Safety alignment in large language models relies on behavioral training that can be overridden when sufficiently strong in-context patterns compete with learned refusal behaviors. We introduce Involuntary In-Context Learning (IICL), an attack class that uses abstract operator framing with few-shot examples to force pattern completion that overrides safety training. Through 3479 probes across 10 OpenAI models, we identify the attack's effective components through a seven-experiment ablation study. Key findings: (1)~semantic operator naming achieves 100\,\% bypass rate (50/50, $p < 0.001$); (2)~the attack requires abstract framing, since identical examples in direct question-and-answer format yield 0\,\%; (3)~example ordering matters strongly (interleaved: 76\,\%, harmful-first: 6\,\%); (4)~temperature has no meaningful effect (46--56\,\% across 0.0--1.0). On the HarmBench benchmark, IICL achieves 24.0\,\% bypass $[18.6\%, 30.4\%]$ against GPT-5.4 with detailed 619-word responses, compared to 0.0\,\% for direct queries.

翻译：大型语言模型的安全对齐依赖于行为训练，但当足够强的上下文模式与习得的拒绝行为竞争时，这种训练可能被覆盖。我们提出“无意上下文学习”（IICL），一种利用抽象算子框架配合少样本示例强制模式补全以覆盖安全训练的攻攻击类别。通过对10个OpenAI模型进行3479次探测，我们通过七实验消融研究识别出该攻击的有效组成部分。关键发现：（1）语义算子命名实现100%绕过率（50/50，$p<0.001$）；（2）该攻击需要抽象框架，因为相同示例以直接问答格式呈现时绕过率为0%；（3）示例顺序影响显著（交错排列：76%，有害优先：6%）；（4）温度参数无明显影响（0.0–1.0范围内为46%–56%）。在HarmBench基准测试中，IICL针对GPT-5.4实现24.0%绕过率[18.6%, 30.4%]，并生成详细619词响应，而直接查询的绕过率为0.0%。