3 ms·DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning7 points by mikhael 1y ago