DeepSeek R1 Teaching an AI to Think - Created Using NotebookLM

https://arxiv.org/abs/2501.12948 This paper introduces the DeepSeek-R1 series of reasoning models, developed by DeepSeek-AI, which leverage reinforcement learning (RL) to enhance the reasoning capabilities of large language models (LLMs). The research explores two main models: DeepSeek-R1-Zero and DeepSeek-R1, alongside several distilled smaller dense models.