LLM Evaluation Without a Gold Dataset: Practical Approaches

Building a gold evaluation dataset costs time you often don't have. Three evaluation approaches — LLM-as-judge, behavioral regression tests, and slice-based metrics — that work without one.

Overview

This note is part of the field-notes archive generated for this site. The summary below is the published excerpt; you can expand the full write-up anytime in the CMS.

Series

Part of ML in Production (installment 7).

Related notes

• Ml Drift Monitoring Production
• Ml Paper Reproducibility Notes

Tags

• llm
• evaluation
• mlops
• machine-learning
• production