The field of visual understanding and reasoning is moving towards more integrated and multimodal approaches, combining computer vision and natural language processing to enable more comprehensive and interpretable understanding of visual data. This is evident in the development of frameworks that unify multimodal reasoning with grounded visual understanding, allowing for more precise segmentation and region-level understanding. These approaches are leveraging the capabilities of large language models and semantic perception to generate structured visual representations and provide robust multi-granularity understanding. Notable papers include: RSVP, which introduces a novel framework for integrating cognitive reasoning with structured visual understanding, and Perceive Anything, which presents a conceptually straightforward and efficient framework for comprehensive region-level visual understanding. Refer to Anything with Vision-Language Prompts also proposes a novel task of omnimodal referring expression segmentation, addressing the limitation of current image segmentation models in providing comprehensive semantic understanding for complex queries.