@wzhao_nlp : Dang, truly impressed by how an academic lab just figured out a lot of mysteries in mid-training to close the RL gap between llama and qwen: * length scheduler plays a key role to stabilize RL * there is some dark magic in prompt template? * the data interaction stuff is really • TwiCopy

Wenting Zhao

@wzhao_nlp

+ Follow

PhD student @cornell_tech
NLP + AI

ID: 1473829704

linkhttps://wenting-zhao.github.io/ calendar_today01-06-2013 05:18:16

343 Tweet

1,1K Takipçi

495 Takip Edilen

Wenting Zhao

@wzhao_nlp

5 months ago

Dang, truly impressed by how an academic lab just figured out a lot of mysteries in mid-training to close the RL gap between llama and qwen: * length scheduler plays a key role to stabilize RL * there is some dark magic in prompt template? * the data interaction stuff is really

thumb_up_off_alt196

chat_bubble_outline3

repeat16

shareShare