JevEval

JevEval is a specialized evaluation metric built atop DeepEval, leveraging Jev to transform explicit, bounded questions and calibrated probability outputs into deterministic, traceable scores for language model evaluation. It separates evaluation logic definition, decision-making, and scoring, enabling precise and inspectable assessments of LLM outputs, including multi-turn conversational agents.