About

Hi,

I’m a Postdoctoral Research Fellow at the MilaNLP group at Bocconi University. My research is generally about how language models behave as information intermediaries and how to embed democratic values in AI systems. More specifically, I investigate how political biases manifest in downstream tasks, and how they are encoded during the several training phases of LLMs and distributed in the training data. Lately, I’m also interested in evaluating LLMs for reliability, diversity, and safety in the context of information seeking.

Besides models themselves, I am also very interested in text analysis. I am co-leading the project MULTIVIEW where we investigate methods for capturing perspectives in texts, how users perceive framing, and how current framing detection methods generalize across topics.

In my free time, I like to bike 🚴 and free-dive 🤿.