Dandruff and flaking scalp

Visible flaking, typically Malassezia-driven seborrhoeic dermatitis.

Position bias

Each pair is shown in a random order. If a model simply prefers whichever formulation it reads first, its ranking is largely noise, so some pairs are deliberately re-run the other way round to measure it. A first-position win rate near 50% is unbiased.

Judge	Pairs judged both ways	Agreement	First-position wins	Verdict
wandb:zai-org/GLM-5.3-Flash	0	–	46%	no pairs judged in both orders yet

Transitivity

Triples where A beats B and B beats C, yet C beats A. Some intransitivity is normal; a lot of it means the comparisons are close to coin flips.

Judge	Violations	Triples checked	Rate
wandb:zai-org/GLM-5.3-Flash	318	4000	8.0%

Cross-judge agreement

Only one judge has ranked this hair type so far. Judge it with a second model to compare.