accuracy

Definition

The fraction of predictions that are correct — the most basic metric there is, and the most easily misread. On imbalanced data it is close to meaningless: a detector that answers no to everything scores 99% when the positive rate is 1%. It also treats every error as equally costly, which is rarely true, so serious evaluation pairs it with precision and recall, per-class breakdowns, or a metric reflecting the real cost of each mistake.

Explainers using this term

• Paper walkthrough: StateM — 95.3% on Terminal-Bench 2.1 and a USD 15 run, without touching a single weightStateM: Reaching 95.3% Raw Accuracy, or a \$15 Frontier Run, on Terminal-Bench 2.1 via Harness Scaling