While Large Language Models (LLMs) have achieved remarkable capabilities, they unintentionally memorize sensitive data, posing critical privacy and security risks. Machine unlearning is pivotal for mitigating these risks, yet existing paradigms face a fundamental dilemma: aggressive unlearning often induces catastrophic forgetting that degrades model utility, whereas conservative strategies risk superficial forgetting, leaving models vulnerable to adversarial recovery. To address this trade-off, we propose $\textbf{AGT$^{AO}$}$ (Adversarial Gating Training with Adaptive Orthogonality), a unified framework designed to reconcile robust erasure with utility preservation. Specifically, our approach introduces $\textbf{Adaptive Orthogonality (AO)}$ to dynamically mitigate geometric gradient conflicts between forgetting and retention objectives, thereby minimizing unintended knowledge degradation. Concurrently, $\textbf{Adversarial Gating Training (AGT)}$ formulates unlearning as a latent-space min-max game, employing a curriculum-based gating mechanism to simulate and counter internal recovery attempts. Extensive experiments demonstrate that $\textbf{AGT$^{AO}$}$ achieves a superior trade-off between unlearning efficacy (KUR $\approx$ 0.01) and model utility (MMLU 58.30). Code is available at https://github.com/TiezMind/AGT-unlearning.

翻译：尽管大型语言模型（LLM）已展现出卓越的能力，但它们会无意中记忆敏感数据，从而带来严重的隐私和安全风险。机器遗忘对于缓解这些风险至关重要，然而现有范式面临一个根本性困境：激进的遗忘策略常常引发灾难性遗忘，导致模型效用下降；而保守的策略则可能导致表面遗忘，使模型容易受到对抗性恢复攻击。为解决这一权衡问题，我们提出了 **AGT$^{AO}$**（具有自适应正交性的对抗门控训练），这是一个旨在协调鲁棒擦除与效用保持的统一框架。具体而言，我们的方法引入了 **自适应正交性（AO）**，以动态缓解遗忘目标与保留目标之间的几何梯度冲突，从而最大限度地减少非预期的知识退化。同时，**对抗门控训练（AGT）** 将遗忘问题表述为一种潜在空间的最小-最大博弈，采用基于课程学习的门控机制来模拟并对抗内部的恢复尝试。大量实验表明，**AGT$^{AO}$** 在遗忘效能（KUR ≈ 0.01）与模型效用（MMLU 58.30）之间实现了更优的权衡。代码发布于 https://github.com/TiezMind/AGT-unlearning。