Vivian Yvonne Nastl

Max Planck Institute for Intelligent Systems, Tübingen, Germany and Eidgenössische Technische Hochschule, Zürich, Switzerland

prof_pic.jpg

Office: G 18

Rämistrasse 101

Zürich, Switzerland

Hi!

I’m a PhD student at the Max Planck ETH Center for Learning Systems, where I’m fortunate to be advised by Moritz Hardt and Nicolai Meinshausen. I’m part of the Social Foundations of Computation Department at MPI for Intelligent Systems (MPI-IS) in Tübingen, Germany, as well as the Seminar of Statistics in the Department of Mathematics at ETH Zürich.

My current research focus is on the responsible use of Machine Learning, encompassing the impact machine learning models have on research and society at large and steering this impact for good.

news

Jan 25, 2025 Our paper on “Limits to scalable evaluation at the frontier: LLM as Judge won’t beat twice the data” was accepted to ICLR 2024 as an oral. :confetti_ball:
Oct 17, 2024 Check out our new insights on “Limits to scalable evaluation at the frontier: LLM as Judge won’t beat twice the data” on arxiv. :sparkles:
Sep 25, 2024 Our paper on “Do causal predictors generalize better to new domains?” was accepted to NeurIPS 2024 as a spotlight. :tada:
Jul 27, 2024 I’m looking forward to presenting our work on causal predictors on the Workshop “Humans, Algorithmic Decision-Making and Society: Modeling Interactions and Impact” at ICML’24. 🎡
Jun 17, 2024 I’m excited to join the Work­shop on Al­go­rithms for Learn­ing and Eco­nom­ics on the beautiful island of Kefalonia and present our work on causal predictors. :whale:

selected publications

  1. Limits to scalable evaluation at the frontier: LLM as Judge won’t beat twice the data
    Florian E. Dorner, Vivian Y. Nastl, and Moritz Hardt
    2025
  2. Do causal predictors generalize better to new domains?
    Vivian Y. Nastl, and Moritz Hardt
    In The Thirty-eighth Annual Conference on Neural Information Processing Systems (NeurIPS), 2024
  3. Causal Inference from Competing Treatments
    Ana-Andreea Stoica, Vivian Y. Nastl, and Moritz Hardt
    In Proceedings of the 41st International Conference on Machine Learning (ICML), 2024