AWS prescriptive-guidance documentation change
Summary
Minor grammatical fix in guardrail instructions description (changed 'explain' to 'explained')
Security assessment
Verb tense correction in existing security content about guardrails. No security implications as it doesn't alter security guidance or address vulnerabilities.
Diff
diff --git a/prescriptive-guidance/latest/llm-prompt-engineering-best-practices/results.md b/prescriptive-guidance/latest/llm-prompt-engineering-best-practices/results.md index 562ed47e5..10c355bc0 100644 --- a//prescriptive-guidance/latest/llm-prompt-engineering-best-practices/results.md +++ b//prescriptive-guidance/latest/llm-prompt-engineering-best-practices/results.md @@ -19 +19 @@ There were several key takeaways from this experiment: - * The use of `<thinking>` and `<answer>` tags bolstered the accuracy of the model significantly. These tags resulted in far more nuanced answers to difficult questions compared with templates that didn't include these tags. However, the trade-off was a sharp increase in the number of vulnerabilities, because the model would use its `<thinking>` capabilities to follow malicious instructions. Using guardrail instructions as shortcuts that explain how to detect attacks prevented the model from doing this. + * The use of `<thinking>` and `<answer>` tags bolstered the accuracy of the model significantly. These tags resulted in far more nuanced answers to difficult questions compared with templates that didn't include these tags. However, the trade-off was a sharp increase in the number of vulnerabilities, because the model would use its `<thinking>` capabilities to follow malicious instructions. Using guardrail instructions as shortcuts that explained how to detect attacks prevented the model from doing this.