RT Generic
T1 Towards Autonomous Web Navigation with LLM-based Agents
A1 Izquierdo Álvarez, Mario
K1 LLM-based
K1 Web navigation
AB Autonomous web navigation remains a complexchallenge, particularly due to the dynamic, diverse and unstructured nature of web environments. Traditional web scraping techniques, while effective, require rigid configurations tied to specificwebsite structures, limiting their generalizability. To addressthese challenges, this work explores the usage of autonomousagents powered by Large Language Models for autonomous webnavigation, focusing on the retrieval of academic publicationsfrom webs of preprint repositories.The proposed solution, based on hyperlink exploration, isdesigned as a component of a potentially broader system forAI-driven paper search assistance. It leverages a multi-agentarchitecture and a structured tree-traversal like approach toexplore and extract relevant documents. Each agent is assigneda specific role, including relevant URL extraction, documentcollection, planning, presentation and quality control. The systemis implemented using AutoGen, which enables flexible agentinteractions and modular design. Unlike traditional web information extraction techniques, this approach generalizes navigationpatterns across different websites without relying on predefinedHTML selectors, allowing its usage on different websites.Experimental results are promising, demonstrating the system’s effectiveness in retrieving relevant academic content. However, challenges such as increased response times and occasionalhallucinations indicate areas for refinement. Future work aimsto enhance interactivity by integrating advanced form-basedsearch capabilities, optimize retrieval efficiency, and implementmore robust evaluation frameworks. These improvements couldcontribute to fully automated AI-driven web exploration, facilitating the development of more generalizable autonomous webnavigation tools.
YR 2025
FD 2025-02-21
LK https://hdl.handle.net/10347/45053
UL https://hdl.handle.net/10347/45053
LA eng
NO Traballo Fin de Máster en Intelixencia Artificial. Curso 20024-2025
DS Minerva
RD 1 may 2026