perturbations专题

【读论文】Learning perturbations to explain time series predictions

文章目录 Abstract1. Introduction2. Background Work3. Method4. Experiments4.1 Hidden Markov model experiment4.2 MIMIC-III experiment 5. ConclusionReferences 论文地址:Learning Perturbations to Explain

Universal adversarial perturbations(2017 CVPR)

Universal adversarial perturbations----《普遍对抗扰动》   通俗UAP算法步骤理解:对于 x i ∈ X {x_i} \in X xi​∈X 的每个采样数据点,比较 k ^ ( x i + v ) \hat k({x_i} + v) k^(xi​+v) 与 k ^ ( x i ) \hat k({x_i}) k^(xi​) ,如果 k ^ ( x

Are Large Language Models Really Robust to Word-Level Perturbations?

本文是LLM系列文章,针对《Are Large Language Models Really Robust to Word-Level Perturbations?》的翻译。 大型语言模型真的对单词级扰动具有鲁棒性吗? 摘要1 引言2 相关工作3 合理稳健性评价的奖励模型(TREvaL)4 LLM的词级鲁棒性评价5 讨论6 结论7 局限性 摘要 大型语言模型(LLM)的规模和功能