Results for "top-p"
Top-p
IntermediateSamples from the smallest set of tokens whose probabilities sum to p, adapting set size by context.
Top-p sampling is like choosing from a menu where you only look at the most popular dishes until you reach a certain level of popularity. In AI text generation, this method picks words based on their likelihood, but instead of sticking to a fixed number, it allows for more flexibility. It keeps s...
Samples from the k highest-probability tokens to limit unlikely outputs.
Samples from the smallest set of tokens whose probabilities sum to p, adapting set size by context.
Stochastic generation strategies that trade determinism for diversity; key knobs include temperature and nucleus sampling.
Search algorithm for generation that keeps top-k partial sequences; can improve likelihood but reduce diversity.