27.10.2026
Runner-Up Prize at AutoML 2026 for Tabular-DC-Bench
A Meta-Dataset for Hyperparameter Optimization in Deep Clustering on Tabular Data
At the International Conference on Automated Machine Learning (AutoML 2026), the paper “Tabular-DC-Bench” received the runner-up prize in the ABCD track (Applications, Benchmarks, Challenges, and Datasets). The work was presented by MCML Junior Member Mamdouh Aljoud. It was co-authored by MCML Junior Member Gabriel Tavares, Sandra Gilhuber, Collin Leiber, and MCML Director Thomas Seidl.
Tabular-DC-Bench is a meta-dataset for hyperparameter optimization in deep clustering on tabular data. It comprises 375,000 runs across 100 datasets and five deep clustering algorithms. The results show that a deep clustering method with its default configuration often performs worse than K-Means, while the best configuration usually outperforms it. At the same time, internal metrics such as the Silhouette score cannot reliably indicate which configuration has been found. The meta-dataset is publicly available on Zenodo and GitHub, so others can build on it.
Congratulations to the authors on this recognition!
Related
12.11.2026
One Toolbox for Many Kinds of Uncertainty
MCML researchers introduce probly, a unified Python package for uncertainty-aware machine learning, presented orally at NeurIPS 2026.
10.11.2026
Enkelejda Kasneci: The Real Value of AI in Education Is More Time for Humans
Enkelejda Kasneci explores how AI can provide personalized learning support while keeping human interaction and student agency at the center.
09.11.2026
Outstanding Paper Award at EMNLP 2026 for Work on Accent Bias in Audio Language Models
A paper co-authored by MCML PI Valentin Hofmann on accent bias in audio language models won an Outstanding Paper Award at EMNLP 2026.