arXiv subjects

Changmin Lee

Publications and source records attributed to Changmin Lee.

At least 19 recordsLinked to original sources

arXiv2026

BiasEdit: A Training-Free Bias-Detect-and-Edit Framework for Learning Fair Visual Classifiers

Visual data from the Web power image classifiers, which often underpin many web services, such as recommendation and content moderation. However, the raw Web data often contain spurious correlations and social biases, and neural networks are known for their tendency to learn biases present in data. This can reinforce unfairness in web services and the web data, leading to a vicious cycle. In the context of image classification, networks learn bias attributes for a specific class when a majority of images contain the same attribute only for a given class. Hence, training a fair and debiased classifier from a biased dataset demands handling an imbalanced problem between a majority of images with bias attributes (bias-aligned samples) and a minority without (bias-conflict samples). In this work, we introduce BiasEdit, a modular framework that automatically detects bias attributes from the original dataset and edits them to construct a debiased dataset. Specifically, BiasEdit first detects unknown bias attributes via statistical dependence and mutual information analysis of visual-linguistic representations, and then explicitly edits those attributes using text-guided image editing to generate realistic bias-conflict samples. Unlike prior works that assume known bias attributes or relies on synthetic mixing, our method operates without manual annotations and can leverage off-the-shelf vision-language and editing models. BiasEdit addresses a fundamental challenge in Web-sourced visual AI, mitigating dataset-induced bias and achieving state-of-the-art debiasing performance even when training data are fully biased.

cs.CV↗

arXiv2026

From Volume to Value: Preference-Aligned Memory Construction for On-Device RAG

With the rapid emergence of personal AI agents based on Large Language Models (LLMs), implementing them on-device has become essential for privacy and responsiveness. To handle the inherently personal and context-dependent nature of real-world requests, such agents must ground their generation in device-resident personal context. However, under tight memory budgets, the core bottleneck is what to store so that retrieval remains aligned with the user. We propose EPIC (Efficient Preference-aligned Index Construction), which focuses on user preferences as a compact and stable form of personal context and integrates them throughout the RAG pipeline. EPIC selectively retains preference-relevant information from raw data and aligns retrieval toward preference-aligned contexts. Across four benchmarks covering conversations, debates, explanations, and recommendations, EPIC reduces indexing memory by 2,404 times, improves preference-following accuracy by 18.79 %p, and achieves 32.17 times lower retrieval latency over the best-performing baseline. In on-device experiments, EPIC maintains under 1 MB memory and achieves 5.21 to 29.35 ms/query latency across three platforms, while supporting streaming updates under preference drift. Our code and data are available at https://github.com/UbiquitousAILab/EPIC.

cs.CL↗

arXiv2026

PhysHanDI: Physics-Based Reconstruction of Hand-Deformable Object Interactions

While existing methods for reconstructing hand-object interactions have made impressive progress, they either focus on rigid or part-wise rigid objects-limiting their ability to model real-world objects (e.g., cloth, stuffed animals) that exhibit highly non-rigid deformations-or model deformable objects without full 3D hand reconstruction. To bridge this gap, we present PhysHanDI (Physics-based Reconstruction of Hand and Deformable Object Interactions), a framework that enables full 3D reconstruction of both interacting hands and non-rigid objects. Our key idea is to physically simulate object deformations driven by forces induced from densely reconstructed 3D hand motions, ensuring that the reconstructed object dynamics are both physically plausible and coherent with the interacting hand movements. Furthermore, we demonstrate that such simulation of object deformations can, in turn, refine and improve hand reconstruction via inverse physics. In experiments, PhysHanDI outperforms the state-of-the-art baseline across reconstruction and future prediction.

cs.CV↗

arXiv2026

StyleID: A Perception-Aware Dataset and Metric for Stylization-Agnostic Facial Identity Recognition

Creative face stylization aims to render portraits in diverse visual idioms such as cartoons, sketches, and paintings while retaining recognizable identity. However, current identity encoders, which are typically trained and calibrated on natural photographs, exhibit severe brittleness under stylization. They often mistake changes in texture or color palette for identity drift or fail to detect geometric exaggerations. This reveals the lack of a style-agnostic framework to evaluate and supervise identity consistency across varying styles and strengths. To address this gap, we introduce StyleID, a human perception-aware dataset and evaluation framework for facial identity under stylization. StyleID comprises two datasets: (i) StyleBench-H, a benchmark that captures human same-different verification judgments across diffusion- and flow-matching-based stylization at multiple style strengths, and (ii) StyleBench-S, a supervision set derived from psychometric recognition-strength curves obtained through controlled two-alternative forced-choice (2AFC) experiments. Leveraging StyleBench-S, we fine-tune existing semantic encoders to align their similarity orderings with human perception across styles and strengths. Experiments demonstrate that our calibrated models yield significantly higher correlation with human judgments and enhanced robustness for out-of-domain, artist drawn portraits. All of our datasets, code, and pretrained models are publicly available at https://kwanyun.github.io/StyleID_page/

cs.GR↗

arXiv2025

MPMAvatar: Learning 3D Gaussian Avatars with Accurate and Robust Physics-Based Dynamics

While there has been significant progress in the field of 3D avatar creation from visual observations, modeling physically plausible dynamics of humans with loose garments remains a challenging problem. Although a few existing works address this problem by leveraging physical simulation, they suffer from limited accuracy or robustness to novel animation inputs. In this work, we present MPMAvatar, a framework for creating 3D human avatars from multi-view videos that supports highly realistic, robust animation, as well as photorealistic rendering from free viewpoints. For accurate and robust dynamics modeling, our key idea is to use a Material Point Method-based simulator, which we carefully tailor to model garments with complex deformations and contact with the underlying body by incorporating an anisotropic constitutive model and a novel collision handling algorithm. We combine this dynamics modeling scheme with our canonical avatar that can be rendered using 3D Gaussian Splatting with quasi-shadowing, enabling high-fidelity rendering for physically realistic animations. In our experiments, we demonstrate that MPMAvatar significantly outperforms the existing state-of-the-art physics-based avatar in terms of (1) dynamics modeling accuracy, (2) rendering accuracy, and (3) robustness and efficiency. Additionally, we present a novel application in which our avatar generalizes to unseen interactions in a zero-shot manner-which was not achievable with previous learning-based methods due to their limited simulation generalizability. Our project page is at: https://KAISTChangmin.github.io/MPMAvatar/

cs.GR↗

arXiv2025

Performance Analysis of IEEE 802.11bn with Coordinated TDMA on Real-Time Applications

Wi-Fi plays a crucial role in connecting electronic devices and providing communication services in everyday life. Recently, there has been a growing demand for services that require low-latency communication, such as real-time applications. The latest amendments to Wi-Fi, IEEE 802.11bn, are being developed to address these demands with technologies such as the multiple access point coordination (MAPC). In this paper, we demonstrate that coordinated TDMA (Co-TDMA), one of the MAPC techniques, effectively reduces the latency of transmitting time-sensitive traffic. In particular, we focus on worst-case latency and jitter, which are key metrics for evaluating the performance of real-time applications. We first introduce a Co-TDMA scheduling strategy. We then investigate how this scheduling strategy impacts latency under varying levels of network congestion and traffic volume characteristics. Finally, we validate our findings through system-level simulations. Our simulation results demonstrate that Co-TDMA effectively mitigates jitter and worst-case latency for low-latency traffic, with the latter exhibiting an improvement of approximately 24%.

cs.IT↗

arXiv2025

Three-Dimensional Channel Modeling for Molecular Communications in Tubular Environments with Heterogeneous Boundary Conditions

Molecular communication (MC), one of the emerging techniques in the field of communication, is entering a new phase following several decades of foundational research. Recently, attention has shifted toward MC in liquid media, particularly within tubular environments, due to novel application scenarios. The spatial constraints of such environments make accurate modeling of molecular movement in tubes more challenging than in traditional free-space channels. In this paper, we propose a three-dimensional channel model for molecular communications with an absorbing ring-shaped receiver in a tubular environment. To the best of our knowledge, this is the first theoretical study to model the impact of an absorbing ring-shaped receiver on the channel response in tube-based MC systems. The problem is formulated as a partial differential equation with heterogeneous boundary conditions, and an approximate solution is derived under flow-dominated conditions. The accuracy of the proposed model is validated through particle-based simulations. We anticipate that the results of this study will contribute to the design of practical MC systems in real-world tubular environments.

cs.IT↗

arXiv2025

Kanana: Compute-efficient Bilingual Language Models

We introduce Kanana, a series of bilingual language models that demonstrate exceeding performance in Korean and competitive performance in English. The computational cost of Kanana is significantly lower than that of state-of-the-art models of similar size. The report details the techniques employed during pre-training to achieve compute-efficient yet competitive models, including high quality data filtering, staged pre-training, depth up-scaling, and pruning and distillation. Furthermore, the report outlines the methodologies utilized during the post-training of the Kanana models, encompassing supervised fine-tuning and preference optimization, aimed at enhancing their capability for seamless interaction with users. Lastly, the report elaborates on plausible approaches used for language model adaptation to specific scenarios, such as embedding, retrieval augmented generation, and function calling. The Kanana model series spans from 2.1B to 32.5B parameters with 2.1B models (base, instruct, embedding) publicly released to promote research on Korean language models.

cs.CL↗

arXiv2025

Light-induced reorientation transition in an antiferromagnetic semiconductor

Due to the lack of a net magnetic moment, antiferromagnets possess a unique robustness to external magnetic fields and are thus predicted to play an important role in future magnetic technologies. However, this robustness also makes them quite difficult to control, and the development of novel methods to manipulate these systems with external stimuli is a fundamental goal of antiferromagnetic spintronics. In this work, we report evidence for a metastable reorientation of the order parameter in an antiferromagnetic semiconductor triggered by an ultrafast quench of the equilibrium order via photoexcitation above the band gap. The metastable state forms less than 10 ps after the excitation pulse, and persists for longer than 150 ps before decaying to the ground state via thermal fluctuations. Importantly, this transition cannot be induced thermodynamically, and requires the system to be driven out of equilibrium. Broadly speaking, this phenomenology is ultimately the result of large magnetoelastic coupling in combination with a relatively low symmetry of the magnetic ground state. Since neither of these properties are particularly uncommon in magnetic materials, the observations presented here imply a generic path toward novel device technology enabled by ultrafast dynamics in antiferromagnets.

cond-mat.mtrl-sci↗

arXiv2024

SIn-NeRF2NeRF: Editing 3D Scenes with Instructions through Segmentation and Inpainting

TL;DR Perform 3D object editing selectively by disentangling it from the background scene. Instruct-NeRF2NeRF (in2n) is a promising method that enables editing of 3D scenes composed of Neural Radiance Field (NeRF) using text prompts. However, it is challenging to perform geometrical modifications such as shrinking, scaling, or moving on both the background and object simultaneously. In this project, we enable geometrical changes of objects within the 3D scene by selectively editing the object after separating it from the scene. We perform object segmentation and background inpainting respectively, and demonstrate various examples of freely resizing or moving disentangled objects within the three-dimensional space.

cs.CV↗

arXiv2024

Controlled Text Generation for Black-box Language Models via Score-based Progressive Editor

Controlled text generation is very important for the practical use of language models because it ensures that the produced text includes only the desired attributes from a specific domain or dataset. Existing methods, however, are inapplicable to black-box models or suffer a significant trade-off between controlling the generated text and maintaining its fluency. This paper introduces the Score-based Progressive Editor (ScoPE), a novel approach designed to overcome these issues. ScoPE modifies the context at the token level during the generation process of a backbone language model. This modification guides the subsequent text to naturally include the target attributes. To facilitate this process, ScoPE employs a training objective that maximizes a target score, thoroughly considering both the ability to guide the text and its fluency. Experimental results on diverse controlled generation tasks demonstrate that ScoPE can effectively regulate the attributes of the generated text while fully utilizing the capability of the backbone large language models. Our codes are available at \url{https://github.com/ysw1021/ScoPE}.

cs.CL↗

arXiv2024

100 Gbps Indoor Access and 4.8 Gbps Outdoor Point-to-Point LiFi Transmission Systems using Laser-based Light Sources

In this paper, we demonstrate the communication capabilities of light-fidelity (LiFi) systems based on highbrightness and high-bandwidth integrated laser-based sources in a surface mount device (SMD) packaging platform. The laserbased source is able to deliver 450 lumens of white light illumination and the resultant light brightness is over 1000 cd mm2. It is demonstrated that a wavelength division multiplexing (WDM) LiFi system with ten parallel channels is able to deliver over 100 Gbps data rate with the assistance of Volterra filter-based nonlinear equalisers. In addition, an aggregated transmission data rate of 4.8 Gbps has been achieved over a link distance of 500 m with the same type of SMD light source. This work demonstrates the scalability of LiFi systems that employ laserbased light sources, particularly in their capacity to enable highspeed short range, as well as long-range data transmission.

eess.SY↗

arXiv2023

Universal spin wavepacket transport in van der Waals antiferromagnets

Antiferromagnets (AFMs) are promising platforms for the transmission of quantum information via magnons (the quanta of spin waves), offering advantages over ferromagnets with regard to dissipation, speed of response, and immunity to external fields. Recently, it was shown that in the insulating van der Waals (vdW) semiconductor, CrSBr, strong spin-exciton coupling enables readout of magnon density and propagation using photons of visible light. This exciting observation came with a puzzle: photogenerated magnons were observed to propagate 10$^3$ times faster than the velocity inferred from neutron scattering, leading to a conjecture that spin wavepackets are carried along by coupling to much faster elastic modes. Here we show, through a combination of theory and experiment, that the propagation mechanism is, instead, coupling within the magnetic degrees of freedom through long range dipole-dipole coupling. This mechanism is an inevitable consequence of Maxwell's equations, and as such, will dominate the propagation of spin at long wavelengths in the entire class of vdW magnets currently under intense investigation. Moreover, identifying the mechanism of spin propagation provides a set of optimization rules, as well as caveats, that are essential for any future applications of these promising systems.

cond-mat.str-el↗

arXiv2022

The Internet of Bio-Nano Things in Blood Vessels: System Design and Prototypes

In this paper, we investigate the Internet of Bio-Nano Things (IoBNT) which relates to networks formed by molecular communications. By providing a means of communication through the ubiquitously connected blood vessels (arteries, veins, and capillaries), molecular communication-based IoBNT enables a host of new eHealth applications. For example, an organ monitoring sensor can transfer internal body signals through the IoBNT for health monitoring applications. We empirically show that blood vessel channels introduce a new set of challenges for the design of molecular communication systems in comparison to free-space channels. We then propose cylindrical duct channel models and discuss the corresponding system designs conforming to the channel characteristics. Furthermore, based on prototype implementations, we confirm that molecular communication techniques can be utilized for composing the IoBNT. We believe that the promising results presented in this work, together with the rich research challenges that lie ahead, are strong indicators that IoBNT with molecular communications can drive novel applications for emerging eHealth systems.

cs.ET↗

arXiv2022

Spin wavepackets in the Kagome ferromagnet Fe$_3$Sn$_2$: propagation and precursors

The propagation of spin waves in magnetically ordered systems has emerged as a potential means to shuttle quantum information over large distances. Conventionally, the arrival time of a spin wavepacket at a distance, $d$, is assumed to be determined by its group velocity, $v_g$. He we report time-resolved optical measurements of wavepacket propagation in the Kagome ferromagnet Fe$_3$Sn$_2$ that demonstrate the arrival of spin information at times significantly less than $d/v_g$. We show that this spin wave "precursor" originates from the interaction of light with the unusual spectrum of magnetostatic modes in Fe$_3$Sn$_2$. Related effects may have far-reaching consequences toward realizing long-range, ultrafast spin wave transport in both ferromagnetic and antiferromagnetic systems.

cond-mat.mtrl-sci↗

arXiv2022

Spin-carrier coupling induced ferromagnetism and giant resistivity peak in EuCd$_2$P$_2$

EuCd$_2$P$_2$ is notable for its unconventional transport: upon cooling the metallic resistivity changes slope and begins to increase, ultimately 100-fold, before returning to its metallic value. Surprisingly, this giant peak occurs at 18K, well above the Néel temperature ($T_N$) of 11.5K. Using a suite of sensitive probes of magnetism, including resonant x-ray scattering and magneto-optical polarimetry, we have discovered that ferromagnetic order onsets above $T_N$ in the temperature range of the resistivity peak. The observation of inverted hysteresis in this regime shows that ferromagnetism is promoted by coupling of localized spins and itinerant carriers. The resulting carrier localization is confirmed by optical conductivity measurements.

cond-mat.str-el↗

arXiv2022

Band Structure Driven Thermoelectric Response of Topological Semiconductor ZrTe$_5$

We report a transport, thermodynamic, and spectroscopic study of the recently identified topological semiconductor ZrTe$_5$ with a focus on elucidating the connections between its band structure and unusual thermoelectric properties. Using time and angle resolved photoemission spectroscopy (tr-ARPES) we observe a small electronic band gap and temperature dependent Fermi level which traverses from a single valence to conduction band with lowering temperature, consistent with previous reports. This low temperature Fermi surface closely matches that derived from quantum oscillations, suggesting it is reflective of the bulk electronic structure. The Seebeck and low field Nernst response is characterized by an unusually large and non-monotonic temperature evolution. We find this can be quantitatively explained using a semiclassical model based on the observed band character and a linear temperature shifting of the Fermi level. Additionally, we observe a large, non-saturating enhancement of both thermoelectric coefficients in magnetic field. We show this can be captured by the Zeeman energy associated with a large effective $g$-factor of 25.8 consistent with that derived from Lifshitz-Kosevich analysis of the quantum oscillations. Together these observations provide a comprehensive picture of ZrTe$_{5}$ as a model high mobility semiconductor and potential platform for significant magnetic field driven thermoelectricity.

cond-mat.mtrl-sci↗

arXiv2022

Observation of a phase transition within the domain walls of ferromagnetic Co3Sn2S2

The ferromagnetic phase of Co$_3$Sn$_2$S$_2$ is widely considered to be a topological Weyl semimetal, with evidence for momentum-space monopoles of Berry curvature from transport and spectroscopic probes. As the bandstructure is highly sensitive to the magnetic order, attention has focused on anomalies in magnetization, susceptibility and transport measurements that are seen well below the Curie temperature, leading to speculation that a "hidden" phase coexists with ferromagnetism. Here we report spatially-resolved measurements by Kerr effect microscopy that identify this phase. We find that the anomalies coincide with a deep minimum in domain wall (DW) mobility, indicating a crossover between two regimes of DW propagation. We demonstrate that this crossover is a manifestation of a 2D phase transition that occurs within the DW, in which the magnetization texture changes from continuous rotation to unidirectional variation. We propose that the existence of this 2D transition deep within the ferromagnetic state of the bulk is a consequence of a giant quality factor for magnetocrystalline anisotropy unique to this compound. This work broadens the horizon of the conventional binary classification of DWs into Bloch and Néel walls, and suggests new strategies for manipulation of domain walls and their role in electron and spin transport.

cond-mat.mes-hall↗