Joris Baan

I am an ELLIS PhD candidate at the University of Amsterdam working with Wilker Aziz, Barbara Plank and Raquel Fernández.

I am interested in understanding how large language models represent uncertainty, and if we can disentangle uncertainty that arises from the data (e.g., underspecified, subjective or ambiguous language) from uncertainty related to a model (e.g., lack of knowledge or reasoning). Ultimately, I think that uncertainty should power better decision-making.

I am part of the Dialogue Modeling Group in Amsterdam and MaiNLP at LMU in Munich. Previously, I interned at the Allen Institute for Artificial Intelligence in Seattle and spent 2.5 years as ML engineer at DPG Media working on large scale news recommendation.

News

• Jan 2025 - New 2-part blogpost series about building and deploying an AI app from scratch (see below).
• March 2024 - Check our new EACL-main position paper being presented in Malta: “Interpreting Predictive Probabilities: Model Confidence or Human Label Variation?”
• December 2023 - I presented our “What Comes Next” paper as an oral talk at EMNLP in Singapore.
• November 2023 - I am teaching assistant for two master AI courses at the UvA: NLP1 and DL1.
• October 2023 - We’re organising the First Workshop on Uncertainty-aware NLP (UncertaiNLP), co-located with EACL in Malta! Check out the website at https://uncertainlp.github.io/ and consider submitting your work!
• October 2023 - One paper accepted at EMNLP (main conference): camera ready coming soon!
• August 2023 - New preprint! “Uncertainty in Natural Language Generation: From Theory to Applications”
• July 2023 - Super excited to join AllenAI in Seattle as a research intern to work with Sarah Wiegreffe!
• May 2023 - New preprint! “What Comes Next? Evaluating Uncertainty in Neural Text Generators Against Human Production Variability”

Blogposts

Occasionally I write blogposts. You can find them below.