Order	Time	ID	Full information
1	15:45-16:00	IT1043	Leveraging Large Language Models for Teaching Reform: An Intelligent Assessment  Practice in Undergraduate Discrete Mathematics Author(s): Dawei Wen, Lifang Xiao Presenter: Dawei Wen, Wuhan Institute of Technology, China Abstract: Against the backdrop of engineering education accreditation, which emphasizes competency development, assessing higher-order thinking skills in foundational courses like discrete mathematics through free-response questions has become imperative. However, this shift imposes a significant grading burden on instructors, hindering timely feedback. This teaching practice study explores the integration of prominent Chinese Large Language Models (LLMs) as intelligent assistants to automate the grading of free-response assignments in an undergraduate discrete mathematics course. We conducted a practical study comparing four LLMs (DeepSeek, Wenxin Yiyan, Kimi, and Xunfei Spark) using 40 student submissions across two assignment types: open-ended logical reasoning problems and structured graph theory problems. Results indicate that DeepSeek demonstrated the strongest alignment with instructor grading, particularly for structured problems. However, all models struggled with open-ended tasks featuring non-unique solutions. The findings lead to concrete instructional recommendations: a hybrid grading model is proposed, where LLMs handle structured problem grading to free up instructor time, allowing them to focus on guiding students through complex, open-ended reasoning processes. This study provides a practical roadmap for leveraging AI to reform assessment practices, ultimately enhancing both instructional efficiency and the depth of student learning in STEM education.
2	16:00-16:15	IT1046	Construction of a Large Language Model-Supported Problem Chain Guided Teaching Model Author(s): Fengdie Cui, Tingting Liu, Chuang Yin Presenter: Fengdie Cui, Beibu Gulf University, China Abstract: Problem chain teaching serves as a critical pathway for advancing students’ cognitive development. In response to existing challenges such as imprecise positioning, limited integration, and weak instructional guidance in current problem chain teaching practices, this study constructs a guided teaching model powered by the knowledge generation and logical reasoning capabilities of large language model. This model emphasizes tripartite collaboration among teachers, students, and technology, with evaluation embedded throughout the instructional process. Using the teaching of Averages as an illustrative case, we systematically elucidate how this framework leverages large language model to create problem scenarios and generate progressive problem chains. It guides students through a complete cognitive process from identifying, formulating and analyzing problems to arriving at solutions, thereby offering a feasible approach to fostering in-depth classroom teaching.
3	16:15-16:30	IT2149	Enhancing Preservice Teachers’ Feedback Literacy through a Large Language Models-Based Multi-Agent System Author(s): Zhiwei Qi, Yuqing Liu, Hongyan Li Presenter: Yuqing Liu, Yunnan University, China Abstract: Feedback literacy is a critical competency for teachers, yet existing teacher education programs often struggle to provide sufficient opportunities to interact with real students for preservice teachers to develop the sophisticated feedback literacy. To address this issue, we propose a novel approach that leverages a large language models (LLMs)-based multi-agent system to enhance preservice teachers’ feedback literacy. We first designed a multi-agent system comprising specialized agents that simulate different roles in the feedback process, including student agent, assessment agent, expert agent, and resource recommendation agent. Then, we recruited 68 preservice teachers and randomly assigned them to experimental group (EG) and control group (CG). Next, we implemented a 5-week intervention where the EG interacted with the multi-agent system to practice feedback literacy, while the CG received traditional instruction. Quantitative data were collected through pre-, post-, and one-month follow-up assessment to measure preservice teachers’ feedback literacy. Qualitative data were gathered via semi-structured interviews to capture their perceptions. Finally, we analyzed the results using independent samples t-tests and thematic analysis to evaluate the system’s effectiveness. The experimental results show that the multi-agent system significantly improves preservice teachers’ feedback literacy compared to the CG. Overall, our study demonstrates the promising potential of integrating a LLMs-based multi-agent system into teacher education and offers a scalable solution for developing feedback literacy.
4	16:30-16:45	IT2171	Overcoming AI Dependency: Practical Pathways and Effectiveness of Educational Agents in Python Instruction Author(s): Fang Xia, Ren Ying Presenter: Xia Fang, Naval Aeronautical University, China Abstract: Abstract. The proliferation of generative artificial intelligence (AI) has given  rise to the phenomenon of "AI dependency" in programming learning, where students tend to directly obtain code solutions, thereby undermining the cultivation  of problem-solving abilities and computational thinking. This study developed  an instructional intervention system centered on "Role-based Pedagogical  Agents" for the Python Programming course, transforming AI from an "answer  provider" into a "thinking support platform." Based on a one-semester quasi-experimental study, students in the experimental group significantly reduced their  behavior of directly requesting complete code. They demonstrated more prominent performance in problem decomposition, algorithm design, and debugging  capabilities, along with a noticeable enhancement in reflection depth. The research confirms that role-based pedagogical agents, through structured process  design and agent intervention, effectively reduce AI dependency, promote the  improvement of students' computational thinking and programming quality, and  provide a practical pathway for programming instruction in the era of artificial  intelligence.
5	16:45-17:00	IT3267	Refining L2 Learner Agency through Multi-role Prompting: A Coordination Framework for AI-Mediated Writing Peer Feedback Author(s): Jie Pan, Yanan Hu, Huimei Chen Presenter: Jie Pan, Shanghai Normal University Tianhua College, China Abstract: While AI feedback has become common in writing instruction, its typically authoritative and prescriptive nature often stifles learner agency. To address this, our study introduces a multi-role prompting framework that shifts students from passive recipients to active decision-makers. Instead of relying on a single, homogenized AI voice, the system coordinates simulated personas—such as specialists in logic, language, and cohesion—to provide diverse perspectives. Crucially, the design does not aim for consensus; instead, it organizes disagreements into visible options by surfacing rhetorical trade-offs and “cognitive frictions” that require students to evaluate competing suggestions. Qualitative interviews with six college students suggest that this approach transforms AI from a directive “corrector” into a mediator of critical inquiry. By implementing strict interaction boundaries that prevent automated rewrites, the system keeps the intellectual work of interpretation and synthesis with the student. The findings support the pedagogical value of persona-driven prompt orchestration in scaffolding metacognitive engagement and helping learners with different proficiency levels make more intentional revision choices in AI-supported environments.
6	17:00-17:15	IT3315	Digital Human Teaching Assistants: LLM-Empowered Construction of First-Class Courses in Higher Education Author(s): Jing Xiong, Xue Zhai, Bingbing Wei, Yuxia Lei, Zhaoan Dong, Yan Yao, Jianguo Liang Presenter: Jing Xiong, Qufu Normal University, China Abstract: The establishment of first-class undergraduate courses, namely “Golden Courses”, stands as a strategic focal point in contemporary higher education. Nevertheless, it encounters impediments in the form of substantial instructor workloads and restricted student-teacher interaction. This paper presents an innovative framework for a Digital Human Teaching Assistant (DHTA) that is enabled by Large Language Models (LLMs) and multi-modal interaction technology. Empirical findings indicate that the DHTA pipeline curtails course production time by 88.5%, concurrently leading to a notable improvement in student performance (the mean score of the Experimental Group is 86.9, compared to 78.7 in the Control Group). The t-test produced a p-value of <0.001, validating the statistically significant influence of the presence of the digital human on knowledge retention and student engagement. This research offers a scalable and efficient model for the intelligent transformation of higher education, integrating technological innovation with pedagogical humanism.
7	17:15-17:30	IT3383	Large Language Model Empowered International Education Resource Sharing and  Personalized Recommendation for Graduate Students Author(s): Jie Zuo, Quan Liu, Wenjun Xu, Wei Meng Presenter: Jie Zuo, Wuhan University of  Technology, China Abstract: With the rapid advancement and widespread deployment of large language models (LLMs) and related artificial intelligence tech-nologies, internationalization of higher education faces unprece-dented opportunities and challenges. This paper addresses prac-tical problems encountered in the internationalized training of postgraduate students in electronic and information engineering, such as fragmentation of teaching and research resources, ineffi-ciencies in cross-cultural collaboration, and coarse-grained ap-proaches to individualized training. Guided by constructivist learning theory and long-tail theory, we propose a three-layer LLM-empowered platform architecture, i.e. data layer-algorithm layer and application layer, for integrating and sharing interna-tional teaching-research resources. A GNN-based entity-relation extraction method is used to construct a domain knowledge graph, and an intelligent recommendation algorithm is intro-duced that implements a three-step mechanism: multi-source data fusion, personalized learning-path planning, and dynamic path adjustment. Experimental evaluation indicates that the plat-form achieves professional domain knowledge coverage and pro-fessional terminology comprehension accuracy both exceeding 95%. The platform therefore has strong potential to improve postgraduate students’ international academic competitiveness, cross-cultural collaborative ability, and practical AI skillset, and offers a transferable solution for cultivating globally competitive, high-caliber innovative talents.
8	17:30-17:45	IT4424	Empowering Writing Assessment and Feedback through Large Language Models: A Hybrid Framework Integrating Local Fine-Tuning and Prompt Engineering Author(s): Huijun Ma, Charuni Samat, Mingjie Li, Jiajun Zhang Presenter: Huijun Ma, Kohn Kaen University, Thailand Abstract: This study investigates the application potential of Large Language Models (LLMs) in English writing assessment and feedback. Experiments were conducted to compare a fine-tuned local small-scale LLM with the web-based DeepSeek-R1 model, optimized via prompt engineering. Results indicate that for composition scoring, the fine-tuned local model achieved moderate consistency with human grading, validating its feasibility. Conversely, regarding instructional feedback, while the local model underperformed, the DeepSeek-R1 model combined with optimized prompt engineering yielded significant improvements, demonstrating that high-quality feedback is attainable through prompt engineering. Consequently, provided privacy risks are controlled, teachers and students can leverage such LLMs to effectively enhance English writing instruction and learning outcomes.
9	17:45-18:00	IT6004	Design and Implementation Path of PBL for Front-End Development Technology Empowered by GAI Author(s): Yunting Song, Sa Li Presenter: Yunting Song, KaShi University, China Abstract: With the increasingly widespread application of Generative Artificial Intelligence(GAI) in higher education, new opportunities for innovation and transformation have emerged for traditional Project-based Learning(PBL) pedagogy. Aiming at the deficiencies of traditional PBL in front-end development technology—such as the lack of efficient content generation, real-time tutoring, and systematic process management—this study conducted a two-semester tracking investigation. It comparatively analyzed the multi-dimensional impacts of GAI on students’ learning outcomes in PBL for front-end development, including development efficiency, code quality, falut rate, and technical documentation quality. A Mann-Whitney U test was performed using Python, and the results show that GAI significantly improves students’ multi-dimensional competencies, enhances their learning interest, and addresses the shortcomings of traditional PBL. This study demonstrates the necessity of empowering university front-end development teaching with GAI and provides new ideas for the advancement of PBL.
10	18:00-18:15	IT3350	Research on Adaptive Teaching for Students in the Initial Primary School Art Class—AIGC-Assisted Teaching Practice in the “Introducing Myself” Lesson Author(s): Meng Jia, He Gao, Xiuli Sun, Jingting Ma Presenter: Meng Jia, Dalian University, China Abstract: Student adaptability plays a crucial role in establishing subject cognition and interest. The first primary school art class serves as a critical starting point for students' artistic learning journey. In response to widespread student adaptation challenges, this study proposes a shift in teaching from passively addressing problems to proactively nurturing potential. An adaptive teaching framework encompassing three dimensions — emotional ice-breaking, behavioral norm-setting, and relational connection-building—is constructed. Subsequently, teaching practices assisted by AIGC in the “Introducing Myself” lesson are implemented, thereby transforming potential “adaptation crisis points” in the first art class into “developmental opportunities”.