Câu 39: NCP-AAI: Agentic AI

You are developing a RAG solution and have decided to use a classifier branch as part of your semantic guardrail system to assess the risk of generated text. Which of the following is a key benefit of using a classifier branch compared to solely relying on prompt filtering?

1. A. Since a classifier branch does not require training, it can identify potentially problematic content.
2. B. Classifier branches primarily focus on detecting factual inaccuracies, rather than stylistic or harmful language.
3. C. Classifier branches can automatically adapt to new forms of harmful language. — đáp án hiện tại
4. D. Classifier branches eliminate the need for human oversight, thereby automating the safety process.