TLDR AI · RESEARCH & INNOVATION
INSTRUCTION HIERARCHY
Summary from the TLDR AI newsletter
OpenAI published research on giving system prompts stronger weighting, which dramatically improves model robustness to jailbreaks and adversarial attacks.
TLDR AI · RESEARCH & INNOVATION
OpenAI published research on giving system prompts stronger weighting, which dramatically improves model robustness to jailbreaks and adversarial attacks.