| 日時(Datetime) | 令和7年6月20日(金)3限 (13:30--15:00), 2025/06/20, Friday, 3rd slot |
|---|---|
| 場所(Location) | パナソニックISセミナーホール（L1）, Panasonic IS Seminar Hall （L1） |
| 司会(Chair) | 嶋利 (Shimari) sensei |
| 講演者(Presenter) | Yusuke Sakai, 自然言語処理学研究室 (Natural Language Processing Research Lab.) |
|---|---|
| 題目(Title) | Toward Sustainable Large Language Model Evaluation |
| 概要(Abstract) | Many large language model (LLM) evaluation benchmarks have been proposed. However, these benchmarks must be frequently updated due to the continuous improvement of LLMs. Nevertheless, this is not sustainable under current benchmark construction practices, due to resource limitations and annotation costs. We also need to consider reusability, such as effective ways of expanding and updating existing datasets, evaluating different aspects using the same benchmark or by reformulating it, and incorporating peripheral evaluation methods, e.g., assessing the robustness of the evaluation framework. In this talk, I discuss my work and future directions from several perspectives related to these challenges, including reasoning ability, knowledge, compositionality, visual understanding, and cultural understanding. |
| 講演者(Presenter) | Chakraborty Dipanita, ネットワークシステム学研究室 (Network Systems Research Lab.) |
|---|---|
| 題目(Title) | Introducing Computer Vision and Artificial Intelligence for 5G and Emerging 6G Wireless Communications |
| 概要(Abstract) | The evolution of 5G and emerging 6G networks demands faster, adaptive, and energy-efficient communication systems. AI-based advancements using Machine Learning (ML) and Deep Learning (DL) algorithms have shown promising results in optimizing data traffic management, interference mitigation, and resource allocation. Within this domain, computer vision, which is a branch of AI that interprets visual data, offers unique opportunities by using visual information from the environment to improve user positioning, obstacle awareness, signal direction, network decision making, and support real-time adaptive control. In light of these developments, this talk presents key concepts, emerging methods, and applications of computer vision and AI for intelligent wireless networks. |
| 講演者(Presenter) | Fan Youmei, ソフトウェア工学研究室 (Software Engineering Research Lab.) |
|---|---|
| 題目(Title) | Evolving Dynamics of Open-Source Software: From Sponsorship to Protestware |
| 概要(Abstract) | This talk explores the multifaceted landscape of open-source software (OSS) development, focusing on how its communities foster sustainable growth and address emerging challenges. We integrate findings from two key areas: the impact of social media on financial sustainability and the dynamics of "protestware" events. Drawing on a study of Twitter/X mentions and GitHub Sponsors, we illustrate how strategic online engagement can significantly bolster developer funding, while also examining the comparative visibility of different donation platforms. Complementing this, we analyze the disruptions caused by "protestware" intentional code changes made for socio-political statements, exemplified by the colors.js and es5-ext incidents. |