The field of multimodal reasoning and large language models is rapidly advancing, with a focus on developing more efficient, accurate, and generalizable models. Recent research has explored the integration of vision and language models to improve reasoning capabilities, particularly in tasks that require complex and structured reasoning. Notably, the use of chain-of-thought (CoT) prompting has been adapted for large vision-language models (LVLMs) to enhance multi-modal reasoning. However, existing LVLMs often ignore the contents of generated rationales in CoT reasoning, highlighting the need for more effective approaches to improve the faithfulness and accuracy of CoT reasoning. To address this challenge, researchers have proposed novel decoding strategies, such as rationale-enhanced decoding (RED), which harmonizes visual and rationale information to improve reasoning accuracy. Additionally, the development of more advanced models, such as Corvid, has demonstrated notable strengths in mathematical reasoning and science problem-solving. Noteworthy papers in this area include Video Event Reasoning and Prediction by Fusing World Knowledge from LLMs with Vision Foundation Models, which proposes a novel framework that synergistically fuses a powerful Vision Foundation Model with a Large Language Model to improve video understanding and reasoning. Another notable paper is CoRE: Enhancing Metacognition with Label-free Self-evaluation in LRMs, which introduces a training-free, label-free self-evaluation framework to detect redundant reasoning patterns and improve reasoning efficiency.