Definition
When a model's competing instructions get re-ranked by whatever's most vivid in context, and the wrong rule wins.
A proposed account of LLM instruction-following failures in which contradictory soft directives compete without enforced priority, and contextual salience determines which directive dominates downstream behavior.