A new Chinese detoxification dataset and 17-model benchmark show that LLMs can remove toxic words but often distort emotional tone, especially for emoji, homophone, and dialogue-based toxicity.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Chinese Toxic Language Mitigation via Sentiment Polarity Consistent Rewrites
A new Chinese detoxification dataset and 17-model benchmark show that LLMs can remove toxic words but often distort emotional tone, especially for emoji, homophone, and dialogue-based toxicity.