. In this study, we meticulously curated features considering dynamic properties and developed machine-learning models to predict the ionic conductivity of solid electrolytes. We compiled 14 phonon-related descriptors from first-principles phonon calculations along with 16 descriptors related to structure and electronic properties. Our logistic regression classifiers exhibit an accuracy of 93 %, while the random forest regression model yields a root mean square error of 1.179 S/cm and $R^2$ of 0.710. Notably, phonon-related features are essential for estimating the ionic conductivity in both models. Furthermore, we applied our prediction model to screen 264 Li-containing materials and identified 11 promising candidates as potential superionic conductors.

cond-mat.mtrl-sci↗

arXiv2024

DeblurGS: Gaussian Splatting for Camera Motion Blur

Although significant progress has been made in reconstructing sharp 3D scenes from motion-blurred images, a transition to real-world applications remains challenging. The primary obstacle stems from the severe blur which leads to inaccuracies in the acquisition of initial camera poses through Structure-from-Motion, a critical aspect often overlooked by previous approaches. To address this challenge, we propose DeblurGS, a method to optimize sharp 3D Gaussian Splatting from motion-blurred images, even with the noisy camera pose initialization. We restore a fine-grained sharp scene by leveraging the remarkable reconstruction capability of 3D Gaussian Splatting. Our approach estimates the 6-Degree-of-Freedom camera motion for each blurry observation and synthesizes corresponding blurry renderings for the optimization process. Furthermore, we propose Gaussian Densification Annealing strategy to prevent the generation of inaccurate Gaussians at erroneous locations during the early training stages when camera motion is still imprecise. Comprehensive experiments demonstrate that our DeblurGS achieves state-of-the-art performance in deblurring and novel view synthesis for real-world and synthetic benchmark datasets, as well as field-captured blurry smartphone videos.

cs.CV↗

arXiv2024

ExBluRF: Efficient Radiance Fields for Extreme Motion Blurred Images

We present ExBluRF, a novel view synthesis method for extreme motion blurred images based on efficient radiance fields optimization. Our approach consists of two main components: 6-DOF camera trajectory-based motion blur formulation and voxel-based radiance fields. From extremely blurred images, we optimize the sharp radiance fields by jointly estimating the camera trajectories that generate the blurry images. In training, multiple rays along the camera trajectory are accumulated to reconstruct single blurry color, which is equivalent to the physical motion blur operation. We minimize the photo-consistency loss on blurred image space and obtain the sharp radiance fields with camera trajectories that explain the blur of all images. The joint optimization on the blurred image space demands painfully increasing computation and resources proportional to the blur size. Our method solves this problem by replacing the MLP-based framework to low-dimensional 6-DOF camera poses and voxel-based radiance fields. Compared with the existing works, our approach restores much sharper 3D scenes from challenging motion blurred views with the order of 10 times less training time and GPU memory consumption.

cs.CV↗

arXiv2022

Conformally mapped black hole effect in elastic curved continuum

We present a black hole effect by strategically leveraging a conformal mapping in elastic continuum with curved-space framework, which is less stringent compared to a Schwarzschild model transformed to isotropic refractive index profiles. In the conformal map approach, the 2D point singularity associated to the black hole effect is accomplished by physical plates with near-to-zero thickness. The analog gravity around the singularity results in highly confined energy and lagged timings within a branch cut of the conformal map. These effects are quantified both numerically and experimentally in reference to control trials in which the thickness is not modulated. The findings would deepen our understanding of the elastic analog in mimicking gravitational phenomena, as well as establish the elastic continuum framework for developing a generic design recipe in the presence of the index singularity. Geometric landscapes with elastically curved surfaces would be applicable in a variety of applications such as sensing, imaging, vibration isolation, and energy harvesting.

physics.class-ph↗

arXiv2022

DATa: Domain Adaptation-Aided Deep Table Detection Using Visual-Lexical Representations

Considerable research attention has been paid to table detection by developing not only rule-based approaches reliant on hand-crafted heuristics but also deep learning approaches. Although recent studies successfully perform table detection with enhanced results, they often experience performance degradation when they are used for transferred domains whose table layout features might differ from the source domain in which the underlying model has been trained. To overcome this problem, we present DATa, a novel Domain Adaptation-aided deep Table detection method that guarantees satisfactory performance in a specific target domain where few trusted labels are available. To this end, we newly design lexical features and an augmented model used for re-training. More specifically, after pre-training one of state-of-the-art vision-based models as our backbone network, we re-train our augmented model, consisting of the vision-based model and the multilayer perceptron (MLP) architecture. Using new confidence scores acquired based on the trained MLP architecture as well as an initial prediction of bounding boxes and their confidence scores, we calculate each confidence score more accurately. To validate the superiority of DATa, we perform experimental evaluations by adopting a real-world benchmark dataset in a source domain and another dataset in our target domain consisting of materials science articles. Experimental results demonstrate that the proposed DATa method substantially outperforms competing methods that only utilize visual representations in the target domain. Such gains are possible owing to the capability of eliminating high false positives or false negatives according to the setting of a confidence score threshold.

cs.LG↗

arXiv2022

Experimental data management platform for data-driven investigation of combinatorial alloy thin films

Experimental materials data are heterogeneous and include a variety of metadata for processing and characterization conditions, making the implementation of data-driven approaches for developing novel materials difficult. In this paper, we introduce the Thin-Film Alloy Database (TFADB), a materials data management platform, designed for combinatorially investigated thin-film alloys through various experimental tools. Using TFADB, researchers can readily upload, edit, and retrieve multidimensional experimental alloy data, such as composition, thickness, X-ray Diffraction, electrical resistivity, nanoindentation, and image data. Furthermore, composition-dependent properties from the database can easily be managed in a format adequate to be preprocessed for machine learning analyses. High flexibility of the software allows management of new types of materials data that can be potentially acquired from new combinatorial experiments.

cond-mat.mtrl-sci↗

arXiv2022

Combinatorial Discovery of Irradiation Damage Tolerant Nano-structured W-based alloys

One of the challenges in fusion reactors is the discovery of plasma facing materials capable of withstanding extreme conditions, such as radiation damage and high heat flux. Development of fusion materials can be a daunting task since vast combinations of microstructures and compositions need to be explored, each of which requires trial-and-error based irradiation experiments and materials characterizations. Here, we utilize combinatorial experiments that allow rapid and systematic characterizations of composition-microstructure dependent irradiation damage behaviors of nanostructured tungsten alloys. The combinatorial materials library of W-Re-Ta alloys was synthesized, followed by the high-throughput experiments for probing irradiation damages to the mechanical, thermal, and structural properties of the alloys. This highly efficient technique allows rapid identification of composition ranges with excellent damage tolerance. We find that the distribution of implanted He clusters can be significantly altered by the addition of Ta and Re, which play a critical role in determining property changes upon irradiation.

cond-mat.mtrl-sci↗