Where Does Toxicity Live? Mechanistic Localization and Targeted Suppression in Language Models

• URL: http://arxiv.org/abs/2605.27997v1
• Date: Wed, 27 May 2026 05:41:19 GMT
• Title: Where Does Toxicity Live? Mechanistic Localization and Targeted Suppression in Language Models
• Abstract summary: Large language models frequently generate toxic, hateful, or harmful content.<n>We introduce Meow2X and TRNE, two complementary retraining-free frameworks that localize toxicity to specific layers and neurons.<n>Our analysis reveals that toxicity is disproportionately encoded in early gradient layers, varies across architectures, and is systematically underestimated by single-evaluator setups.
• Score: 1.2040549525117266
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language models frequently generate toxic, hateful, or harmful content, yet existing mitigation methods rely on costly retraining or output-level filtering with no mechanistic insight into where toxicity originates internally. We introduce Meow2X and TRNE, two complementary retraining-free frameworks that localize toxicity to specific layers and neurons by analyzing activation differentials between toxic and neutral prompts, then suppress them via inference-time scaling or minimal rank-one weight edits -- without any gradient descent. Evaluations across five LMs, two benchmarks, and 90 configurations using dual safety evaluators demonstrate consistent toxicity reduction while preserving language modeling quality. Our analysis reveals that toxicity is disproportionately encoded in early MLP layers, varies across architectures, and is systematically underestimated by single-evaluator setups -- underscoring the need for multi-evaluator safety assessment. By bridging mechanistic interpretability with practical detoxification, our framework offers a principled path toward safer, more transparent language models.

Related papers

• 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)
• 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)
• Redefining Experts: Interpretable Decomposition of Language Models for Toxicity Mitigation [12.58703387927632]
  We investigate the stability of neuron-level toxicity indicators, the advantages of structural (layer-wise) representations, and the interpretability of mechanisms driving toxic generation.<n>We propose a novel principled intervention technique, EigenShift, based on eigen-decomposition of the language model's final output layer.
  arXiv  Detail & Related papers  (2025-09-20T12:21:52Z)
• Breaking Bad Tokens: Detoxification of LLMs Using Sparse Autoencoders [9.203871622144863]
  Large language models (LLMs) are now ubiquitous in user-facing applications, yet they still generate undesirable toxic outputs.<n>We leverage sparse autoencoders (SAEs) to identify toxicity-related directions in the residual stream of models and perform targeted activation steering.
  arXiv  Detail & Related papers  (2025-05-20T15:55:31Z)
• Aligned Probing: Relating Toxic Behavior and Model Internals [78.20380492883022]
  We introduce aligned probing, a novel interpretability framework that aligns the behavior of language models (LMs)<n>Using this framework, we examine over 20 OLMo, Llama, and Mistral models, bridging behavioral and internal perspectives for toxicity for the first time.<n>Our results show that LMs strongly encode information about the toxicity level of inputs and subsequent outputs, particularly in lower layers.
  arXiv  Detail & Related papers  (2025-03-17T17:23:50Z)
• 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)
• 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)
• 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.