The paper investigates how language models fail to detect harmful intent after polarity-altering transformations, a vulnerability called polarity blindness. It examines grammatical negation, antonym replacement, and synonym replacement as controls, and tests the impact of adversarial wrappers. Across 260 prompts, 10 LLMs, and three languages, polarity blindness is widespread, with the most vulnerable model scoring an average harm level of 4.12 on a 0‑5 scale.
arXiv:2606. 28843v1 Announce Type: cross Abstract: Fine-tuning a large language model is a ubiquitous method for enhancing its capability on a specific downstream task.
The paper identifies a vulnerability in large language models where harmful intent can be hidden within benign narratives, a phenomenon termed Semantic Camouflage. By examining latent activation patterns across several small language model families, the authors discover an "Intent Horizon"—a layer depth where harmful intent representations collapse. They propose Latent Intent Verification (LIV), a lightweight probing defense that detects harmful intent in early layers and outperforms existing guardrails on the PKU-SafeRLHF dataset.
arXiv:2608. 14577v1 Announce Type: cross Abstract: Frontier large language models (LLMs) safety evaluation has largely treated harmful generation as an attack outcome rather than as an object of analysis.
The paper presents an instruction‑tuned large language model (LLM) based on Qwen3 that is fine‑tuned for hate speech mitigation by unifying 36 English hate speech datasets. The authors show that this generalist LLM achieves state‑of‑the‑art performance on in‑domain benchmarks and delivers significant gains in cross‑domain and cross‑lingual generalization, outperforming specialist encoder‑based classifiers.
The paper examines how automated safety systems unevenly flag harmful content, disproportionately labeling African American English as offensive and misclassifying content from drag queens and white nationalists. It identifies seven failure modes rooted in vocabulary-based harm detection and proposes an evaluation protocol that highlights gaps in performance across languages. Applied to five open systems in eleven languages, the study finds that none achieve native-speaker accuracy, and higher capability often increases discrimination, widening the disparity between best- and worst-served language communities.
arXiv:2607. 02079v1 Announce Type: cross Abstract: We present HaloGuard 1.
arXiv:2604. 09544v2 Announce Type: replace-cross Abstract: Large language models (LLMs) undergo alignment training to avoid harmful behaviors, yet the resulting safeguards remain brittle: jailbreaks routinely bypass them, and fine-tuning on narrow domains can induce ``emergent misalignment'' that generalizes broadly.
arXiv:2503. 11832v5 Announce Type: replace Abstract: Recent vision language models (VLMs) have made remarkable strides in generative modeling with multimodal inputs, particularly text and images.
The paper introduces Semantic Confusion to assess how consistently large language models refuse similar prompts. It presents ParaGuard, a 10k‑prompt corpus of controlled paraphrase clusters, and proposes three token‑level metrics—Confusion Index, Confusion Rate, and Confusion Depth—to measure contradictory refusal decisions across meaning‑preserving paraphrases. Experiments show that global false rejection rates can mask local inconsistencies, revealing that refusal evaluation must consider both frequency and consistency across nearby paraphrases.
arXiv:2607. 28862v1 Announce Type: cross Abstract: The rapid development of Large Language Models (LLMs) has led to significant advances across a wide range of language tasks, while simultaneously raising growing concerns about unauthorized data exploitation and privacy leakage.
Safety alignment in large language models (LLMs) is largely developed in English, assuming these safeguards generalize across multilingual settings. However, this assumption remains underexplored and exposes a vulnerability in low-resource languages.
arXiv:2608. 09510v1 Announce Type: cross Abstract: Detecting machine-generated disinformation on social media is increasingly difficult as large language models (LLMs) make it easier to generate and rewrite misleading content at scale.