机器之心

MeanFlow再下一城，北大提出机器人学习新范式MP1，实现速度与成功率双SOTA

作者

北大学者提出MP1框架，引入MeanFlow范式提升机器人动作生成速度和泛化能力。该方法实现毫秒级推理速度，且无需外部一致性约束，大幅提高任务成功率和数据效率。

ICCV高分论文｜可灵ReCamMaster在海外爆火，带你从全新角度看好莱坞大片

作者

浙江大学博士生团队提出ReCamMaster，一种可以沿新相机轨迹重新运镜的视频生成模型。该模型性能显著优于baseline方法，并发布高质量多相机同步拍摄数据集MultiCamVideo-Dataset。

这才是IMO奥赛战神：满分，5战3金，刚被MIT录取

作者

Canada’s Warren Bei is one of five IMO contestants who scored a perfect 42/42. He won three gold medals and two silver medals over the past few years, showcasing his exceptional talent in mathematics. His success was influenced by an early passion for logic puzzles and science, leading him to participate in math competitions. Despite being praised for his achievements, he is still undecided about pursuing academic research in future.