pawbench

Definition

A coined benchmark name. The -Bench suffix is a convention: it signals a packaged evaluation — a task set, its data, and a scoring procedure — aimed at one capability. Everything specific, including what the first element abbreviates, what is measured and how it is scored, has to come from the paper introducing it rather than from the name.

Explainers using this term

• PAWBench Explained — Can Video Generators Get the Odds Right, Not Just the Physics?PAWBench: How Far Are We from Probabilistically Aligned World Modeling?