Project Title: A Framework for Monitoring Multidimensional Reasoning Quality Under Conceptual Shift Conditions in LLM Query Flows

Project Leader: Assoc. Prof. Dr. Ali ŞENOL
Faculty of Engineering
Department of Computer Engineering

Project Duration: 24 months

The reliability of large language models depends not only on their ability to generate correct responses but also on their capacity to maintain consistent and robust reasoning capabilities in the face of changing user queries and usage conditions. The aim of this project is to develop a framework that monitors the reasoning quality of large language models in a multidimensional manner under conditions where the topic, difficulty, and content characteristics of queries change over time. The project aims to detect early changes in quality that emerge in the models’ responses and reasoning processes. Through the open-source software and evaluation dataset to be developed, the project aims to contribute to the continuous monitoring of the reliability of AI systems and the development of more reliable applications.

News: Office of Information and Communication