Osaka Institute of Technology Achieves 9.5% Acceptance Rate in MEXT's AI for Science Challenge Research Program

Osaka Institute of Technology (President: Shinya Inoue) announced that 4 out of 42 research proposals submitted to the first call for proposals of the FY2026 'AI for Science萌芽的挑戦研究創出事業 (SPReAD)' program, implemented by the Ministry of Education, Culture, Sports, Science and Technology (MEXT), have been accepted. With an acceptance rate of 9.5%, significantly surpassing the national average of 2.9%, the university ranked first among institutions that submitted more than 40 proposals.

Key Highlights:

● 4 research projects accepted in the first round of the 'AI for Science萌芽的挑戦研究創出事業 (SPReAD)'

● 9.5% acceptance rate—the highest among institutions submitting over 40 proposals

● 2 of the 61 student-led projects accepted nationwide are from OIT graduate students

SPReAD is a MEXT initiative supporting innovative research ideas that leverage AI to advance and accelerate scientific research. Launched this year amid rapid AI advancements and societal integration, it targets university researchers, students, and private-sector researchers across all disciplines. The first call received 15,868 applications from 787 universities and institutions nationwide, with 456 projects accepted.

OIT submitted 42 proposals and secured 4 acceptances. This places the university third among private universities nationwide and first among private universities in the Kansai region (Osaka, Kyoto, Hyogo, Shiga, Nara, and Wakayama). Among student submissions, 2,624 were received nationwide, with 61 accepted—2 of which are from OIT graduate students.

The university attributes this success to its ongoing efforts to promote interdisciplinary collaboration and build a research environment that integrates AI and information technologies into domain-specific expertise. OIT will continue strengthening its research infrastructure and collaborative networks to drive innovation and real-world implementation.

The accepted research projects (principal investigator, title, and summary) are as follows:

1. Haruna Takagi, Master’s Program Year 2, Department of Electrical, Electronic, and Mechanical Engineering, Graduate School of Engineering

"Quantitative Magnetic Field Data Conversion from Magneto-Optical Images via AI and Construction of a Non-Contact Measurement Framework"

For magneto-optical imaging (MOI) data, where magnetic field strength and direction are difficult to read directly, this project aims to develop an AI-assisted non-contact measurement framework using a custom deep learning model to convert MOI images into quantitative magnetic field data, advancing MOI from observation to measurement.

2. Naoya Okumura, Doctoral Program Year 1, Department of Robotics & Design Engineering, Graduate School of Engineering

"Gait Acquisition for Morphing Legged Robots via Sim2Real"

Targeting robots capable of transforming between wheeled and legged configurations, this project uses AI-based reinforcement learning within a physics simulator to achieve stable locomotion control on steps and stairs, enabling real-world deployment through Sim2Real transfer.

3. Professor Toshihisa Nishiguchi, Department of Real-World Information, Faculty of Information Science

"Creation of Real-World Information Representation for Safe Human and Robot Navigation Using VLM, RAG, and SAT/ILP Formal Verification"

This project will build a small-scale proof-of-concept that links 3D point clouds and color images from depth cameras with linguistic knowledge from specifications and regulations using Vision-Language Models (VLM) and Retrieval-Augmented Generation (RAG), enabling safe navigation for visually impaired individuals and robots.

4. Associate Professor Tetsu Ochi, Department of Intelligent Information, Faculty of Information Science

"AI Analysis of Well-Intentioned Misinformation Spread under Social Anxiety"

Focusing on SNS content related to healthcare, disasters, and crime prevention—topics prone to social anxiety—this project uses generative AI to analyze how general users unintentionally share misinformation, believing it to be beneficial ('well-intentioned misinformation spread').