Misha Laskin
@MishaLaskin
What's interesting about the DeepSeek IMO paper is that it achieves an incredible result with no new ideas.
It combines two known ideas (GRPO reasoning + process reward models) with careful execution at scale.
Need more open work like this at the frontier coming from the West.
It combines two known ideas (GRPO reasoning + process reward models) with careful execution at scale.
Need more open work like this at the frontier coming from the West.
0 72