| Resumen |
With the advent and proliferation of advanced Large Language Models (LLMs) such as BLOOM, GPT series, and ChatGPT, there is a growing concern regarding the potential misuse of this technology. Consequently, it has become imperative to develop machine learning techniques that can discern whether a given text has been generated by an LLM or authored by a human. In this paper, we present our approach in the AuTexTification shared task, where we fine-tuned BERT-based models and GPT-2 Small. Remarkably, GPT-2 Small achieved the highest F1-macro score in the validation set, prompting us to evaluate its performance on the testing set. We achieved an F1-macro score of 0.74134, securing the third position in the benchmark. Furthermore, we extended our fine-tuning efforts to the model attribution subtask, yielding a F1-macro score of 0.52282. © 2023 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). |
