Large Language Models (LLMs), especially their compact efficiency-oriented variants, remain susceptible to jailbreak attacks that can elicit harmful outputs despite extensive alignment efforts. Existing adversarial prompt generation techniques often rely on manual engineering or rudimentary obfuscation, producing low-quality or incoherent text that is easily flagged by perplexity-based filters. We present an automated red-teaming framework that evolves semantically meaningful and stealthy jailbreak prompts for aligned compact LLMs. The approach employs a multi-stage evolutionary search, where candidate prompts are iteratively refined using a population-based strategy augmented with temperature-controlled variability to balance exploration and coherence preservation. This enables the systematic discovery of prompts capable of bypassing alignment safeguards while maintaining natural language fluency. We evaluate our method on benchmarks in English (In-The-Wild Jailbreak Prompts on LLMs), and a newly curated Arabic one derived from In-The-Wild Jailbreak Prompts on LLMs and annotated by native Arabic linguists, enabling multilingual assessment.

We propose a novel certified defense method for Image Quality Assessment (IQA) models based on randomized smoothing with noise applied in the feature space rather than the input space. Unlike prior approaches that inject Gaussian noise directly into input images, often degrading visual quality, our method preserves image fidelity while providing robustness guarantees. To formally connect noise levels in the feature space with corresponding input-space perturbations, we analyze the maximum singular value of the backbone network's Jacobian. Our approach supports both full-reference (FR) and no-reference (NR) IQA models without requiring any architectural modifications, suitable for various scenarios. It is also computationally efficient, requiring a single backbone forward pass per image. Compared to previous methods, it reduces inference time by 99.5% without certification and by 20.6% when certification is applied. We validate our method with extensive experiments on two benchmark datasets, involving six widely-used FR and NR IQA models and comparisons against five state-of-the-art certified defenses. Our results demonstrate consistent improvements in correlation with subjective quality scores by up to 30.9%.