arXiv subjects

Yusuke Sakai

Publications and source records attributed to Yusuke Sakai.

At least 19 recordsLinked to original sources

arXiv2026

ExpArt-KG: Artwork Image Description Generation through Iterative Exploration of Knowledge Graphs

Large Vision-Language Models (LVLMs) achieve strong performance on image-grounded text generation and visual question answering. However, it remains difficult for them to comprehensively and accurately describe the factual relations among the entities and concepts associated with the objects depicted in an image. In this work, we propose a framework that efficiently exploits factual information from a knowledge graph via retrieval-augmented generation (RAG), with the goal of enabling LVLMs to generate detailed and accurate image explanations. Specifically, our method alternates between answer generation and knowledge-graph retrieval, and controls the search using a correctness judgment, thereby acquiring the necessary and sufficient factual information efficiently. We also construct a knowledge graph for the artwork domain (ExpArt-KG), in which the correspondence between images and entities is unambiguous. Applying the proposed method to this knowledge graph, we show experimentally that it improves the level of detail of artwork explanations and reduces the retrieval cost of external knowledge while maintaining generation quality comparable to that of iterating a fixed number of times.

cs.CL↗

arXiv2026

Overfitting Mitigation via Singular Value Decomposition in Minimum Bayes Risk Decoding

Minimum Bayes Risk (MBR) decoding enables high-quality text generation by selecting the hypothesis that maximizes a utility metric over sampled pseudo-references. However, it is highly susceptible to metric overfitting: it can irregularly inflate the chosen utility metric at the direct expense of other unoptimized evaluation metrics. To mitigate this, we introduce SVD-MBR, which frames the pairwise utility matrix as a noisy information signal. By computing a low-rank approximation via Singular Value Decomposition (SVD) and retaining only the top-$k$ components, we effectively decouple true consensus from metric noise. Experiments demonstrate that SVD-MBR successfully regularizes decoding, yielding substantial gains across a range of generalized metrics. Furthermore, we reveal that this denoising is metric-dependent: neural metrics encode a robust low-rank consensus ideal for SVD, whereas surface-level metrics struggle to separate signal from metric noise.

cs.CL↗

arXiv2026

Attention-Guided Layer Selection for Contrastive Decoding in Large Language Models

Contrastive decoding methods such as DoLa improve the factuality of Large Language Models (LLMs) by contrasting the output distributions of mature and premature layers. However, DoLa's dynamic layer selection relies solely on divergences in output vocabulary distributions. In this work, we propose three attention-guided strategies: Attention-JSD, Attention-Entropy-Max, and Attention-Entropy-Min, which leverage structural information carried by internal self-attention mechanisms as a signal for layer selection. Experimental results on TruthfulQA demonstrate that our strategies, particularly Attention-JSD and Attention-Entropy-Min, consistently outperform the original DoLa. We observe significant gains on multi-answer metrics (MC2 and MC3), suggesting that attention distributions can provide a more sensitive signal for resolving factual knowledge than output vocabulary distributions.

cs.CL↗

arXiv2026

Chandra X-Ray Imaging and Spatially Resolved Spectroscopy of SN 1987A: Energy-Dependent Morphology of the Equatorial Ring

We present a systematic imaging and spatially resolved spectral study of SN 1987A using Chandra observations obtained between 1999 and 2025. By combining multiepoch ACIS and HETG data, we investigate the long-term evolution of the remnant in both the soft and hard X-ray bands. To characterize the radial structure, we model the projected emission with a torus profile and derive its radius and width on the image plane. We find an energy dependence in the ring morphology: while the soft and hard bands exhibit similar structures at early epochs, the soft-band emission becomes systematically broader than the hard-band emission after the early 2010s. Furthermore, when considering the radius and width together, the soft-band emission shows an inward extension, suggesting an increasing contribution from interior and/or high-latitude emission components. The flux evolution of the Fe K line is consistent with previous XMM-Newton results, and we detect its presence already in earlier epochs (~2007-2009) using combined Chandra spectra. Spatially resolved analysis further indicates that the Fe K emission is enhanced in the eastern region. These results provide a unified view of the long-term morphological and spectral evolution of SN 1987A and highlight the emergence of energy-dependent radial structure as a key feature in its late-time evolution.

astro-ph.HE↗

arXiv2026

Noisy-Channel Minimum Bayes Risk Decoding

Minimum Bayes Risk (MBR) decoding yields more robust and higher-quality text generation than maximum a posteriori (MAP) decoding by selecting hypotheses that maximize expected utility over sampled pseudo-references. However, there exists a discrepancy in the design: hypothesis selection calculates expected utility scores conditioned on given pseudo-references, while commonly used evaluation metrics, e.g., BLEU and COMET, are asymmetric. Therefore, it is important to consider both hypothesis-to-reference and reference-to-hypothesis directional effects. In this study, we introduce a noisy channel decomposition of MBR decoding that naturally incorporates bidirectional effects to account for these asymmetries. We decompose MBR decoding into four interacting components: hypothesis-to-reference likelihood, reference-to-hypothesis likelihood, hypothesis prior, and reference prior. This decomposition provides a unified interpretation of existing MBR variants and enables metric- and task-specific interpretability by isolating the contribution of each channel. Our comprehensive analysis reveals that channel-wise contributions exhibit distinct characteristics across metrics while remaining consistent across tasks, and suggests that appropriate channel weighting may lead to improvements over original MBR decoding.

cs.LG↗

arXiv2026

Closed-Form Steepest Descent Direction toward Flat Minima: Reducing Upper Bounds on the Loss Hessian Eigenspectrum in Neural Networks

The flatness hypothesis suggests that flatness of the loss landscape, as measured by the eigenvalues of the loss Hessian, correlates with better neural network generalization. While various algorithms reduce these eigenvalues, most focus on procedural design, leaving it unclear how data distributions and NN parameters structurally determine directions toward flat minima. Characterizing these directions analytically is generally intractable. To overcome this mathematical difficulty, recent studies derived the Wolkowicz-Styan (WS) upper bound on the maximum eigenvalue of the cross-entropy loss Hessian in three-layer NNs. Although this upper bound is differentiable, its gradient was not derived. Therefore, we analytically derive the gradient of the WS upper bound to characterize directions leading to flat minima. Based on this, we propose Hessian Spectral Range (HSR) Regularization, which updates parameters along the steepest descent direction of the WS bound. Experiments demonstrate that HSR Regularization narrows the Hessian eigenvalue spectrum, avoids sharp minima and saddle points, and promotes convergence to flat minima. Although the applicability of this method is currently limited to cross-entropy loss and three-layer architectures, to the best of the authors' knowledge, this is the first study to report a closed-form gradient that promotes convergence to flat minima without numerical approximations. Therefore, the theoretical analysis of this gradient is expected to contribute to the further development of NNs.

cs.LG↗

arXiv2026

Wavelet-Based Extraction of Transient Noise in Gravitational-Wave Interferometers using a Saliency-Guided Learning Architecture

Gravitational-wave interferometers exhibit a wide variety of short-duration non-Gaussian transients, commonly referred to as glitches, that complicate the detection of astrophysical signals, bias parameter estimation, and detector characterisation. Existing machine-learning approaches classify glitch morphologies but do not provide a complete mechanism to segment and extract these disturbances from the strain data. We introduce a wavelet-based, saliency-guided framework for the supervised extraction of transient noise. Candidates are first pre-tagged using Uniform Manifold Approximation and Projection, which is also used as a diagnostic of the learned representations. A traditional learning model operating on Continuous Wavelet Transform spectrograms then identifies relevant time-frequency regions through saliency maps. These saliency patterns are transferred to an invertible multiresolution representation via the Discrete Wavelet Transform, where adaptive coefficient masking enables exact reconstruction of both glitch-only and glitch-suppressed waveforms. We demonstrate effective extraction across several representative glitch families, including 'Whistle' and 'Scattered-Light' transients, and show robustness in challenging regimes such as low signal-to-noise events and partially overlapping structures, where classical thresholding or band-limited filtering methods typically fail or introduce leakage. The proposed framework offers an interpretable and computationally efficient approach to transient-noise extraction, establishing a foundation for scalable applications to larger glitch catalogs and future observing runs.

astro-ph.IM↗

arXiv2026

Multilinguality of Large Language Models From a Structural Perspective

Large language models (LLMs) have excelled in processing multiple languages through pre- and post-training on multilingual data, even though English dominates the training data. Prior work focusing on token representations has revealed how those LLMs process non-English text. Although these analyses have provided insightful findings, they fail to capture a structural view, which is an inherent property of language. In this study, we explore the multilinguality of LLMs through representational structural analysis. Our findings reveal that low-resource languages are structurally more different from English than high- and mid-resource languages, and that language-specific post-training alters their structures while preserving inter-language relationships.

cs.CL↗

arXiv2026

Enhancing Factuality through Consensus and Consistency in Summarization Using Minimum Bayes Risk Decoding

Improving the quality of model-generated summaries, especially factuality, the accuracy of a summary with respect to its source content, remains a challenge. While reranking could select the optimal output from multiple generated candidates, it is limited to only using the source as guidance, resulting in unreliable summaries. To address this limitation, we propose ConSUM that reranks candidate summaries by considering two factors: consistency to the source document and consensus among the other candidates. Consensus is established using Minimum Bayes Risk (MBR) decoding over the set of generated summaries, while ensuring consistency by employing factuality-aware metrics that compare the summary against the source. Rigorous testing demonstrates that our system is competitive with existing methods, with human evaluations further confirming that its generated summaries are preferred over those from other systems. Our code is available at https://github.com/naist-nlp/ConSUM .

cs.CL↗

arXiv2026

Edit-level Majority Voting Mitigates Over-Correction in LLM-based Grammatical Error Correction

Grammatical error correction using large language models often suffers from the over-correction issue. To mitigate this, we propose a training-free inference method that performs edit-level majority voting over multiple candidates generated by a single model, without requiring model modifications or additional training. Across nine benchmarks covering English, Czech, German, Ukrainian, Korean, Hindi, and Romanian, the proposed method outperforms both greedy and MBR decoding in most cases. Moreover, it yields stable correction quality regardless of the instruction prompts used. We release two repository supporting GEC datasets loading and LLM inference.

cs.CL↗

arXiv2026

One Single Hub Text Breaks CLIP: Identifying Vulnerabilities in Cross-Modal Encoders via Hubness

The hubness problem, in which hub embeddings are close to many unrelated examples, occurs often in high-dimensional embedding spaces and may pose a practical threat for purposes such as information retrieval and automatic evaluation metrics. In particular, since cross-modal similarity between text and images cannot be calculated by direct comparisons, such as string matching, cross-modal encoders that project different modalities into a shared space are helpful for various cross-modal applications, and thus, the existence of hubs may pose practical threats. To reveal the vulnerabilities of cross-modal encoders, we propose a method for identifying the hub embedding and its corresponding hub text. Experiments on image captioning evaluation in MSCOCO and nocaps along with image-to-text retrieval tasks in MSCOCO and Flickr30k showed that our method can identify a single hub text that unreasonably achieves comparable or higher similarity scores than human-written reference captions in many images, thereby revealing the vulnerabilities in cross-modal encoders.

cs.CL↗

arXiv2026

HalluCiteChecker: A Lightweight Toolkit for Hallucinated Citation Detection and Verification in the Era of AI Scientists

We introduce HalluCiteChecker, a toolkit for detecting and verifying hallucinated citations in scientific papers. While AI assistant technologies have transformed the academic writing process, including citation recommendation, they have also led to the emergence of hallucinated citations that do not correspond to any existing work. Such citations not only undermine the credibility of scientific papers but also impose an additional burden on reviewers and authors, who must manually verify their validity during the review process. In this study, we formalize hallucinated citation detection as an NLP task and provide a corresponding toolkit as a practical foundation for addressing this problem. Our package is lightweight and can perform verification in seconds on a standard laptop. It can also be executed entirely offline and runs efficiently using only CPUs. We hope that HalluCiteChecker will help reduce reviewer workload and support organizers by enabling systematic pre-review and publication checks. Our code is released under the Apache 2.0 license on GitHub and is distributed as an installable package via PyPI. A demonstration video is available on YouTube.

cs.CL↗

arXiv2026

CArtBench: Evaluating Vision-Language Models on Chinese Art Understanding, Interpretation, and Authenticity

We introduce CARTBENCH, a museum-grounded benchmark for evaluating vision-language models (VLMs) on Chinese artworks beyond short-form recognition and QA. CARTBENCH comprises four subtasks: CURATORQA for evidence-grounded recognition and reasoning, CATALOGCAPTION for structured four-section expert-style appreciation, REINTERPRET for defensible reinterpretation with expert ratings, and CONNOISSEURPAIRS for diagnostic authenticity discrimination under visually similar confounds. CARTBENCH is built by aligning image-bearing Palace Museum objects from Wikidata with authoritative catalog pages, spanning five art categories across multiple dynasties. Across nine representative VLMs, we find that high overall CURATORQA accuracy can mask sharp drops on hard evidence linking and style-to-period inference; long-form appreciation remains far from expert references; and authenticity-oriented diagnostic discrimination stays near chance, underscoring the difficulty of connoisseur-level reasoning for current models.

cs.CL↗

arXiv2026

Wolkowicz-Styan Upper Bound on the Hessian Eigenspectrum for Cross-Entropy Loss in Nonlinear Smooth Neural Networks

Neural networks (NNs) are central to modern machine learning and achieve state-of-the-art results in many applications. However, the relationship between loss geometry and generalization is still not well understood. The local geometry of the loss function near a critical point is well-approximated by its quadratic form, obtained through a second-order Taylor expansion. The coefficients of the quadratic term correspond to the Hessian matrix, whose eigenspectrum allows us to evaluate the sharpness of the loss at the critical point. Extensive research suggests flat critical points generalize better, while sharp ones lead to higher generalization error. However, sharpness requires the Hessian eigenspectrum, but general matrix characteristic equations have no closed-form solution. Therefore, most existing studies on evaluating loss sharpness rely on numerical approximation methods. Existing closed-form analyses of the eigenspectrum are primarily limited to simplified architectures, such as linear or ReLU-activated networks; consequently, theoretical analysis of smooth nonlinear multilayer neural networks remains limited. Against this background, this study focuses on nonlinear, smooth multilayer neural networks and derives a closed-form upper bound for the maximum eigenvalue of the Hessian with respect to the cross-entropy loss by leveraging the Wolkowicz-Styan bound. Specifically, the derived upper bound is expressed as a function of the affine transformation parameters, hidden layer dimensions, and the degree of orthogonality among the training samples. The primary contribution of this paper is an analytical characterization of loss sharpness in smooth nonlinear multilayer neural networks via a closed-form expression, avoiding explicit numerical eigenspectrum computation. We hope that this work provides a small yet meaningful step toward unraveling the mysteries of deep learning.

cs.LG↗

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

Formation of first star clusters under the supersonic gas flow -- III. Environmental trends and halo-to-halo scatter in the Pop III IMF

The first generations of stars ionised and enriched their host galaxies and seeded the growth of massive black holes. Models often assume that Pop III stellar masses in different minihaloes are stochastic realisations of a single universal initial mass function (IMF). We use 138 cosmological zoom-in hydrodynamics simulations to test this assumption and to map the first-star IMF across redshift, halo mass, and baryon-dark matter streaming velocity (SV). We construct a dense-cloud merger tree and assign first-star masses by mapping the radial gas accretion-rate profile to stellar mass, yielding per-halo stellar mass functions without imposing any a priori IMF. The high-mass tail and multiplicity increase systematically with redshift, halo mass, and SV. Low-mass, low-SV haloes form only one or a few first stars, whereas massive, high-SV haloes host rich first star clusters and commonly produce very massive ($\gtrsim10^3$-$10^4\,{\rm M}_\odot$) first stars. Even in a fixed environment, halo-to-halo scatter remains substantial. Nevertheless, group-averaged IMFs converge to well-defined forms, ruling out a single universal IMF at the halo level across the range of environments probed here. Mapping our seeds onto the redshift-mass plane, we show that high-SV and massive haloes preferentially populate the heavy-seed regime relevant to luminous high-redshift sources. At the same time, low-SV environments are consistent with single/few-event enrichment signatures in metal-poor stars. Our results deliver a practical, physically motivated prescription for per-halo IMF.

astro-ph.GA↗

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

StructLens: A Structural Lens for Language Models via Maximum Spanning Trees

Language exhibits inherent structures, a property that explains both language acquisition and language change. Given this characteristic, we expect language models to manifest their own internal structures as well. While interpretability research has investigated how models compute representations mechanistically through attention patterns and Sparse AutoEncoders, the organization of the resulting representations is overlooked. To address this gap, we introduce StructLens, a framework to analyze representations through a holistic structural view. StructLens constructs maximum spanning trees based on the semantic representations in residual streams, inspired by tree representation in dependency parsing, and provides summaries of token relationships in representation space. We analyze how contiguous tokens are also nearby in representation space and find that middle layers show the strongest local-span organization. Moreover, analysis of pre-training checkpoints reveals that smaller local units become detectable earlier in pre-training, and larger units later. Our findings demonstrate that StructLens provides insights into how models organize token representations across layers and training. Our code is available at https://github.com/naist-nlp/structlens.

cs.CL↗