arXiv subjects

Wonjun Lee

Publications and source records attributed to Wonjun Lee.

At least 37 records · Page 2Linked to original sources

arXiv2026

Characterization of the 20-inch Photomultiplier Tubes for RENE Detector

To address the Reactor Antineutrino Anomaly (RAA) observed in neutrino experiments, the Reactor Experiment for Neutrino and Exotics (RENE) has been initiated using a liquid scintillation detector. In this study, we investigate the characteristics of two 20-inch Hamamatsu R12860 photomultiplier tubes (PMTs) intended for installation in the RENE detector. The charge and timing responses of the PMTs were evaluated at both the nominal and target gains expected during actual operation. In particular, gain non-uniformity arising from the large-diameter photocathode with a box-and-line type dynode structure was examined, and the maximum gain variation was measured. The occurrence rate, timing, and charge distributions of late pulses and afterpulses were also investigated to characterize the specific response features of the R12860 PMT. The results reported in this study will aid in the interpretation of signals from the RENE detector and serve as a reference for estimating potential systematic uncertainties in RENE data. Furthermore, these findings are expected to provide valuable information for other experiments employing the same type of PMTs.

physics.ins-det↗

arXiv2026

Deep Kinetic JKO schemes for Vlasov-Fokker-Planck Equations

We introduce a deep neural network-based numerical method for solving kinetic Fokker Planck equations, including both linear and nonlinear cases. Building upon the conservative dissipative structure of Vlasov-type equations, we formulate a class of generalized minimizing movement schemes as iterative constrained minimization problems: the conservative part determines the constraint set, while the dissipative part defines the objective functional. This leads to an analog of the classical Jordan-Kinderlehrer-Otto (JKO) scheme for Wasserstein gradient flows, and we refer to it as the kinetic JKO scheme. To compute each step of the kinetic JKO iteration, we introduce a particle-based approximation in which the velocity field is parameterized by deep neural networks. The resulting algorithm can be interpreted as a kinetic-oriented neural differential equation that enables the representation of high-dimensional kinetic dynamics while preserving the essential variational and structural properties of the underlying PDE. We validate the method with extensive numerical experiments and demonstrate that the proposed kinetic JKO-neural ODE framework is effective for high-dimensional numerical simulations.

math.NA↗

arXiv2026

IgPose: A Generative Data-Augmented Pipeline for Robust Immunoglobulin-Antigen Binding Prediction

Predicting immunoglobulin-antigen (Ig-Ag) binding remains a significant challenge due to the paucity of experimentally-resolved complexes and the limited accuracy of de novo Ig structure prediction. We introduce IgPose, a generalizable framework for Ig-Ag pose identification and scoring, built on a generative data-augmentation pipeline. To mitigate data scarcity, we constructed the Structural Immunoglobulin Decoy Database (SIDD), a comprehensive repository of high-fidelity synthetic decoys. IgPose integrates equivariant graph neural networks, ESM-2 embeddings, and gated recurrent units to synergistically capture both geometric and evolutionary features. We implemented interface-focused k-hop sampling with biologically guided pooling to enhance generalization across diverse interfaces. The framework comprises two sub-networks--IgPoseClassifier for binding pose discrimination and IgPoseScore for DockQ score estimation--and achieves robust performance on curated internal test sets and the CASP-16 benchmark compared to physics and deep learning baselines. IgPose serves as a versatile computational tool for high-throughput antibody discovery pipelines by providing accurate pose filtering and ranking. IgPose is available on GitHub (https://github.com/arontier/igpose).

cs.LG↗

arXiv2026

Mixture-of-Experts with Intermediate CTC Supervision for Accented Speech Recognition

Accented speech remains a persistent challenge for automatic speech recognition (ASR), as most models are trained on data dominated by a few high-resource English varieties, leading to substantial performance degradation for other accents. Accent-agnostic approaches improve robustness yet struggle with heavily accented or unseen varieties, while accent-specific methods rely on limited and often noisy labels. We introduce Moe-Ctc, a Mixture-of-Experts architecture with intermediate CTC supervision that jointly promotes expert specialization and generalization. During training, accent-aware routing encourages experts to capture accent-specific patterns, which gradually transitions to label-free routing for inference. Each expert is equipped with its own CTC head to align routing with transcription quality, and a routing-augmented loss further stabilizes optimization. Experiments on the Mcv-Accent benchmark demonstrate consistent gains across both seen and unseen accents in low- and high-resource conditions, achieving up to 29.3% relative WER reduction over strong FastConformer baselines.

cs.CL↗

arXiv2026

Gluing Randomness via Entanglement: Tight Bound from Second Rényi Entropy

The efficient generation of random quantum states is a long-standing challenge, motivated by their diverse applications in quantum information processing tasks. In this work, we identify entanglement as the key resource that enables local random unitaries to generate global random states by effectively gluing randomness across the system. Specifically, we demonstrate that approximate random states can be produced from an entangled state $|ψ\rangle$ through the application of local random unitaries. We show that the resulting ensemble forms an approximate state design with an error saturating as $Θ(e^{-\mathcal{N}_2(ψ)})$, where $\mathcal{N}_2(ψ)$ is the second Rényi entanglement entropy of $|ψ\rangle$. Furthermore, we prove that this tight bound also applies to the second Rényi entropy of coherence when the ensemble is constructed using coherence-free operations. These results imply that, when restricted to resource-free gates, the quality of the generated random states is determined entirely by the resource content of the initial state. Notably, we find that among all $α$-Rényi entropeis, the second Rényi entropy yields the tightest bounds. Consequently, these second Rényi entropies can be interpreted as the maximal capacities for generating randomness using resource-free operations. Finally, moving beyond approximate state designs, we utilize this entanglement-assisted gluing mechanism to present a novel method for generating pseudorandom states in multipartite systems from a locally entangled state via pseudorandom unitaries in each of parties.

quant-ph↗

arXiv2025

Shallow quantum circuit for generating extremely low-entangled approximate state designs

Random quantum states have various applications in quantum information science. We discover a new ensemble of quantum states that serve as an $ε$-approximate state $t$-design while possessing extremely low entanglement, magic, and coherence. These resources can reach their theoretical lower bounds, $Ω(\log (t/ε))$, which are also proven in this work. This implies that for fixed $t$ and $ε$, entanglement, magic, and coherence do not scale with the system size, i.e., $O(1)$ with respect to the total number of qubits $n$. Moreover, we explicitly construct an ancilla-free shallow quantum circuit for generating such states by transforming $k$-qubit approximate state designs into $n$-qubit ones without increasing the support size. The depth of such a quantum circuit, $O(t [\log t]^3 \log n \log(1/ε))$, is the most efficient among existing algorithms without ancilla qubits. A class of quantum circuits proposed in our work offers reduced cost for classical simulation of random quantum states, leading to potential applications in quantum information processing. As a concrete example, we propose classical shadow tomography using an estimator with superpositions between only two states, from which almost all quantum states can be efficiently certified by requiring only $O(1)$ measurements and classical post-processing time.

quant-ph↗

arXiv2025

Speak & Spell: LLM-Driven Controllable Phonetic Error Augmentation for Robust Dialogue State Tracking

Dialogue State Tracking (DST) is a key part of task-oriented dialogue systems, identifying important information in conversations. However, its accuracy drops significantly in spoken dialogue environments due to named entity errors from Automatic Speech Recognition (ASR) systems. We introduce a simple yet effective data augmentation method that targets those entities to improve the robustness of DST model. Our novel method can control the placement of errors using keyword-highlighted prompts while introducing phonetically similar errors. As a result, our method generated sufficient error patterns on keywords, leading to improved accuracy in noised and low-accuracy ASR environments.

cs.CL↗

arXiv2025

RENE experiment for the sterile neutrino search using reactor neutrinos

This paper summarizes the details of the Reactor Experiment for Neutrinos and Exotics (RENE) experiment. It covers the detector construction, Monte Carlo (MC) simulation study, and physics expectations. The primary goal of the RENE project is to investigate the sterile neutrino oscillation at $Δ{m}^{2}_{41}\sim 2\,{\rm{eV}^{2}}$. which overlap with the allowed region predicted by the Reactor Antineutrino Anomaly (RAA). On the other hand, the STEREO and PROSPECT experiments have excluded certain regions of the parameter space with 95 \% confidence level (C.L.), while the joint study conducted by RENO and NEOS suggests possible indications of sterile neutrinos at $Δ{m}^{2}_{41}\sim2.4\,{\rm{eV}^{2}}$ and $\sim{1.7}{\,\rm{eV}^{2}}$ with sin$^{2}θ_{41} < 0.01$. Accordingly, a more meticulous investigation of these remaining regions continues to be a scientifically valuable endeavor. This paper reports the technical details of the detector and physics objectives.

hep-ex↗

arXiv2025

ELITE: Enhanced Language-Image Toxicity Evaluation for Safety

Current Vision Language Models (VLMs) remain vulnerable to malicious prompts that induce harmful outputs. Existing safety benchmarks for VLMs primarily rely on automated evaluation methods, but these methods struggle to detect implicit harmful content or produce inaccurate evaluations. Therefore, we found that existing benchmarks have low levels of harmfulness, ambiguous data, and limited diversity in image-text pair combinations. To address these issues, we propose the ELITE benchmark, a high-quality safety evaluation benchmark for VLMs, underpinned by our enhanced evaluation method, the ELITE evaluator. The ELITE evaluator explicitly incorporates a toxicity score to accurately assess harmfulness in multimodal contexts, where VLMs often provide specific, convincing, but unharmful descriptions of images. We filter out ambiguous and low-quality image-text pairs from existing benchmarks using the ELITE evaluator and generate diverse combinations of safe and unsafe image-text pairs. Our experiments demonstrate that the ELITE evaluator achieves superior alignment with human evaluations compared to prior automated methods, and the ELITE benchmark offers enhanced benchmark quality and diversity. By introducing ELITE, we pave the way for safer, more robust VLMs, contributing essential tools for evaluating and mitigating safety risks in real-world applications.

cs.CV↗

arXiv2025

Pseudochaotic Many-Body Dynamics as a Pseudorandom State Generator

Quantum chaos is central to understanding quantum dynamics and is crucial for generating random quantum states, a key resource for quantum information tasks. In this work, we introduce a new class of quantum many-body dynamics, termed pseudochaotic dynamics. Although distinct from chaotic dynamics, out-of-time-ordered correlators, the key indicators of quantum chaos, fail to distinguish them. Moreover, pseudochaotic dynamics generates pseudorandom states that are computationally indistinguishable from Haar-random states. We construct pseudochaotic dynamics by embedding a smaller $k$-qubit subsystem into a larger $n$-qubit system. We demonstrate that a subsystem of size $k=ω(\log n)$ is sufficient to induce pseudochaotic behavior in the entire $n$-qubit system. Furthermore, we construct a quantum circuit exhibiting pseudochaotic dynamics and demonstrate that it generates pseudorandom states within $\mathrm{polylog}(n)$ depth. In summary, our results constitute the discovery of new quantum dynamics that are computationally indistinguishable from genuine quantum chaos, which provides efficient routes to generate useful pseudorandom states.

quant-ph↗

arXiv2025

SELFI: Selective Fusion of Identity for Generalizable Deepfake Detection

Face identity provides a powerful signal for deepfake detection. Prior studies show that even when not explicitly modeled, classifiers often learn identity features implicitly. This has led to conflicting views: some suppress identity cues to reduce bias, while others rely on them as forensic evidence. To reconcile these views, we analyze two hypotheses: (1) whether face identity alone is discriminative for detecting deepfakes, and (2) whether such identity features generalize poorly across manipulation methods. Our experiments confirm that identity is informative but context-dependent. While some manipulations preserve identity-consistent artifacts, others distort identity cues and harm generalization. We argue that identity features should neither be blindly suppressed nor relied upon, but instead be explicitly modeled and adaptively controlled based on per-sample relevance. We propose \textbf{SELFI} (\textbf{SEL}ective \textbf{F}usion of \textbf{I}dentity), a generalizable detection framework that dynamically modulates identity usage. SELFI consists of: (1) a Forgery-Aware Identity Adapter (FAIA) that extracts identity embeddings from a frozen face recognition model and projects them into a forgery-relevant space via auxiliary supervision; and (2) an Identity-Aware Fusion Module (IAFM) that selectively integrates identity and visual features using a relevance-guided fusion mechanism. Experiments on four benchmarks show that SELFI improves cross-manipulation generalization, outperforming prior methods by an average of 3.1\% AUC. On the challenging DFDC dataset, SELFI exceeds the previous best by 6\%. Code will be released upon paper acceptance.

cs.CV↗

arXiv2025

DeRAGEC: Denoising Named Entity Candidates with Synthetic Rationale for ASR Error Correction

We present DeRAGEC, a method for improving Named Entity (NE) correction in Automatic Speech Recognition (ASR) systems. By extending the Retrieval-Augmented Generative Error Correction (RAGEC) framework, DeRAGEC employs synthetic denoising rationales to filter out noisy NE candidates before correction. By leveraging phonetic similarity and augmented definitions, it refines noisy retrieved NEs using in-context learning, requiring no additional training. Experimental results on CommonVoice and STOP datasets show significant improvements in Word Error Rate (WER) and NE hit ratio, outperforming baseline ASR and RAGEC methods. Specifically, we achieved a 28% relative reduction in WER compared to ASR without postprocessing. Our source code is publicly available at: https://github.com/solee0022/deragec

cs.CL↗

arXiv2025

Benign-to-Toxic Jailbreaking: Inducing Harmful Responses from Harmless Prompts

Optimization-based jailbreaks typically adopt the Toxic-Continuation setting in large vision-language models (LVLMs), following the standard next-token prediction objective. In this setting, an adversarial image is optimized to make the model predict the next token of a toxic prompt. However, we find that the Toxic-Continuation paradigm is effective at continuing already-toxic inputs, but struggles to induce safety misalignment when explicit toxic signals are absent. We propose a new paradigm: Benign-to-Toxic (B2T) jailbreak. Unlike prior work, we optimize adversarial images to induce toxic outputs from benign conditioning. Since benign conditioning contains no safety violations, the image alone must break the model's safety mechanisms. Our method outperforms prior approaches, transfers in black-box settings, and complements text-based jailbreaks. These results reveal an underexplored vulnerability in multimodal alignment and introduce a fundamentally new direction for jailbreak approaches.

cs.CV↗

arXiv2025

Construction and analysis of guiding center distributions for tokamak plasmas with ambient radial electric field

The contribution of a time-independent toroidally-symmetric radial electric field $E_r$ is implemented in VisualStart [Comp. Phys. Comm. 275 (2022) 108305; arXiv:2111.08224], a code whose purposes include the construction of guiding center (GC) drift orbit databases for the study of plasma instabilities in tokamaks. $E_r$ is important for the thermal part of the velocity distribution and for fast particle resonances in the kHz frequency range. KSTAR, JT-60U and ITER tokamak cases are used as working examples to test our methods and discuss practical issues connected with $E_r$. Two points are worth noting: First, the GC orbit space is sampled in the magnetic midplane as before, and we find that in the presence of $E_r$, midplane-based coordinates are not only equivalent but superior to conventional constants of motion, allowing to attain high numerical accuracy and efficiency with a relatively simple mesh. Second, the periodic parallel acceleration and deceleration of GCs via the mirror force is modulated by $E_r$. Although this parallel electric acceleration averages to zero during a poloidal transit (or bounce) period, it has important consequences, one being the known shift of the trapped-passing boundary. Another consequence is that electric frequency shifts depend on the chosen reference point, so that some care is required when evaluating the $E_r$-dependence of transit frequencies for resonance analyses.

physics.plasm-ph↗

arXiv2025

Super-Polynomial Growth of the Generalized Persistence Diagram

The Generalized Persistence Diagram (GPD) for multi-parameter persistence naturally extends the classical notion of persistence diagram for one-parameter persistence. However, unlike its classical counterpart, computing the GPD remains a significant challenge. The main hurdle is that, while the GPD is defined as the Möbius inversion of the Generalized Rank Invariant (GRI), computing the GRI is intractable due to the formidable size of its domain, i.e., the set of all connected and convex subsets in a finite grid in $\mathbb{R}^d$ with $d \geq 2$. This computational intractability suggests seeking alternative approaches to computing the GPD. In order to study the complexity associated to computing the GPD, it is useful to consider its classical one-parameter counterpart, where for a filtration of a simplicial complex with $n$ simplices, its persistence diagram contains at most $n$ points. This observation leads to the question: 'Given a $d$-parameter simplicial filtration, could the cardinality of its GPD (specifically, the support of the GPD) also be bounded by a polynomial in the number of simplices in the filtration?' This is the case for $d=1$, where we compute the persistence diagram directly at the simplicial filtration level. If this were also the case for $d\geq2$, it might be possible to compute the GPD directly and much more efficiently without relying on the GRI. We show that the answer to the question above is negative, demonstrating the inherent difficulty of computing the GPD. More specifically, we construct a sequence of $d$-parameter simplicial filtrations where the cardinalities of their GPDs are not bounded by any polynomial in the the number of simplices. Furthermore, we show that several commonly used methods for constructing multi-parameter filtrations can give rise to such "wild" filtrations.

math.AT↗

arXiv2025

Maximizing the Position Embedding for Vision Transformers with Global Average Pooling

In vision transformers, position embedding (PE) plays a crucial role in capturing the order of tokens. However, in vision transformer structures, there is a limitation in the expressiveness of PE due to the structure where position embedding is simply added to the token embedding. A layer-wise method that delivers PE to each layer and applies independent Layer Normalizations for token embedding and PE has been adopted to overcome this limitation. In this paper, we identify the conflicting result that occurs in a layer-wise structure when using the global average pooling (GAP) method instead of the class token. To overcome this problem, we propose MPVG, which maximizes the effectiveness of PE in a layer-wise structure with GAP. Specifically, we identify that PE counterbalances token embedding values at each layer in a layer-wise structure. Furthermore, we recognize that the counterbalancing role of PE is insufficient in the layer-wise structure, and we address this by maximizing the effectiveness of PE through MPVG. Through experiments, we demonstrate that PE performs a counterbalancing role and that maintaining this counterbalancing directionality significantly impacts vision transformers. As a result, the experimental results show that MPVG outperforms existing methods across vision transformers on various tasks.

cs.CV↗

arXiv2025

DyPCL: Dynamic Phoneme-level Contrastive Learning for Dysarthric Speech Recognition

Dysarthric speech recognition often suffers from performance degradation due to the intrinsic diversity of dysarthric severity and extrinsic disparity from normal speech. To bridge these gaps, we propose a Dynamic Phoneme-level Contrastive Learning (DyPCL) method, which leads to obtaining invariant representations across diverse speakers. We decompose the speech utterance into phoneme segments for phoneme-level contrastive learning, leveraging dynamic connectionist temporal classification alignment. Unlike prior studies focusing on utterance-level embeddings, our granular learning allows discrimination of subtle parts of speech. In addition, we introduce dynamic curriculum learning, which progressively transitions from easy negative samples to difficult-to-distinguishable negative samples based on phonetic similarity of phoneme. Our approach to training by difficulty levels alleviates the inherent variability of speakers, better identifying challenging speeches. Evaluated on the UASpeech dataset, DyPCL outperforms baseline models, achieving an average 22.10\% relative reduction in word error rate (WER) across the overall dysarthria group.

cs.CL↗

arXiv2024

FRIDAY: Mitigating Unintentional Facial Identity in Deepfake Detectors Guided by Facial Recognizers

Previous Deepfake detection methods perform well within their training domains, but their effectiveness diminishes significantly with new synthesis techniques. Recent studies have revealed that detection models often create decision boundaries based on facial identity rather than synthetic artifacts, resulting in poor performance on cross-domain datasets. To address this limitation, we propose Facial Recognition Identity Attenuation (FRIDAY), a novel training method that mitigates facial identity influence using a face recognizer. Specifically, we first train a face recognizer using the same backbone as the Deepfake detector. The recognizer is then frozen and employed during the detector's training to reduce facial identity information. This is achieved by feeding input images into both the recognizer and the detector, and minimizing the similarity of their feature embeddings through our Facial Identity Attenuating loss. This process encourages the detector to generate embeddings distinct from the recognizer, effectively reducing the impact of facial identity. Extensive experiments demonstrate that our approach significantly enhances detection performance on both in-domain and cross-domain datasets.

cs.CV↗