通过轨迹控制边界实现持续遗忘，防止大模型在推理时重新激活受限知识。
ST$^2$U: Stateful Test-Time Unlearning via Restricted Knowledge Boundary Control
- 基于低维可逆坐标建模受限知识边界，动态修正生成过程中的隐藏状态。
- 实验显示比基线减少46.5%-59.1%的受限知识重入，且保持良好非目标能力。
- 适合需安全部署、避免重复训练的LLM应用场景，如内容过滤与合规生成。
控制大语言模型中的受限知识对模型对齐与安全部署至关重要。测试时遗忘无需代价高昂的再训练和参数更新，仅在推理阶段干预。然而，现有激活编辑方法仅进行孤立的点状修正，忽略了自回归生成中隐藏状态持续从提示、缓存和已生成前缀重构的过程。因此，即使局部修正成功，后续状态仍可能返回受限知识区域，导致知识重入。本文提出状态感知的测试时遗忘方法——ST²U，将测试时遗忘建模为轨迹级边界控制。ST²U首先在低维可逆坐标中建模受限知识边界，同时保持正交非目标分量不变。推理过程中，监控轨迹风险，以上下文锚定方式施加最小边界修正，并跨标记传播历史修正状态，缓解知识重入问题。该轨迹级控制实现了更持久的遗忘，同时保留非目标能力并限制推理开销。在三个基准和三个模型族上，ST²U表现出最佳综合平衡：保留性能达最优或第二优，遗忘效果具有竞争力，且受限知识重入率显著低于测试时基线（13.76%-19.84% 对比 46.50%-59.10%）。
原文摘要 · Abstract (English)
Controlling restricted knowledge in large language models is essential for model alignment and safe deployment. Test-time unlearning avoids costly retraining and parameter updates by intervening only during inference. However, existing activation-editing methods apply isolated pointwise corrections, overlooking how autoregressive generation continually reconstructs hidden states from the prompt, cache, and generated prefix. Consequently, later states may return to restricted knowledge regions after a locally successful correction, causing restricted knowledge re-entry. In this work, we propose Stateful Test-Time Unlearning via restricted knowledge boundary control (ST$^2$U), which formulates test-time unlearning as trajectory-wide boundary control. ST$^2$U first models restricted knowledge boundaries in low-dimensional invertible coordinates while leaving orthogonal non-target components unchanged. During inference, ST$^2$U monitors risk along the trajectory, applies minimal boundary corrections with contextual anchoring, and propagates historical correction states across tokens to mitigate knowledge re-entry. This trajectory-wide control enables more persistent forgetting while preserving non-target capabilities and limiting inference overhead. Across three benchmarks and three model families, ST$^2$U delivers the strongest overall balance, combining best or second-best retention with competitive forgetting and substantially less restricted-knowledge re-entry than test-time baselines (13.76%-19.84% versus 46.50%-59.10%).
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。