DataFlow provides three model evaluation options “from easy to advanced”, covering needs from quick start to research-grade benchmark evaluation. You only need to choose and read ONE of the following documents to complete your evaluation (these are different entry points; you do not need to learn all of them).

How to Choose

Which user are you?	What you want	Recommended Reading
👶 Beginner - want to get started fast	Evaluate directly via CLI (for QA data, works out of the box)	Model Evaluation (QA Quickstart)
🧑‍💻 Beginner+ - simple parameter tuning, model before/after comparison	Modify pipeline script parameters (more straightforward)	Model Evaluation (Beginner Edition)
🧪 Researcher - academic, standardized benchmark metrics	Unified benchmark evaluation framework (task types + full evaluation parameters)	Model Evaluation (Research Edition)

Document Entries

• Model Evaluation (QA Quickstart): CLI-based, beginner-friendly, suitable for quick evaluation on QA-style datasets.
• Model Evaluation (Beginner Edition): pipeline-code based, for beginner/intermediate users, adjust evaluation settings by editing script parameters.
• Model Evaluation (Research Edition): research-grade evaluation, for users who need to pass full evaluation parameters to evaluate specific benchmarks.