Measuring and Mitigating Toxicity in Large Language Models: A Comprehensive Replication Study

• URL: http://arxiv.org/abs/2605.14087v2
• Date: Fri, 15 May 2026 01:03:24 GMT
• Title: Measuring and Mitigating Toxicity in Large Language Models: A Comprehensive Replication Study
• Abstract summary: Large Language Models (LLMs) trained on web-scale corpora inherently absorb toxic patterns from their training data.<n>Decoding-time Experts (DExperts) is an inference-time mitigation technique that steers generation without requiring model retraining.<n>This study contributes to the growing body of work on AI safety by highlighting the gap between explicit and implicit toxicity mitigation.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large Language Models (LLMs) trained on web-scale corpora inherently absorb toxic patterns from their training data. This leads to toxic degeneration where even innocuous prompts can trigger harmful outputs. This phenomenon poses significant risks for real-world deployments. Thus, necessitating effective mitigation strategies that should maintain model utility while ensuring safety. In this comprehensive replication study, we evaluate the efficacy of DExperts (Decoding-time Experts), which is an inference-time mitigation technique that steers generation without requiring model retraining. We structured our research into three systematic phases: (1) establishing baseline toxicity measurements using RealToxicityPrompts on standard GPT-2 models; then (2) implementing and evaluating DExperts to mitigate explicit toxicity; and finally (3) stress-testing the method against implicit hate speech using the adversarial ToxiGen dataset. Our empirical results confirm that while DExperts achieves near-perfect safety rates (100%) on explicit toxicity benchmarks, it exhibits brittleness against adversarial, implicit hate speech, with safety rates dropping to 98.5%. Furthermore, we quantify a critical trade-off. The method introduces a 10x latency penalty (from 0.2s to 2.0s per generation), posing challenges for real-time deployment scenarios. This study contributes to the growing body of work on AI safety by highlighting the robustness gap between explicit and implicit toxicity mitigation. We emphasize the need for more sophisticated approaches that generalize across diverse hate-speech patterns without incurring prohibitive computational costs.

Related papers

• Detoxification for LLM: From Dataset Itself [69.09071947137926]
  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.
  arXiv  Detail & Related papers  (2026-04-21T06:12:29Z)
• Do Prompts Guarantee Safety? Mitigating Toxicity from LLM Generations through Subspace Intervention [6.808534332444413]
  Large Language Models (LLMs) are powerful text generators.<n>LLMs can produce toxic or harmful content even when given seemingly harmless prompts.<n>This presents a serious safety challenge and can cause real-world harm.
  arXiv  Detail & Related papers  (2026-02-06T11:33:17Z)
• Cleansing the Artificial Mind: A Self-Reflective Detoxification Framework for Large Language Models [14.566005698357747]
  Large Language Models (LLMs) have revealed remarkable generative capabilities and emerging self-regulatory mechanisms.<n>We introduce a fully self-reflective detoxification framework that harnesses the inherent capacities of LLMs to detect, correct toxic content.<n>Our findings underscore the potential for truly self-regulated language models, paving the way for more responsible and ethically guided text generation systems.
  arXiv  Detail & Related papers  (2026-01-16T21:01:26Z)
• The Eminence in Shadow: Exploiting Feature Boundary Ambiguity for Robust Backdoor Attacks [51.468144272905135]
  Deep neural networks (DNNs) underpin critical applications yet remain vulnerable to backdoor attacks.<n>We provide a theoretical analysis targeting backdoor attacks, focusing on how sparse decision boundaries enable disproportionate model manipulation.<n>We propose Eminence, an explainable and robust black-box backdoor framework with provable theoretical guarantees and inherent stealth properties.
  arXiv  Detail & Related papers  (2025-12-11T08:09:07Z)
• Associative Poisoning to Generative Machine Learning [5.094623170336122]
  We introduce a novel data poisoning technique called associative poisoning.<n>It compromises fine-grained features of the generated data without requiring control of the training process.<n>This attack perturbs only the training data to manipulate statistical associations between specific feature pairs in the generated outputs.
  arXiv  Detail & Related papers  (2025-11-07T11:47:33Z)
• Rethinking Toxicity Evaluation in Large Language Models: A Multi-Label Perspective [104.09817371557476]
  Large language models (LLMs) have achieved impressive results across a range of natural language processing tasks.<n>Their potential to generate harmful content has raised serious safety concerns.<n>We introduce three novel multi-label benchmarks for toxicity detection.
  arXiv  Detail & Related papers  (2025-10-16T06:50:33Z)
• 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)
• LoGU: Long-form Generation with Uncertainty Expressions [49.76417603761989]
  We introduce the task of Long-form Generation with Uncertainty(LoGU)<n>We identify two key challenges: Uncertainty Suppression and Uncertainty Misalignment.<n>Our framework adopts a divide-and-conquer strategy, refining uncertainty based on atomic claims.<n>Experiments on three long-form instruction following datasets show that our method significantly improves accuracy, reduces hallucinations, and maintains the comprehensiveness of responses.
  arXiv  Detail & Related papers  (2024-10-18T09:15:35Z)
• Contrastive Perplexity for Controlled Generation: An Application in Detoxifying Large Language Models [21.341749351654453]
  The generation of toxic content by large language models (LLMs) remains a critical challenge for the safe deployment of language technology.<n>We propose a novel framework for implicit knowledge editing and controlled text generation by fine-tuning LLMs with a prototype-based contrastive perplexity objective.
  arXiv  Detail & Related papers  (2024-01-16T16:49:39Z)
• On Practical Aspects of Aggregation Defenses against Data Poisoning Attacks [58.718697580177356]
  Attacks on deep learning models with malicious training samples are known as data poisoning. Recent advances in defense strategies against data poisoning have highlighted the effectiveness of aggregation schemes in achieving certified poisoning robustness. Here we focus on Deep Partition Aggregation, a representative aggregation defense, and assess its practical aspects, including efficiency, performance, and robustness.
  arXiv  Detail & Related papers  (2023-06-28T17:59:35Z)
• Adding Instructions during Pretraining: Effective Way of Controlling Toxicity in Language Models [29.505176809305095]
  We propose two novel pretraining data augmentation strategies that significantly reduce model toxicity without compromising its utility. Our two strategies are: (1) MEDA: adds raw toxicity score as meta-data to the pretraining samples, and (2) INST: adds instructions to those samples indicating their toxicity. Our results indicate that our best performing strategy (INST) substantially reduces the toxicity probability up to 61% while preserving the accuracy on five benchmark NLP tasks.
  arXiv  Detail & Related papers  (2023-02-14T23:00:42Z)
• Constructing Highly Inductive Contexts for Dialogue Safety through Controllable Reverse Generation [65.48908724440047]
  We propose a method called emphreverse generation to construct adversarial contexts conditioned on a given response. We test three popular pretrained dialogue models (Blender, DialoGPT, and Plato2) and find that BAD+ can largely expose their safety problems.
  arXiv  Detail & Related papers  (2022-12-04T12:23:41Z)
• 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.