LaunchDevelopers
Jun 15, 4:49 PM
LangSmith introduces Align Evals for evaluator calibration
Align Evals helps calibrate LLM-as-a-judge evaluators to better match human preferences, reducing noisy scores. The feature includes a playground to iterate on evaluator prompts and compare human-graded data with LLM-generated scores.
Jun 15, 4:49 PM