Seungyoon Lee, Minhyuk Kim, Seongtae Hong, Youngjoon Jang, Dongsuk Oh, Heuiseok Lim. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Large Language Models (LLMs) have rapidly advanced, with domain-specific expert models emerging to handle specialized tasks across various fields. However, the predominant focus on English-centric models demands extensive data, making it challenging to develop comparable models for middle and low-resource languages. To address this limitation, we introduce Migrate, a novel method that leverages open-source static embedding models and up to 3 million tokens of code-switching data to facilitate the seamless transfer of embeddings to target languages. Migrate enables effective cross-lingual adaptation without requiring large-scale domain-specific corpora in the target language, promoting the accessibility of expert LLMs to a diverse range of linguistic communities. Our experimental results demonstrate that Migrate significantly enhances model performance in target languages, outperforming baseline and existing cross-lingual transfer methods. This approach provides a practical and efficient solution for extending the capabilities of domain-specific expert models.

The approaches that guide Large Language Models (LLMs) to emulate human reasoning during response generation have emerged as an effective method for enabling them to solve complex problems in a step-by-step manner, thereby achieving superior performance.However, most existing approaches using few-shot prompts to generate responses heavily depend on the provided examples, limiting the utilization of the model's inherent reasoning capabilities.Moreover, constructing task-specific few-shot prompts is often costly and may lead to inconsistencies across different tasks.In this work, we introduce Template-Oriented Reasoning (TORSO), which elicits the model to utilize internal reasoning abilities to generate proper responses across various tasks without the need for manually crafted few-shot examples.Our experimental results demonstrate that TORSO achieves strong performance on diverse LLMs benchmarks with reasonable rationales.

Counter-narrative generation, i.e., the generation of fact-based responses to hate speech with the aim of correcting discriminatory beliefs, has been demonstrated to be an effective method to combat hate speech. However, its effectiveness is limited by the resource-intensive nature of dataset construction processes and only focuses on the primary language. To alleviate this problem, we propose a Korean Hate Speech Counter Punch (KHSCP), a cost-effective counter-narrative generation method in the Korean language. To this end, we release the first counter-narrative generation dataset in Korean and pose two research questions. Under the questions, we propose an effective augmentation method and investigate the reasonability of a large language model to overcome data scarcity in low-resource environments by leveraging existing resources. In this regard, we conduct several experiments to verify the effectiveness of the proposed method. Our results reveal that applying pre-existing resources can improve the generation performance by a significant margin. Through deep analysis on these experiments, this work proposes the possibility of overcoming the challenges of generating counter-narratives in low-resource environments.

This work-in-progress research paper investigated where and how international students seek career support and perceived barriers. Existing literature has examined career readiness and preparation, particularly focused on international students' diaspora patterns and cultural adjustment, but lacks holistic insights into their career decision-forming and support systems from a mixed-method social network perspective. We employed a mixed-method research approach and leveraged data collected from the international undergraduate student body at a U.S. Midwestern university via interviews and surveys sequentially. This WIP paper primarily focused on the qualitative section of this project based on the data from the nine interview transcripts. The data were analyzed by both egocentric social network and thematic analysis. Our preliminary findings identified the most common source of support for international students. We also recognized barriers existing in multiple forms and dimensions and a lack of support in three major areas for international students, particularly access, human connection, and clarity. Our unique contribution in this study highlighted the interconnection among students' perceived (or lack of) career support and barriers from a social network perspective. Based on our preliminary findings, we offer suggestions to inform practices and policy change congruent to the actual needs of international students from multiple stakeholders, including, but not limited to, home institutions, faculty, and staff members.

Abstract Understanding how information use contributes to uncertainties surrounding evacuation decisions is crucial during disasters. While literature increasingly establishes that people consult multiple information sources in disaster situations, little is known about the patterns in which multiple media and personal network sources are combined simultaneously and sequentially across decision‐making phases. We address this gap using survey data collected from households in Jacksonville, Florida affected by 2016's Hurricane Matthew. Results direct attention to perceived consistency of information as a key predictor of uncertainty regarding hurricane impact and evacuation logistics. Frequently utilizing National Weather Service, national and local TV channels, and personal network contacts contributed to higher perceived consistency of information, while the use of other local and online sources was associated with lower perceived consistency. Furthermore, combining a larger number of media and official sources predicted higher levels of perceived information consistency. One's perception of information amount did not significantly explain uncertainty. This study contributes to the theorizing of individuals' information environment from the perspective of media and network multiplexity and provides practical implications regarding the need of information coordination for improved evacuation decision‐making.

Social network analysis (SNA) is becoming increasingly concerned not only with actors and their relations, but also with distinguishing between different types of such entities. For example, social scientists may want to investigate asymmetric relations in organizations with strict chains of command, or incorporate non-actors such as conferences and projects when analyzing coauthorship patterns. Multimodal social networks are those where actors and relations belong to different types, or modes, and multimodal social network analysis (mSNA) is accordingly SNA for such networks. In this paper, we present a design study that we conducted with several social scientist collaborators on how to support mSNA using visual analytics tools. Based on an openended, formative design process, we devised a visual representation called parallel node-link bands (PNLBs) that splits modes into separate bands and renders connections between adjacent ones, similar to the list view in Jigsaw. We then used the tool in a qualitative evaluation involving five social scientists whose feedback informed a second design phase that incorporated additional network metrics. Finally, we conducted a second qualitative evaluation with our social scientist collaborators that provided further insights on the utility of the PNLBs representation and the potential of visual analytics for mSNA.

Visualization plays a crucial role in understanding dynamic social networks at many different levels (i.e., group, subgroup, and individual). Node-link-based visualization techniques are currently widely used for these tasks and have been demonstrated to be effective, but it was found that they also have limitations in representing temporal changes, particularly at the individual and subgroup levels. To overcome these limitations, this article presents a new network visualization technique, called “TimeMatrix,” based on a matrix representation. Interaction techniques, such as overlay controls, a temporal range slider, semantic zooming, and integrated network statistical measures, support analysts in studying temporal social networks. To validate the design, the article presents a user study involving three social scientists analyzing inter-organizational collaboration data. The study demonstrates how TimeMatrix may help analysts gain insights about the temporal aspects of network data that can be subsequently tested with network analytic methods.