arXiv subjects

Taehee Kim

Publications and source records attributed to Taehee Kim.

At least 19 recordsLinked to original sources

arXiv2026

Retrieve Only Relevant Tables Whether Few or Many: Adaptive Table Retrieval Method

Retrieving relevant tables from extensive databases for a given natural language query is essential for accurately answering questions in tasks such as text-to-SQL. Existing table retrieval approaches select a pre-determined set of k tables with the highest similarity to the query. However, the number of required tables varies across queries and cannot be known in advance. Enforcing a fixed number of retrieved tables regardless of the query may either retrieve an undersized set, failing to obtain all necessary evidence, or retrieve an oversized pool, including irrelevant tables. To address this issue, we propose an adaptive table retrieval method that adjusts the number of tables retrieved according to the requirements of each query. Specifically, we utilize an adaptive thresholding mechanism to selectively retrieve tables and integrate a sliding-window reranking algorithm to efficiently process a large table corpus. Extensive experiments on Spider, BIRD, and Spider 2.0 demonstrate that our method effectively addresses the limitations of the top-k retrieval strategy, improving performance in retrieval and downstream tasks. Our code and data are available at https://github.com/sbY99/Adaptive-Table-Retrieval.

cs.IR↗

arXiv2026

Searching for New Physics Inside Jets with the Herwig 7 Generalised Parton Shower

This study investigates parton shower evolution incorporating both Standard Model (SM) and beyond-the-Standard-Model (BSM) radiation, focusing on the phenomenology of a massive $Z'$ boson. While traditional approaches typically assume direct $Z'$ production in the hard process, the possibility of $Z'$ production within jets, enabled by subsequent emissions in the parton shower, offers a complementary opportunity to probe new physics through jet substructure and event topology. The newly developed Herwig 7 framework supporting BSM parton showers enables efficient simulation of $Z'$ production in the logarithmically enhanced regime. Using a simple BSM benchmark, the minimal $U(1)_{B{\rm -}L}$ extension of the SM, the interplay between the SM and BSM showers is evaluated to identify kinematic features that distinguish $Z'$-induced jets from conventional signatures. BSM-radiation signatures are contrasted with SM backgrounds such as QCD, top-quark, and Drell-Yan production, identifying potential discriminants for experimental searches. Experimental sensitivity at the LHC and prospective future colliders is estimated via statistical-significance projections. We find that $Z'$ bosons produced through parton shower radiation populate non-isolated regions inside jets, providing an avenue for new-physics searches overlooked in traditional analyses.

hep-ph↗

arXiv2026

Doubly slicing knots and embedding 3-manifolds in 4-manifolds

For a knot $K$ in the 3-sphere and a simply connected closed 4-manifold $X$, we define the $X$-double slice genus of $K$, extending the notion from the case when $X$ is the 4-sphere. We show that for each integer $n$, there exists an algebraically doubly slice and ribbon knot $K$ whose $X$-double slice genus is greater than $n$. Our arguments use new $L^2$-signature obstructions to embedding closed 3-manifolds with infinite cyclic first homology into closed 4-manifolds with infinite cyclic fundamental group, in a way that preserves first homology. We also extend the concept of the superslice genus of a knot to simply connected 4-manifolds and show that there exist doubly slice knots whose generalized superslice genera are arbitrarily large. Furthermore, we define the double stabilizing number of a knot, extending the stabilizing number introduced by Conway and Nagel, and show that this invariant can also be arbitrarily large.

math.GT↗

arXiv2025

The Collective Turing Test: Large Language Models Can Generate Realistic Multi-User Discussions

Large Language Models (LLMs) offer new avenues to simulate online communities and social media. Potential applications range from testing the design of content recommendation algorithms to estimating the effects of content policies and interventions. However, the validity of using LLMs to simulate conversations between various users remains largely untested. We evaluated whether LLMs can convincingly mimic human group conversations on social media. We collected authentic human conversations from Reddit and generated artificial conversations on the same topic with two LLMs: Llama 3 70B and GPT-4o. When presented side-by-side to study participants, LLM-generated conversations were mistaken for human-created content 39\% of the time. In particular, when evaluating conversations generated by Llama 3, participants correctly identified them as AI-generated only 56\% of the time, barely better than random chance. Our study demonstrates that LLMs can generate social media conversations sufficiently realistic to deceive humans when reading them, highlighting both a promising potential for social simulation and a warning message about the potential misuse of LLMs to generate new inauthentic social media content.

cs.CL↗

arXiv2025

Improving Electroencephalogram-Based Deception Detection in Concealed Information Test under Low Stimulus Heterogeneity

The concealed information test (CIT) is widely used for detecting deception in criminal investigations, primarily leveraging the P300 component of electroencephalogram (EEG) signals. However, the traditional bootstrapped amplitude difference (BAD) method struggles to accurately differentiate deceptive individuals from innocent ones when irrelevant stimuli carry familiarity or inherent meaning, thus limiting its practical applicability in real-world investigations. This study aimed to enhance the deception detection capability of the P300-based CIT, particularly under conditions of low stimulus heterogeneity. To closely simulate realistic investigative scenarios, we designed a realistic mock-crime setup in which participants were familiarized with all CIT stimuli except the target stimulus. EEG data acquired during CIT sessions were analyzed using the BAD method, machine learning algorithms, and deep learning (DL) methods (ShallowNet and EEGNet). Among these techniques, EEGNet demonstrated the highest deception detection accuracy at 86.67%, when employing our proposed data augmentation approach. Overall, DL methods could significantly improve the accuracy of deception detection under challenging conditions of low stimulus heterogeneity by effectively capturing subtle cognitive responses not accessible through handcrafted features. To the best of our knowledge, this is the first study that employed DL approaches for subject-independent deception classification using the CIT paradigm.

q-bio.NC↗

arXiv2025

Compact and robust optical frequency reference module based on reproducible and redistributable optical design

Stabilized optical frequency references (OFRs) are indispensable for atom-based quantum technologies, optical communications, and precision metrology. As these systems become more sophisticated, demands for compactness, robustness, and straightforward reproduction have grown. In this work, we present a robust 19-inch rack-mountable OFR module designed via a web-based CAD workflow that allows straightforward redistribution and reproduction. Its optical subsystem, designed based on a modeled laser beam path, places optical elements with sub-millimeter accuracy on a custom-machined aluminum plate, allowing straightforward assembly without extensive alignment and providing high mechanical stability. The module maintains frequency-stable operation for several months without user intervention and exhibits high robustness to mechanical vibrations up to 4g. All design files, including mechanical and optical metadata, are openly shared for straightforward reproduction and adaptation.

physics.atom-ph↗

arXiv2024

Can Tool-augmented Large Language Models be Aware of Incomplete Conditions?

Recent advancements in integrating large language models (LLMs) with tools have allowed the models to interact with real-world environments. However, these tool-augmented LLMs often encounter incomplete scenarios when users provide partial information or the necessary tools are unavailable. Recognizing and managing such scenarios is crucial for LLMs to ensure their reliability, but this exploration remains understudied. This study examines whether LLMs can identify incomplete conditions and appropriately determine when to refrain from using tools. To quantitatively evaluate this capability, we construct a new benchmark dataset where instances are systematically altered to simulate the ambiguous and incomplete conditions common in real-world interactions. Our experiments reveal that even state-of-the-art LLMs often struggle to identify these conditions, attempting to use tools without sufficient information or when the correct tool is unavailable. To better understand these limitations, we conduct a detailed behavioral analysis across various conditions, including implicit evaluation and scenarios where models receive feedback from previous tool invocations. Based on this analysis, we propose a novel prompting-based reasoning strategy that explicitly instructs models to assess the sufficiency of information and the availability of tools. Our proposed approach significantly enhances the models' ability to recognize incomplete conditions, resulting in more informed and contextually appropriate tool-use decisions. We believe our research contributes to advancing the reliability of LLMs, especially in real-world applications where incomplete or ambiguous information is common. Our dataset is available at https://huggingface.co/datasets/ddehun/ICT.

cs.CL↗

arXiv2024

Circularly Polarized Luminescence Without External Magnetic Fields from Individual CsPbBr3 Perovskite Quantum Dots

Lead halide perovskite quantum dots (QDs), the latest generation of colloidal QD family, exhibit outstanding optical properties which are now exploited as both classical and quantum light sources. Most of their rather exceptional properties are related to the peculiar exciton fine-structure of band-edge states which can support unique bright triplet excitons. The degeneracy of the bright triplet excitons is lifted with energetic splitting in the order of millielectronvolts, which can be resolved by the photoluminescence (PL) measurements of single QDs at cryogenic temperatures. Each bright exciton fine-structure-state (FSS) exhibits a dominantly linear polarization, in line with several theoretical models based on the sole crystal field, exchange interaction and shape anisotropy. Here, we show that in addition to a high degree of linear polarization, the individual exciton FSS can exhibit a non-negligible degree of circular polarization even without external magnetic fields by investigating the four Stokes parameters of the exciton fine-structure in individual CsPbBr3 QDs through Stokes polarimetric measurements. We observe a degree of circular polarization up to ~38%, which could not be detected by using the conventional polarimetric technique. In addition, we found a consistent transition from left- to right-hand circular polarization within the fine-structure triplet manifold, which was observed in magnetic field dependent experiments. Our optical investigation provides deeper insights into the nature of the exciton fine-structures and thereby drives the yet-incomplete understanding of the unique photophysical properties of this novel class of QDs, potentially opening new scenarios in chiral quantum optics.

physics.app-ph↗

arXiv2024

Pretraining Vision-Language Model for Difference Visual Question Answering in Longitudinal Chest X-rays

Difference visual question answering (diff-VQA) is a challenging task that requires answering complex questions based on differences between a pair of images. This task is particularly important in reading chest X-ray images because radiologists often compare multiple images of the same patient taken at different times to track disease progression and changes in its severity in their clinical practice. However, previous works focused on designing specific network architectures for the diff-VQA task, missing opportunities to enhance the model's performance using a pretrained vision-language model (VLM). Here, we introduce a novel VLM called PLURAL, which is pretrained on natural and longitudinal chest X-ray data for the diff-VQA task. The model is developed using a step-by-step approach, starting with being pretrained on natural images and texts, followed by being trained using longitudinal chest X-ray data. The longitudinal data consist of pairs of X-ray images, along with question-answer sets and radiologist's reports that describe the changes in lung abnormalities and diseases over time. Our experimental results show that the PLURAL model outperforms state-of-the-art methods not only in diff-VQA for longitudinal X-rays but also in conventional VQA for a single X-ray image. Through extensive experiments, we demonstrate the effectiveness of the proposed VLM architecture and pretraining method in improving the model's performance.

cs.CV↗

arXiv2024

Iterated satellite operators on the knot concordance group

We show that for a winding number zero satellite operator $P$ on the knot concordance group, if the axis of $P$ has nontrivial self-pairing under the Blanchfield form of the pattern, then the image of the iteration $P^n$ generates an infinite rank subgroup for each $n$. Furthermore, the graded quotients of the filtration of the knot concordance group associated with $P$ have infinite rank at all levels. This gives an affirmative answer to a question of Hedden and Pinz\'{o}n-Caicedo in many cases. We also show that under the same hypotheses, $P^n$ is not a homomorphism on the knot concordance group for each $n$. We use amenable $L^2$-signatures to prove these results.

math.GT↗

arXiv2024

Generalizing Visual Question Answering from Synthetic to Human-Written Questions via a Chain of QA with a Large Language Model

Visual question answering (VQA) is a task where an image is given, and a series of questions are asked about the image. To build an efficient VQA algorithm, a large amount of QA data is required which is very expensive. Generating synthetic QA pairs based on templates is a practical way to obtain data. However, VQA models trained on those data do not perform well on complex, human-written questions. To address this issue, we propose a new method called {\it chain of QA for human-written questions} (CoQAH). CoQAH utilizes a sequence of QA interactions between a large language model and a VQA model trained on synthetic data to reason and derive logical answers for human-written questions. We tested the effectiveness of CoQAH on two types of human-written VQA datasets for 3D-rendered and chest X-ray images and found that it achieved state-of-the-art accuracy in both types of data. Notably, CoQAH outperformed general vision-language models, VQA models, and medical foundation models with no finetuning.

cs.CL↗

arXiv2023

Fast and accurate sparse-view CBCT reconstruction using meta-learned neural attenuation field and hash-encoding regularization

Cone beam computed tomography (CBCT) is an emerging medical imaging technique to visualize the internal anatomical structures of patients. During a CBCT scan, several projection images of different angles or views are collectively utilized to reconstruct a tomographic image. However, reducing the number of projections in a CBCT scan while preserving the quality of a reconstructed image is challenging due to the nature of an ill-posed inverse problem. Recently, a neural attenuation field (NAF) method was proposed by adopting a neural radiance field algorithm as a new way for CBCT reconstruction, demonstrating fast and promising results using only 50 views. However, decreasing the number of projections is still preferable to reduce potential radiation exposure, and a faster reconstruction time is required considering a typical scan time. In this work, we propose a fast and accurate sparse-view CBCT reconstruction (FACT) method to provide better reconstruction quality and faster optimization speed in the minimal number of view acquisitions ($<$ 50 views). In the FACT method, we meta-trained a neural network and a hash-encoder using a few scans (= 15), and a new regularization technique is utilized to reconstruct the details of an anatomical structure. In conclusion, we have shown that the FACT method produced better, and faster reconstruction results over the other conventional algorithms based on CBCT scans of different body parts (chest, head, and abdomen) and CT vendors (Siemens, Phillips, and GE).

eess.IV↗

arXiv2023

Correlation-Driven Multi-Level Multimodal Learning for Anomaly Detection on Multiple Energy Sources

Advanced metering infrastructure (AMI) has been widely used as an intelligent energy consumption measurement system. Electric power was the representative energy source that can be collected by AMI; most existing studies to detect abnormal energy consumption have focused on a single energy source, i.e., power. Recently, other energy sources such as water, gas, and heating have also been actively collected. As a result, it is necessary to develop a unified methodology for anomaly detection across multiple energy sources; however, research efforts have rarely been made to tackle this issue. The inherent difficulty with this issue stems from the fact that anomalies are not usually annotated. Moreover, existing works of anomaly definition depend on only individual energy sources. In this paper, we first propose a method for defining anomalies considering not only individual energy sources but also correlations between them. Then, we propose a new Correlation-driven Multi-Level Multimodal Learning model for anomaly detection on multiple energy sources. The distinguishing property of the model incorporates multiple energy sources in multi-levels based on the strengths of the correlations between them. Furthermore, we generalize the proposed model in order to integrate arbitrary new energy sources with further performance improvement, considering not only correlated but also non-correlated sources. Through extensive experiments on real-world datasets consisting of three to five energy sources, we demonstrate that the proposed model clearly outperforms the existing multimodal learning and recent time-series anomaly detection models, and we observe that our model makes further the performance improvement as more correlated or non-correlated energy sources are integrated.

cs.LG↗

arXiv2022

PePe: Personalized Post-editing Model utilizing User-generated Post-edits

Incorporating personal preference is crucial in advanced machine translation tasks. Despite the recent advancement of machine translation, it remains a demanding task to properly reflect personal style. In this paper, we introduce a personalized automatic post-editing framework to address this challenge, which effectively generates sentences considering distinct personal behaviors. To build this framework, we first collect post-editing data that connotes the user preference from a live machine translation system. Specifically, real-world users enter source sentences for translation and edit the machine-translated outputs according to the user's preferred style. We then propose a model that combines a discriminator module and user-specific parameters on the APE framework. Experimental results show that the proposed method outperforms other baseline models on four different metrics (i.e., BLEU, TER, YiSi-1, and human evaluation).

cs.CL↗

arXiv2022

Reweighting Strategy based on Synthetic Data Identification for Sentence Similarity

Semantically meaningful sentence embeddings are important for numerous tasks in natural language processing. To obtain such embeddings, recent studies explored the idea of utilizing synthetically generated data from pretrained language models (PLMs) as a training corpus. However, PLMs often generate sentences much different from the ones written by human. We hypothesize that treating all these synthetic examples equally for training deep neural networks can have an adverse effect on learning semantically meaningful embeddings. To analyze this, we first train a classifier that identifies machine-written sentences, and observe that the linguistic features of the sentences identified as written by a machine are significantly different from those of human-written sentences. Based on this, we propose a novel approach that first trains the classifier to measure the importance of each sentence. The distilled information from the classifier is then used to train a reliable sentence embedding model. Through extensive evaluation on four real-world datasets, we demonstrate that our model trained on synthetic data generalizes well and outperforms the existing baselines. Our implementation is publicly available at https://github.com/ddehun/coling2022_reweighting_sts.

cs.CL↗

arXiv2022

Knot reversal and rational concordance

We give an infinite family of knots that are not rationally concordant to their reverses. More precisely, if R denotes the involution of the rational knot concordance group QC induced by string reversal and Fix(R) denotes the subgroup of knots fixed under R in QC, then QC/Fix(R) contains an infinite rank subgroup. As a corollary, we show that there exists a knot K such that for every pair of coprime integers p and q, the (p,q)-cable of K is not concordant to the reverse of the (p,q)-cable of K.

math.GT↗

arXiv2021

AVocaDo: Strategy for Adapting Vocabulary to Downstream Domain

During the fine-tuning phase of transfer learning, the pretrained vocabulary remains unchanged, while model parameters are updated. The vocabulary generated based on the pretrained data is suboptimal for downstream data when domain discrepancy exists. We propose to consider the vocabulary as an optimizable parameter, allowing us to update the vocabulary by expanding it with domain-specific vocabulary based on a tokenization statistic. Furthermore, we preserve the embeddings of the added words from overfitting to downstream data by utilizing knowledge learned from a pretrained language model with a regularization term. Our method achieved consistent performance improvements on diverse domains (i.e., biomedical, computer science, news, and reviews).

cs.CL↗

arXiv2020

Unsupervised Neural Machine Translation for Low-Resource Domains via Meta-Learning

Unsupervised machine translation, which utilizes unpaired monolingual corpora as training data, has achieved comparable performance against supervised machine translation. However, it still suffers from data-scarce domains. To address this issue, this paper presents a novel meta-learning algorithm for unsupervised neural machine translation (UNMT) that trains the model to adapt to another domain by utilizing only a small amount of training data. We assume that domain-general knowledge is a significant factor in handling data-scarce domains. Hence, we extend the meta-learning algorithm, which utilizes knowledge learned from high-resource domains, to boost the performance of low-resource UNMT. Our model surpasses a transfer learning-based approach by up to 2-4 BLEU scores. Extensive experimental results show that our proposed algorithm is pertinent for fast adaptation and consistently outperforms other baseline models.

cs.CL↗