Proactive-Self-Refine-in-LLMs. [ICLR26] An open-source project dedicated to training language models to proactively refine their outputs during generation via reinforcement learning.