arXiv subjects

Jiwon Park

Publications and source records attributed to Jiwon Park.

At least 19 recordsLinked to original sources

arXiv2026

DEO: Training-Free Direct Embedding Optimization for Negation-Aware Retrieval

Recent advances in Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) have enabled diverse retrieval methods. However, existing retrieval methods often fail to accurately retrieve results for negation and exclusion queries. To address this limitation, prior approaches rely on embedding adaptation or fine-tuning, which introduce additional computational cost and deployment complexity. We propose Direct Embedding Optimization (DEO), a training-free method for negation-aware text and multimodal retrieval. DEO decomposes queries into positive and negative components and optimizes the query embedding with a contrastive objective. Without additional training data or model updates, DEO outperforms baselines on NegConstraint, with gains of +0.0738 nDCG@10 and +0.1028 MAP@100, while improving Recall@5 by +6\% over OpenAI CLIP in multimodal retrieval. These results demonstrate the practicality of DEO for negation- and exclusion-aware retrieval in real-world settings.

cs.CL↗

arXiv2026

DreamFlow: Local Navigation Beyond Observation via Conditional Flow Matching in the Latent Space

Local navigation in cluttered environments often suffers from dense obstacles and frequent local minima. Conventional local planners rely on heuristics and are prone to failure, while deep reinforcement learning(DRL)based approaches provide adaptability but are constrained by limited onboard sensing. These limitations lead to navigation failures because the robot cannot perceive structures outside its field of view. In this paper, we propose DreamFlow, a DRL-based local navigation framework that extends the robot's perceptual horizon through conditional flow matching(CFM). The proposed CFM based prediction module learns probabilistic mapping between local height map latent representation and broader spatial representation conditioned on navigation context. This enables the navigation policy to predict unobserved environmental features and proactively avoid potential local minima. Experimental results demonstrate that DreamFlow outperforms existing methods in terms of latent prediction accuracy and navigation performance in simulation. The proposed method was further validated in cluttered real world environments with a quadrupedal robot. The project page is available at https://dreamflow-icra.github.io.

cs.RO↗

arXiv2025

Controllable Superconductivity in Suspended van der Waals Materials

Tunable superconductors provide a versatile platform for advancing next-generation quantum technologies. Here, we demonstrate controllable superconductivity in suspended NbSe2 thin layers, achieved through local strain and thermal modulation of the superconducting state. Our results show that suspended NbSe2 structures enable strain modulation of the critical temperature by up to approximately 0.92 K (about 12.5% of the critical temperature) and allow the realization of gate-tunable superconducting critical currents. We further demonstrate configurable hysteretic transport characteristics exhibiting multistability and negative differential resistance, providing easily reconfigurable, spatially dependent superconducting states. These phenomena are well explained by calculations of electron-phonon coupling using density functional theory, together with time-dependent Ginzburg-Landau dynamics coupled to the thermal diffusion equation. Our work provides profound insight into strain and thermal modulation of van der Waals superconductors and opens new opportunities for tunable on-chip superconductor devices, integrated superconducting circuits, and quantum simulators.

cond-mat.supr-con↗

arXiv2025

Exploring Diffusion Models for Generative Forecasting of Financial Charts

Recent advances in generative models have enabled significant progress in tasks such as generating and editing images from text, as well as creating videos from text prompts, and these methods are being applied across various fields. However, in the financial domain, there may still be a reliance on time-series data and a continued focus on transformer models, rather than on diverse applications of generative models. In this paper, we propose a novel approach that leverages text-to-image model by treating time-series data as a single image pattern, thereby enabling the prediction of stock price trends. Unlike prior methods that focus on learning and classifying chart patterns using architectures such as ResNet or ViT, we experiment with generating the next chart image from the current chart image and an instruction prompt using diffusion models. Furthermore, we introduce a simple method for evaluating the generated chart image against ground truth image. We highlight the potential of leveraging text-to-image generative models in the financial domain, and our findings motivate further research to address the current limitations and expand their applicability.

cs.AI↗

arXiv2025

DocHop-QA: Towards Multi-Hop Reasoning over Multimodal Document Collections

Despite rapid progress in large language models (LLMs), current QA benchmarks still overlook the core challenge of real-world scientific information seeking: synthesizing multimodal evidence scattered across multiple documents and structural formats. Existing QAs remain narrow in scope, relying on unimodal text and short-span reasoning that fail to capture the complexity of real information-seeking. We introduce DocHop-QA, a benchmark of 11,379 instances for evaluating multimodal, multi-document, multi-hop scientific QA. Built from publicly available PubMed articles, DocHop-QA incorporates textual passages, tables, and layout cues, enabling cross-document inference without explicit hyperlinks. To scale realistic QA construction, we develop an LLM-driven generation pipeline grounded in 11 scientific reasoning concepts, producing diverse and coherent question-answer pairs. To highlight the utility and versatility of the dataset, we propose a task-driven evaluation framework spanning four settings, including generative answering, multimodal evidence integration and structured index prediction. Experiments show that current models struggle with DocHop-QA's long-context, multi-evidence demands, establishing it as a rigorous testbed for advancing next-generation scientific QA systems.

cs.CL↗

arXiv2025

Structure-Preserving Zero-Shot Image Editing via Stage-Wise Latent Injection in Diffusion Models

We propose a diffusion-based framework for zero-shot image editing that unifies text-guided and reference-guided approaches without requiring fine-tuning. Our method leverages diffusion inversion and timestep-specific null-text embeddings to preserve the structural integrity of the source image. By introducing a stage-wise latent injection strategy-shape injection in early steps and attribute injection in later steps-we enable precise, fine-grained modifications while maintaining global consistency. Cross-attention with reference latents facilitates semantic alignment between the source and reference. Extensive experiments across expression transfer, texture transformation, and style infusion demonstrate state-of-the-art performance, confirming the method's scalability and adaptability to diverse image editing scenarios.

cs.CV↗

arXiv2025

Constraining Anisotropic Universe Through Big Bang Nucleosynthesis: A Case Study of The Bianchi Type-I Universe

The isotropy and homogeneity of our Universe are the cardinal principles of modern cosmology built on the definition of metric through the prescription by Friedmann-Lema$\hat{i}$tre-Robertson-Walker (FLRW). From the aspects of geometry, the presence of anisotropy, inhomogeneity, or both are allowed in the metrics defined as the Bianchi type I and V metrics. In this letter, the Big Bang Nucleosynthesis (BBN) formalism, and the latest observational constraints on nuclear abundances are being used to put bounds on the global anisotropy offered in the Bianchi type I metrics, providing a new path to explore in the background of global anisotropy.

astro-ph.CO↗

arXiv2024

Reasoning Abilities of Large Language Models: In-Depth Analysis on the Abstraction and Reasoning Corpus

The existing methods for evaluating the inference abilities of Large Language Models (LLMs) have been predominantly results-centric, making it challenging to assess the inference process comprehensively. We introduce a novel approach using the Abstraction and Reasoning Corpus (ARC) benchmark to evaluate the inference and contextual understanding abilities of LLMs in a process-centric manner, focusing on three key components from the Language of Thought Hypothesis (LoTH): Logical Coherence, Compositionality, and Productivity. Our carefully designed experiments reveal that while LLMs demonstrate some inference capabilities, they still significantly lag behind human-level reasoning in these three aspects. The main contribution of this paper lies in introducing the LoTH perspective, which provides a method for evaluating the reasoning process that conventional results-oriented approaches fail to capture, thereby offering new insights into the development of human-level reasoning in artificial intelligence systems.

cs.CL↗

arXiv2023

Mass Spectra Prediction with Structural Motif-based Graph Neural Networks

Mass spectra, which are agglomerations of ionized fragments from targeted molecules, play a crucial role across various fields for the identification of molecular structures. A prevalent analysis method involves spectral library searches,where unknown spectra are cross-referenced with a database. The effectiveness of such search-based approaches, however, is restricted by the scope of the existing mass spectra database, underscoring the need to expand the database via mass spectra prediction. In this research, we propose the Motif-based Mass Spectrum Prediction Network (MoMS-Net), a system that predicts mass spectra using the information derived from structural motifs and the implementation of Graph Neural Networks (GNNs). We have tested our model across diverse mass spectra and have observed its superiority over other existing models. MoMS-Net considers substructure at the graph level, which facilitates the incorporation of long-range dependencies while using less memory compared to the graph transformer model.

cs.LG↗

arXiv2022

$f(R)$ gravity with broken Weyl gauge symmetry, cosmological backreaction, and its effects on CMB anisotropy

We propose a new class of $f(R)$ theory where its Weyl gauge symmetry is broken in the primordial era of the universe. This symmetry forces one to adopt a new scalar field, namely a Weyl field and a gauge vector boson. Furthermore, an equivalent form of the Einstein-Hilbert Lagrangian with a non-minimally coupled scalar field corresponding to the function $f(R)$ is found. Due to the geometrical feature of the Weyl field, it turns out that the symmetry breaking induces a non-minimal coupling, which cannot be expected in the standard $f(R)$ theories. We explain how this affects the evolution of the universe at cosmological scales. It is shown that there may be a value shift in the Planck constant and the cosmological constant. This can be regarded as a genuine exemplification of the cosmological backreaction. Furthermore, one also finds new features in the evolution of perturbational variables and cosmic microwave background anisotropy. Moreover, we prove that when a specific $f(R)$ model invokes inflation, the amplitude of the primordial gravitational waves affects the evolution of scalar perturbation due to the new non-minimal coupling. As a case study, we explain how this can be embodied in the Starobinsky inflation. Finally, we discuss some impacts that this physics can bear and the possibility of giving a new restriction of the estimation of cosmological variables such as the gravitational wave amplitude with experiments.

gr-qc↗

arXiv2022

In-Plane and Out-of-Plane Excitonic Coupling in 2D Molecular Crystals

Understanding the nature of molecular excitons in low-dimensional molecular solids is of paramount importance in fundamental photophysics and various applications such as energy harvesting, switching electronics and display devices. Despite this, the spatial evolution of molecular excitons and their transition dipoles have not been captured in the precision of molecular length scales. Here we show in-plane and out-of-plane excitonic evolution in quasilayered two-dimensional (2D) perylene-3, 4, 9, 10-tetracarboxylic dianhydride (PTCDA) crystals assembly-grown on hexagonal boron nitride (BN) crystals. Complete lattice constants with orientations of two herringbone-configured basis molecules are determined with polarization-resolved spectroscopy and electron diffraction methods. In the truly 2D limit of single layers, two Frenkel emissions Davydov-split by Kasha-type intralayer coupling exhibit energy inversion with decreasing temperature, which enhances excitonic coherence. As the thickness increases, the transition dipole moments of newly emerging charge transfer excitons are reoriented because of mixing with the Frenkel states. The current spatial anatomy of 2D molecular excitons will inspire a deeper understanding and groundbreaking applications of low-dimensional molecular systems.

cond-mat.mtrl-sci↗

arXiv2021

GeoT: A Geometry-aware Transformer for Reliable Molecular Property Prediction and Chemically Interpretable Representation Learning

In recent years, molecular representation learning has emerged as a key area of focus in various chemical tasks. However, many existing models fail to fully consider the geometric information of molecular structures, resulting in less intuitive representations. Moreover, the widely used message-passing mechanism is limited to provide the interpretation of experimental results from a chemical perspective. To address these challenges, we introduce a novel Transformer-based framework for molecular representation learning, named the Geometry-aware Transformer (GeoT). GeoT learns molecular graph structures through attention-based mechanisms specifically designed to offer reliable interpretability, as well as molecular property prediction. Consequently, GeoT can generate attention maps of interatomic relationships associated with training objectives. In addition, GeoT demonstrates comparable performance to MPNN-based models while achieving reduced computational complexity. Our comprehensive experiments, including an empirical simulation, reveal that GeoT effectively learns the chemical insights into molecular structures, bridging the gap between artificial intelligence and molecular sciences.

cs.LG↗

arXiv2021

Proceedings of the second MadAnalysis 5 workshop on LHC recasting in Korea

We document the activities performed during the second MadAnalysis 5 workshop on LHC recasting, that was organised in KIAS (Seoul, Korea) on February 12-20, 2020. We detail the implementation of 12 new ATLAS and CMS searches in the MadAnalysis 5 Public Analysis Database, and the associated validation procedures. Those searches probe the production of extra gauge and scalar/pseudoscalar bosons, supersymmetry, seesaw models and deviations from the Standard Model in four-top production.

hep-ph↗

arXiv2019

Identification of additional jets in the t-tbar b-bbar events using a deep neural network

In the top quark pair production in association with the Higgs boson decaying to a b quark pair t-tbar H (b-bbar), the final state has an irreducible nonresonant background from the production of a top quark pair in association with a b quark pair t-tbar b-bbar. Therefore, understanding of the t-tbar b-bbar process precisely in particular differential cross-section as functions of the properties of the additional b jets not from the top quark decay is essential for improving the sensitivity of a search for the t-tbar H b-bbar process. The two additional b jets can be identified by using various approaches. In this paper, the performances are compared quantitatively in the lepton+jets decay channel in terms of the matching efficiency of assigning two additional b jets as a figure of merit. This study provides valuable information towards the precise measurement of differential cross-sections as a function of properties of the additional b jets in the t-tbar b-bbar events. We showed that a matching efficiency of around 40% could be achieved using a deep neural network method. In the events with at least 4 b jets, this performance is 8% better than that achieved using minimum Delta R(b,bbar) method. This is consistent with the boosted decision tree method within its statistical uncertainty.

hep-ex↗

arXiv2019

Probing Primordial Symmetry Breaking with Cosmic Microwave Background Anisotropy

There have been vigorous research attempts to test various modified gravity theories by usingphysics of the cosmic microwave background (CMB). Meanwhile, symmetry breaking such as Higgsmechanism is one of the most important phenomena in physics but there have been not so muchresearches to make them contact with cosmological observations. In this article, with the CMBpower spectra we try to distinguish two different scenarios of spontaneous symmetry breaking inprimordial era of the universe. The first model is based on a broken symmetric theory of gravity,which was suggested by A. Zee in 1979. The second model is an application of Palatini formalismto the first model. Perturbation equations are computed and they show differences originated fromthe property of symmetry. Furthermore, it turns out that two models have different features ofCMB power spectra with the same potential scale. This fact enables us to verify distinct kinds ofprimordial symmetry breaking with CMB physics.

gr-qc↗

arXiv2018

Correlation between $R_{D^{(*)}}$ and top quark FCNC decays in leptoquark models

Some interpretations of $R_{D^{(*)}}$ anomaly in $B$ meson decay using leptoquark (LQ) models can also generate top quark decays through Flavor Changing Neutral Current (FCNC). In this work we focus on two LQs, i.e. scalar $S_1$ and vector $U_1$ which are both singlet under the $SU(2)_L$ gauge group in the Standard Model (SM). We investigate their implications on the 3-body top FCNC decays $t\to c \ell_i \ell_j$ at tree level and the 2-body $t\to c V$ at one-loop level, with $\ell$ being the SM leptons and $V=\gamma, Z, g$ being the SM gauge bosons. We utilize the $2\sigma$ parameter fitting ranges of the LQ models and find that $Br(t\to c \ell_i \ell_j)$ at tree level can reach $\mathcal{O}(10^{-6})$ and $Br(t\to c V)$ at one-loop level can reach $\mathcal{O}(10^{-10})$. Some quick collider search prospects are also analyzed.

hep-ph↗

arXiv2018

Dynamical analysis of Brans-Dicke Universe with inverse power-law effective potential

We study Brans-Dicke cosmology with an inverse power-law effective potential. By using dynamical analyses, we search for fixed points corresponding to the radiation-like matter and dark energy-dominated era of our Universe, and the stability of fixed points is also investigated. We find phase space trajectories which are attracted to the stable point of the dark energy-dominated era from unstable fixed points like matter-dominated era of the Universe. The dark energy comes from effective potentials of the Brans-Dicke field, whose variation (related to the time-variation of the gravitational coupling constant) is shown to be in good agreement with observational data.

gr-qc↗

arXiv2018

Proceedings of the first MadAnalysis 5 workshop on LHC recasting in Korea

We present the activities performed during the first MadAnalysis 5 workshop on LHC recasting that has been organized at High 1 (Gangwon privince, Korea) on August 20-27, 2017. This report includes details on the implementation in the MadAnalysis 5 framework of eight ATLAS and CMS analyses, as well as a description of the corresponding validation and the various issues that have been observed.

hep-ph↗