安全干预时机影响模型鲁棒性，早介入更利安全推理，晚介入更强泛化。
When Should We Introduce Safety Interventions During Pretraining?
- 调整安全干预启动时间（0%、20%、60%）测试效果差异。
- 20%-60%阶段后干预，在下游微调后提升最明显。
- 早期干预更利于安全生成控制，重塑模型内部表征。
已有研究证明，预训练中引入安全干预（如删除和重写有害内容）可显著提升模型鲁棒性。本文首次探讨被忽视的核心问题：‘安全干预应在预训练何时引入？’ 在固定数据源与干预方式的前提下，我们测试了三种启动时机（0%、20%、60%预训练词数后）。结果表明，最优时机并非通用：使用标准top-k解码时，经过短时纯安全预训练（20%-60%）后再引入干预，能带来最强鲁棒性，优势在下游良性微调后显现；而针对安全感知推理，从一开始就引入干预则更利于生成安全输出。此外，早期干预会重塑内部表示：线性探测显示安全与有害样本的分离更清晰。本研究首次将干预时机确立为安全预训练中的关键课程设计要素。
原文摘要 · Abstract (English)
Prior work has shown that safety interventions applied during pretraining, such as removing and rephrasing harmful content, can substantially improve the robustness of the resulting models. In this paper, we study the fundamental question that prior work has overlooked: "When during pretraining should safety interventions be introduced?" We keep the underlying data sources and pretraining interventions fixed, varying the intervention start time (after 0%, 20%, or 60% of pretraining tokens). We find that the optimal start time is not one-size-fits-all: with standard top-k decoding, introducing interventions after a short initial phase of safe-only pretraining (20%-60%) often yields the strongest robustness, with the clearest benefits emerging after downstream, benign finetuning. In contrast, for safety-aware inference, interventions starting from the beginning improve steerability towards safer generations. Finally, we observe that earlier interventions reshape internal representations: linear probes more cleanly separate safe vs harmful examples. Our results are the first to establish intervention timing as a key curriculum design choice for safety.
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。