← back to paper
arxiv: 2605.02196 · 2 revisions
DurableUn: Quantization-Induced Recovery Attacks in Machine Unlearning