arXiv subjects

Ichiro Sakata

Publications and source records attributed to Ichiro Sakata.

18 recordsLinked to original sources

arXiv2026

Preprints Without Curation Are Increasingly Cited by Journals

The recent strain on peer review systems necessitates the publish-review-curate model, where the public dissemination of manuscripts is decoupled from formal peer review and later editorial processes. Citation norms traditionally pointed to the curated version; those that subsequently appear in peer-reviewed journals. Citations to preprints are therefore generally expected to occur only to those eventually curated. Using metadata from six major preprint servers linked to the world's largest bibliographic reference database, and focusing on citations from long-established journals with rigorous editorial workflows, we compare observed citation rates with a null model reflecting the expected rate if authors were indifferent between preprints and journal articles. The growth is evident in both citation intensity and the breadth of adoption: at its peak in 2021, nearly one in five journal articles cited at least one preprint. The preference to cite preprints has grown exponentially since the mid-2010s, with key inflection points coinciding with the launch of bioRxiv (2013) and the introduction of a preprint content type in Crossref's DOI system (2016). An increasing share of these citations refer to non-curated preprints and this shift cannot be explained by confounding factors: the growing stock of preprints, faster publishing cycles, or the outsized influence of a small set of highly cited preprints. These findings reveal a structural shift in citing behavior: preprints are increasingly treated as legitimate, autonomous objects of citation, independent of their later publication status, motivating renewed attention to how citation conventions and journal policies should accommodate the expanding role of preprints in scholarly communication. The COVID-19 pandemic did not amplify this trend but instead coincided with a deceleration in preprint citation growth.

cs.DL↗

arXiv2026

Large Language Models Exhibit Normative Conformity

The conformity bias exhibited by large language models (LLMs) can pose a significant challenge to decision-making in LLM-based multi-agent systems (LLM-MAS). While many prior studies have treated "conformity" simply as a matter of opinion change, this study introduces the social psychological distinction between informational conformity and normative conformity in order to understand LLM conformity at the mechanism level. Specifically, we design new tasks to distinguish between informational conformity, in which participants in a discussion are motivated to make accurate judgments, and normative conformity, in which participants are motivated to avoid conflict or gain acceptance within a group. We then conduct experiments based on these task settings. The experimental results show that, among the six LLMs evaluated, up to five exhibited tendencies toward not only informational conformity but also normative conformity. Furthermore, intriguingly, we demonstrate that by manipulating subtle aspects of the social context, it may be possible to control the target toward which a particular LLM directs its normative conformity. These findings suggest that decision-making in LLM-MAS may be vulnerable to manipulation by a small number of malicious users. In addition, through analysis of internal vectors associated with informational and normative conformity, we suggest that although both behaviors appear externally as the same form of "conformity," they may in fact be driven by distinct internal mechanisms. Taken together, these results may serve as an initial milestone toward understanding how "norms" are implemented in LLMs and how they influence group dynamics.

cs.AI↗

arXiv2026

Position: No Retroactive Cure for Infringement during Training

As generative AI faces intensifying legal challenges, the machine learning community has increasingly relied on post-hoc mitigation -- especially machine unlearning and inference-time guardrails -- to argue for compliance. This paper argues that such post-hoc mitigation methods cannot retroactively cure liability from unlawful acquisition and training, because compliance hinges on data lineage, not the outputs. Our argument has three parts. First, unauthorized copying/ingestion can be a legally complete completed act, and model weights may operate as fixed copies that retain training-derived expressive value, making later filtering beside the point for infringement. Second, contract and tort/unfair-competition rules -- via licenses, terms of service, and anti-free-riding principles -- can independently restrict access and use, often bypassing copyright defenses (e.g., fair use or TDM exceptions). Third, since value from protected inputs can persist in weights, remedies such as unjust enrichment and disgorgement may require stripping gains and, in some cases, reaching the model itself. We therefore argue for a shift from Post-Hoc Sanitization to verifiable Ex-Ante Process Compliance.

cs.CR↗

arXiv2026

Cooperation Breakdown in LLM Agents Under Communication Delays

LLM-based multi-agent systems (LLM-MAS), in which autonomous AI agents cooperate to solve tasks, are gaining increasing attention. For such systems to be deployed in society, agents must be able to establish cooperation and coordination under real-world computational and communication constraints. We propose the FLCOA framework (Five Layers for Cooperation/Coordination among Autonomous Agents) to conceptualize how cooperation and coordination emerge in groups of autonomous agents, and highlight that the influence of lower-layer factors - especially computational and communication resources - has been largely overlooked. To examine the effect of communication delay, we introduce a Continuous Prisoner's Dilemma with Communication Delay and conduct simulations with LLM-based agents. As delay increases, agents begin to exploit slower responses even without explicit instructions. Interestingly, excessive delay reduces cycles of exploitation, yielding a U-shaped relationship between delay magnitude and mutual cooperation. These results suggest that fostering cooperation requires attention not only to high-level institutional design but also to lower-layer factors such as communication delay and resource allocation, pointing to new directions for MAS research.

cs.MA↗

arXiv2025

Unveiling the Listener Structure Underlying K-pop's Global Success: A Large-Scale Listening Data Analysis

From the mid-2000s to the 2010s, K-pop moved beyond its status as a regionally popular genre in Asia and established itself as a global music genre with enthusiastic fans around the world. However, little is known about how the vast number of music listeners across the globe have listened to and perceived K-pop. This study addresses this question by analyzing a large-scale listening dataset from Last.fm. An analysis of the distribution of play counts reveals that K-pop experienced a significant increase in plays between 2005 and 2019, largely supported by a small group of heavy listeners. The Gini coefficient in play counts is notably greater than that of existing mainstream genres and other growing niche genres. Furthermore, an analysis based on user-assigned genre tags quantitatively demonstrates that between 2005 and 2010, K-pop shed its status as a local Asian genre and established itself as a distinct music genre in its own right.

cs.SI↗

arXiv2025

Asymmetric Impact of Basic Scientists during Applied Shift

Despite broad acclaim for basic research, science is undergoing an applied shift that marginalizes basic scientists. This gap reflects an incomplete understanding of their distinctive roles, which prevents translating philosophical appreciation into effective support. We introduce a scalable metric--the application score--to position research along the basic-applied spectrum and apply it to 62 million publications (1970-2023) to reveal the distinctive contributions of basic scientists. We find a structural asymmetry: involvement of basic scientists substantially increases citation impact, even more so in applied contexts, while applied scientists show no such effect in basic domains. This asymmetric effect arises from their intellectual leadership in conceptualization, writing, and experimental design, amplified in large, multidisciplinary, and intermediate career teams. Yet basic scientists remain concentrated in historically prestigious institutions, while new entrants shift toward applied work, indicating critical undersupply. These findings provide large-scale evidence for the indispensable role of basic scientists, guiding policy and institutional strategy to sustain the foundations of discovery and innovation.

cs.DL↗

arXiv2025

Advancement of Circular Economy Through Interdisciplinary Collaboration: A Bibliometric Approach

Since the European Union introduced its Circular Economy (CE) Action Plan in 2015, CE research has expanded rapidly. However, the structure of this emerging field - both in terms of its constituent disciplines and researcher dynamics - remains poorly understood. To address this gap, we analyze over 25,000 CE-related publications from Scopus by combining conventional bibliometric approaches with advanced machine learning techniques, including text embeddings and clustering. This hybrid method enables both a macro-level mapping of research domains and a micro-level investigation of individual researchers' disciplinary backgrounds and collaborations. We classify CE research into 16 distinct clusters, identifying the original disciplines of researchers and visualizing patterns of interdisciplinary collaboration. Building on this foundation, we ask: Which CE-related research domains receive the most attention in academic and policy contexts? And how are different types of interdisciplinary collaboration associated with research impact? Our findings show that research in business and management attracts substantial academic and policy attention, while engineering research - though less visible - tends to achieve higher funding success. This suggests a positive dynamic in which the former draws attention to CE issues and the latter secures the economic resources necessary to realize them. We further demonstrate that CE papers co-authored by researchers from different disciplines tend to show higher research impact than intradisciplinary work. Qualitative case analyses also highlight this tendency. Centered particularly on collaborations between business-oriented and engineering-oriented disciplines, our findings underscore the importance of interdisciplinary efforts in CE research and offer insights for guiding future cross-disciplinary engagement in the field.

cs.SI↗

arXiv2025

From Rapid Release to Reinforced Elite: Citation Inequality Is Stronger in Preprints than Journals

Preprints have been considered primarily as a supplement to journal-based systems for the rapid dissemination of relevant scientific knowledge and have historically been supported by studies indicating that preprints and published reports have comparable authorship, references, and quality. However, as preprints increasingly serve as an independent medium for scholarly communication rather than precursors to the version of record, it remains uncertain how preprint usage is shaping scientific discourse. Our research revealed that the preprint citations exhibit significantly higher inequality than journal citations, consistently among categories. This trend persisted even when controlling for age and the mean citation count of the journal matched to each of the preprint categories. We also found that the citation inequality in preprints is not solely driven by a few highly cited papers or those with no impact, but rather reflects a broader systemic effect. Whether the preprint is subsequently published in a journal or not does not significantly affect the citation inequality. Further analyses of the structural factors show that preferential attachment does not significantly contribute to citation inequality in preprints, whereas author prestige plays a substantial role. Notably, the gap in citation inequality between the preprint category and the journal is more pronounced in fields where preprints are more established, such as mathematics, physics, and high-energy physics. This highlights a potential vulnerability in preprint ecosystems where reputation-driven citation may hinder scientific diversity.

cs.DL↗

arXiv2025

Influential scientists shape knowledge flows between science and IGO policy

Intergovernmental organizations (IGOs) increasingly rely on scientific evidence, yet the pathways through which scientific research enters policy remain opaque. By linking 230,737 scientific papers cited in IGO policy documents (2015-2023) to their authors and collaboration networks, we identify a small group of policy-influential scientists (PI-Sci) who dominate this knowledge flow. These scientists form tightly interconnected, internationally spanning co-authorship networks and achieve policy citations shortly after publication, a distinctive feature of cumulative advantage at the science-policy interface. The concentration of influence varies by field: tightly clustered in established domains like climate modeling, and more dispersed in emerging areas like AI governance. Many PI-Sci serve on high-level advisory bodies (e.g., IPCC), and major IGOs frequently co-cite the same PI-Sci papers, indicating synchronized knowledge diffusion through shared expert networks. These findings reveal how network structure and elite brokerage shape the translation of research into global policy, highlighting opportunities to broaden the scope of knowledge that informs policy.

cs.DL↗

arXiv2025

UniDetox: Universal Detoxification of Large Language Models via Dataset Distillation

We present UniDetox, a universally applicable method designed to mitigate toxicity across various large language models (LLMs). Previous detoxification methods are typically model-specific, addressing only individual models or model families, and require careful hyperparameter tuning due to the trade-off between detoxification efficacy and language modeling performance. In contrast, UniDetox provides a detoxification technique that can be universally applied to a wide range of LLMs without the need for separate model-specific tuning. Specifically, we propose a novel and efficient dataset distillation technique for detoxification using contrastive decoding. This approach distills detoxifying representations in the form of synthetic text data, enabling universal detoxification of any LLM through fine-tuning with the distilled text. Our experiments demonstrate that the detoxifying text distilled from GPT-2 can effectively detoxify larger models, including OPT, Falcon, and LLaMA-2. Furthermore, UniDetox eliminates the need for separate hyperparameter tuning for each model, as a single hyperparameter configuration can be seamlessly applied across different models. Additionally, analysis of the detoxifying text reveals a reduction in politically biased content, providing insights into the attributes necessary for effective detoxification of LLMs.

cs.CL↗

arXiv2025

Do Researchers Benefit Career-wise from Involvement in International Policy Guideline Development?

Researchers are no longer limited to producing knowledge; in today's complex world, they also address societal challenges by engaging in policymaking. Although involvement in policymaking has expanded, direct empirical evidence of its career benefits remains underexplored. Prior survey-based studies suggest potential advantages-such as broader professional networks and enhanced opportunities-yet raise concerns about insufficient institutional support. Here, we examine the 2021 WHO global air quality guideline-a science-based regulatory guideline-as a case study. To evaluate the impact of guideline development on research outcomes, we match guideline researchers with a control group of peers sharing similar research topics and prior performance. Our analysis reveals that guideline researchers attain higher future citation counts in both academic and policy domains. New collaborations formed during development yield publications with higher citation impact and the disruptive index. Moreover, about half the guideline's references are derived from guideline researchers' papers, highlighting their central role in shaping the evidence base. These results provide empirical support for the career benefits of policy engagement. Our findings indicate that engaging in international guideline development offers tangible career incentives for researchers, and that institutions can enhance research impact and promote innovative scientific progress by actively supporting their researchers' participation in such initiatives.

cs.SI↗

arXiv2024

Towards Transfer Unlearning: Empirical Evidence of Cross-Domain Bias Mitigation

Large language models (LLMs) often inherit biases from vast amounts of training corpora. Traditional debiasing methods, while effective to some extent, do not completely eliminate memorized biases and toxicity in LLMs. In this paper, we study an unlearning-based approach to debiasing in LLMs by performing gradient ascent on hate speech against minority groups, i.e., minimizing the likelihood of biased or toxic content. Specifically, we propose a mask language modeling unlearning technique, which unlearns the harmful part of the text. This method enables LLMs to selectively forget and disassociate from biased and harmful content. Experimental results demonstrate the effectiveness of our approach in diminishing bias while maintaining the language modeling abilities. Surprisingly, the results also unveil an unexpected potential for cross-domain transfer unlearning: debiasing in one bias form (e.g. gender) may contribute to mitigating others (e.g. race and religion).

cs.CL↗

arXiv2024

Mid-career pitfall of consecutive success in science

The creativity of scientists often manifests as localized hot streaks of significant success. Understanding the underlying mechanisms of these influential phases can enhance the effectiveness of support systems and funding allocation, fostering groundbreaking discoveries worthy of accolades. Historically, analyses have suggested that hot streaks occur randomly over time. However, our research, through meticulous examination, reveals that these phases are not flatly distributed but are more frequent at the early and late stages of scientists' careers. Notably, both early and late hot streaks are marked by dense tie collaborations, with the former typically involving close partnerships with particular authors and the latter being characterized by involvement in large-scale projects compared with single-top or ordinary papers. This pattern indicates that mid-career researchers lack both intimate relations and resources to keep big projects, leading to``mid-career pitfal'' of consecutive success. This insight holds profound implications for the development of policies and initiatives aimed at bolstering innovative research and discovery.

cs.SI↗

arXiv2023

Differentiable Instruction Optimization for Cross-Task Generalization

Instruction tuning has been attracting much attention to achieve generalization ability across a wide variety of tasks. Although various types of instructions have been manually created for instruction tuning, it is still unclear what kind of instruction is optimal to obtain cross-task generalization ability. This work presents instruction optimization, which optimizes training instructions with respect to generalization ability. Rather than manually tuning instructions, we introduce learnable instructions and optimize them with gradient descent by leveraging bilevel optimization. Experimental results show that the learned instruction enhances the diversity of instructions and improves the generalization ability compared to using only manually created instructions.

cs.CL↗

arXiv2023

SciReviewGen: A Large-scale Dataset for Automatic Literature Review Generation

Automatic literature review generation is one of the most challenging tasks in natural language processing. Although large language models have tackled literature review generation, the absence of large-scale datasets has been a stumbling block to the progress. We release SciReviewGen, consisting of over 10,000 literature reviews and 690,000 papers cited in the reviews. Based on the dataset, we evaluate recent transformer-based summarization models on the literature review generation task, including Fusion-in-Decoder extended for literature review generation. Human evaluation results show that some machine-generated summaries are comparable to human-written reviews, while revealing the challenges of automatic literature review generation such as hallucinations and a lack of detailed information. Our dataset and code are available at https://github.com/tetsu9923/SciReviewGen.

cs.CL↗

arXiv2022

Storyteller: The papers co-citing Sleeping Beauty and Prince before awakening

In the Cumulative Advantage(CA) model, which is one of the most fundamental approaches to understand the mechanism of citation dynamics, papers receive citations depending on how much they have been already cited. On the other hand, a substantial effect not included in CA is that some surprising discoveries suddenly acquire citations after a long time from publishing. This phenomenon is known as Sleeping Beauty(SB). Since disrupting discoveries need long-time discussion by the research community to accept, SBs can capture innovative findings and reveal the nature of disruptive scientific knowledge production. To research SBs citation burst mechanism, bibliometricians consider the existence of the Prince(PR) for each SBs, which can be the trigger of SBs awakeness. For example, the discovery of Green Fluorescent Protein(GFP), which got Nobel prize in chemistry, had been overlooked for 30 years until Chalfie and Tsien, who also received the prize, developed a method to use GFP as a marker protein in genetic engineering. However, how does Chalfies and Tsiens research relight the hidden knowledge in the research community? If we can clarify such a mechanism rediscovering from nearly nothing, it can be helpful in science support and policy decision-making. This study proposes a Storyteller that focuses on the connection between SB and PR before SB gets citation burst by co-citation. PR is found to be the paper awakening SB in retrospect, but it is not easy to detect it as the trigger of SBs awakeness at the time of PR submission. We named the papers which co-cites SB and PR before the citation burst of SB as Storyteller(ST) and analyze (1) how ST contributes to broadening the novelty of SB&PR connections and (2) how much ST leads the citation burst after awakening.

cs.DL↗

arXiv2021

Unsupervised Abstractive Opinion Summarization by Generating Sentences with Tree-Structured Topic Guidance

This paper presents a novel unsupervised abstractive summarization method for opinionated texts. While the basic variational autoencoder-based models assume a unimodal Gaussian prior for the latent code of sentences, we alternate it with a recursive Gaussian mixture, where each mixture component corresponds to the latent code of a topic sentence and is mixed by a tree-structured topic distribution. By decoding each Gaussian component, we generate sentences with tree-structured topic guidance, where the root sentence conveys generic content, and the leaf sentences describe specific topics. Experimental results demonstrate that the generated topic sentences are appropriate as a summary of opinionated texts, which are more informative and cover more input contents than those generated by the recent unsupervised summarization model (Bražinskas et al., 2020). Furthermore, we demonstrate that the variance of latent Gaussians represents the granularity of sentences, analogous to Gaussian word embedding (Vilnis and McCallum, 2015).

cs.CL↗

arXiv2019

Unsupervised Neural Single-Document Summarization of Reviews via Learning Latent Discourse Structure and its Ranking

This paper focuses on the end-to-end abstractive summarization of a single product review without supervision. We assume that a review can be described as a discourse tree, in which the summary is the root, and the child sentences explain their parent in detail. By recursively estimating a parent from its children, our model learns the latent discourse tree without an external parser and generates a concise summary. We also introduce an architecture that ranks the importance of each sentence on the tree to support summary generation focusing on the main review point. The experimental results demonstrate that our model is competitive with or outperforms other unsupervised approaches. In particular, for relatively long reviews, it achieves a competitive or better performance than supervised models. The induced tree shows that the child sentences provide additional information about their parent, and the generated summary abstracts the entire review.

cs.CL↗