4 Matching Annotations
- Feb 2023
-
arxiv.org arxiv.org
-
Webgpt
-
-
www.arxiv-vanity.com www.arxiv-vanity.com
-
It does not do fine tuning, only in context learning
-
-
arxiv.org arxiv.org
-
Ul2 poses lots of experiments comparing different pretraining approaches. Sequential, span corruption, to r,s,x denoiser Lots of ablation study
-
-
arxiv.org arxiv.org
-
- Generate instruction via llm
- on gpt3
- with good experiment data
-