research

Indirect Prompt Injection in Small Language Models attack taxonomy and size analysis for 0.5B-3B models