ianarawjo/evalstats

Statistical inference for AI evaluations, from model comparisons to corrections for LLM judge bias, optimized for small sample sizes. All defaults battle-tested in Monte Carlo simulations.