Fig.1

Concept

Test-time scaling

Test-time scaling means spending more compute at inference to improve output quality, without retraining or changing the model weights. The anchor most engineers know is beam search in decoding: instead of taking one greedy sample, you explore several candidates and keep the…

The rest of “Test-time scaling” is a premium feature: every concept in the library gets a precise, practitioner-focused write-up like this one, cross-linked straight from the paper summaries that use it.

Log in to unlock

← Back to the library

Test-time scaling, explained · Fig. 1