We investigated the impact of data selection on machine translation fine-tuning for open LLMs. Using Japanese-English corpora, we compare five selectors: TF-IDF, COMET Kiwi, QuRate, FD-Score, and random selection, under controlled training conditions. We observed that semantic selectors consistently outperform lexical and geometry-based heuristics, and that even when the selected data differ by less than 3%, the impact on model performance is substantial, underscoring the sensitivity of fine-tuning to data quality.

This Work-in-Progress in the Research to Practice Category paper will present our faculty-led challenges in preparing IT-majored students for the global environment. The global IT human resource development in Japan in recent years is always being challenged specially with the Japanese younger generation's `inward tendency' despite the ever-growing demand for such resources equipped with comprehensive skills. The Japanese government took initiative to encourage higher education institutes to put efforts on fostering human resources who can succeed in the current global community. We, College of Information Science and Engineering, Ritsumeikan University, were focusing on globalisation in IT field from an early stage and trying to establish a path for students to be prepared for a global environment. In 2012, we were selected for the Japanese government funding project. Through the five-year project, we built a foundation for faculty-led global IT human resource development environment. In addition, we have been organising faculty-led study abroad programs for many years. Questionnaire surveys were carried out in 2018 and 2019 for evaluation of outcomes of study abroad programs. The survey results enable us to conclude that our study abroad programs increase students' motivation in studying English and academic subjects. This paper shares our faculty's approach and contributes to all the other universities which have the same challenge in preparing information science and engineering students for the global environment.

Creation is important process in learning programming for chil-dren. It is said that children can learn well by actually creatingprograms. In a programming workshop for children, an exampleprogram is illustrated to children prior to the creation activities. Itis assumed that children can create programs using the illustratedprogram. However, there are many cases in which it is difficultfor children to use the example programs, and they can not createprograms. This paper proposes a system that supports creationactivities by exemplifying programs from search by visual queries.Visual queries composed of sprites and emoji were applied in thissystem as children are familiar with visual environment.

In this paper, we propose a programming learning system which incorporates a concept of small steps to nurture the computational thinking. While the project-based programming environment allows children to reach an unlimited goal of their achievements, most of learners often suffer from finding a meaningful project to work and coming up with a way of building programs for their project. In order to scaffold their programming experience, our system gives a step-by-step procedure to build a complete meaningful project, where the entire project is decomposed into small steps in advance. In addition, the system requires learners to ask their peer to check if an ongoing step is correctly completed before proceeding to the next step. In this way, our system allows them to experience the repeated cycle of the computational thinking process while encouraging them to interact each other. We also show our empirical findings obtained by applying this system to a programming workshop. In this workshop, 11 small steps to build a game project were provided to participants with our iPad application. As a result, we observed that they could complete the given project regardless of their programming experience and the system gave them an opportunity to interact with others while they were doing programming.

Maximizing the quality of service experienced by passengers of public transportation systems is of great importance to service operators. Accurate and timely information about passenger load, both at stops and on board vehicles, can help in many ways to improve the quality of service. This information is currently not readily available using data collection techniques that are transparent to passengers. We describe a method and feasibility study for estimating transit passenger population at two bus stops by passively monitoring Wi-Fi network activity of mobile devices. The collected data is filtered and analyzed to distinguish passengers' mobile devices from non-passenger devices. Our experimental results show that filtered Wi-Fi activity correlates well with observed passenger population at bus stops and has the potential to provide near real-time information.

In this paper, we propose a new framework to synthesize rapid single flux quantum (RSFQ) logic circuits. In our framework, we construct a virtual cell, which we call "2-AND/XOR," from the RSFQ logic primitives. By using 2-AND/XOR cells, we can successfully adopt the conventional logic design techniques into our framework, and thus we can successfully generate RSFQ circuits in reasonable time even for large benchmark circuits that have not been reported in the existing researches

This paper proposes the example expansion which supports control system development. In the example expansion, real-time system engineers familiar with programming for real-time computing develop a small example for maintaining the timing consistency. Since the example is small, its development is easy. The example is expanded according to requirements specified by control engineers who are unfamiliar with real-time computing but have expertise in the regulation of a specific plant. Since formal rules and tools expand the example, a mechanism for maintaining the timing consistency is correctly transferred to a target system. The example expansion allows the two kinds of engineers to concentrate on their own work. A control system for a steel mill plant has been developed using this method, and the system has been developed 2.7 times faster than the time taken with a conventional method.

In order to solve the problem of inopportune preemption in multiprogrammed multiprocessor synchronization, two strategies which are applicable to real-time systems have been investigated: preemption-safe locking and wait-free synchronization. Both of them, however, have a problem for use in the implementation of a real-time kernel. Preemption-safe locking has the drawback that the preemption cost becomes large; while wait-free operations on complex data structures are generally very inefficient and are not practical. We propose a novel approach to multiprogrammed multiprocessor synchronization, called the SPEPP (Spinning Processor Executes for Preempted Processors) synchronization, with which the preemption cost can be reduced to almost zero, while operations on complex data structures can be realized with reasonable efficiency. This paper presents the two algorithms of the SPEPP synchronization and their extensions, and demonstrates its effectiveness through the performance measurements of real-time kernels implemented with the SPEPP synchronization algorithms.