Ahmed Ahmed's banner

Ahmed Ahmed

@AhmedSQRD • 1,017 subscribers

CS PhD @Stanford w/ @percyliang @sanmikoyejo Funding @KnightHennessy @NSF- 🇸🇩

Shorts

Synthetic rewrites are now standard to improve data quality and when data is limited. But how much compute should you spend rewriting? And can a model trained on rewrites beat the rewriting model? We fit scaling laws to answer both: 1/🧵

Synthetic rewrites are now standard to improve data quality and when data is limited. But how much compute should you spend rewriting? And can a model trained on rewrites beat the rewriting model? We fit scaling laws to answer both: 1/🧵

32,055 views