Definition
Plain language
A fixed starting number that makes a randomized setup repeatable, so the same scene comes out identically every time.
As stated in the literature
An initialization value that pins a pseudo-random process (object positions, distractors, starting state); ASPIRE trains on debug seeds and evaluates on disjoint held-out seeds to enforce genuine generalization.
Also called: seed, seeds
Why it matters: It makes randomized experiments repeatable and lets researchers train on one set of scenes and test on fresh ones to prove the skill truly generalizes.
For example, using seed 42 places the blocks in the exact same spots every run, while a different seed reshuffles where everything starts.
Heard on the show
“Dead flat, across five rounds and three random seeds.”Episode 207 — An AI Graded Its Own Math Test 94 Percent — It Actually Scored 20