What metric checks whether model output is identical to a reference answer character-for-character?

What metric checks whether model output is identical to a reference answer character-for-character?

• A. AI Agents perform actions autonomously
• B. Demographic Parity
• C. Exact Match Metrics
• D. Hallucinations

Answer: (C)

Exact Match metrics measure whether the model's output is identical to the reference answer on every character, including punctuation, spacing, and casing. This strict criterion means a response only counts as correct if it exactly matches the reference string. It’s ideal when you need precise replication of a known answer, where even a small difference would be unacceptable. In practice, you compare the produced text to the reference, and an exact, byte-for-byte match yields a positive result; any deviation is treated as incorrect. This contrasts with other ideas listed, which are not about text similarity: one describes a model’s independent action capability, another is a fairness metric, and the last refers to a phenomenon where models sometimes generate ungrounded content.

Exact Match metrics measure whether the model's output is identical to the reference answer on every character, including punctuation, spacing, and casing. This strict criterion means a response only counts as correct if it exactly matches the reference string. It’s ideal when you need precise replication of a known answer, where even a small difference would be unacceptable. In practice, you compare the produced text to the reference, and an exact, byte-for-byte match yields a positive result; any deviation is treated as incorrect. This contrasts with other ideas listed, which are not about text similarity: one describes a model’s independent action capability, another is a fairness metric, and the last refers to a phenomenon where models sometimes generate ungrounded content.