Word
top p
Definition
n. a text generation setting that picks words only from the smallest group whose combined probability reaches a chosen cutoff.
n. a dynamic token sampling strategy (nucleus sampling) that restricts generation to the smallest subset of candidate tokens whose cumulative probability exceeds threshold p.
Examples
“Using a top-p setting of zero point nine cuts off the least likely words from being chosen.”
“Combining temperature scaling with top-p sampling produces fluent responses while eliminating nonsensical vocabulary.”
“Unlike fixed-k filtering, top-p sampling dynamically adapts the candidate pool size according to the flatness of the underlying next-token distribution.”
Examples
simple
“Using a top-p setting of zero point nine cuts off the least likely words from being chosen.”
contextual
“Combining temperature scaling with top-p sampling produces fluent responses while eliminating nonsensical vocabulary.”
complex
“Unlike fixed-k filtering, top-p sampling dynamically adapts the candidate pool size according to the flatness of the underlying next-token distribution.”
Real-World Examples
“In top-p sampling, a probability threshold p is set, and the next item in a sequence is sampled only from the smallest possible set of high-probability candidates whose cumulative probability exceeds p.” Etymology
Abbreviation for top-probability; introduced under the name 'nucleus sampling' by Ari Holtzman et al. in their influential 2019 paper on neural text generation.
Etymology adapted from Wiktionary, available under CC BY-SA 4.0.
Domains
Scan code
englishreference.com/q/top-p