/

Testing & Simulation

/

Semantic Accuracy

Semantic Accuracy

/ semantic-accuracy /

Whether a response is correct in meaning — not just fluent. The bar agent evaluations should measure against.

Whether a response is correct in meaning — not just fluent. The bar agent evaluations should measure against.

Why it matters

Fluency is cheap; correctness is not. Evals must judge whether answers are right in meaning, or agents get rewarded for sounding confident.

Related — Testing & Simulation