Evaluation & Quality
Win rate
The percentage of head-to-head comparisons in which one model or system is preferred.
Example
Reviewers prefer one model's answer in 62% of head-to-head comparisons.
Why it matters
It helps compare alternatives when a direct preference is more useful than an isolated score.
You'll hear it
“People preferred this version in most of the comparisons.”
Can you control it?
Set by the team
The people building or running the AI choose this setup. An everyday user generally needs their help to change how this part works.
Common questions
- Does winning mean the answer was factually correct?
- No. A more appealing response can still contain a mistake.
- Can the choice of opponent affect the rate?
- Yes. Beating a weak alternative is different from beating a strong one.
- Can presentation influence which answer wins?
- Yes. Order, labels or formatting can affect preferences if the comparison is not handled carefully.