In this research, we have constructed a model of fraudulent financial reporting using an action decomposition tree. Previously, we presented action models of accounting and auditing standards at the JSAI 2022 Annual Conference. Building on this, the current study aims to clarify the implicit assumptions in auditing standards, particularly regarding the structure and process of fraudulent financial reporting.

The Knowledge Graph Reasoning Challenge provides a systematically structured dataset in which human annotators have organized the knowledge graphs necessary for inference in some Sherlock Holmes stories. These knowledge graphs capture the logical reasoning process required to derive conclusions from the original texts.In this paper, we propose a story generation system that extracts Chain-of-Thought (CoT) representations corresponding to the plots of detective fiction and utilizes them to generate mystery narratives.By preserving the logical reasoning structure inherent in CoT, the proposed system generates diverse expressions of detective stories while maintaining logical consistency.This approach explores a text generation method that leverages the diversity in the correspondence between knowledge graphs and natural language descriptions.

In this study, we investigated how transforming a knowledge graph and varying input formats affect QA tasks as a way to effectively integrate external knowledge into LLMs (large-scale language models). Specifically, using a QA dataset generated from the DBpedia infobox, we compared five different input methods, including natural language and JSON formats. The results revealed that a clear key-value structure facilitates the accurate extraction of required information by the LLM. On the other hand, approaches employing natural language formats or containing excessive information tended to show reduced accuracy due to redundancy and missing information.