# Learnings - Workflow (F011)
- For fair method benchmarking, evaluate all conditions with shared controls (`SEED`, `N_EVAL_EPISODES`) and only render comparison outputs from a single merged `all_results` collection. *(F011)*