Reinforcement learning 10 DeepSeekR1 = CoT + RL(GRPO)

[@TheTruthOfAI]

finally a video that walks the notation of the GRPO and decomposes it properly.. unlike the 99.9% of the other videos that talks about DeepSeek-R1 .. this one is the one that truly highlights the reward/policy forward.

[@vietchuxuan8789]

Thank you, these are some good notes.