{"ID":2834134,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2512.03001","arxiv_id":"2512.03001","title":"Invasive Context Engineering to Control Large Language Models","abstract":"Current research on operator control of Large Language Models improves model robustness against adversarial attacks and misbehavior by training on preference examples, prompting, and input/output filtering. Despite good results, LLMs remain susceptible to abuse, and jailbreak probability increases with context length. There is a need for robust LLM security guarantees in long-context situations. We propose control sentences inserted into the LLM context as invasive context engineering to partially solve the problem. We suggest this technique can be generalized to the Chain-of-Thought process to prevent scheming. Invasive Context Engineering does not rely on LLM training, avoiding data shortage pitfalls which arise in training models for long context situations.","short_abstract":"Current research on operator control of Large Language Models improves model robustness against adversarial attacks and misbehavior by training on preference examples, prompting, and input/output filtering. Despite good results, LLMs remain susceptible to abuse, and jailbreak probability increases with context length....","url_abs":"https://arxiv.org/abs/2512.03001","url_pdf":"https://arxiv.org/pdf/2512.03001v1","authors":"[\"Thomas Rivasseau\"]","published":"2025-12-02T18:25:55Z","proceeding":"cs.AI","tasks":"[\"cs.AI\"]","methods":"[\"Large Language Model\",\"Language Model\"]","has_code":false}