Your question is Batch vs Streaming Model Evaluation. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
You're working on a pipeline that computes model evaluation results from application outputs and human or automated labels. Before choosing an implementation, you want to compare whether the evaluation system should run in scheduled batch jobs, continuous streams, or a mix of both.
Explain the trade-offs between batch processing and streaming architectures in the context of model evaluation.