RAGEN is the first open-source reproduction of DeepSeek-R1 for training agentic models via reinforcement learning. - View it on GitHub