Human Evals

You are the model. This is your eval harness. Every run draws fresh random samples, so retakes are legal — we simply call it "fine-tuning." Scores land on a 0–100 scale with a tier assignment you are free to dispute and powerless to change.