[3Xin2] Poster session 1

[3Xin2-28] Proposal of qualitative reasoning system that has more compatibilitywith large language model based on physical laws and principles.

〇Masaharu Yoshioka¹, Shinpei Suzuki¹ (1.Hokkaido University)

Keywords:Qualitative Reasoning, Physical laws and principles, Large Language Model

Recent development of Large Language Model (LLM) allows the user to obtain the explanation of physical phenomenon as text. However, the output of LLM may contain inappropriate explanation, and it is not easy to justify the appropriateness of the generated text. To solve this problem, we have proposed a framework to construct a model for qualitative reasoning system based on the text analysis results of LLM (e.g., a list of objects and physical phenomena). In this paper, we propose a qualitative reasoning system that has more compatibility to interpret the output of LLM and demonstrate the system with simple physical behavior simulation.