The field of vision-language models is moving towards unified architectures that can handle both visual understanding and generation tasks. Recent developments have focused on enabling image editing capabilities within these models, with a emphasis on training-free methods and iterative refinement processes. This has led to significant improvements in image generation and editing quality, with some models achieving state-of-the-art performance on benchmarks. Notably, the integration of vision-language models with autoregressive modeling and diffusion-based methods has shown great promise. The use of large-scale datasets and carefully designed training schedules has also been crucial in achieving high-fidelity multimodal integration. Overall, the field is advancing towards more practical and deployable multimodal AI systems. Noteworthy papers include: UniEdit-I, which introduces a novel training-free framework for image editing, and Skywork UniPic, which demonstrates a unified autoregressive model for visual understanding and generation. LumiGen is also notable for its LVLM-enhanced iterative framework for fine-grained text-to-image generation.