The field of 3D object detection is moving towards more robust and efficient methods that can handle diverse modalities and scenarios. Recent developments have focused on integrating multiple sensors and modalities, such as LiDAR, camera, and radar, to improve detection accuracy and reliability. Notable progress has been made in developing frameworks that can adapt to different environments and conditions, such as sensor dropout or unseen modality class combinations. Additionally, there is a growing interest in on-device learning and multi-modal fusion techniques that can enable efficient and accurate detection on edge devices. Some papers are particularly noteworthy, including PEFT-DML, which achieves significant training efficiency and robustness to fast motion and weather variability. BEVDilation is also notable for its LiDAR-centric multi-modal fusion approach, which prioritizes LiDAR information and effectively alleviates spatial misalignment caused by image depth estimation errors. GraphFusion3D introduces a unified framework combining multi-modal fusion with advanced feature learning, and demonstrates substantial performance improvement over existing approaches.