The field of natural language processing is moving towards more effective integration of structured knowledge and improved text segmentation techniques. Recent developments have focused on leveraging reinforcement learning to generate and adapt structural formats for multi-step reasoning, as well as optimizing layout understanding for document parsing. Additionally, there has been a push towards more efficient and accurate text segmentation methods, including the use of reinforced boundary generation and verifiable rewards. These advancements have the potential to significantly improve the performance of large language models and other NLP systems. Noteworthy papers include: Structure-R1, which proposes a novel framework for dynamically leveraging structural knowledge in LLM reasoning through reinforcement learning. Infinity Parser, which introduces a reinforcement learning framework for layout-aware document parsing. olmOCR 2, which presents a powerful OCR system powered by a specialized vision language model trained using reinforcement learning with verifiable rewards. BoundRL, which proposes an efficient approach to structured text segmentation through reinforced boundary generation.