Vivian Nastl wins 2025 MPI-IS Outstanding Female Doctoral Student Prize

Tübingen – Vivian Nastl has won the 2025 Outstanding Female Doctoral Student Prize, an award initiated in 2023 by the Max Planck Institute for Intelligent Systems (MPI-IS) to honor one outstanding female doctoral researcher each year for her scientific achievements and contributions to the research community. This year’s winner, Vivian Nastl, is co-supervised by Moritz Hardt, Director of the Social Foundations of Computation Department at MPI-IS, and by Nicolai Meinshausen, a professor of statistics at ETH Zurich. In addition, three other doctoral researchers received honorable mentions for this prize: Natalia Sanchez-Tamayo, Melis Bal, and Maria-Paola Forte.

“I am truly honored to receive the MPI-IS Outstanding Female Doctoral Student Prize, and it motivates me as I continue on my research journey,” said Nastl. “I am deeply grateful for the invaluable mentorship of my advisors, Moritz Hardt and Nicolai Meinshausen, who have greatly influenced and shaped the path of my research. I feel fortunate to be part of a community of warm, supportive colleagues whose collaborative spirit fosters curiosity, rigorous inquiry, and shared progress.”

Nastl is part of the Max Planck ETH Center for Learning Systems (CLS) doctoral program, a joint academic program between ETH Zurich and the Max Planck Society. With a background in financial mathematics, she studies statistical methods for applied machine learning, with a focus on causal inference and evaluation. 

Nastl’s work showcases a deep understanding of the theory of causality and its practical applications. In her first-author work published at NeurIPS 2024, “Do Causal Predictors Generalize Better to New Domains?”, she studied a recent hypothesis stating that causal features improve domain generalization. Vivian showed that, across many models, datasets and domains, models trained on all features (regardless of causal relationships) generalize better on new domains than models trained on only causal features. Her work blends cutting-edge techniques such as causal discovery algorithms and state-of-the-art deep learning models in a principled and extensive experimental design.  

In her recent work, “Limits to Scalable Evaluation at The Frontier: LLM as Judge Won’t Beat Twice the Data”, Vivian Nastl studies the statistical limits and promises of annotations – increasingly a bottleneck in the evaluation of large language models (LLMs). Researchers try to use models as judges (LLM as judge) to evaluate other models. Unfortunately, LLM judges have numerous biases that limit their success in practice. Recent methods promise to debias model evaluations from a small number of human ground-truth evaluations. In a remarkable result with Florian Dorner, Nastl showed that any such method can never be better than using twice as many human evaluations. Awarded an oral presentation at ICLR 2025, the result is as surprising as it is timely. 

In another paper titled “Causal Inference from Competing Treatments”, Vivian Nastl and her collaborator Ana Stoica study a common, yet overlooked issue when applying causal inference on digital platforms. At any point in time, multiple experimenters will be working with the same candidate pool. An example is that of a group of advertisers trying to estimate the effectiveness of their campaigns. Due to the competition between ads (i.e., treatments) on screen, the treatment choices of one experimenter compete with those of the others. Stoica and Nastl characterize the optimal causal inference strategy at equilibrium when experimenters act strategically. The work addresses an important problem with experimentation on digital platforms, relevant to any team working on A/B testing for online services. This work weaves together tools from economics and causality, bringing game-theoretical concepts into the standard statistical inference toolkit. “Working with Vivian opened new pathways at the intersection of Economics and Machine Learning. Her deep understanding of both domains showed up in every step of our project”, said Ana Stoica.  

The MPI-IS Outstanding Female Doctoral Student Prize was introduced as part of our institute’s first Gender Equality Plan (GEP) and continues under the second GEP (2024–2026), aiming to increase gender equality in both MPI-IS sites, Stuttgart and Tübingen. The award also aims to inspire the research community to look for potentially unrecognized excellence among the institute’s female doctoral students. The winner receives up to €2,000 to support career-building activities of her choice, such as attending a workshop or conference. An external scientific selection committee determines the winner and the honorable mentions, taking particular care to avoid conflicts of interest with the nominees and their advisors. The 2025 selection committee consisted of:

•    Georgia Chalvatzaki, Professor of Interactive Robot Perception and Learning at Technical University of Darmstadt
•    Asja Fischer, Professor of Machine Learning at Ruhr-University Bochum
•    Josie Hughes, Assistant Professor of Computational Robot Design and Fabrication at Ecole polytechnique fédérale de Lausanne (EPFL)
•    Hilde Kühne, Professor for Multimodal Learning at the Tübingen AI Center and affiliated professor at the MIT-IBM Watson AI Lab

At the MPI-IS Special Symposium and Summer Party on June 27, 2025, the winner and three honorable mentions were announced by the award organizing team, led by Katherine J. Kuchenbecker, Director of the Haptic Intelligence Department, and including Birgül Akolpoglu, Florian Hartmann, and Wieland Brendel.

“We created this prize to honor the achievements of our amazing women doctoral researchers, and to give them career opportunities that they might not otherwise be able to access,” Katherine J. Kuchenbecker explained. “Our institute is pursuing world-leading research in the fields of robotics and AI, which are both traditionally quite male-dominated. Many prestigious and financially rewarding prizes in academia are dominated by male nominees and recipients, which can reduce fairness, recognition, and career opportunities for female scientists. Indeed, research has shown that advisors of all genders tend to think first of nominating men for honors. Thus, this prize aims to raise awareness about potential biases in the nomination and selection of awardees and to increase support for female scientists early in their careers.”

“The groundbreaking research at our institute continually inspires me, and this award highlights the remarkable contributions that female doctoral researchers make to advancing this work,” Vivian Nastl continued. “I sincerely thank the award organizers and the selection committee for creating and maintaining this important platform for female talents to push and promote. I am proud to stand alongside the exceptional recipients of honorable mentions – Paola, Melis, and Natalia – and congratulate them on their outstanding achievements.”

Nastl grew up near Stuttgart and studied financial mathematics at the University of Konstanz. During her master's, she shifted her focus to statistics, benefiting from the interdisciplinary course of study in mathematics and economics, and she also earned a second bachelor's degree in mathematics.

The recipients of the three honorable mentions awarded in 2025 are:

Melis Ilayda Bal, a doctoral student in the Learning and Dynamical Systems Group in Tübingen, led by Michael Mühlebach. 
Melis is a highly talented and dedicated researcher with a track record of excellence. Her work stands out for its creativity and interdisciplinary breadth, spanning operations research, Bayesian optimization for protein design, and the robust training of large-scale machine learning models. For example, in one of her ICLR 2025 papers, Melis developed a novel adversarial training method to improve robustness against label poisoning, a scenario where training labels are maliciously corrupted. Her most recent work, currently under review at NeurIPS 2025, builds on this foundation, introducing a method for selective token-level pretraining of large language models that improves both computational efficiency and downstream performance. Melis joined the CS@maxplanck doctoral program after ranking first in her master’s program in Operations Research at Middle East Technical University in Turkey. She has conducted research at ETH Zurich and MPI-SWS and is currently an Amazon–Max Planck Science Hub Fellow and a doctoral student at EPFL.

Maria-Paola Forte, an IMPRS-IS doctoral student in the Haptic Intelligence (HI) and Perceiving Systems (PS) Departments, co-advised by Katherine J. Kuchenbecker and Michael J. Black. 
One line of Paola Forte’s research blends modern computer vision of humans with novel wearable sensing for the worthy and difficult application of capturing sign language. In a first-author paper published at CVPR 2023, Paola showed SGNify, her powerful method for reconstructing highly realistic signing avatars from low-resolution frontal videos, which are widespread online. Forte educated herself about the linguistics of sign language, i.e., the rules that govern hand movements and shapes across all sign languages. She read deeply to be able to formulate specific rules such as left- and right-hand poses being identical or a hand pose changing linearly over time, both of which greatly increase the number of pixels that can be used to infer hand pose for a particular frame. Forte also takes part in the EXIST Women program, which supports and trains promising female scientists who are interested in entrepreneurship. Through this program, she is exploring commercial pathways to transform her research into products that enhance accessibility and real-world impact. Forte received her B.Sc. in Biomedical Engineering from the University of Genova, Italy, followed by an M.Sc. in Bioengineering Technologies for Electronics from Politecnico di Milano. She conducted her master’s thesis in the HI Department starting in late 2017. After completing her degree, she worked in the same department as a research engineer for 1.5 years before beginning her doctorate at the end of 2019.

Natalia Sanchez-Tamayo, an IMPRS-IS doctoral student co-advised by Katherine J. Kuchenbecker, Director of the Haptic Intelligence Department, and Christoph Keplinger, Director of the Robotic Materials Department.

Natalia Sanchez-Tamayo specializes in soft actuators and sensors for haptic feedback and robotic applications. Her current projects focus on designing versatile wearable haptic devices using electrohydraulic actuation, studying user perception of tactile feedback, and developing machine learning-enhanced soft tactile sensors for robotic manipulation. One of her key contributions is leading the development of Cutaneous Electrohydraulic (CUTE) devices—compact, wrist-worn devices that provide rich, expressive tactile sensations that go far beyond the buzzing of typical consumer devices. Each device is powered by a stack of ten electrohydraulic actuators: oil-filled pouches that expand when voltage is applied, pressing against the skin with up to several newtons of force. These devices enable a wide range of tactile experiences, from gentle presses to calming touch and high-frequency vibrations. Sanchez-Tamayo conducted a user study demonstrating approximately 98% haptic cue recognition, with participants rating nearly all haptic sensations as pleasant. In fall 2024, she was a Research Scientist Intern at Meta. Prior to joining MPI-IS, she completed a master’s degree in Industrial Engineering at Purdue University in 2020, after earning dual bachelor’s degrees in Mechanical Engineering and Civil Engineering from Universidad de Los Andes in Colombia.