LLM Evaluation Metrics BLEU, ROUGE, perplexity which quantify the similarity between generated text and reference outputs. LLM