The paper “Formal Reasoning for Intelligent QA Systems: A Case Study in the Educational Domain” by members of the URA Research Group has been accepted at AIQAM ’25 (The 2nd ACM Workshop on AI-powered Question & Answering Systems), held in conjunction with the prestigious 33rd ACM International Conference on Multimedia (ACMMM 2025). This marks a significant milestone for the group, especially considering that the majority of the authors are first-year students, supported by senior researchers and professors at HCMUT.
The paper introduces a novel framework named MCFR (Model Checking for Formal Reasoning), which integrates model checking techniques with large language models (LLMs) to enhance the accuracy and reliability of question answering (QA) systems. The approach is particularly suited for domains with strict policy and procedural compliance, such as education. To evaluate their method, the team created EduMCQA, a dataset derived from real-world academic procedures at HCMUT, simulating student inquiries and formal rule-based responses. The system shows promise in delivering both correct answers and verifiable reasoning chains, contributing toward more explainable and trustworthy QA systems.
The AIQAM workshop, hosted under the umbrella of ACMMM, gathers leading researchers in AI-powered QA, NLP, and multimedia systems, providing an excellent venue for exchanging cutting-edge ideas and technologies. This success not only highlights the technical capability and ambition of the URA Research Group but also serves as a powerful inspiration for its members to further pursue rigorous, explainable AI research.