arXiv subjects

Sungwoo Park

Publications and source records attributed to Sungwoo Park.

At least 19 recordsLinked to original sources

arXiv2026

Conformal Prediction and Conditional Coverage for Tabular Foundation Models

Tabular foundation models (TFMs) provide predictive distributions for regression, but their prediction regions can exhibit undercoverage or overcoverage even when point predictions are accurate. We introduce C-USIM (Conditionally-Uniformized Score Integration Method), a lightweight application of highest predictive density split conformal prediction that accommodates multimodal predictions. Given calibration and test outputs, it requires no additional training or model inference. It provides finite-sample marginal validity under our assumptions. We bound conditional-marginal coverage gaps using distribution-estimation error and score discreteness, and examine coverage heterogeneity through percentile rank-score plots. Experiments with TabPFN and TabICL show improved marginal coverage accuracy and lower average conditional and group coverage errors. Under a fixed data budget, allocating more observations to calibration can reduce marginal coverage error despite less accurate point predictions.

stat.ML↗

arXiv2026

Context-Continuous Preference Learning for Exoskeleton Personalization

Personalizing exoskeleton assistance across operating conditions is constrained by the time and physical effort required to collect user feedback. We examined whether a user's preference landscape varies smoothly across operating conditions and when this continuity supports learning from limited feedback. We propose Context-Continuous Preference Learning (CCPL), a Gaussian-process preference model that shares observations across nearby contexts while retaining context-specific utility estimates. We evaluated CCPL through simulations and retrospective analyses of ankle and elbow exoskeleton preference data from nine healthy adults. In simulations, CCPL improved reconstruction and preference-based Bayesian optimization relative to independent learning when preferences varied smoothly, but showed negative transfer when continuity was weak. In both human studies, full-data reference landscapes estimated separately for each participant and context tended to be more similar between nearby operating conditions. With five exposures per context, CCPL increased mean reconstruction correlation with these references from 0.644 to 0.720 for ankle assistance and from 0.476 to 0.526 for elbow assistance relative to independent learning. The five-exposure budget was approximately 37% lower for ankle and 17% lower for elbow than the estimated independent-learning budgets needed to match these correlations. CCPL also improved held-out response prediction relative to independent learning, while benefits over pooled learning varied. These findings support context continuity as a basis for sharing preference observations under limited feedback, although benefits for online personalization in humans remain to be established.

cs.LG↗

arXiv2026

The Momentum Fraction, Helicity and Transversity Isovector Moments of Nucleons from \texorpdfstring{$2+1$}{2+1}-flavor Lattice QCD

Results for the isovector momentum fraction, $\langle x \rangle_{u-d}$, helicity moment, $\langle x \rangle_{Δu-Δd}$, and the transversity moment, $\langle x\rangle_{δu-δd}$, of the nucleon are presented using high-statistics data on thirteen NME ensembles of gauge configurations generated by the JLab/W\&M/LANL/MIT/Marseille collaborations using $2+1$-flavors of dynamical Wilson-clover quarks. The much higher statistics facilitated better control over all systematics compared to our previous lattice calculation. The least controlled systematic---excited-state contamination---is quantified by studying the variation of the results as a function of three estimates of the mass gap of the first excited state, obtained from two- and three-point correlation functions. The final results are obtained using a simultaneous fit to extrapolate in the lattice spacing, $a$, pion and kaon masses, $M_π$ and $M_K$, and the finite volume parameter, $M_πL$. The data show no significant finite-volume correction, and some dependence on the lattice spacing and the renormalization factors. The largest systematic uncertainty is due to possible remaining excited states contributions. Our final results, in the $\overline{\rm MS}$ scheme at 2~GeV, are $\langle x \rangle_{u-d} = 0.154(10)(9)$, $\langle x \rangle_{Δu-Δd} = 0.177(10)(15)$ and $\langle x \rangle_{δu-δd} = 0.197(12)(18)$, where the first error is the overall statistical uncertainty and the second represents the various systematic uncertainties added in quadrature. Results for the momentum fraction and helicity moment are consistent with phenomenological global fit values, while the transversity moment is a prediction.

hep-lat↗

arXiv2026

KineFuse: Kinematic-Aware Haptic Fusion for In-Hand Occluded-Object Pose Tracking

Dexterous in-hand manipulation requires continuous 6D pose tracking, yet the manipulating fingers inevitably occlude the object from the camera. We study how to structure the sparse haptic signals already available on multi-fingered hands, including proprioception, proximal force/torque, and binary contact, to complement a pretrained visual pose tracker under occlusion. We propose a kinematic-aware finger-level encoder and systematically compare it against four alternative designs through three levels of evaluation: per-frame refinement, sequential open-loop tracking, and closed-loop manipulation. Our experiments reveal that (i) per-frame evaluation cannot distinguish encoder quality, while sequential tracking amplifies architectural differences by up to 15 times; (ii) the structured encoder learns task-specific cross-modal gating, using vision exclusively for translation and dedicating one attention head to haptics for rotation, without explicit supervision; and (iii) compact finger-level tokenization with 4 tokens outperforms both flat fusion and joint-level representations, which suppress vision through norm dominance. We validate that improved tracking yields higher success in a downstream reorientation task and provide qualitative real-world demonstrations. Our project page is available at https://cold-young.github.io/kine-fuse/.

cs.RO↗

arXiv2026

cuRPQ: A High-Performance GPU-Based Framework for Processing Regular and Conjunctive Regular Path Queries

Regular path queries (RPQs) are fundamental for path-constrained reachability analysis, and more complex variants such as conjunctive regular path queries (CRPQs) are increasingly used in graph analytics. Evaluating these queries is computationally expensive, but to the best of our knowledge, no prior work has explored GPU acceleration. In this paper, we propose cuRPQ, a high-performance GPU-optimized framework for processing RPQs and CRPQs. cuRPQ addresses the key GPU challenges through a novel traversal algorithm, an efficient visited-set management scheme, and a concurrent exploration-materialization strategy. Extensive experiments show that cuRPQ outperforms state-of-the-art methods by orders of magnitude, without out-of-memory errors.

cs.DB↗

arXiv2026

The Spectrum and Scale Setting on 2+1-flavor NME Lattices

This paper describes the thirteen ensembles, named NME, generated with 2+1-flavor Wilson-clover fermions by the JLab/W\&M/LANL/MIT/Marseille collaborations, and presents an analysis of the meson and baryon spectrum, decay constants $f_π$ and $f_K$, flow scales $t_0$ and $w_0$, and time histories of the $Θ$ and Weinberg operators under gradient flow. Using these quantities, the physical point values of the two flow scales, ${t_0^{\rm Phy}}$ and ${w_0^{\rm Phy}}$, and the ratio $\mathop{f_K / f_π}^{\rm Phy}$ are determined. The masses of the octet and decuplet baryons are analyzed using both the next-to-leading order (NLO) and the next-next-to-leading order (NNLO) ansatz from heavy baryon chiral perturbation theory (HB$χ$PT). The NNLO fit to the octet baryons, $M_N$, $M_Σ$, $M_Λ$ and $M_Ξ$, is preferred while the corresponding fits to the decuplet Omega mass, $M_Ω$, are not distinguished. We also present a study of the autocorrelations in the data and show that there is no evidence, even at large flow time, of the freezing of the topological charge or the Weinberg three-gluon operator.

hep-lat↗

arXiv2025

Flavor diagonal nucleon charges using clover fermions on MILC HISQ ensembles

We present lattice results for the flavor diagonal charges of the proton from the analysis of eight ensembles generated using 2+1+1-flavors of highly improved staggered quarks (HISQ) by the MILC collaboration. The calculation includes all the needed connected and disconnected contributions to nucleon three-point function. For extracting matrix elements using fits to the spectral decomposition of these correlation functions, two strategies to remove excited state contributions are employed and compared. To renormalize these charges, the 2+1-flavor mixing matrix is calculated in the RI-sMOM intermediate scheme on the lattice. The final results are presented in the $\overline{\text{MS}}$ scheme at scale 2GeV. The axial charges for the proton are $g_A^u = 0.781(25)$, $g_A^d = -0.440(39)$, and $g_A^s = -0.055(9)$; the tensor charges are $g_T^u = 0.782(28)$, $g_T^d = -0.195(16)$, and $g_T^s = -0.0016(12)$; and the scalar charges are $g_S^u = 9.39(88)$, $g_S^d = 8.84(93)$, and $g_S^s = 0.37(14)$. Results for the neutron are given by the $u \leftrightarrow d$ interchange. Results for the sigma terms are $σ_{πN}|_{\rm standard} = 42(6)~{\rm MeV}$ from a "standard" analysis and $σ_{πN}|_{N π} = 61(6)~{\rm MeV}$ from a "$Nπ$" analysis that includes the contributions of multihadron $Nπ$ excited states as motivated by chiral perturbation theory. Our preferred value $σ_{πN}|_{N π}$ is consistent with the phenomenological extraction from $π- N$ scattering data. The strangeness content of the proton, for which the "standard" analysis is appropriate, is $σ_{s}|_{\rm standard} = 35(13)~{\rm MeV}$.

hep-lat↗

arXiv2025

Linear Bandits with Partially Observable Features

We study the linear bandit problem that accounts for partially observable features. Without proper handling, unobserved features can lead to linear regret in the decision horizon $T$, as their influence on rewards is unknown. To tackle this challenge, we propose a novel theoretical framework and an algorithm with sublinear regret guarantees. The core of our algorithm consists of (i) feature augmentation, by appending basis vectors that are orthogonal to the row space of the observed features; and (ii) the introduction of a doubly robust estimator. Our approach achieves a regret bound of $\tilde{O}(\sqrt{(d + d_h)T})$, where $d$ is the dimension of the observed features and $d_h$ depends on the extent to which the unobserved feature space is contained in the observed one, thereby capturing the intrinsic difficulty of the problem. Notably, our algorithm requires no prior knowledge of the unobserved feature space, which may expand as more features become hidden. Numerical experiments confirm that our algorithm outperforms both non-contextual multi-armed bandits and linear bandit algorithms depending solely on observed features.

stat.ML↗

arXiv2025

2025 update on $\varepsilon_K$ in the Standard Model with lattice QCD inputs

We present theoretical results for the indirect CP violation parameter, $ |\varepsilon_K| $, evaluated directly within the Standard Model using lattice QCD inputs including $ B_K $, $ |V_{cb}| $, $ |V_{us}| $, $ |V_{ud}| $, $ ξ_0 $, $ ξ_2 $, $ F_K $, and the charm quark mass $ m_c $. Our analysis reveals a significant tension at the $ \sim 5σ$ level ($ 4.6σ$ to $ 5.2σ$) between the Standard Model prediction and the experimental value of $ |\varepsilon_K| $. The Standard Model prediction, informed by lattice QCD inputs, accounts for only approximately 65% of the experimental value, leaving a 35% discrepancy unexplained. Notably, this tension vanishes when using the inclusive determination of $ |V_{cb}| $, derived from heavy quark expansion. The discrepancy is therefore tied to the well-known tension between the exclusive and inclusive determinations of $ |V_{cb}| $. We further report results for $ |\varepsilon_K| $ obtained using the Brod--Gorbahn--Stamou (BGS) method based on $ u\text{--}t $ unitarity, which yields an even stronger discrepancy, in the range of $ 4.9σ$ to $ 5.5σ$, when combined with lattice QCD inputs.

hep-lat↗

arXiv2025

2024 Update on $\varepsilon_K$ with lattice QCD inputs

We report recent progress on $\varepsilon_K$ evaluated directly from the standard model (SM) with lattice QCD inputs such as $\hat{B}_K$, exclusive $|V_{cb}|$, $|V_{us}|$, $|V_{ud}|$, $ξ_0$, $ξ_2$, $ξ_\text{LD}$, $f_K$, and $m_c$. We find that the standard model with exclusive $|V_{cb}|$ and lattice QCD inputs describes only $2/3 \cong 65\%$ of the experimental value of $|\varepsilon_K|$ and does not explain its remaining 35\%, which represents a strong tension in $|\varepsilon_K|$ at the $5.1σ\sim 4.1σ$ level between the SM theory and experiment. We also find that this tension disappears when we use the inclusive value of $|V_{cb}|$ obtained using the heavy quark expansion based on the QCD sum rule approach. We also report results for $|\varepsilon_K|$ obtained using the Brod-Gorbahn-Stamou (BGS) method for $η_i$ of $u-t$ unitarity, which leads to even a stronger tension of $5.7σ\sim 4.2σ$ with lattice QCD inputs.

hep-lat↗

arXiv2025

Finite Temperature Transition in Hyper Stealth Dark Matter using Möbius Domain Wall Fermions

The first-order confinement transition of a strongly coupled composite dark matter theory can provide a possible source of gravitational waves in the early universe. In this work, on behalf of the Lattice Strong Dynamics (LSD) Collaboration, we present our recent investigation on the finite temperature confinement transition of the one-flavor SU(4) dark gauge theory named Hyper Stealth Dark Matter (HSDM). The dark matter candidate in this theory is a composite bosonic baryon and can have a remarkably low mass of a few GeV. We expect the finite temperature transition to be first-order over at least in some finite range of fermionic masses and to be a potential source of observable gravitational radiation. The finite temperature simulation of one-flavor SU(4) is done by using Möbius Domain wall fermions. The order of the transition and its fermionic mass dependence are explored by monitoring the Polyakov loop, chiral condensate and topological charge using three lattice volumes at $N_t=8$.

hep-lat↗

arXiv2025

Gradient flow of the Weinberg operator

We present preliminary results on the susceptibilities involving the CP-violating (CPV) Weinberg three-gluon operator and the topological $Θ$ term using the gradient flow scheme, and study their continuum and chiral extrapolations. These are used to provide an estimate of the $Θ$ induced by the Weinberg operator in theories with the Peccei-Quinn (PQ) mechanism. Combined with the calculations of the matrix elements (MEs) of quark-bilinears between nucleon states, such calculations will enable estimates of the electric dipole moments (EDMs) and CPV pion-nucleon couplings due to the Weinberg operator, thereby providing robust constraints on beyond the standard model (BSM) physics.

hep-lat↗

arXiv2024

Mean-field Chaos Diffusion Models

In this paper, we introduce a new class of score-based generative models (SGMs) designed to handle high-cardinality data distributions by leveraging concepts from mean-field theory. We present mean-field chaos diffusion models (MF-CDMs), which address the curse of dimensionality inherent in high-cardinality data by utilizing the propagation of chaos property of interacting particles. By treating high-cardinality data as a large stochastic system of interacting particles, we develop a novel score-matching method for infinite-dimensional chaotic particle systems and propose an approximation scheme that employs a subdivision strategy for efficient training. Our theoretical and empirical results demonstrate the scalability and effectiveness of MF-CDMs for managing large high-cardinality data structures, such as 3D point clouds.

cs.LG↗

arXiv2024

InstructCV: Instruction-Tuned Text-to-Image Diffusion Models as Vision Generalists

Recent advances in generative diffusion models have enabled text-controlled synthesis of realistic and diverse images with impressive quality. Despite these remarkable advances, the application of text-to-image generative models in computer vision for standard visual recognition tasks remains limited. The current de facto approach for these tasks is to design model architectures and loss functions that are tailored to the task at hand. In this paper, we develop a unified language interface for computer vision tasks that abstracts away task-specific design choices and enables task execution by following natural language instructions. Our approach involves casting multiple computer vision tasks as text-to-image generation problems. Here, the text represents an instruction describing the task, and the resulting image is a visually-encoded task output. To train our model, we pool commonly-used computer vision datasets covering a range of tasks, including segmentation, object detection, depth estimation, and classification. We then use a large language model to paraphrase prompt templates that convey the specific tasks to be conducted on each image, and through this process, we create a multi-modal and multi-task training dataset comprising input and output images along with annotated instructions. Following the InstructPix2Pix architecture, we apply instruction-tuning to a text-to-image diffusion model using our constructed dataset, steering its functionality from a generative model to an instruction-guided multi-task vision learner. Experiments demonstrate that our model, dubbed InstructCV, performs competitively compared to other generalist and task-specific vision models. Moreover, it exhibits compelling generalization capabilities to unseen data, categories, and user instructions.

cs.CV↗

arXiv2024

Parameter-Free Algorithms for Performative Regret Minimization under Decision-Dependent Distributions

This paper studies performative risk minimization, a formulation of stochastic optimization under decision-dependent distributions. We consider the general case where the performative risk can be non-convex, for which we develop efficient parameter-free optimistic optimization-based methods. Our algorithms significantly improve upon the existing Lipschitz bandit-based method in many aspects. In particular, our framework does not require knowledge about the sensitivity parameter of the distribution map and the Lipshitz constant of the loss function. This makes our framework practically favorable, together with the efficient optimistic optimization-based tree-search mechanism. We provide experimental results that demonstrate the numerical superiority of our algorithms over the existing method and other black-box optimistic optimization methods.

cs.LG↗

arXiv2024

Progress report on testing robustness of the Newton method in data analysis on 2-point correlation function using a MILC HISQ ensemble

We report recent progress in data analysis on the two point correlation functions which will be prerequisite to obtain semileptonic form factors for the $B_{(s)} \to D_{(s)}\ellν$ decays. We use a MILC HISQ ensemble for the measurement. We use the HISQ action for light quarks, and the Oktay-Kronfeld (OK) action for the heavy quarks ($b$ and $c$). We used a sequential Bayesian method for the data analysis. Here we test the new fitting methodology of Benjamin J.~Choi in a completely independent manner.

hep-lat↗

arXiv2024

Current progress on the semileptonic form factors for $\bar{B} \to D^{\ast} \ell \barν$ decay using the Oktay-Kronfeld action

We present recent progress in calculating the semileptonic form factors $h_{A_1}(w)$ for the $\bar{B} \to D^{\ast} \ell \barν$ decays. We use the Oktay-Kronfeld (OK) action for the charm and bottom valence quarks and the HISQ action for light quarks. We adopt the Newton method combined with the scanning method to find a good initial guess for the $χ^2$ minimizer in the fitting of the 2pt correlation functions. The main advantage is that the Newton method lets us to consume all the time slices allowed by the physical positivity. We report the first, reliable, but preliminary results for $h_{A_1}(w)/ρ_{A_1}$ at zero recoil ($w=1$). Here we use a MILC HISQ ensemble ($a = 0.12$ fm, $M_π$ = 220 MeV, and $N_f = 2 + 1 + 1$ flavors).

hep-lat↗

arXiv2024

Update on flavor diagonal nucleon charges from clover fermions

We present a summary of the full calculation of the axial, scalar and tensor flavor diagonal charges of the nucleon carried out using Wilson-clover fermions on eight ensembles generated using 2+1+1-flavors of highly improved staggered quarks (HISQ) by the MILC collaboration. We also give results for the $3\times 3$ matrix of renormalization factors between the RI-sMOM and $\overline{\rm MS}$ scheme for the 2+1 flavor theory that include flavor mixing. Preliminary results for $g_{A,S,T}^{u,d,s}$ are presented in the $\overline{\rm MS}$ scheme at scale 2 GeV.

hep-lat↗