: Vocabulaire du TAL: prompt tuning

NLP methods and tools > model training > prompt tuning

prompt tuning

The process of adapting LLMs to new tasks by training a small set of parameters, or prompts. (Shah, Prompt Tuning: A Powerful Technique for Adapting LLMs to New Tasks, on medium.com, 2023)

Although prompt tuning has been proven to be an efficient tuning paradigm for commonsense reasoning tasks it requires further exploration. (Huang, Li, Xu, Zhang, Gan, Zhang & Wang, 2023)
Prompt Tuning necessitates approximately 0.2 times the training cost of vanilla fine-tuning utilizing only 0.0067% of the total parameters as trainable. (Somayajula, Liang, Zhang, Singh & Xie, 2024)
Under the latest paradigm of prompt-tuning with generative PLMs they have not considered the explicit interaction between prompts and documents. (Zhang, Yang, Zhu, Lin, Xu & Liu, 2024)

http://data.loterre.fr/ark:/67375/8LP-CRW81XDS-M

RDF/XML TURTLE JSON-LD Dernière modification le 13/06/2024

Loterre