|
gptkbp:instanceOf
|
large language model
|
|
gptkbp:abbreviation
|
gptkb:PaLM
|
|
gptkbp:abilities
|
gptkb:public_speaker
translator
code generation
summarization
commonsense reasoning
multi-step reasoning
explanation generation
math reasoning
|
|
gptkbp:affiliation
|
gptkb:Google_Research
gptkb:Google_Brain
|
|
gptkbp:architecture
|
gptkb:transformation
|
|
gptkbp:author
|
gptkb:Aakanksha_Chowdhery
|
|
gptkbp:basedOn
|
gptkb:Pathways_system
|
|
gptkbp:citation
|
high (hundreds)
|
|
gptkbp:developedBy
|
gptkb:Google
|
|
gptkbp:hasDemo
|
no
|
|
https://www.w3.org/2000/01/rdf-schema#label
|
Pathways Language Model
|
|
gptkbp:input
|
gptkb:text
|
|
gptkbp:introducedIn
|
2022
|
|
gptkbp:license
|
proprietary
|
|
gptkbp:notableAchievement
|
improved reasoning over GPT-3
outperformed previous models on many benchmarks
strong few-shot performance
|
|
gptkbp:notableFor
|
few-shot learning
improved reasoning
scaling to 540B parameters
|
|
gptkbp:notablePublication
|
gptkb:PaLM:_Scaling_Language_Modeling_with_Pathways
|
|
gptkbp:output
|
gptkb:text
|
|
gptkbp:parameter
|
540 billion
|
|
gptkbp:relatedTo
|
gptkb:GPT-3
gptkb:LaMDA
gptkb:BERT
|
|
gptkbp:scalingLaw
|
applies
|
|
gptkbp:supports
|
English
multilingual tasks
|
|
gptkbp:trainer
|
gptkb:Wikipedia
books
self-supervised learning
unsupervised learning
conversational data
massive web corpus
|
|
gptkbp:usedFor
|
natural language processing
question answering
text generation
reasoning tasks
|
|
gptkbp:website
|
https://ai.googleblog.com/2022/04/pathways-language-model-palm-scaling-to.html
|
|
gptkbp:bfsParent
|
gptkb:PaLM
|
|
gptkbp:bfsLayer
|
5
|