arXiv subjects

Donghoon Kim

Publications and source records attributed to Donghoon Kim.

At least 37 records · Page 2Linked to original sources

arXiv2026

GazeShift: Unsupervised Gaze Estimation and Dataset for VR

Gaze estimation is instrumental in modern virtual reality (VR) systems. Despite significant progress in remote-camera gaze estimation, VR gaze research remains constrained by data scarcity, particularly the lack of large-scale, accurately labeled datasets captured with the off-axis camera configurations typical of modern headsets. Gaze annotation is difficult since fixation on intended targets cannot be guaranteed. To address these challenges, we introduce VRGaze, the first large-scale off-axis gaze estimation dataset for VR, comprising 2.1 million near-eye infrared images collected from 68 participants. We further propose GazeShift, an attention-guided unsupervised framework for learning gaze representations without labeled data. Unlike prior redirection-based methods that rely on multi-view or 3D geometry, GazeShift is tailored to near-eye imagery, achieving effective gaze-appearance disentanglement in a compact, real-time model. GazeShift embeddings can be optionally adapted to individual users via lightweight few-shot calibration, achieving a 1.84° mean error on VRGaze. On the remote-camera MPIIGaze dataset, the model achieves a 7.15° person-agnostic error, doing so with 10x fewer parameters and 35x fewer FLOPs than baseline methods. Deployed natively on a VR headset GPU, inference takes only 5 ms. Combined with demonstrated robustness to illumination changes, these results highlight GazeShift as a label-efficient, real-time solution for VR gaze tracking. Project code and the VRGaze dataset are released at https://github.com/gazeshift3/gazeshift

cs.CV↗

arXiv2026

Universal purification dynamics in real non-unitary quantum processes

We study purification dynamics in monitored quantum processes governed by ensembles of quantum circuits in different random-matrix symmetry classes. We analyze the universal aspects that emerge away from the measurement induced phase transition and inside the volume/weak measurement phase and in the scaling limit of large time and Hilbert space dimension. We present two toy models that reveal two complementary visions and provide quantitative access to universal scaling: i) a discrete-time dynamic in which each time step corresponds to multiplication by a Gaussian random matrix; ii) weak continuous-time monitoring that induces a Dyson brownian motion of the eigenvalues of the density matrix. The first approach provides an algebraic characterization based on rotational invariance emerging in Kraus's operator space, focusing in particular on the unitary and orthogonal cases, respectively $β=2$ and $β=1$, with $β$ the Dyson random-matrix index. The second approach, on the other hand, allows for a unified treatment for any $β$, thanks to the mapping of the Fokker-Planck evolution of eigenvalues onto the Calogero-Sutherland integrable Hamiltonian diagonalized in terms of Jack polynomials. We provide explicit expressions for the universal decrease of Rényi entropies. We show that, approaching the universal scaling limit, numerical simulations of different models agree with each other and with our theoretical predictions. Our results clarify the existence of different classes of universality for the purification process in hybrid quantum systems, accessible in random circuit architectures and weak measurement protocols.

quant-ph↗

arXiv2026

Adaptive Capacity Allocation for Vision Language Action Fine-tuning

Vision language action models (VLAs) are increasingly used for Physical AI, but deploying a pre-trained VLA model to unseen environments, embodiments, or tasks still requires adaptation. Parameter-efficient fine-tuning (PEFT), especially LoRA, is common for VLA policies, yet the exposed capacity knob, the rank, does not transfer uniformly: robotics transfer exhibits a higher and task-varying intrinsic rank than language fine-tuning. Small ranks suffice for LLMs (e.g., $r \in \{4, 8\}$), while spectral analyses indicate VLAs may require much larger ranks (e.g., $r \approx 128$) or near-full rank, a mismatch that worsens in multi-task settings. We present LoRA-SP (Select-Prune), a rank-adaptive fine-tuning method that replaces fixed-rank updates with input- and layer-wise capacity. LoRA-SP uses an SVD-style parameterization with a small router whose nonnegative scores act as singular values over a shared vector bank. The active set is chosen by an energy target on the cumulative squared scores $E(k) \ge η$, providing a direct link to approximation error via our spectral analysis. During training, $η$ concentrates energy on a few directions and teaches the router to rely on fewer vectors while preserving accuracy. This yields compact adapters that reduce cross-task interference and improve generalization. On four real-robot manipulation tasks collected on an unseen AgileX PiPER arm, across two VLA backbones ($π_0$ and SmolVLA), LoRA-SP matches or exceeds full fine-tuning with far fewer trainable parameters, and improves multi-task success by up to 31.6% over standard LoRA while remaining robust to rank choice.

cs.RO↗

arXiv2026

Variational preparation and characterization of chiral spin liquids in quantum circuits

Quantum circuits have been shown to be a fertile ground for realizing long-range entangled phases of matter. While various quantum double models with non-chiral topological order have been theoretically investigated and experimentally implemented, the realization and characterization of chiral topological phases have remained less explored. Here we show that chiral topological phases in spin systems, i.e., chiral spin liquids, can be prepared in quantum circuits using the variational quantum eigensolver (VQE) framework. On top of the VQE ground state, signatures of the chiral topological order are revealed using the recently proposed tangent space excitation ansatz for quantum circuits. We show that, both topological ground state degeneracy and the chiral edge mode can be faithfully captured by this approach. We demonstrate our approach using the Kitaev honeycomb model, finding excellent agreement of low-energy excitation spectrum on quantum circuits with exact solution in all topological sectors. Further applying this approach to a non-exactly solvable chiral spin liquid model on square lattice, the results suggest this approach works well even when the topological sectors are not exactly known.

cond-mat.str-el↗

arXiv2025

Spectral Small-Incremental-Entangling: Breaking Quasi-Polynomial Complexity Barriers in Long-Range Interacting Systems

How the detailed structure of quantum complexity emerges from quantum dynamics remains a fundamental challenge highlighted by advances in quantum simulators and information processing. The celebrated Small-Incremental-Entangling (SIE) theorem provides a universal constraint on the rate of entanglement generation, yet it leaves open the problem of fully characterizing fine entanglement structures. Here we introduce the concept of Spectral-Entangling strength, which captures the structural entangling power of an operator, and establish a spectral SIE theorem: a universal speed limit for R'enyi entanglement growth at $α\ge 1/2$, revealing a robust $1/s^2$ decay threshold in the entanglement spectrum. Remarkably, our bound at $α=1/2$ is both qualitatively and quantitatively optimal, defining the universal threshold beyond which entanglement growth becomes unbounded. This exposes the detailed structure of Schmidt coefficients and enables rigorous truncation-based error control, linking entanglement structure to computational complexity. Building on this, we derive a generalized entanglement area law under an adiabatic-path condition, extending a central principle of quantum many-body physics to general interactions. As a concrete application, we show that one-dimensional long-range interacting systems admit polynomial bond-dimension approximations for ground, time-evolved, and thermal states, thereby closing the long-standing quasi-polynomial gap and demonstrating that such systems can be simulated efficiently with tensor-network methods. By explicitly controlling R'enyi entanglement, we obtain a rigorous, a priori error guarantee for the time-dependent density-matrix renormalization-group algorithm. Overall, our results extend the SIE theorem to the spectral domain and establish a unified framework that unveils the detailed and universal structure underlying quantum complexity.

quant-ph↗

arXiv2025

Multimodal Safety Evaluation in Generative Agent Social Simulations

Can generative agents be trusted in multimodal environments? Despite advances in large language and vision-language models that enable agents to act autonomously and pursue goals in rich settings, their ability to reason about safety, coherence, and trust across modalities remains limited. We introduce a reproducible simulation framework for evaluating agents along three dimensions: (1) safety improvement over time, including iterative plan revisions in text-visual scenarios; (2) detection of unsafe activities across multiple categories of social situations; and (3) social dynamics, measured as interaction counts and acceptance ratios of social exchanges. Agents are equipped with layered memory, dynamic planning, multimodal perception, and are instrumented with SocialMetrics, a suite of behavioral and structural metrics that quantifies plan revisions, unsafe-to-safe conversions, and information diffusion across networks. Experiments show that while agents can detect direct multimodal contradictions, they often fail to align local revisions with global safety, reaching only a 55 percent success rate in correcting unsafe plans. Across eight simulation runs with three models - Claude, GPT-4o mini, and Qwen-VL - five agents achieved average unsafe-to-safe conversion rates of 75, 55, and 58 percent, respectively. Overall performance ranged from 20 percent in multi-risk scenarios with GPT-4o mini to 98 percent in localized contexts such as fire/heat with Claude. Notably, 45 percent of unsafe actions were accepted when paired with misleading visuals, showing a strong tendency to overtrust images. These findings expose critical limitations in current architectures and provide a reproducible platform for studying multimodal safety, coherence, and social dynamics.

cs.AI↗

arXiv2025

Post-Training Quantization via Residual Truncation and Zero Suppression for Diffusion Models

Diffusion models achieve high-quality image generation but face deployment challenges due to their high computational requirements. Although 8-bit outlier-aware post-training quantization (PTQ) matches full-precision performance, extending PTQ to 4 bits remains challenging. Larger step sizes in 4-bit quantization amplify rounding errors in dense, low-magnitude activations, leading to the loss of fine-grained textures. We hypothesize that not only outliers but also small activations are critical for texture fidelity. To this end, we propose Quantization via Residual Truncation and Zero Suppression (QuaRTZ), a 4-bit PTQ scheme for diffusion models. QuaRTZ applies 8-bit min-max quantization for outlier handling and compresses to 4 bits via leading-zero suppression to retain LSBs, thereby preserving texture details. Our approach reduces rounding errors and improves quantization efficiency by balancing outlier preservation and LSB precision. Both theoretical derivations and empirical evaluations demonstrate the generalizability of QuaRTZ across diverse activation distributions. Notably, 4-bit QuaRTZ achieves an FID of 6.98 on FLUX.1-schnell, outperforming SVDQuant that requires auxiliary FP16 branches.

cs.CV↗

arXiv2025

Fuzzy-Based Control Method for Autonomous Spacecraft Inspection with Minimal Fuel Consumption

This study explores an energy-efficient control strategy for spacecraft inspection using a fuzzy inference system combined with a bio-inspired optimization technique to incorporate learning capability into the control process. The optimized fuzzy controller produces a minimally fuel-consuming force while maintaining reliable inspection within constraints, such as illumination, restricted field of view, thrust limits, and safe regions. The performance of the proposed control strategy is validated through Monte Carlo simulations.

eess.SY↗

arXiv2025

Modeling and Control Framework for Autonomous Space Manipulator Handover Operations

Autonomous space robotics is poised to play a vital role in future space missions, particularly for In-space Servicing, Assembly, and Manufacturing (ISAM). A key capability in such missions is the Robot-to-Robot (R2R) handover of mission-critical objects. This work presents a dynamic model of a dual-arm space manipulator system and compares various tracking control laws. The key contributions of this work are the development of a cooperative manipulator dynamic model and the comparative analysis of control laws to support autonomous R2R handovers in ISAM scenarios.

cs.RO↗

arXiv2025

Understanding and Utilizing Dynamic Coupling in Free-Floating Space Manipulators for On-Orbit Servicing

This study proposes a dynamic coupling-informed trajectory optimization algorithm for free-floating space manipulator systems (SMSs). Dynamic coupling between the base and the manipulator arms plays a critical role in influencing the system's behavior. While prior research has predominantly focused on minimizing this coupling, often overlooking its potential advantages, this work investigates how dynamic coupling can instead be leveraged to improve trajectory planning. Singular value decomposition (SVD) of the dynamic coupling matrix is employed to identify the dominant components governing coupling behavior. A quantitative metric is then formulated to characterize the strength and directionality of the coupling and is incorporated into a trajectory optimization framework. To assess the feasibility of the optimized trajectory, a sliding mode control-based tracking controller is designed to generate the required joint torque inputs. Simulation results demonstrate that explicitly accounting for dynamic coupling in trajectory planning enables more informed and potentially more efficient operation, offering new directions for the control of free-floating SMSs.

cs.RO↗

arXiv2025

Adapting Biological Reflexes for Dynamic Reorientation in Space Manipulator Systems

Robotic arms mounted on spacecraft, known as space manipulator systems (SMSs), are critical for enabling on-orbit assembly, satellite servicing, and debris removal. However, controlling these systems in microgravity remains a significant challenge due to the dynamic coupling between the manipulator and the spacecraft base. This study explores the potential of using biological inspiration to address this issue, focusing on animals, particularly lizards, that exhibit mid-air righting reflexes. Based on similarities between SMSs and these animals in terms of behavior, morphology, and environment, their air-righting motion trajectories are extracted from high-speed video recordings using computer vision techniques. These trajectories are analyzed within a multi-objective optimization framework to identify the key behavioral goals and assess their relative importance. The resulting motion profiles are then applied as reference trajectories for SMS control, with baseline controllers used to track them. The findings provide a step toward translating evolved animal behaviors into interpretable, adaptive control strategies for space robotics, with implications for improving maneuverability and robustness in future missions.

cs.RO↗

arXiv2025

Thermal Area Law in Long-Range Interacting Systems

The area law of the bipartite information measure characterizes one of the most fundamental aspects of quantum many-body physics. In thermal equilibrium, the area law for the mutual information universally holds at arbitrary temperatures as long as the systems have short-range interactions. In systems with power-law decaying interactions, $r^{-α}$ ($r$: distance), conditions for the thermal area law are elusive. In this work, we aim to clarify the optimal condition $α> α_c$ such that the thermal area law universally holds. A standard approach to considering the conditions is to focus on the magnitude of the boundary interaction between two subsystems. However, we find here that the thermal area law is more robust than this conventional argument suggests. We show the optimal threshold for the thermal area law by $α_c= (D+1)/2$ ($D$: the spatial dimension of the lattice), assuming a power-law decay of the clustering for the bipartite correlations. Remarkably, this condition encompasses even the thermodynamically unstable regimes $α< D$. We verify this condition numerically, finding that it is qualitatively accurate for both integrable and non-integrable systems. Unconditional proof of the thermal area law is possible by developing the power-law clustering theorem for $α> D$ above a threshold temperature. Furthermore, the numerical calculation for the logarithmic negativity shows that the same criterion $α> (D+1)/2$ applies to the thermal area law for quantum entanglement.

quant-ph↗

arXiv2025

Visually Guided Decoding: Gradient-Free Hard Prompt Inversion with Language Models

Text-to-image generative models like DALL-E and Stable Diffusion have revolutionized visual content creation across various applications, including advertising, personalized media, and design prototyping. However, crafting effective textual prompts to guide these models remains challenging, often requiring extensive trial and error. Existing prompt inversion approaches, such as soft and hard prompt techniques, are not so effective due to the limited interpretability and incoherent prompt generation. To address these issues, we propose Visually Guided Decoding (VGD), a gradient-free approach that leverages large language models (LLMs) and CLIP-based guidance to generate coherent and semantically aligned prompts. In essence, VGD utilizes the robust text generation capabilities of LLMs to produce human-readable prompts. Further, by employing CLIP scores to ensure alignment with user-specified visual concepts, VGD enhances the interpretability, generalization, and flexibility of prompt generation without the need for additional training. Our experiments demonstrate that VGD outperforms existing prompt inversion techniques in generating understandable and contextually relevant prompts, facilitating more intuitive and controllable interactions with text-to-image models.

cs.AI↗

arXiv2025

A Novel Analysis Framework for Microstructural Characterization of Ferroelectric Hafnia: Experimental Validation and Application

Herein, we present a novel analysis framework for grain size profile of ferroelectric hafnia to tackle critical shortcomings inherent in the current microstructural analysis. We vastly enhanced visibility of grains with ion beam treatment and performed accurate grain segmentation using deep neural network (DNN). By leveraging our new method, we discovered unexpected discrepancies that contradict previous results, such as deposition temperature (Tdep) and post-metallization annealing (PMA) dependence of grain size statistics, prompting us to reassess earlier interpretations. Combining microstructural analysis with electrical tests, we found that grain size reduction had both positive and negative outcomes: it caused significant diminishing of die-to-die variation (~68 % decrease in standard deviation) in coercive field (Ec), while triggering an upsurge in leakage current. These uncovered results signify robustness of our method in characterization of ferroelectric hafnia for in-depth examination of both device variability and reliability.

cond-mat.mtrl-sci↗

arXiv2025

Provably Efficient Simulation of 1D Long-Range Interacting Systems at Any Temperature

We introduce a method that ensures efficient computation of one-dimensional quantum systems with long-range interactions across all temperatures. Our algorithm operates within a quasi-polynomial runtime for inverse temperatures up to $β={\rm poly}(\ln(n))$. At the core of our approach is the Density Matrix Renormalization Group algorithm, which typically does not guarantee efficiency. We have created a new truncation scheme for the matrix product operator of the quantum Gibbs states, which allows us to control the error analytically. Additionally, our method can be applied to simulate the time evolution of systems with long-range interactions, achieving significantly better precision than that offered by the Lieb-Robinson bound.

quant-ph↗

arXiv2025

Study on Real-Time Road Surface Reconstruction Using Stereo Vision

Road surface reconstruction plays a crucial role in autonomous driving, providing essential information for safe and smooth navigation. This paper enhances the RoadBEV [1] framework for real-time inference on edge devices by optimizing both efficiency and accuracy. To achieve this, we proposed to apply Isomorphic Global Structured Pruning to the stereo feature extraction backbone, reducing network complexity while maintaining performance. Additionally, the head network is redesigned with an optimized hourglass structure, dynamic attention heads, reduced feature channels, mixed precision inference, and efficient probability volume computation. Our approach improves inference speed while achieving lower reconstruction error, making it well-suited for real-time road surface reconstruction in autonomous driving.

cs.CV↗

arXiv2025

Electrical control of a Kondo spin screening cloud

Quantitative analysis of quantum many-body systems, consisting of numerous itinerant electrons that interact with localized spins or electrons, is a long-standing issue. The Kondo cloud, a quantum many-body object of conduction electrons that screens a single localized spin, is the building block of such strongly correlated electronic systems. While quantitative analysis of the Kondo cloud associated with a single magnetic impurity is well established for uniform conduction electrons, the fundamental properties of a deformed Kondo cloud influenced by conduction electrons with a modulated density of states remain unsolved. Here we report engineering of the Kondo cloud deformation by confining a part of the cloud into a quantum box called the Kondo box that mimics realistic material systems. We demonstrate quantitative control of the Kondo cloud by developing a way of tuning quantum interference in the box and monitoring the Kondo entanglement. The temperature dependence of the entanglement reveals counterintuitively that the cloud shape is altered mainly outside the box although the quantum interference in the box is tuned. Our work provides a way to simulate various strongly correlated systems by integrating the Kondo cloud, which is not possible in the current theoretical framework.

cond-mat.mes-hall↗

arXiv2024

Deep Learning-Based Prediction of PET Amyloid Status Using Multi-Contrast MRI

Identifying amyloid-beta positive patients is crucial for determining eligibility for Alzheimer's disease (AD) clinical trials and new disease-modifying treatments, but currently requires PET or CSF sampling. Previous MRI-based deep learning models for predicting amyloid positivity, using only T1w sequences, have shown moderate performance. We trained deep learning models to predict amyloid PET positivity and evaluated whether multi-contrast inputs improve performance. A total of 4,058 exams with multi-contrast MRI and PET-based quantitative amyloid deposition were obtained from three public datasets: the Alzheimer's Disease Neuroimaging Initiative (ADNI), the Open Access Series of Imaging Studies 3 (OASIS3), and the Anti-Amyloid Treatment in Asymptomatic Alzheimer's Disease (A4). Two separate EfficientNet models were trained for amyloid positivity prediction: one with only T1w images and the other with both T1w and T2-FLAIR images as network inputs. The area under the curve (AUC), accuracy, sensitivity, and specificity were determined using an internal held-out test set. The trained models were further evaluated using an external test set. In the held-out test sets, the T1w and T1w+T2FLAIR models demonstrated AUCs of 0.62 (95% CI: 0.60, 0.64) and 0.67 (95% CI: 0.64, 0.70) (p = 0.006); accuracies were 61% (95% CI: 60%, 63%) and 64% (95% CI: 62%, 66%) (p = 0.008); sensitivities were 0.88 and 0.71; and specificities were 0.23 and 0.53, respectively. The trained models showed similar performance in the external test set. Performance of the current model on both test sets exceeded that of the publicly available model. In conclusion, the use of multi-contrast MRI, specifically incorporating T2-FLAIR in addition to T1w images, significantly improved the predictive accuracy of PET-determined amyloid status from MRI scans using a deep learning approach.

eess.IV↗