一项新的研究论文质疑了 Transformer 模型在网络入侵检测中的有效性，特别是在 CIC-IDS2017 数据集上。研究发现，评估方法，特别是填充约定和数据拆分，显著影响了报告的性能，常常高估了 Transformer 的能力。在没有填充的、现实的、无泄漏的条件下进行评估时，Transformer 的性能会大幅下降，这表明架构选择的重要性不如严格的评估实践。 AI
影响 强调了 AI 安全研究中标准化、无泄漏评估协议的关键需求，以准确评估模型能力。
排序理由 评估特定数据集上 Transformer 模型性能的研究论文。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
arXiv:2606.11098v1 Announce Type: cross Abstract: Recent deep learning approaches for network intrusion detection increasingly incorporate temporal architectures such as recurrent networks and Transformers, often reporting near-perfect performance on CIC-IDS2017. However, many ex…
Recent deep learning approaches for network intrusion detection increasingly incorporate temporal architectures such as recurrent networks and Transformers, often reporting near-perfect performance on CIC-IDS2017. However, many existing studies neither supply their temporal modul…
Recent deep learning approaches for network intrusion detection increasingly incorporate temporal architectures such as recurrent networks and Transformers, often reporting near-perfect performance on CIC-IDS2017. However, many existing studies neither supply their temporal modul…