Eugene Berta

I am a third year PhD student in the SIERRA team at INRIA Paris and École Normale Supérieure. I am advised by Francis Bach and Michael I. Jordan.

Prior to that, I graduated from Télécom Paris and got my Master's degrees in Mathematics, Vision and Learning at Université Paris Saclay.

Email  /  Scholar  /  Github  /  LinkedIn

While machine learning systems can be incredibly powerful, it is of crucial interest to understand how confident we can be in their outputs. My work focuses on providing statistical guarantees for these predictions. In particular, I do research on classifier calibration, proper scoring rules, and conformal prediction.

I am developing the probmetrics Python package. The package provides efficient implementations for various post-hoc calibration methods, as well as classification metrics, especially metrics for assessing the quality of probabilistic predictions.

I co-organised the workshop Towards Trustworthy Predictions: Theory and Applications of Calibration for Modern AI at AISTATS 2026. The event was held on May 5, 2026 in Tangier, Morocco. A list of accepted papers can be found on the workshop website or on OpenReview.

I am helping organize the ELLIS UnConference 2026. The event will be held on December 8, 2026 in Paris, France.

I regularly review for the AISTATS, ICML, and NeurIPS conferences.

2025-2026: Teaching assistant for the second year course "Mathématiques fondamentales" in the "Cycle Pluridisciplinaire d’Études Supérieures" (CPES) of Lycée Henri IV and PSL university.

Air travel is generally the largest part of the work-related carbon footprint of a researcher (by far!). For example, a single return flight from Paris to the US emits roughly 2 tonnes of CO₂-equivalent per passenger, which is about the yearly per-person budget compatible with the goals of the Paris agreement.

While I do not expect (nor wish) everybody to stop flying at once, I believe that we should prioritize longer research stays and drastically reduce flying for short trips.

As a scientific community, it is our responsibility to listen to what our colleagues are documenting and to lead by example.

Since starting my PhD, I have chosen not to fly. Despite this constraint, I managed to attend several international conferences and workshops, for instance travelling from Paris to AISTATS in Valencia (2024) and in Tangier (2026) by train and ferry. While the journeys are longer, they are great for reading, spending time with other motivated colleagues, and realizing how far you are going.

I am very grateful to the researchers who organise local editions of major conferences, such as the NeurIPS satellite in Paris this year. This makes it much easier for everyone to connect with the community while reducing distances travelled.

For a broader analysis of the problem and concrete solution proposals I recommend having a look at Pierre Marion's website.