Text detoxification has the potential to mitigate the harms of toxicity by
rephrasing text to remove offensive meaning, but subtle toxicity remains
challenging to tackle. We introduce MaRCo, a detoxification algorithm that
combines controllable generation and text rewriting methods using a Product of
Experts with autoencoder language models (LMs). MaRCo uses likelihoods under a
non-toxic LM (expert) and a toxic LM (anti-expert) to find candidate words to
mask and potentially replace. We evaluate our method on several subtle toxicity
and microaggressions datasets, and show that it not only outperforms baselines
on automatic metrics, but MaRCo's rewrites are preferred 2.1 $\times$ more in
human evaluation. Its applicability to instances of subtle toxicity is
especially promising, demonstrating a path forward for addressing increasingly
elusive online hate.

介绍了 MaRCo 算法，该算法采用 Product of Experts 和自编码语言模型相结合的方式，通过在非有毒和有毒语言模型下评估可能要屏蔽和替换的候选单词，能够有效地解决细微的毒性挑战，即使在人类评估中也要比基线方法更受欢迎，证明了其在解决日益逃避的网络仇恨问题上具有广泛的应用前景。