This thesis explores how to build conversational assistants based on Large Language Models (LLMs) to support the understanding and use of complex autonomous systems, such as robotic vehicles or intelligent boats. The project starts from a first dataset of technical documents made available through the BlueBoat project, developed in collaboration with NTNU – Norwegian University of Science and Technology. The aim is to design a pipeline for training, customizing, and effectively interacting with a domain-specific LLM. The final goal is to create an accessible and reusable system that helps human users interact effectively with autonomous systems.
Questa tesi esplora come costruire assistenti conversazionali basati su modelli linguistici di grandi dimensioni (LLM) per facilitare la comprensione e l’utilizzo di sistemi autonomi complessi, come veicoli robotici o imbarcazioni intelligenti. Il lavoro prende avvio da un primo insieme di documenti tecnici forniti nell’ambito del progetto BlueBoat, sviluppato in collaborazione con la NTNU – Norwegian University of Science and Technology. L’obiettivo è progettare una pipeline che consenta l’addestramento, la personalizzazione e l'interazione efficace con un LLM specializzato. Il fine ultimo è la creazione di un sistema accessibile e riutilizzabile, utile per supportare utenti nell’interazione con i sistemi autonomi.
Conversare con i Robot: Come Creare Assistenti LLM per Capire e Usare Sistemi Autonomi
PIZZATO, FRANCESCO
2024/2025
Abstract
This thesis explores how to build conversational assistants based on Large Language Models (LLMs) to support the understanding and use of complex autonomous systems, such as robotic vehicles or intelligent boats. The project starts from a first dataset of technical documents made available through the BlueBoat project, developed in collaboration with NTNU – Norwegian University of Science and Technology. The aim is to design a pipeline for training, customizing, and effectively interacting with a domain-specific LLM. The final goal is to create an accessible and reusable system that helps human users interact effectively with autonomous systems.| File | Dimensione | Formato | |
|---|---|---|---|
|
Pizzato_Francesco.pdf
accesso aperto
Dimensione
4.55 MB
Formato
Adobe PDF
|
4.55 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/91686