前言:Facebook提出了一种NLP通用的攻打办法,而且能够通过梯度优化,论文发表在EMNLP2021,名为Gradient-based Adversarial Attacks against Text Transformers,源码在facebookresearch/text-adversarial-attack
这篇文章次要是对EMNLP2021上的论文Raise a Child in Large Language Model: Towards Effective and Generalizable Fine-tuning进行解说。论文题目有些形象,然而用作者的话来说,这篇论文的思维能够归结为两个词:Child Tuning