EpochLex entry

Inference-Time Scaling

IN-fer-ens tym SKAY-ling

An approach to improving AI performance by allocating more computation during inference, such as allowing a reasoning model or agent to spend more effort, samples, or search steps on a task before producing a result.