J. V. van Rijn

Does Target Class Matter? A Reproducibility Study of Plausible Counterfactual Explanations for Image Classification

Exploring Solver-Level Warmstarting for Neural Network Verification

Neural network verification has become a key tool for providing formal guarantees on the behaviour of neural networks. However, many verification problems remain computationally intractable in the worst case: even for common adversarial robustness specifications, verification is NP-complete. Here, we explore the applic...

Investigating Local Robustness of TabPFN on Small Numerical Binary Classification Tasks

It is discovered that TabPFN is more robust than all baselines on synthetic data under single-step FGSM attacks for moderate-to-large perturbation budgets, but that this advantage largely disappears under iterative PGD, suggesting that TabPFN’s gradient landscape obstructs single-step attacks.

Evolving Executable Pipeline Programs for AutoML with Language Models

This work presents LACE, an AutoML framework that instead searches over complete executable pipeline programs: an evolutionary loop maintains a population of scikit-learn-compatible Python classes, and a large language model acts as the variation operator.