Ainhoa Vivel Couso

In 2021, I was awarded a CiTIUS Summer Scholarship, where I began my research in Natural Language Technologies. In 2022, I participated in Proxecto Nós, contributing to the creation of textual linguistic resources for Galician, aimed at integrating the language into Natural Language Processing (NLP) tools. In 2023, I received the IV Edition Altia Premia Award for my Bachelor's Final Project, which focused on the automatic generation of natural language descriptions to justify estimated chronological age using a deep learning–based method.

From 2023 to 2024, I worked at the CiTIUS research center as a Research Collaborator, contributing to the REPROHUM and DeepR3 projects. I am currently pursuing a PhD, where my research centers on the reproducibility of human evaluations in NLP and the development of efficient methods to extend, reuse, and adapt pre-trained language models for new domains, genres, and languages.

In my research group, we focus on topics such as reproducibility, repeatability, knowledge transfer, reusability, and energy efficiency in large language models, including mT5 (Google), GPT-2 (OpenAI), and LLaMA 2 (Meta). My current work involves conducting state-of-the-art reviews by searching, summarizing, classifying, and analyzing scientific publications. Additionally, I reproduce experiments and train language models using knowledge transfer techniques.