⚡️ Reverse Engineering OpenAI's Training Data — Pratyush Maini, Datology
Автор: Latent Space
Загружено: 2026-02-09
Просмотров: 2468
Описание:
Should you add reasoning traces to your pretraining data? There’s been a surge of academic work speculating its advantages. But do frontier labs actually do this? Turns out, we can answer confidently, by forensically analyzing how models respond to a single question about an emoji.
This is exactly what today's guest, Pratyush, did: https://x.com/pratyushmaini/status/20...
And we convened a quick ⚡️ pod to talk about it!
https://pratyushmaini.substack.com/p/...
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке: