吃了的,不经意又吐出来
训练数据提取攻击
困惑度:GPT-2模型的困惑度(perplexity)
Small:小型GPT-2模型和大型GPT-2模型的交叉熵比值
Medium:中型GPT-2模型和大型GPT-2模型的交叉熵比值
zlib:GPT-2困惑度(或交叉熵)和压缩算法熵(通过压缩文本计算)的比值
Lowercase:GPT-2模型在原始样本和小写字母样本上的困惑度比例
Window:在最大型GPT-2上,任意滑动窗口圈住的50个字能达到的最小困惑度
大语言模型全军覆没?
如果在数据集中进行任意的单次替换的影响足够小,那么查询结果就不能用来推断任何单个个体的信息,因此保证了隐私。
https://arxiv.org/abs/2012.07805
https://ai.googleblog.com/2020/12/privacy-considerations-in-large.html
https://venturebeat.com/2020/12/16/google-apple-and-others-show-large-language-models-trained-on-public-data-expose-personal-information/
https://www.reddit.com/r/MachineLearning/comments/ke01x4/r_extracting_training_data_from_large_language/
<pre><section style="letter-spacing: 0.544px;white-space: normal;font-family: -apple-system-font, system-ui, "Helvetica Neue", "PingFang SC", "Hiragino Sans GB", "Microsoft YaHei UI", "Microsoft YaHei", Arial, sans-serif;"><section powered-by="xiumi.us"><section style="margin-top: 15px;margin-bottom: 25px;opacity: 0.8;"><section><section style="letter-spacing: 0.544px;"><section powered-by="xiumi.us"><section style="margin-top: 15px;margin-bottom: 25px;opacity: 0.8;"><section><section style="margin-bottom: 15px;padding-right: 0em;padding-left: 0em;color: rgb(127, 127, 127);font-size: 12px;font-family: sans-serif;line-height: 25.5938px;letter-spacing: 3px;text-align: center;"><span style="color: rgb(0, 0, 0);"><strong><span style="font-size: 16px;font-family: 微软雅黑;caret-color: red;">为您推荐</span></strong></span></section><p style="margin: 5px 16px;padding-right: 0em;padding-left: 0em;font-family: sans-serif;letter-spacing: 0px;opacity: 0.8;line-height: normal;text-align: center;">一个算法工程师的日常是怎样的?</p><p style="margin: 5px 16px;padding-right: 0em;padding-left: 0em;font-family: sans-serif;letter-spacing: 0px;opacity: 0.8;line-height: normal;text-align: center;">彻底搞懂机器学习中的正则化<br /></p><section style="margin-top: 5px;margin-bottom: 5px;padding-right: 0em;padding-left: 0em;min-height: 1em;font-family: sans-serif;letter-spacing: 0px;opacity: 0.8;line-height: normal;text-align: center;"><span style="font-size: 14px;">13个算法工程师必须掌握的PyTorch Tricks</span></section><section style="margin-top: 5px;margin-bottom: 5px;padding-right: 0em;padding-left: 0em;min-height: 1em;font-family: sans-serif;letter-spacing: 0px;opacity: 0.8;line-height: normal;text-align: center;"><span style="font-size: 14px;">吴恩达上新:生成对抗网络(GAN)专项课程</span></section><section style="margin-top: 5px;margin-bottom: 5px;padding-right: 0em;padding-left: 0em;min-height: 1em;font-family: sans-serif;letter-spacing: 0px;opacity: 0.8;line-height: normal;text-align: center;">从SGD到NadaMax,十种优化算法原理及实现</section></section></section></section></section></section></section></section></section>
本篇文章来源于: 深度学习这件小事
本文为原创文章,版权归知行编程网所有,欢迎分享本文,转载请保留出处!
内容反馈