Toxic HallucinAItions: Perturbing Prompts and Tracing LLM Circuits

• URL: http://arxiv.org/abs/2605.30913v1
• Date: Fri, 29 May 2026 06:58:47 GMT
• Title: Toxic HallucinAItions: Perturbing Prompts and Tracing LLM Circuits
• Abstract summary: We study how lexical and tone-based prompt perturbations affect the factual reliability of large language models (LLMs)<n>We find that toxic lexical perturbations consistently reduce factual accuracy and increase uncertainty, while polite phrasing yields limited and inconsistent changes.
• Score: 17.92293259974243
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large language models (LLMs) are increasingly deployed in conversational settings where user tone ranges from polite to adversarial or toxic, yet less is known about whether toxic language in otherwise semantically equivalent prompts can degrade factual reliability. We study how lexical and tone-based prompt perturbations affect the factual reliability of LLMs. Using controlled prompt variations across polite, random, and three toxicity levels, we evaluate five LLMs on ARC-Easy, GSM8K, and MMLU. We find that toxic lexical perturbations consistently reduce factual accuracy and increase uncertainty, while polite phrasing yields limited and inconsistent changes. To examine whether these answer inconsistencies correspond to internal changes, we conduct attribution-graph analyses of model activations and influences. We find that increasing toxicity selectively amplifies perturbation-sensitive variant nodes while relatively stable core reasoning nodes remain more invariant. These findings position prompt tone as a critical dimension of LLM reliability and provide behavioral and mechanistic evidence that surface-level lexical variation can alter factual outputs and internal computation.

Related papers

• Evaluating and Explaining Prompt Sensitivity of LLMs Using Interactions [10.595427068095397]
  Even subtle and semantically irrelevant changes in prompts can cause dramatic fluctuations in performance.<n>We propose an Interaction-based Prompt Sensitivity metric by quantifying changes in interactions when we introduce subtle changes to prompts.
  arXiv  Detail & Related papers  (2026-08-19T04:45:59Z)
• Same Meaning, Different Scores: Lexical and Syntactic Sensitivity in LLM Evaluation [40.210132040677]
  This paper examines how controlled, truth-conditionally equivalent lexical and syntactic perturbations affect the absolute performance and relative ranking of 23 contemporary Large Language Models (LLMs)<n>Results show that lexical perturbations consistently induce substantial, statistically significant performance degradation across nearly all models and tasks, while syntactic perturbations have more heterogeneous effects, occasionally improving results.
  arXiv  Detail & Related papers  (2026-02-19T12:24:42Z)
• Multi-stage Prompt Refinement for Mitigating Hallucinations in Large Language Models [49.435669307386156]
  Multi-stage Prompt Refinement (MPR) is a framework designed to systematically improve ill-formed prompts across multiple stages.<n>MPR iteratively enhances the clarity of prompts with additional context and employs a self-reflection mechanism with ranking to prioritize the most relevant input.<n>Results on hallucination benchmarks show that MPR achieve over an 85% win rate compared to their original forms.
  arXiv  Detail & Related papers  (2025-10-14T00:31:36Z)
• 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)
• Human-Aligned Faithfulness in Toxicity Explanations of LLMs [23.860142056633233]
  We develop a novel criterion to measure the extent to which free-form toxicity explanations align with those of a rational human under ideal conditions.<n>We conduct experiments on three Llama models and an 8B Ministral model on five diverse toxicity datasets.<n>Our results show that while LLMs generate plausible explanations to simple prompts, their reasoning about toxicity breaks down when prompted about the nuanced relations between the complete set of reasons, the individual reasons, and their toxicity stances.
  arXiv  Detail & Related papers  (2025-06-23T20:41:45Z)
• Adaptive Detoxification: Safeguarding General Capabilities of LLMs through Toxicity-Aware Knowledge Editing [49.85884082568318]
  ToxEdit is a toxicity-aware knowledge editing approach.<n>It dynamically detects toxic activation patterns during forward propagation.<n>It then routes computations through adaptive inter-layer pathways to mitigate toxicity effectively.
  arXiv  Detail & Related papers  (2025-05-28T12:37:06Z)
• Factual Self-Awareness in Language Models: Representation, Robustness, and Scaling [56.26834106704781]
  Factual incorrectness in generated content is one of the primary concerns in ubiquitous deployment of large language models (LLMs)<n>We provide evidence supporting the presence of LLMs' internal compass that dictate the correctness of factual recall at the time of generation.<n>Scaling experiments across model sizes and training dynamics highlight that self-awareness emerges rapidly during training and peaks in intermediate layers.
  arXiv  Detail & Related papers  (2025-05-27T16:24:02Z)
• 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)
• Dialectal Toxicity Detection: Evaluating LLM-as-a-Judge Consistency Across Language Varieties [23.777874316083984]
  There has been little systematic study on how dialectal differences affect toxicity detection by modern LLMs. We create a multi-dialect dataset through synthetic transformations and human-assisted translations, covering 10 language clusters and 60 varieties. We then evaluated three LLMs on their ability to assess toxicity across multilingual, dialectal, and LLM-human consistency.
  arXiv  Detail & Related papers  (2024-11-17T03:53:24Z)
• 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)

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.