Introducing the Instruction Hierarchy, our latest safety research to advance robustness for prompt injections and other ways of tricking LLMs into executing unsafe actions. More details:
Instruction Hierarchy Advances LLM Robustness Against Prompt Injections
By
–