Natural Language Inference (NLI) via LLMs.

Natural Language Inference (NLI) is a fundamental task in Natural Language Processing that aims to determine whether a hypothesis can be inferred from a given premise. Although numerous experiments have been conducted to evaluate this task using various language models, most of these efforts have focused on the English language, leaving Spanish relatively unexplored. This thesis evaluates the state of the art of NLI task using LLMs for the Spanish language, implementing prompting strategies to maximize the performance of LLMs (Large Language Models) in this task. Finally, LLM performance in NLI is determinated for Few-Shot and Zero-Shot scenarios. The Spanish corpus was tested with GPT-4o-Mini which achieved an accuracy of 60% in a human validated subset of the Spanish Corpus, where it was concluded that LLMs struggle to identify the Entailment label and show a preference for selecting the Reasoning label over the others.

• Autores: Pérez Terán, Nicolás
• Tipo de recurso: Trabajo de grado de pregrado
• Fecha de publicación: 2025
• Institución: Universidad de los Andes
• Repositorio: Séneca: repositorio Uniandes
• Idioma: eng
• OAI Identifier: oai:repositorio.uniandes.edu.co:1992/75670
• Acceso en línea: https://hdl.handle.net/1992/75670
• Palabra clave: Recognizing Textual Entailment; Natural Language Processing; IA; Research; Natural Language Inference; Large Language Models; Human Validation; Ingeniería
• License: Attribution-ShareAlike 4.0 International