Projects
Here is a list of projects I am currently interested in, I am open/looking for collaborations, do not hesitate to contact me !
- Under manipulations, are there models harder to audit? link
- Metrics between models in output space: a benchmark.
- Auditia reading group. link.
You can find a list of older projects here.
Research interests
- Algorithmic auditing of remote algorithms
- ML theory and auditing theory
- Distribution testing
Personal
When I am not working, I like to travel France and Europe by bike (it’s called biketouring !). I took the picture on the left in Grenoble during the 2020 winter.
News
[2023-12] WIDE @ EPFL
I will be presenting our results on the auditability of large models in the presence of platform manipulation to the IC school at EPFL on Dec. 9th at 11AM.
[2023-10] SaTML 2023: Under manipulations, are there models harder to audit?
We submitted our results on the auditability of large models in the presence of platform manipulation to SaTML 2023. Preprint on its way.
[2023-07] PFIA 2023: Change-Relaxed Active Fairness Auditing
I will be presenting an early version of our work on the black-box auditability of ML models at PFIA. You can find the paper here, more to come on that subject!
[2023-01] Journal publication of the reworked CNA 2021 study
Our work on the effect of collaborative filtering on the diversity of users' exposure has been accepted for publication in Applied Network Science ! More info here.
[2022-11] 🥈 2nd place at PeREN hackathon with Jade !
The « Segmente moi si tu peux ! » hackathon, was organized by the PeREN. The goal was to detect the type of pricing algorithm behind a delivery app by generating black-box queries to the app. We reached the second place, together with Jade Garcia-Bourrée and were awarded our price by the French Minister for Digital Transition and Telecommunications Jean-Noël Barrot.
[2022-11] Off to Rennes !
I am starting my PhD on "Auditing the mutations of online AI models" at WIDE !
[2022-04] NLE internship
Grenoble, here I come! Starting my masters final internship at NaverLabs Europe with Jean-Michel Renders on Diversity in Search and Recommendation systems.
[2021-11] CNA 2021
I will be presenting our work "Recommender systems increase exposure diversity. Or do they? A complex networks approach., Augustin Godinot, Fabien Tarissan" at the Complex Networks and their Applications conference ! Slides and info here.
[2021-05] Submitted our article to RecSys !
We submitted our Measuring the effect of collaborative filtering on the diversity of users’ attention article together with Fabien!