Improving instruction hierarchy in frontier LLMs

Company news OpenAI Source language: English

IH-Challenge trains models to prioritize trusted instructions, improving instruction hierarchy, safety steerability, and resistance to prompt injection attacks.

Excerpt

IH-Challenge trains models to prioritize trusted instructions, improving instruction hierarchy, safety steerability, and resistance to prompt injection attacks.

Short excerpt from the source — open the original for the full text.

Sources

← Back to the daily brief