AI Eval Metrics Cheat Sheet — Every Metric You Need to Evaluate LLMs & Agents

Description

Save this before you ship your next AI app. 12 evaluation metrics across 4 categories — RAG, Agents, General, and Reference-based — in one cheat sheet. What's covered: → Faithfulness, Context Precision, Context Recall — for RAG pipelines → Task Completion, Tool Accuracy, Step Efficiency — for agents → Hallucination, G-Eval, Toxicity — for any LLM app → BLEU, ROUGE, BERTScore — for summarisation Most teams pick one metric and call it done. Production AI needs all of these working together.