arXiv:2605.22737v3 Announce Type: replace 
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 high-value examples, and a teacher-side defense template that suppresses outputs most useful for distillation. From a cheap proxy for example value, we derive Product-of-Experts (PoE), a simple forward-pass-only defense that combines the teacher with a proxy student during generation. Empirically, adaptive evaluation reveals a large passive--adaptive gap: on state-of-the-art defenses, adaptive students recover substantially more capability than passive evaluation suggests on GSM8K and MATH. Under this stronger evaluation, the apparent robustness gap between expensive defenses and PoE narrows considerably, while PoE remains substantially cheaper and preserves higher-quality reasoning traces. Overall, our results suggest that strong distillation remains difficult to stop, and that progress on antidistillation should be judged against adaptive students rather than passive ones. Our code is available at: https://github.com/ysfalh/distillation-game.

تدرس الدراسة المعنونة 'The Distillation Game: Adaptive Attacks & Efficient Defenses' التوازن الذي يواجهه مقدمو النماذج بين تعزيز فائدة النموذج وخطر التقليد من خلال هجمات التقطير. تقدم البحث إطار عمل لعبة مينيمكس يتضمن معلمًا مقيدًا بفائدة وطالبًا تكيفيًا، مما يؤدي إلى استراتيجيات دفاع فعالة ضد هذه الهجمات.

El estudio titulado 'The Distillation Game: Adaptive Attacks & Efficient Defenses' explora el compromiso que enfrentan los proveedores de modelos entre mejorar la utilidad del modelo y el riesgo de imitación a través de ataques de destilación. La investigación introduce un marco de juego minimax que involucra a un maestro restringido por la utilidad y a un estudiante adaptativo, lo que lleva a estrategias de defensa efectivas contra tales ataques.

L'étude intitulée 'The Distillation Game: Adaptive Attacks & Efficient Defenses' examine le compromis auquel sont confrontés les fournisseurs de modèles entre l'amélioration de l'utilité du modèle et le risque d'imitation par le biais d'attaques de distillation. La recherche introduit un cadre de jeu minimax impliquant un enseignant contraint par l'utilité et un étudiant adaptatif, menant à des stratégies de défense efficaces contre de telles attaques.

The study titled 'The Distillation Game: Adaptive Attacks & Efficient Defenses' explores the trade-off faced by model providers between enhancing model utility and the risk of imitation through distillation attacks. The research introduces a minimax game framework involving a utility-constrained teacher and an adaptive student, leading to effective defense strategies against such attacks.