The field of artificial intelligence is witnessing significant developments in vision-language models and physics reasoning. Recent research has focused on improving the performance of large vision-language models (VLMs) in various tasks, including physics problem-solving, image generation, and coreference resolution. The introduction of novel frameworks and benchmarks has enabled the evaluation of VLMs' capabilities in interactive grounding contexts, semantic drift, and physics reasoning. Notably, some studies have demonstrated the potential of VLMs in high-energy physics applications, such as neutrino event classification. Furthermore, research has explored the use of reinforcement learning and verifiable rewards to improve LLMs' ability to generate symbolic graphics programs. The development of new metrics and evaluation protocols has also facilitated a deeper understanding of VLMs' strengths and limitations. Some noteworthy papers in this area include: Physics Supernova, which introduces an AI agent that matches elite gold medalists at the International Physics Olympiad, and Interpretable Physics Reasoning and Performance Taxonomy in Vision-Language Models, which presents a framework for evaluating VLMs' understanding of 2D physics. Additionally, Adapting Vision-Language Models for Neutrino Event Classification in High-Energy Physics demonstrates the potential of VLMs in physics event classification, while Augmenting speech transcripts of VR recordings with gaze, pointing, and visual context for multimodal coreference resolution presents a system for improving coreference resolution accuracy in multimodal conversations.