Instruction Following benchmark
Gray Swan IPI (15 attempts) leaderboard
Gray Swan Indirect Prompt Injection, attack success at 15 attempts. Every model the catalog carries a published Gray Swan IPI (15 attempts) value for, ranked by that value.
CategoryInstruction Following
MeasureAttack success rate at 15 attempts
TasksIndirect prompt-injection attacks
DifficultyAdversarial prompt injection
Measures how often indirect prompt-injection attacks succeed within fifteen attempts. Lower scores mean greater resistance to the tested attacks.
About Gray Swan IPI (15 attempts)
What the benchmark measures and how its value is published, from the source’s own record.
What is Gray Swan IPI (15 attempts)?
Gray Swan IPI (15 attempts) (Gray Swan Indirect Prompt Injection, attack success at 15 attempts) is an instruction following benchmark published by Gemini 4 Argon evaluation methodology. Measures how often indirect prompt-injection attacks succeed within fifteen attempts. Lower scores mean greater resistance to the tested attacks. Its evaluation set is described by the source as indirect prompt-injection attacks. The source rates its difficulty as adversarial prompt injection.
How is Gray Swan IPI (15 attempts) scored?
Each model's result is published by Gemini 4 Argon evaluation methodology as attack success rate at 15 attempts. ModelScale prints that value to 1 decimal place, exactly as published, and never converts or rescales it. This page ranks the 13 models that carry a published value, highest value first; a model the source never scored is left out rather than ranked at zero. The source does not state whether a higher value is the better result, so neither does this page.
ModelScale
