trends.barnas.net

method · first seen 8 Oct 2026

synthetic data for pretraining

Using machine-generated tasks and examples as part of a language model's pretraining corpus to improve downstream abilities.

Mentions per week

0119 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

data mixture scaling lawsatomic fact extraction