A Hybrid LLM–CBR Approach for Explainable and Stable Assessment of Academic Theses

The assessment of master’s theses is a complex and time-consuming process, often burdened by a significant degree of subjectivity. In recent years, Large Language Models (LLMs) have demonstrated high effectiveness in analyzing academic texts; however, their application to the evaluation of degree theses raises concerns about inconsistent decisions and limited ability to objectively verify content. This article proposes a hybrid approach that integrates LLMs with the Case-Based Reasoning (CBR) methodology to support the assessment of master’s theses by leveraging analogies to previously evaluated cases. The paper presents the system architecture, the formalization of a case representation, the integration of LLMs with a case base, and methods for evaluating the quality of assessments and their justifications. The analysis indicates that the proposed approach can enhance assessment consistency and improve transparency while preserving the role of the human evaluator as the final decision-maker.