arXiv subjects

Jinsoo Kim

Publications and source records attributed to Jinsoo Kim.

12 recordsLinked to original sources

arXiv2026

StreamAlign: Streaming Text-Aligned Speech Tokenization

Text-aligned speech tokenization methods have emerged to better align speech tokens with LLM token spaces, enabling more effective utilization of pretrained LLMs. However, they rely on offline automatic speech recognition (ASR), leading to two key limitations: (i) the need for complete utterances before tokenization, precluding real-time streaming, and (ii) vocabulary mismatch between ASR and LLMs, which reduces acoustic granularity from the subword to the word level. We introduce StreamAlign, a text-aligned speech tokenization framework that enables streaming tokenization for real-time speech-text joint modeling. StreamAlign performs online speech-text alignment by combining character-level RNN-Transducer alignment with word-level ASR guidance, mitigating ASR-LLM vocabulary mismatch while preserving recognition accuracy. A proactive word boundary classifier anticipates word completion at chunk boundaries, reducing tokenization latency from 560 ms to 270 ms. On LibriSpeech, StreamAlign achieves the lowest WER and highest UTMOS among evaluated tokenizers. Furthermore, StreamAlign-SLM, a spoken language model trained on StreamAlign units, outperforms other end-to-end spoken language models in speech continuation while achieving the strongest overall consistency on SALMon and spoken StoryCloze.

cs.CL↗

arXiv2026

Axial-Centric Cross-Plane Attention for 3D Medical Image Classification

Abridged: Clinicians commonly interpret 3D medical images by examining multiple anatomical planes rather than relying on volumetric views. In clinical CT workflows, the axial plane often serves as the primary diagnostic reference, while the auxiliary planes provide complementary spatial context. However, many existing 3D deep learning approaches either process volumetric data holistically or assign equal importance to all planes, failing to reflect this asymmetric, axial-centric interpretation strategy. To address this, we propose an axial-centric cross-plane attention architecture for 3D medical image classification that models asymmetric dependencies between anatomical planes. The architecture employs large-scale axial CT images pretrained MedDINOv3 as a frozen feature extractor for axial, coronal, and sagittal planes. RICA blocks and intra-plane transformer encoders capture plane-specific positional and contextual information, while axial-centric cross-plane transformer encoders selectively condition axial representations on complementary auxiliary representations. Experiments on six datasets from the MedMNIST3D benchmark show that the proposed method consistently outperforms existing 3D and multi-plane models in ACC and AUC. A lightweight variant, AC-Tiny, achieves competitive performance with substantially fewer trainable parameters, suggesting that architectural design contributes more to performance gains than increased model scale. Ablation studies further validate the importance of axial-centric querying, QKV allocation, directional cross-plane fusion, residual-free cross-attention, and classification head design. Slice-level Grad-CAM visualizations demonstrate that the model identifies diagnostically relevant regions across all planes. These findings highlight the value of aligning architectural design with clinical interpretation workflows for robust 3D medical image analysis.

cs.CV↗

arXiv2025

The RAG Paradox: A Black-Box Attack Exploiting Unintentional Vulnerabilities in Retrieval-Augmented Generation Systems

With the growing adoption of retrieval-augmented generation (RAG) systems, various attack methods have been proposed to degrade their performance. However, most existing approaches rely on unrealistic assumptions in which external attackers have access to internal components such as the retriever. To address this issue, we introduce a realistic black-box attack based on the RAG paradox, a structural vulnerability arising from the system's effort to enhance trust by revealing both the retrieved documents and their sources to users. This transparency enables attackers to observe which sources are used and how information is phrased, allowing them to craft poisoned documents that are more likely to be retrieved and upload them to the identified sources. Moreover, as RAG systems directly provide retrieved content to users, these documents must not only be retrievable but also appear natural and credible to maintain user confidence in the search results. Unlike prior work that focuses solely on improving document retrievability, our attack method explicitly considers both retrievability and user trust in the retrieved content. Both offline and online experiments demonstrate that our method significantly degrades system performance without internal access, while generating natural-looking poisoned documents.

cs.CR↗

arXiv2024

RICAU-Net: Residual-block Inspired Coordinate Attention U-Net for Segmentation of Small and Sparse Calcium Lesions in Cardiac CT

The Agatston score, which is the sum of the calcification in the four main coronary arteries, has been widely used in the diagnosis of coronary artery disease (CAD). However, many studies have emphasized the importance of the vessel-specific Agatston score, as calcification in a specific vessel is significantly correlated with the occurrence of coronary heart disease (CHD). In this paper, we propose the Residual-block Inspired Coordinate Attention U-Net (RICAU-Net), which incorporates coordinate attention in two distinct manners and a customized combo loss function for lesion-specific coronary artery calcium (CAC) segmentation. This approach aims to tackle the high class-imbalance issue associated with small and sparse CAC lesions. Experimental results and the ablation study demonstrate that the proposed method outperforms the five other U-Net based methods used in medical applications, by achieving the highest per-lesion Dice scores across all four lesions.

eess.IV↗

arXiv2023

D-Score: A Synapse-Inspired Approach for Filter Pruning

This paper introduces a new aspect for determining the rank of the unimportant filters for filter pruning on convolutional neural networks (CNNs). In the human synaptic system, there are two important channels known as excitatory and inhibitory neurotransmitters that transmit a signal from a neuron to a cell. Adopting the neuroscientific perspective, we propose a synapse-inspired filter pruning method, namely Dynamic Score (D-Score). D-Score analyzes the independent importance of positive and negative weights in the filters and ranks the independent importance by assigning scores. Filters having low overall scores, and thus low impact on the accuracy of neural networks are pruned. The experimental results on CIFAR-10 and ImageNet datasets demonstrate the effectiveness of our proposed method by reducing notable amounts of FLOPs and Params without significant Acc. Drop.

cs.NE↗

arXiv2004

Quantum key distribution between two groups using secret sharing

In this paper, we investigate properties of some multi-particle entangled states and, from the properties applying the secret sharing present a new type of quantum key distribution protocols as generalization of quantum key distribution between two persons. In the protocols each group can retrieve the secure key string, only if all members in each group should cooperate with one another. We also show that the protocols are secure against an external eavesdropper using the intercept/resend strategy.

quant-ph↗

arXiv2002

Photoemission and x-ray absorption study of MgC_(1-x)Ni_3

We investigated electronic structure of MgC_(1-x)Ni_3 with photoemission and x-ray absorption spectroscopy. Both results show that overall band structure is in reasonable agreement with band structure calculations including the existence of von Hove singularity (vHs)near E_F. However, we find that the sharp vHs peak theoretically predicted near the E_F is substantially suppressed. As for the Ni core level and absorption spectrum, there exist the satellites of Ni 2p which have a little larger energy separation and reduced intensity compared to the case of Ni-metal. These facts indicate that correlation effects among Ni 3d electrons may be important to understand various physical properties.

cond-mat.supr-con↗

arXiv2000

Function-dependent Phase Transform in Quantum Computing

We construct a quantum algorithm that performs function-dependent phase transform and requires no initialization of an ancillary register. The algorithm recovers the initial state of an ancillary register regardless of whether its state is pure or mixed. Thus we can use any qubits as an ancillary register even though they are entangled with others and are occupied by other computational process. We also show that our algorithm is optimal in the sense of the number of function evaluations.

quant-ph↗

arXiv2000

Quantum Algorithm for Generalized Deutsch-Jozsa Problem

We generalize the Deutsch-Jozsa problem and present a quantum algorithm that can solve the generalized Deutsch-Jozsa problem by a single evaluation of a given function. We discuss the initialization of an auxiliary register and present a generalized Deutsch-Jozsa algorithm that requires no initialization of an auxiliary register.

quant-ph↗

arXiv2000

The f-conditioned Phase Transform

We present a quantum algorithm for the f-conditioned phase transform which does not require any initialization of ancillary register. We also develop a quantum algorithm that can solve the generalized Deutsch-Jozsa problem by a single evaluation of a function.

quant-ph↗

arXiv1997

Quantum Database Searching by a Single Query

In this paper we give a quantum mechanical algorithm that can search a database by a single query, when the number of solutions is more than a quarter. It utilizes modified Grover operator of arbitrary phase.

quant-ph↗

arXiv1997

A Polynomial-Time Quantum Algorithm for Collision Problem

In this paper, we give a quantum algorithm which solves collision problem in an expected polynomial time. Especially, when the function is two-to-one, we present a quantum algorithm which can find a collision with certainty in a worst-case polynomial time. We also give a quantum algorithm which solves claw problem with certainty in a worst-case polynomial time.

quant-ph↗