Francesco Salvi

prof_pic.jpg

Hi there! I am a first-year PhD student in Computer Science at Princeton University, advised by Manoel Horta Ribeiro. I am affiliated with Princeton’s Center for Information Technology Policy (CITP) and supported by the Gordon Wu Fellowship.

Previously, I received my MSc at EPFL’s DLAB, working with Robert West, and my BSc in Physics at the University of Bologna. During my studies, I also spent some time as a Visiting Fellow at FBK and interning at Bain & Company and DemoSquare.

My research broadly focuses on connecting computational methods and machine learning with society, analyzing the societal impacts of LLMs and drawing insights on human behavior. I am particularly interested in the persuasive capabilities of frontier models and in political phenomena such as online polarization and misinformation.

Keywords: Computational Social Science, Social Computing, Natural Language Processing, Persuasion.

Outside of work, I enjoy hiking, playing board games, and being a psephology geek.

Selected Publications

  1. platform.jpg
    On the conversational persuasiveness of GPT-4
    Francesco Salvi, Manoel Horta Ribeiro, Riccardo Gallotti, and Robert West
    Nature Human Behaviour

    Media Coverage:  The Guardian   The Washington Post   The Economist   Nature   Il Corriere della Sera   El País   Público   Frankfurter Allgemeine Zeitung   ABC   BresciaOggi   New Scientist   MIT Technology Review   Ansa   Der Spiegel   SWI   The Register   Vanity Fair   Veja   Gizmodo   La Vanguardia   Folha de S.Paulo  

  2. meditron.jpg
    MEDITRON-70B: Scaling Medical Pretraining for Large Language Models
    Zeming Chen, Alejandro Hernández-Cano, Angelika Romanou, Antoine Bonnet, Kyle Matoba, Francesco Salvi, and 14 more authors
    arXiv preprint, Nov 2023

    Media Coverage:  EPFL   Meta   TechXplore  

  3. lobby_heatmap_v2.jpg
    Studying Lobby Influence in the European Parliament
    Aswin Suresh, Lazar Radojevic*, Francesco Salvi*, Antoine Magron, Victor Kristof, and Matthias Grossglauser
    Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL 2024)