Detoxification for LLM: From Dataset Itself
For LLM practitioners, this provides a source-level detoxification method that reduces inherent toxicity without sacrificing data utility, potentially lowering the cost of downstream behavior correction.
The authors propose HSPD, a pipeline that detoxifies raw pretraining corpora by rewriting toxic spans while preserving semantics, achieving state-of-the-art detoxification on GPT2-XL (TP from 0.42 to 0.18, EMT from 0.43 to 0.20) and consistent improvements on LLaMA2-7B, OPT-6.7B, and Falcon-7B.
Existing detoxification methods for large language models mainly focus on post-training stage or inference time, while few tackle the source of toxicity, namely, the dataset itself. Such training-based or controllable decoding approaches cannot completely suppress the model's inherent toxicity, whereas detoxifying the pretraining dataset can fundamentally reduce the toxicity that the model learns during training. Hence, we attempt to detoxify directly on raw corpora with SoCD (Soft Contrastive Decoding), which guides an LLM to localize and rewrite toxic spans in raw data while preserving semantics, in our proposed HSPD (Hierarchical Semantic-Preserving Detoxification) pipeline, yielding a detoxified corpus that can drop-in replace the original for fine-tuning or other training. On GPT2-XL, HSPD attains state-of-the-art detoxification, reducing Toxicity Probability (TP) from 0.42 to 0.18 and Expected Maximum Toxicity (EMT) from 0.43 to 0.20. We further validate consistent best-in-class results on LLaMA2-7B, OPT-6.7B, and Falcon-7B. These findings show that semantics-preserving, corpus-level rewriting with HSPD effectively suppresses downstream toxicity while retaining data utility and allowing seamless source-level mitigation, thereby reducing the cost of later model behavior adjustment. (Code is available at: https://github.com/ntsw2001/data_detox_for_llm)