Learn how to build effective task-specific scorecards for evaluating LLMs. Discover why ROUGE fails for summarization, how to use BERTScore and Ragas, and best practices for judging Q&A and extraction tasks.