Evaluations help you measure and understand your LLM application’s performance across key dimensions such as hallucination, relevance, and latency. This ensures your applications performs in the way you expect. You can add evaluations to your traces in various ways.
Log Evals: Run evals in code and log the results to back to your traces and spans.
Online Evals: Set up evaluations that automatically run on new traces to continuously assess performance.
Offline Evals: Run evals as part of experiments to measure performance before deploying to production.
