PulseAugur
实时 07:41:28
English(EN) S-GBT: Smooth Growth Bound Tensor for Certified Robustness Against Word Substitution Attacks in NLP

新的 S-GBT 方法提高了 NLP 模型对词语攻击的鲁棒性

研究人员开发了一种名为平滑增长界张量 (S-GBT) 的新方法,以增强自然语言处理 (NLP) 模型在面对词语替换攻击时的鲁棒性。与以往关注一阶敏感度的方法不同,S-GBT 通过逐元素约束 Hessian 来利用二阶信息。这种方法被整合到训练目标中,旨在同时控制梯度及其变化,从而获得更紧密的认证鲁棒性界限。在基准数据集上的评估表明,与现有技术相比,S-GBT 可将认证鲁棒准确率提高高达 23.4%,同时保持具有竞争力的干净准确率。 AI

影响 增强了 NLP 模型对对抗性攻击的韧性,有可能提高在敏感应用中的可靠性。

排序理由 该集群包含一篇详细介绍 NLP 模型鲁棒性新方法的论文。

在 arXiv cs.CL 阅读 →

AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →

报道来源 [2]

  1. arXiv cs.CL TIER_1 English(EN) · Mohammed Bouri, Mohammed Erradi, Adnane Saoud ·

    S-GBT: Smooth Growth Bound Tensor for Certified Robustness Against Word Substitution Attacks in NLP

    arXiv:2606.13439v1 Announce Type: new Abstract: Despite recent progress in Natural Language Processing (NLP), models remain vulnerable to word substitution attacks. Most existing defenses focus on first order sensitivity and measure how much the output changes when the input is s…

  2. arXiv cs.CL TIER_1 English(EN) · Adnane Saoud ·

    S-GBT: Smooth Growth Bound Tensor for Certified Robustness Against Word Substitution Attacks in NLP

    Despite recent progress in Natural Language Processing (NLP), models remain vulnerable to word substitution attacks. Most existing defenses focus on first order sensitivity and measure how much the output changes when the input is slightly perturbed. However, they ignore how this…