Full Professor, Ludwig-Maximilians-Universität München
4 papers at NeurIPS 2025
We extend Quantification Learning to graphs and propose two new structure-based extensions to the Adjusted Count approach.
ResponseRank enables data-efficient learning of distance-aware reward models through stratified comparison strength rankings.
This paper proposes a principled method to construct credal sets using relative likelihood and ensemble learning, achieving improved uncertainty representation without sacrificing predictive accuracy.
We introduce faithful interaction explanations of CLIP and SigLIP models (FIxLIP), offering a unique perspective on interpreting image–text similarity predictions.
