The field of autonomous driving is rapidly evolving, with a focus on developing more efficient, safe, and interpretable systems. Recent research has emphasized the importance of integrating Vision-Language Models (VLMs) and Reinforcement Learning (RL) to enhance decision-making and generalization in complex scenarios. Notably, the use of domain-specific VLMs, such as PlanGPT-VL, has shown significant improvements in urban planning map analysis and interpretation. Furthermore, the development of novel RL algorithms, like QC-SAC and HCRMP, has enabled more effective handling of oversteer control and collision avoidance. The incorporation of tool-aware reasoning, as seen in AgentThink, has also demonstrated promising results in boosting overall reasoning scores and answer accuracy. Additionally, the proposal of frameworks like VERDI and SOLVE has highlighted the potential of distilling VLM reasoning into modular AD stacks and synergizing VLMs with end-to-end models for enhanced autonomous vehicle planning. Some noteworthy papers include: PlanGPT-VL, which significantly outperforms general-purpose state-of-the-art VLMs in specialized planning map interpretation tasks, and iPad, which achieves state-of-the-art performance in end-to-end autonomous driving while being significantly more efficient than prior leading methods.