Breadcrumb
- Home
- AI Hub For Education
- Research Study Repository
- Evidence Explorer
- AI Tools
- Custom-built Research Tool
The Repository
Search and Filter
Custom-built research tool
Studies tagged with AI from their full text or abstract; see each study for details.
Showing 361 - 390 of 531 studies
Narrowing the Gap: Supervised Fine-Tuning of Open-Source LLMs as a Viable Alternative to Proprietary Models for Pedagogical Tools
Lorenzo Lee Solano, Charles Koutcheme, Juho Leinonen, Alexandra Vassar, Jake Renzella. (07/2025). arXiv.
Fine-tuned smaller models significantly improved over their base versions, winning 58% to 74% of expert head-to-head comparisons across compile- and run-time errors.
Do Students Write Better Post-AI Support? Effects of Generative AI Literacy and Chatbot Interaction Strategies on Multimodal Academic Writing
Yueqiao Jin, Kaixun Yang, Roberto Martinez-Maldonado, Dragan Gašević, Lixiang Yan. (07/2025). arXiv.
Proactive AI dialogues fostered stronger connections among conceptual reasoning, evidence use, and constructive engagement, whereas reactive interactions were more factual and procedural.
From Query to Explanation: Uni-RAG for Multi-Modal Retrieval-Augmented Learning in STEM
Xinyi Wu, Yanhao Jia, Luwei Xiao, Shuai Zhao, Fengkuang Chiang, Erik Cambria. (07/2025). arXiv.
Uni-RAG achieved superior retrieval accuracy on the SER dataset, reaching an 84.1% top-1 score on text-to-image queries and outperforming baseline models.
Text Production and Comprehension by Human and Artificial Intelligence: Interdisciplinary Workshop Report
Patrick Bolger, Morten H. Christiansen, Bill Cope, Pablo Contreras Kallens, Mary Kalantzis, Kyle Mahowald, Martin Pickering, Michael Ramscar, Alina Reznitskaya, Leila Wehbe, Wei Xu, Evgeny Chukharev, Emily Dux Speltz, Mark Torrance, Kimberly Becker, Phuong Nguyen, Dan Song. (07/2025). arXiv.
Large language models offer hypotheses about human language processing through uncertainty reduction and statistical learning, but their real-world functional competence remains uneven compared to formal grammar rules.
Designing an Adaptive Storytelling Platform to Promote Civic Education in Politically Polarized Learning Environments
Christopher M. Wegemer, Edward Halim, Jeff Burke. (06/2025). arXiv.
The AI storytelling prototype successfully aggregated real-time affective signals from facial emotion tracking and adjusted narrative language beat by beat to sustain user emotional engagement.
Leveraging a Multi-Agent LLM-Based System to Educate Teachers in Hate Incidents Management
Ewelina Gajewska, Michal Wawer, Katarzyna Budzynska, Jarosław A. Chudziak. (06/2025). arXiv.
A multi-agent setup combined with retrieval-augmented generation achieved the highest classification accuracy on both explicit hate speech (79%) and implicit hate speech (60%).
That's Not the Feedback I Need! - Student Engagement with GenAI Feedback in the Tutor Kai
Sven Jacobs, Maurice Kempf, Natalie Kiesler. (06/2025). arXiv.
Generative AI feedback advanced problem-solving in 49.7% of requests, proving more helpful for experienced students (60.9%) than inexperienced students (43.0%).
CogGen: A Learner-Centered Generative AI Architecture for Intelligent Tutoring with Programming Videos
Wengxi Li, Roy Pea, Nick Haber, Hari Subramonyam. (06/2025). arXiv.
The full CogGen pipeline achieved the highest TrueSkill rating (30.14), significantly outperforming knowledge-only, method-only, and unconstrained baseline conditions.
Confucius3-Math: A Lightweight High-Performance Reasoning LLM for Chinese K-12 Mathematics Learning
Lixin Wu, Na Cai, Qiao Cheng, Jiachen Wang, Yitao Duan. (06/2025). arXiv.
Confucius3-Math achieved higher accuracy on Chinese K-12 benchmarks than larger models, scoring 96.24% on CK12-MATH and 98.46% on GAOKAO-Bench.
Can theory-driven learning analytics dashboard enhance human-AI collaboration in writing learning? Insights from an empirical experiment
Angxuan Chen, Jingjing Lian, Xinran Kuang, Jiyou Jia. (06/2025). arXiv.
Students using the dashboard achieved significantly higher academic writing knowledge scores on the post-test than students writing without it (90.00 versus 82.70 points).
CoachGPT: A Scaffolding-based Academic Writing Assistant
Fumian Chen, Sotheara Veng, Joshua Wilson, Xiaoming Li, Hui Fang. (06/2025). arXiv.
Users reported that the assistant provided personalized feedback that made them feel heard, understood, attentive, and supported throughout the writing process.
Exploring the Usage of Generative AI for Group Project-Based Offline Art Courses in Elementary Schools
Zhiqing Wang, Haoxiang Fan, Shiwei Wu, Qiaoyi Chen, Yongqi Liang, Zhenhui Peng. (06/2025). arXiv.
Generative AI provided elementary students with background information, creative inspiration, and personalized guidance during their art projects.
Combining Log Data and Collaborative Dialogue Features to Predict Project Quality in Middle School AI Education
Conrad Borchers, Xiaoyi Tian, Kristy Elizabeth Boyer, Maya Israel. (06/2025). arXiv.
System log data predicted student productivity (training phrase count) best, achieving an AUC of 0.8053 compared to 0.5971 for dialogue alone.
COGENT: A Curriculum-oriented Framework for Generating Grade-appropriate Educational Content
Zhengyuan Liu, Stella Xin Yin, Dion Hoe-Lian Goh, Nancy F. Chen. (06/2025). arXiv.
The COGENT framework generated passages with significantly higher curriculum alignment scores (4.62) than baseline generation (4.08).
SRLAgent: Enhancing Self-Regulated Learning Skills through Gamification and LLM Assistance
Wentao Ge, Yuqing Sun, Ziyan Wang, Haoyue Zheng, Weiyang He, PiaoHong Wang, Qianyu Zhu, Benyou Wang. (06/2025). arXiv.
Undergraduate students using SRLAgent significantly improved their self-regulated learning skills from pretest to post-test, whereas control and non-scaffolded baseline groups showed no significant gains.
Educators' Perceptions of Large Language Models as Tutors: Comparing Human and AI Tutors in a Blind Text-only Setting
Sankalan Pal Chowdhury, Terry Jingchen Zhang, Donya Rooein, Dirk Hovy, Tanja Kasser, Mrinmaya Sachan. (06/2025). arXiv.
Educators perceived the large language model tutor as outperforming human tutors across all four metrics, with statistically significant advantages in conciseness, empathy, and scaffolding.
Exploring Effective Strategies for Building a Customised GPT Agent for Coding Classroom Dialogues
Luwei Bai, Dongkeun Han, Sara Hennessy. (06/2025). arXiv.
Targeted prompt engineering and iterative refinement produced an AI coding assistant reaching 64.9% turn-level agreement with human coders.
T-TEXTS (Teaching Text Expansion for Teacher Scaffolding): Enhancing Text Selection in High School Literature through Knowledge Graph-Based Recommendation
Nirmal Gelal, Chloe Snow, Ambyr Rios, Hande Kuçuk McGinty. (06/2025). arXiv.
Node2Vec achieved the highest link prediction accuracy across all dataset sizes (0.9642 to 0.9750) and had the strongest ranking performance on larger text sets.
Generating Pedagogically Meaningful Visuals for Math Word Problems: A New Benchmark and Analysis of Text-to-Image Models
Junling Wang, Anna Rutkiewicz, April Yi Wang, Mrinmaya Sachan. (06/2025). arXiv.
The MATH2VISUAL framework combined with large language models outperformed standard text-to-image models across accuracy, completeness, clarity, and cognitive load optimization.
SingaKids: A Multilingual Multimodal Dialogic Tutor for Language Learning
Zhengyuan Liu, Geyu Lin, Yanfeng Lu, Hui Li Tan, Huayun Zhang, Xiaoxue Gao, Stella Xin Yin, He Sun, Hock Huan Goh, Lung Hsiang Wong, Nancy F. Chen. (06/2025). arXiv.
High-performing students received more feedback (69% versus 43%) and explanations (21% versus 9%), whereas lower-performing students received more hints (12% versus 5%) and social-emotional support (31% versus 17%).
Towards Generating Controllable and Solvable Geometry Problem by Leveraging Symbolic Deduction Engine
Zhuoxuan Jiang, Tianyang Zhang, Peiyan Peng, Jing Chen, Yinong Xun, Haotian Zhang, Lichi Li, Yong Li, Shaohua Zhang. (06/2025). arXiv.
The proposed framework with qualification checking generated solvable problems on 100% of tested cases on JGEX-AG-231 and 96% on GeoQA, outperforming GPT-4o.
TestAgent: An Adaptive and Intelligent Expert for Human Assessment
Junhao Yu, Yan Zhuang, YuXuan Sun, Weibo Gao, Qi Liu, Mingyue Cheng, Zhenya Huang, Enhong Chen. (06/2025). arXiv.
TestAgent achieved the same ability estimation accuracy as traditional methods with 20% fewer questions across test datasets.
K12Vista: Exploring the Boundaries of MLLMs in K-12 Education
Chong Li, Chenglin Zhu, Tao Zhang, Mingan Lin, Zenan Zhou, Jian Xie. (06/2025). arXiv.
Gemini2-thinking achieved the highest overall performance, scoring 55.47% in direct answering and 57.36% in step-by-step reasoning across Chinese K-12 science subjects.
A Practical Guide for Supporting Formative Assessment and Feedback Using Generative AI
Sapolnach Prompiengchai, Charith Narreddy, Steve Joordens. (06/2025). arXiv.
Most existing research primarily evaluates AI-generated feedback as a single task-level measure of accuracy, overlooking process-level and self-regulation-level feedback.
CO-OPERA: A Human-AI Collaborative Playwriting Tool to Support Creative Storytelling for Interdisciplinary Drama Education
Xuejiao Ma, Haibo Zhao, Zinuo Guo, Yijie Guo, Guanhong Liu, Bo Jiang. (06/2025). arXiv.
Interviews with 13 teachers revealed that current general-purpose AI tools cannot effectively assist teachers and students during playwriting.
Learn Like Feynman: Developing and Testing an AI-Driven Feynman Bot
Akshaya Rajesh, Sumbul Khan. (05/2025). arXiv.
Learners using the Feynman Bot achieved higher mean learning gains (3.83 points) than passive learners (2.17 points), with more consistent improvement across participants.
A Human-Centric Approach to Explainable AI for Personalized Education
Vinitra Swamy. (05/2025). arXiv.
Existing post-hoc explainers show systematic disagreements, highlighting a need for inherently interpretable model architectures in education.
Particle Builder - Learn about the Standard Model while playing against an AI*
Mohammad Attar, Andrew Carse, Yeming Chen, Thomas Green, Jeong-Yeon Ha, Yanbai Jin, Amy McWilliams, Theirry Panggabean, Zhengyu Peng, Lujin Sun, Jing Ru, Jiacheng She, Jialin Wang, Zilun Wei, Jiayuan Zhu, Lachlan McGinness. (05/2025). arXiv.
Students' understanding of particle physics concepts improved significantly, with an average gain of 0.18 and a normalised gain of 0.23.
RATAS: A Generative AI Framework for Explainable and Scalable Automated Answer Grading
Masoud Safilian, Amin Beheshti, Stephen Elbourn. (05/2025). arXiv.
RATAS achieved high grading accuracy with a mean absolute error of 0.0309 and an R-squared of 0.9627, substantially outperforming direct GPT-4o.
CoderAgent: Simulating Student Behavior for Personalized Programming Learning with Large Language Models
Yi Zhan, Qi Liu, Weibo Gao, Zheng Zhang, Tianfu Wang, Shuanghong Shen, Junyu Lu, Zhenya Huang. (05/2025). arXiv.
CoderAgent outperformed baseline models in predicting next modification intentions, edit locations, and subsequent code submissions, achieving CodeBLEU scores up to 0.7698 on CSEDM.
