Definition
Plain language
One overall number that blends several different measures of how well something did.
As stated in the literature
A weighted aggregate of component metrics; in FM-Bench it is computed arithmetically from engine-tracked quantities, with no LLM or human judge in the loop.
Also called: composite
Why it matters: It makes many-dimensional performance comparable at a glance, but it also bakes in someone's opinion about which dimensions count for how much.
For example, a single final number might combine league finishes, financial health, and squad quality into one figure you can rank models by.
Heard on the show
“" Another might be a multi-step composite like "open Chrome language settings.”Episode 066 — Why Giving an AI Agent More Tools Can Make It Worse at Using a Computer