arXiv subjects

Yusuke Sakai

Publications and source records attributed to Yusuke Sakai.

At least 37 records · Page 2Linked to original sources

arXiv2026

The Wisdom of Many Queries: Complexity-Diversity Principle for Dense Retriever Training

Synthetic query generation has become essential for training dense retrievers, yet prior methods generate one query per document, focusing solely on query quality. We are the first to systematically study multi-query synthesis and discover a quality-diversity trade-off: high-quality queries benefit in-domain tasks, while diverse queries benefit out-of-domain (OOD) generalization. Through controlled experiments on 4 benchmark types across Contriever, RetroMAE, and Qwen3-Embedding, we find that diversity benefit strongly correlates with query complexity (r$\geq$0.95, p<0.05), approximated by content words (CW). We formalize this as the Complexity-Diversity Principle (CDP): query complexity determines optimal diversity. Based on CDP, we propose complexity-aware training: multi-query synthesis for high-complexity tasks and CW-weighted training for existing data. Both strategies improve OOD performance on reasoning-intensive benchmarks, with compounded gains when combined.

cs.IR↗

arXiv2026

Spectral and photometric variability of SS 433 observed with XRISM and simultaneous optical and near-infrared telescopes

We present results from coordinated multiwavelength observations of the SS 433, obtained with XRISM, optical telescopes, and near-infrared camera during 2024 April and 2025 March. The XRISM exposures amounted to ~200 ks in 2024 and ~100 ks in 2025. With XRISM/Resolve's high spectral resolution and large effective area, we clearly resolved numerous emission lines even in short time segments, achieving improved accuracy in Doppler-shift measurements relative to earlier observations. The simultaneously obtained X-ray and optical Doppler shifts suggest a possible tendency for the optical emission to lag slightly behind the X-rays. In the Resolve data, the Doppler shifts of the two jet components exhibited apparent asymmetries, with jet speeds fluctuating around ~0.26$\pm$0.01$c$ in 2024 and ~0.30$\pm$0.01$c$ in 2025. The velocity variations indicated modulations on a timescale of ~6.3 d, with a phase offset of about -90$^{\circ}$ relative to the nutation cycle. The observed line widths and flux of the approaching and receding jets appear consistent with the expected geometrical effects, indicating systematically larger line widths in the inner regions of the jets, as proposed by Shidatsu et al. (2025). Optical light curves show flares of ~400 s in 2024 and ~1600 s in 2025, with amplitudes up to ~15% during out-of-eclipse intervals, while the XRISM/Xtend light curves show no significant variability within the overlapping intervals and given the statistical uncertainties. Near-infrared photometry in 2024, obtained during an out-of-eclipse interval at a different epoch from the optical observations, showed no flare-like variability, and the X-ray band also remained constant within uncertainties. These coordinated observations provide a foundation for future XRISM studies aimed at probing the dynamical properties of the relativistic jets in SS 433.

astro-ph.HE↗

arXiv2026

Grammatical Error Correction Evaluation by Optimally Transporting Edit Representation

Automatic evaluation in grammatical error correction (GEC) is crucial for selecting the best-performing systems. Currently, reference-based metrics are a popular choice, which basically measure the similarity between hypothesis and reference sentences. However, similarity measures based on embeddings, such as BERTScore, are often ineffective, since many words in the source sentences remain unchanged in both the hypothesis and the reference. This study focuses on edits specifically designed for GEC, i.e., ERRANT, and computes similarity measured over the edits from the source sentence. To this end, we propose edit vector, a representation for an edit, and introduce a new metric, UOT-ERRANT, which transports these edit vectors from hypothesis to reference using unbalanced optimal transport. Experiments with SEEDA meta-evaluation show that UOT-ERRANT improves evaluation performance, particularly in the +Fluency domain where many edits occur. Moreover, our method is highly interpretable because the transport plan can be interpreted as a soft edit alignment, making UOT-ERRANT a useful metric for both system ranking and analyzing GEC systems. Our code is available from https://github.com/gotutiyan/uot-errant.

cs.CL↗

arXiv2026

Towards Artwork Explanation in Large-scale Vision Language Models

Large-scale Vision-Language Models (LVLMs) output text from images and instructions, demonstrating capabilities in text generation and comprehension. However, it has not been clarified to what extent LVLMs possess the ability to understand the knowledge necessary for explaining images, the complex relationships between various pieces of knowledge, and how they integrate these understandings into their explanations. To address this issue, we propose a new task: the artwork explanation generation task, along with its evaluation dataset and metrics for quantitatively assessing the understanding and utilization of knowledge about artworks. This task is apt for image description based on the premise that LVLMs are expected to have pre-existing knowledge of artworks, which are often subjects of wide recognition and documented information. It consists of two parts: generating explanations from images and titles of artworks, and generating explanations using only images, thus evaluating the LVLMs' language-based and vision-based knowledge. Alongside, we release a training dataset for LVLMs to learn explanations that incorporate knowledge about artworks. Our findings indicate that LVLMs not only struggle with integrating language and visual information but also exhibit a more pronounced limitation in acquiring knowledge from images alone.

cs.CV↗

arXiv2026

Microscopic Determination of the c-axis-Oriented Antiferromagnetic Structure in LaMnSi by $^{55}$Mn and $^{139}$La NMR

We report a microscopic investigation of the magnetic structure and electronic properties of LaMnSi in its antiferromagnetic (AFM) state using nuclear magnetic resonance (NMR). Field-swept $^{55}$Mn- and $^{139}$La-NMR spectra, as well as zero-field 55Mn-NMR (ZFNMR) spectra, reveal that the Mn ordered moments are parallel to the tetragonal c axis, consistent with the C-type AFM structure and the realization of an odd-parity multipole order. The internal field at the Mn site is determined to be 19.64 T at 4.2 K, corresponding to a hyperfine coupling constant of Ahf = 6.0 T/uB. Nuclear spin-lattice relaxation rate 1/T1 exhibits a characteristic behavior of itinerant antiferromagnetism, showing metallic behavior at low temperatures and magnon-induced enhancement upon approaching the Neel temperature (TN = 295 K). These results show LaMnSi as an ideal compound to study 3d electron magnetism and odd-parity multipole order in the RT Si (R = rare-earth, T = transition metal) system, free of the complexities of 4f electrons.

cond-mat.str-el↗

arXiv2026

Diagnosing Vision Language Models' Perception by Leveraging Human Methods for Color Vision Deficiencies

Large-scale Vision-Language Models (LVLMs) are being deployed in real-world settings that require visual inference. As capabilities improve, applications in navigation, education, and accessibility are becoming practical. These settings require accommodation of perceptual variation rather than assuming a uniform visual experience. Color perception illustrates this requirement: it is central to visual understanding yet varies across individuals due to Color Vision Deficiencies, an aspect largely ignored in multimodal AI. In this work, we examine whether LVLMs can account for variation in color perception using the Ishihara Test. We evaluate model behavior through generation, confidence, and internal representation, using Ishihara plates as controlled stimuli that expose perceptual differences. Although models possess factual knowledge about color vision deficiencies and can describe the test, they fail to reproduce the perceptual outcomes experienced by affected individuals and instead default to normative color perception. These results indicate that current systems lack mechanisms for representing alternative perceptual experiences, raising concerns for accessibility and inclusive deployment in multimodal settings.

cs.CV↗

arXiv2026

HalluCitation Matters: Revealing the Impact of Hallucinated References with 300 Hallucinated Papers in ACL Conferences

Recently, we have often observed hallucinated citations or references that do not correspond to any existing work in papers under review, preprints, or published papers. Such hallucinated citations pose a serious concern to scientific reliability. When they appear in accepted papers, they may also negatively affect the credibility of conferences. In this study, we refer to hallucinated citations as "HalluCitation" and systematically investigate their prevalence and impact. We analyze all papers published at ACL, NAACL, and EMNLP in 2024 and 2025, including main conference, Findings, and workshop papers. Our analysis reveals that nearly 300 papers contain at least one HalluCitation, most of which were published in 2025. Notably, half of these papers were identified at EMNLP 2025, the most recent conference, indicating that this issue is rapidly increasing. Moreover, more than 100 such papers were accepted as main conference and Findings papers at EMNLP 2025, affecting the credibility.

cs.CL↗

arXiv2026

Hacking Neural Evaluation Metrics with Single Hub Text

Strongly human-correlated evaluation metrics serve as an essential compass for the development and improvement of generation models and must be highly reliable and robust. Recent embedding-based neural text evaluation metrics, such as COMET for translation tasks, are widely used in both research and development fields. However, there is no guarantee that they yield reliable evaluation results due to the black-box nature of neural networks. To raise concerns about the reliability and safety of such metrics, we propose a method for finding a single adversarial text in the discrete space that is consistently evaluated as high-quality, regardless of the test cases, to identify the vulnerabilities in evaluation metrics. The single hub text found with our method achieved 79.1 COMET% and 67.8 COMET% in the WMT'24 English-to-Japanese (En--Ja) and English-to-German (En--De) translation tasks, respectively, outperforming translations generated individually for each source sentence by using M2M100, a general translation model. Furthermore, we also confirmed that the hub text found with our method generalizes across multiple language pairs such as Ja--En and De--En.

cs.CL↗

arXiv2025

Agreement-Constrained Probabilistic Minimum Bayes Risk Decoding

Minimum Bayes risk (MBR) decoding generates high-quality translations by maximizing the expected utility of output candidates, but it evaluates all pairwise scores over the candidate set; hence, it takes quadratic time with respect to the number of candidates. To reduce the number of utility function calls, probabilistic MBR (PMBR) decoding partially evaluates quality scores using sampled pairs of candidates and completes the missing scores with a matrix completion algorithm. Nevertheless, it degrades the translation quality as the number of utility function calls is reduced. Therefore, to improve the trade-off between quality and cost, we propose agreement-constrained PMBR (AC-PMBR) decoding, which leverages a knowledge distilled model to guide the completion of the score matrix. Our AC-PMBR decoding improved approximation errors of matrix completion by up to 3 times and achieved higher translation quality compared with PMBR decoding at a comparable computational cost on the WMT'23 En$\leftrightarrow$De translation tasks.

cs.CL↗

arXiv2025

XRISM High-resolution Spectroscopy of SS 433: Evidence of Decreasing Line-of-Sight Velocity Dispersion along the Jet

We report on the jet structure in SS 433 based on X-ray high resolution spectroscopy with the XRISM/Resolve. The source was observed over 5 days covering both inside and outside an eclipse of the compact object by the companion star. Doppler-shifted, ionized Fe and Ni K emission lines were resolved, as well as lower-energy lines including Si and S K lines. Time-resolved spectral analysis showed that Fe and Ni K line widths were $1020 \pm 40$ km s$^{-1}$ (corresponding the 1$σ$ width) in the eclipse phase, gradually increased during the egress, and reached $1740 \pm 30$ km s$^{-1}$ outside the eclipse. A time-averaged spectrum outside the eclipse confirmed that the Fe and Ni K lines in 5.5-9 keV are significantly broader than the Si and S K$α$ emission lines in 2--4 keV. Specifically, the width in 5.5-9 keV was measured to be $1900 \pm 80$ km s$^{-1}$, whereas the width in 2-4 keV is $1300^{+300}_{-400}$ km s$^{-1}$ for the approaching (blueshifted) jet component. These results indicate that radial velocity dispersion of the jet plasma in SS 433 decreases as it moves outward. We interpret this variation as progressive jet collimation along its axis, as suggested by Namiki et al. (2003), or a decrease in turbulence in the jet plasma flow within the X-ray emitting region. We also detected a clear difference in velocity dispersion between the approaching and receding (redshifted) jet components in the 5.5-9 keV band outside eclipse. The receding jet exhibited a smaller velocity dispersion ($1400 \pm 200$ km s$^{-1}$) than the approaching jet. Since the observation was conducted when the approaching jet was tilted toward the observer, this may suggest that the receding jet was more extensively occulted by the accretion disk.

astro-ph.HE↗

arXiv2025

Decadal brightening in the northeastern non-thermal filament of Cassiopeia A

We present the decadal brightening of non-thermal emission flux in the northeastern filament of the young supernova remnant Cassiopeia A (Cas A), which highlights dynamic processes in the forward shock. This filament, characterized by the highest particle acceleration rate among Cas A's outer shells, offers an exceptional opportunity to investigate underlying astrophysical mechanisms. Since 2000, the non-thermal flux has increased by several tens of percent before plateauing, while the spectral shape has remained largely unchanged. Over the past two decades, the filament's morphology has evolved significantly, splitting into two distinct sections. Detailed analysis reveals contrasting behaviors: one section shows a flux increase followed by saturation, while the other maintains a steady flux. These differences likely arise from nonlinear effects, including magnetohydrodynamic interactions influenced by magnetic field orientation, interactions with surrounding material, and complex fluid dynamics associated with young supernova remnants. The localized evolution of this filament, captured with high spatial resolution, provides critical insights into the temporal dynamics of non-thermal particles and the generation of cosmic rays from asymmetric supernova explosions.

astro-ph.HE↗

arXiv2025

Glitch noise classification in KAGRA O3GK observing data using unsupervised machine learning

Gravitational wave interferometers are disrupted by various types of nonstationary noise, referred to as glitch noise, that affect data analysis and interferometer sensitivity. The accurate identification and classification of glitch noise are essential for improving the reliability of gravitational wave observations. In this study, we demonstrated the effectiveness of unsupervised machine learning for classifying images with nonstationary noise in the KAGRA O3GK data. Using a variational autoencoder (VAE) combined with spectral clustering, we identified eight distinct glitch noise categories. The latent variables obtained from VAE were dimensionally compressed, visualized in three-dimensional space, and classified using spectral clustering to better understand the glitch noise characteristics of KAGRA during the O3GK period. Our results highlight the potential of unsupervised learning for efficient glitch noise classification, which may in turn potentially facilitate interferometer upgrades and the development of future third-generation gravitational wave observatories.

gr-qc↗

arXiv2025

Regression of Suspension Violin Modes in KAGRA O3GK Data with Kalman Filters

Suspension thermal modes in interferometric gravitational-wave detectors produce narrow, high-Q spectral lines that can contaminate gravitational searches and bias parameter estimation. In KAGRA, cryogenic mirrors are held by thick suspension fibers, designed to sustain such a low-temperature environment, which may further affect inharmonicity modes, fiber dimensions, and mechanical behavior compared to typical interferometers. As these modes remain a prominent source of narrowband contamination, we implement a Kalman filter to model and track violin lines, building on the methodology introduced in [1], and apply subtraction to KAGRA O3GK data. Using gravitational-wave template injections, we validate that the subtraction preserves matched-filter SNR while effectively suppressing line power. Comparisons of power spectral densities and residual analyses confirm that the method removes deterministic line contributions without introducing waveform distortions. This approach provides a cleaner strain channel for searches and parameter estimation and will become increasingly important for future low-temperature detectors with higher-Q suspensions, such as the Einstein Telescope.

gr-qc↗

arXiv2025

Reliability Crisis of Reference-free Metrics for Grammatical Error Correction

Reference-free evaluation metrics for grammatical error correction (GEC) have achieved high correlation with human judgments. However, these metrics are not designed to evaluate adversarial systems that aim to obtain unjustifiably high scores. The existence of such systems undermines the reliability of automatic evaluation, as it can mislead users in selecting appropriate GEC systems. In this study, we propose adversarial attack strategies for four reference-free metrics: SOME, Scribendi, IMPARA, and LLM-based metrics, and demonstrate that our adversarial systems outperform the current state-of-the-art. These findings highlight the need for more robust evaluation methods.

cs.CL↗

arXiv2025

Multilingual Dialogue Generation and Localization with Dialogue Act Scripting

Non-English dialogue datasets are scarce, and models are often trained or evaluated on translations of English-language dialogues, an approach which can introduce artifacts that reduce their naturalness and cultural appropriateness. This work proposes Dialogue Act Script (DAS), a structured framework for encoding, localizing, and generating multilingual dialogues from abstract intent representations. Rather than translating dialogue utterances directly, DAS enables the generation of new dialogues in the target language that are culturally and contextually appropriate. By using structured dialogue act representations, DAS supports flexible localization across languages, mitigating translationese and enabling more fluent, naturalistic conversations. Human evaluations across Italian, German, and Chinese show that DAS-generated dialogues consistently outperform those produced by both machine and human translators on measures of cultural relevance, coherence, and situational appropriateness.

cs.CL↗

arXiv2025

Gravitational Wave Memory from Accelerating Relativistic Jets in Multiple Thick Shell Scenarios

Gravitational wave (GW) memory, a permanent distortion of the space-time metric, is anticipated during the acceleration of relativistic jets in gamma-ray bursts (GRBs). While the precise mechanism behind GRBs is not yet fully understood, detecting GW memory may contribute to clarifying their nature. In this paper, we consider various scenarios of GW memory emission, including both single and multiple shells with thin- and thick-shells. In particular, the memory spectrum for each scenario is compared with the sensitivity of next-generation detectors, namely DECIGO and ET-D. Physical properties spread over a broad-band region, emphasizing the importance of combined and wide-band observations. We also simulate GW memory based on nearby, realistic scenarios and demonstrate its detectability.

astro-ph.HE↗

arXiv2025

SinhalaMMLU: A Comprehensive Benchmark for Evaluating Multitask Language Understanding in Sinhala

Large Language Models (LLMs) demonstrate impressive general knowledge and reasoning abilities, yet their evaluation has predominantly focused on global or anglocentric subjects, often neglecting low-resource languages and culturally specific content. While recent multilingual benchmarks attempt to bridge this gap, many rely on automatic translation, which can introduce errors and misrepresent the original cultural context. To address this, we introduce SinhalaMMLU, the first multiple-choice question answering benchmark designed specifically for Sinhala, a low-resource language. The dataset includes over 7,000 questions spanning secondary to collegiate education levels, aligned with the Sri Lankan national curriculum, and covers six domains and 30 subjects, encompassing both general academic topics and culturally grounded knowledge. We evaluate 26 LLMs on SinhalaMMLU and observe that, while Claude 3.5 sonnet and GPT-4o achieve the highest average accuracies at 67% and 62% respectively, overall model performance remains limited. In particular, models struggle in culturally rich domains such as the Humanities, revealing substantial room for improvement in adapting LLMs to low-resource and culturally specific contexts.

cs.CL↗

arXiv2025

Languages Still Left Behind: Toward a Better Multilingual Machine Translation Benchmark

Multilingual machine translation (MT) benchmarks play a central role in evaluating the capabilities of modern MT systems. Among them, the FLORES+ benchmark is widely used, offering English-to-many translation data for over 200 languages, curated with strict quality control protocols. However, we study data in four languages (Asante Twi, Japanese, Jinghpaw, and South Azerbaijani) and uncover critical shortcomings in the benchmark's suitability for truly multilingual evaluation. Human assessments reveal that many translations fall below the claimed 90% quality standard, and the annotators report that source sentences are often too domain-specific and culturally biased toward the English-speaking world. We further demonstrate that simple heuristics, such as copying named entities, can yield non-trivial BLEU scores, suggesting vulnerabilities in the evaluation protocol. Notably, we show that MT models trained on high-quality, naturalistic data perform poorly on FLORES+ while achieving significant gains on our domain-relevant evaluation set. Based on these findings, we advocate for multilingual MT benchmarks that use domain-general and culturally neutral source texts rely less on named entities, in order to better reflect real-world translation challenges.

cs.CL↗