· Xiaojing Yang · Statistics
Multiple Comparisons in AI Experiments
When we try many models, prompts, seeds, and metrics, false discoveries become easier than they look.
When we try many models, prompts, seeds, and metrics, false discoveries become easier than they look.