← All questions
EvaluationMedium· System Comparison· Mezzo Analysis· Metrics

Two candidate systems land on the exact same macro F1....

When macro metrics are tied, the choice should be driven by the traffic mix. One system might excel at complex multi-hop queries while the other performs better on simple single-hop queries. By aligning the system’s strengths with the actual query distribution of your users, you ensure the chosen model provides the most value in production.

Code playground (Python)