s-nlp/

pseudoparadetox_llama3_8b_10shot_patched

	Dataset Card for Dataset Name











	


	Dataset Details

Using this data, we got the highlighted results using BART sequence-to-sequence model. The configs and code for fine-tuning can be found on github

This is a PseudoParaDetox dataset with real source toxic data and generated neutral detoxification by a patched LLama 3 8B with 10-shot. This dataset is based on the ParaDetox dataset for English texts detoxification.

Curated by:… See the full description on the dataset page: https://huggingface.co/datasets/s-nlp/pseudoparadetox_llama3_8b_10shot_patched.