The use of LLM to identify goals and objectives from the text of degree thesis using fine-tuning and prompt-engineering methods

This research examines the application of Large Language Models (LLMs) fine-tuned on specially prepared datasets to solve the task of identifying goals and objectives from introductions of student degree theses. Models Qwen2.5, Qwen3, and Gemma3 were considered, each fine-tuned using LoRA and DoRA methods. The obtained results confirm that LLMs have potential for effective analysis of complex structured texts containing both scientific and practice-oriented academic materials. The Qwen3 model fine-tuned using the DoRA method demonstrated the best results in extracting specific semantic fragments from complex structured and voluminous texts.