The Beginning of My LLM Journey

Under the guidance of Prof. An-Zi Yen, I delved into a lecture-abstract-based question generation system project, where we fine-tuned an LLM on self-collected datasets. Our demo system’s performance has outperformed state-of-the-art models, and I led the authorship of a paper accepted at the International Joint Conference on Artificial Intelligence (IJCAI). Recognizing LLMs’ challenges in performing complex reasoning and answering questions, we focused on error detection in math problems and achieved improvements through our proposed self-refinement framework.

These experiences show my ability to conduct independent research and stimulate my research interest in enhancing model reasoning capabilities and robustness, mainly for educational purposes.