Introduction of Linguistic Quantitative Reasoning into Large Language Models

Abstract

The purpose of this research is to incorporate quantitative inference (especially fuzzy inference) functions corresponding to the real world into large language models. Given the state of the environment and the conditions of an object when two objects collide, we show that the post-collision state corresponding to the conditions can be inferred in natural language. The physical state of the consequence inferred as the post-collision state is expressed in natural language, and at the same time, by capturing the context using a large-scale language model, it is possible to estimate the membership function corresponding to the linguistic expression of the quantitative state change.