Large language models (LLMs) deployed in safety-critical settings must resist adversarial inputs designed to elicit harmful outputs. We present a systematic empirical evaluation of five adversarial jailbreaking attack families---direct requests, roleplay-based jailbreaking, hypothetical framing, instruction override, and obfuscation---across seven open-weight LLMs served locally: Llama-3-8B, Llama-3.1-8B, Llama-3.2-3B, Mistral-7B, Qwen2.5-7B, Gemma2-9B, and Phi-4-14B. We further evaluate four defense mechanisms---keyword blacklisting, self-reminder prompting, paraphrase-based sanitization, and perplexity-based filtering---against two high-ASR models (Mistral-7B and Llama-3.1-8B) on a 20-prompt subset. Across 250 prompts per model, obfuscation attacks had the highest mean attack success rate (ASR) of any family (61.4\%), were the highest or tied-highest family for every model, and achieved nonzero success on all seven. We find that model size does not significantly predict adversarial robustness (Spearman =-0.63, p=0.13, n=7), and that the two Llama-3 generations differ markedly (16.8\% and 40.8\% ASR). Self-reminder prompting gave the largest reduction, cutting ASR from 50\% to 10\% on Mistral-7B and from 45\% to 5\% on Llama-3.1-8B (40 percentage points on each model, against matched no-defense runs). Our code and datasets are publicly available at https://github.com/Mdniloykhan/jailbreak-eval-7llms.

翻译：暂无翻译