III Taller de Inteligencia Artificial, Ciencia de Datos y Aplicaciones Criptográficas

Large Language Models as NLP Algorithm Replacements in AutoML Pipelines

No programado

20m

Facultad de Matemática y Computación

Ponente

Gabriel Hernández Rodríguez (Universidad de la Habana)

Descripción

The convergence of Large Language Models (LLMs) and Automated Machine Learning (AutoML) offers transformative potential for streamlining NLP pipeline design, traditionally reliant on manual algorithm selection and feature engineering. This research investigates how LLM inference capabilities—leveraged as dynamic, self-contained modules—can automate and optimize NLP task integration within AutoML systems. By framing LLMs as trainable inference engines, we explore their role in replacing rigid, handcrafted NLP components (e.g., tokenizers, parsers, or classifiers) with adaptable, prompt-driven solutions that reduce human intervention in pipeline configuration.