Can I read Detoxifying Language Models: Challenges on EtoBox?

Detoxifying Language Models: Challenges by 328697760 is a document available to read on EtoBox.

What is Detoxifying Language Models: Challenges about?

The document discusses the challenges of detoxifying language models (LMs) and evaluates various toxicity mitigation strategies. It highlights the trade-offs between reducing toxicity and maintaining model quality, particularly the negative impact on the representation of marginalized groups. The findings indicate that while automatic metrics can optimize toxicity scores, they may not align with human evaluations and can exacerbate social biases.

Author

328697760

Language

EN