Adrian Sauter

Hey, I’m Adrian, welcome to my website. :)

I’m a third-year MSc student in Artificial Intelligence at the University of Amsterdam. I am broadly interested in the interpretability, safety, and alignment of AI and believe that these topics are crucial for developing trustworthy and beneficial AI systems.

Most recently, I worked on a project exploring how visual grounding affects both speech- and text-based language models (submitted to ICASSP 2026, preprint on arXiv), and completed an internship at KPN, where I applied concept bottleneck models with LLMs to customer call transcripts to develop an explainable-by-design framework for identifying key drivers of customer churn. I am now excited to be pursuing my thesis on “Contextual Sensitivity in Moral Judgements of Large Language Models”.

Before my MSc, I completed a BSc in Cognitive Science at the University of Tübingen, where I explored the intersection of human cognition and intelligent systems and wrote my thesis on “Video Background Extraction with a Masked-Autoencoder-Based Neural Network”, which contributed to a publication at ICANN 2024.

Feel free to reach out if you’d like to discuss AI, cognitive science, or any exciting interdisciplinary projects. I’m always open to new opportunities and collaborations! You can reach me at adrian dot a dot sauter at student dot uva dot nl or adriansauter07 dot as at gmail dot com.

news

Sep 01, 2025	Today, I started the third year of my AI Master’s at the University of Amsterdam. Over the next six months, I’ll be working on my thesis, “Contextual Sensitivity in Moral Judgements of Large Language Models”, while continuing my role as a teaching assistant for first-year master’s students.
Feb 03, 2025	I recently started a role as an XAI intern at KPN in Amsterdam. Over the next six months, I’ll be researching Explainable Customer Call Classification, applying Concept Bottleneck Models to text data with the use of LLMs.
Oct 10, 2024	I’m excited to announce that my co-authors and I will be presenting our recent paper, ““Studying How to Efficiently and Effectively Guide Models with Explanations” - A Reproducibility Study”, at NeurIPS in Vancouver this December.
Sep 13, 2024	Our paper ““Studying How to Efficiently and Effectively Guide Models with Explanations” - A Reproducibility Study” has been accepted at the ML Reproducibility Challenge 2023. You can find it here on OpenReview.
Sep 02, 2024	I began my second year in the MSc Artificial Intelligence program at the University of Amsterdam and took up a position as a Teaching Assistant for the courses Computer Vision 1 and Natural Language Processing 1, taught to first-year MSc AI students.

selected publications

Loci-Segmented: Improving Scene Segmentation Learning

Manuel Traub, Frederic Becker, Adrian Sauter, and 2 more authors

arXiv preprint arXiv:2310.10410, 2023

Abs Bib PDF

Current slot-oriented approaches for compositional scene segmentation from images and videos rely on provided background information or slot assignments. We present a segmented location and identity tracking system, Loci-Segmented (Loci-s), which does not require either of this information. It learns to dynamically segment scenes into interpretable background and slotbased object encodings, separating rgb, mask, location, and depth information for each. The results reveal largely superior video decomposition performance in the MOVi datasets and in another established dataset collection targeting scene segmentation. The system’s well-interpretable, compositional latent encodings may serve as a foundation model for downstream tasks.
@article{traub2023loci, title = {Loci-Segmented: Improving Scene Segmentation Learning}, author = {Traub, Manuel and Becker, Frederic and Sauter, Adrian and Otte, Sebastian and Butz, Martin V}, journal = {arXiv preprint arXiv:2310.10410}, year = {2023}, }
“Studying How to Efficiently and Effectively Guide Models with Explanations” - A Reproducibility Study

Adrian Sauter, Milan Miletić, Ryan Ott, and 1 more author

Transactions on Machine Learning Research, 2024

Abs Bib PDF Poster

Model guidance describes the approach of regularizing the explanations of a deep neural network model towards highlighting the correct features to ensure that the model is “right for the right reasons”. Rao et al. (2023) conducted an in-depth evaluation of effective and efficient model guidance for object classification across various loss functions, attributions methods, models, and ’guidance depths’ to study the effectiveness of different methods. Our work aims to (1) reproduce the main results obtained by Rao et al. (2023), and (2) propose several extensions to their research. We conclude that the major part of the original work is reproducible, with certain minor exceptions, which we discuss in this paper. In our extended work, we point to an issue with the Energy Pointing Game (EPG) metric used for evaluation and propose an extension for increasing its robustness. In addition, we observe the EPG metric’s predisposition towards favoring larger bounding boxes, a bias we address by incorporating a corrective penalty term into the original Energy loss function. Furthermore, we revisit the feasibility of using segmentation masks in light of the original study’s finding that minimal annotated data can significantly boost model performance. Our findings suggests that Energy loss inherently guides models to on-object features without the requirement for segmentation masks. Finally, we explore the role of contextual information in object detection and, contrary to the assumption that focusing solely on object-specific features suffices for accurate classification, our findings suggest the importance of contextual cues in certain scenarios. Code available at: https://github.com/ryan-ott/model-guidance-reproducibility.
@article{sauter2024studying, title = {{\textquotedblleft}Studying How to Efficiently and Effectively Guide Models with Explanations{\textquotedblright} - A Reproducibility Study}, author = {Sauter, Adrian and Miletić, Milan and Ott, Ryan and Prabakaran, Rohith Saai Pemmasani}, journal = {Transactions on Machine Learning Research}, issn = {2835-8856}, year = {2024}, url = {https://openreview.net/forum?id=9ZzASCVhDF}, }