Câu 7: UiAAAv1: UiPath Agentic Automation Associate v1.0

You are evaluating an agent that processes structured input and generates outputs containing exact numerical values, boolean fields, and predefined strings. The outputs must match the expected structure and values precisely for the solution to function correctly. Which evaluator type is best suited for this scenario?

1. A. LLM-as-a-Judge, as it evaluates semantic similarity and reasoning, and you can instruct it to check for an exact match between the input and expected output.
2. B. JSON similarity, but only for assessing open-ended natural language responses in complex scenarios.
3. C. LLM-as-a-Judge, as it guarantees exact alignment between numerical and boolean fields in structured outputs.
4. D. Deterministic evaluators (Exact match or JSON similarity), as they ensure precise alignment between the generated output and the expected structure and values. — đáp án hiện tại