2 papers
cs.LG2026
Prevalence calibration as shortcut mitigation
Mohamed Amine Kina, Eike Petersen
Shortcut learning denotes the widespread situation in which a classifier exploits spurious correlations rather than diagnostic features. Existing mitigation strategies mostly aim t…
cs.LG2026
On the Failure of Boundary-Seeking Distillation in Bottlenecked Generative Architectures
Mohamed Amine Kina
Data-free knowledge distillation transfers the knowledge encoded in a teacher model to a student model without access to the original training data. Prior work such as Contrastive…