Detoxification for LLM: From Dataset Itself

• URL: http://arxiv.org/abs/2604.19124v1
• Date: Tue, 21 Apr 2026 06:12:29 GMT
• Title: Detoxification for LLM: From Dataset Itself
• Abstract summary: 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.<n>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.<n>On GPT2-XL, HSPD attains state-of-the-art detoxification, reducing Toxicity Probability (TP) from 0.42 to 0.18 and Maximum Expected Toxicity (EMT) from 0.43 to 0.20.
• Score: 69.09071947137926
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: 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)

Related papers

• Detoxifying Large Language Models via Autoregressive Reward Guided Representation Editing [77.75609817898035]
  Large Language Models (LLMs) have demonstrated impressive performance across various tasks, yet they remain vulnerable to generating toxic content.<n>We propose textscAutoregressive textscReward textscGuided textscRe presentation textscEditing (ARGRE)<n>ARGRE explicitly models toxicity transitions within the latent representation space, enabling stable and precise reward-guided editing.
  arXiv  Detail & Related papers  (2025-09-24T03:40:32Z)
• <think> So let's replace this phrase with insult... </think> Lessons learned from generation of toxic texts with LLMs [60.169913160819]
  This paper explores the possibility of using synthetic toxic data as an alternative to human-generated data for training models for detoxification.<n>Experiments show that models fine-tuned on synthetic data consistently perform worse than those trained on human data.<n>The root cause is identified as a critical lexical diversity gap: LLMs generate toxic content using a small, repetitive vocabulary of insults that fails to capture the nuances and variety of human toxicity.
  arXiv  Detail & Related papers  (2025-09-10T07:48:24Z)
• Detoxification of Large Language Models through Output-layer Fusion with a Calibration Model [15.394714537797183]
  Existing approaches for Large language model (LLM) detoxification rely on training on large-scale non-toxic or human-annotated preference data.<n>We propose a compact, pre-trained calibration model that guides the detoxification process of a target LLM via a lightweight intervention in its generation pipeline.
  arXiv  Detail & Related papers  (2025-06-02T02:36:32Z)
• UniDetox: Universal Detoxification of Large Language Models via Dataset Distillation [18.150899267807965]
  UniDetox is a method designed to mitigate toxicity across various large language models (LLMs)<n>We propose a novel and efficient dataset distillation technique for detoxification using contrastive decoding.<n>Experiments demonstrate that the detoxifying text distilled from GPT-2 can effectively detoxify larger models, including OPT, Falcon, and LLaMA-2.
  arXiv  Detail & Related papers  (2025-04-29T07:40:00Z)
• Large Language Models can be Strong Self-Detoxifiers [82.6594169242814]
  Self-disciplined Autoregressive Sampling (SASA) is a lightweight controlled decoding algorithm for toxicity reduction of large language models (LLMs) SASA tracks the margin of the current output to steer the generation away from the toxic subspace, by adjusting the autoregressive sampling strategy. evaluated on LLMs of different scale and nature, namely Llama-3.1-Instruct (8B), Llama-2 (7B), and GPT2-L models with the RealToxicityPrompts, BOLD, and AttaQ benchmarks.
  arXiv  Detail & Related papers  (2024-10-04T17:45:15Z)
• Model Editing as a Robust and Denoised variant of DPO: A Case Study on Toxicity [6.786565820048478]
  We introduce a tuning-free alignment alternative, ProFS, and demonstrate its effectiveness under the use case of toxicity reduction.<n>ProFS identifies a toxic subspace in the model parameter space and reduces model toxicity by projecting away the detected subspace.<n>We show that ProFS is more sample-efficient than DPO, further showcasing greater robustness to noisy data.
  arXiv  Detail & Related papers  (2024-05-22T20:08:48Z)
• Detoxifying Large Language Models via Knowledge Editing [57.0669577257301]
  This paper investigates using knowledge editing techniques to detoxify Large Language Models (LLMs) We construct a benchmark, SafeEdit, which covers nine unsafe categories with various powerful attack prompts. We conduct experiments with several knowledge editing approaches, indicating that knowledge editing has the potential to detoxify LLMs with a limited impact on general performance efficiently.
  arXiv  Detail & Related papers  (2024-03-21T15:18:30Z)
• Unveiling the Implicit Toxicity in Large Language Models [77.90933074675543]
  The open-endedness of large language models (LLMs) combined with their impressive capabilities may lead to new safety issues when being exploited for malicious use. We show that LLMs can generate diverse implicit toxic outputs that are exceptionally difficult to detect via simply zero-shot prompting. We propose a reinforcement learning (RL) based attacking method to further induce the implicit toxicity in LLMs.
  arXiv  Detail & Related papers  (2023-11-29T06:42:36Z)
• CMD: a framework for Context-aware Model self-Detoxification [22.842468869653818]
  Text detoxification aims to minimize the risk of language models producing toxic content. Existing detoxification methods fail to achieve a decent balance between detoxification effectiveness and generation quality. We introduce a Context-aware Model self-Detoxification(CMD) framework that pays attention to both the context and the detoxification process.
  arXiv  Detail & Related papers  (2023-08-16T11:50:38Z)
• RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models [93.151822563361]
  Pretrained neural language models (LMs) are prone to generating racist, sexist, or otherwise toxic language which hinders their safe deployment. We investigate the extent to which pretrained LMs can be prompted to generate toxic language, and the effectiveness of controllable text generation algorithms at preventing such toxic degeneration.
  arXiv  Detail & Related papers  (2020-09-24T03:17:19Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.