Invasive Context Engineering to Control Large Language Models

Thomas Rivasseau

Current research on operator control of Large Language Models improves model robustness against adversarial attacks and misbehavior by training on preference examples, prompting, a…