The Distillation Game: Adaptive Attacks & Efficient Defenses 事件

The Distillation Game: Adaptive Attacks & Efficient Defenses arXiv:2605.22737v2 Announce Type: replace-cross Abstract: Distillation attacks create a deployment trade-off for model providers: the same outputs that make a model more useful can also make it easier to imitate. We study this trade-off through a minimax game between a utility-constrained teacher and an adaptive student. Our framework yields tractable one-sided response rules: an adaptive evaluation rule in which the student reweights