Scholar
Sebastian Bordt
Google Scholar ID: 6PnL3BgAAAAJ
University of Tübingen
Citations & Impact
Citations
426
H-index
10
i10-index
10
Publications
17
Co-authors
12
list available
Publications
Resume (English only)
Academic Achievements
- NeurIPS 2025 Spotlight paper: 'On the Surprising Effectiveness of Large Learning Rates under Standard Width Scaling'
- Two papers accepted at ICML 2025: 'How Much Can We Forget about Data Contamination?' and 'Rethinking Explainable Machine Learning as Applied Statistics'
- COLM 2024 paper: 'Elephants Never Forget: Memorization and Learning of Tabular Data in Large Language Models'
- NeurIPS 2023 Spotlight paper: 'Which Models have Perceptually-Aligned Gradients? An Explanation via Off-Manifold Robustness'
- AISTATS 2023 paper: 'From Shapley Values to Generalized Additive Models and back'
- Preprint (Sep 2025): 'Train Once, Answer All: Many Pretraining Experiments for the Cost of One'
- Preprint (Aug 2025): 'Informative Post-Hoc Explanations Only Exist for Simple Functions'
- Invited talk at Banff Workshop 'New Directions in Machine Learning Theory' (Oct 2024)
Background
- Postdoctoral researcher in the theory of machine learning group at the University of Tübingen
- Research interests include large language models (LLMs) and interpretability
- Currently working on a systematic understanding of pre-training
- Previously conducted black-box evaluations of LLMs
- At Microsoft Research, ran experiments combining GPT-4 and Generalized Additive Models (GAMs) in healthcare settings
- During PhD, explored various topics in explainable machine learning, including connections between post-hoc explanation methods and inherently interpretable models, and the regulatory suitability of explanation algorithms
