The pseudo-randomly chosen subset of the vocabulary that the sampler favors at each step; which tokens are green depends on a hash of the preceding context.
Continue to AI University →