Any4D: Open-Prompt 4D Generation from Natural Language and Images
Any4D: 从自然语言和图像生成开放提示的4D生成
专题命中 通用世界模型 :world model(summary_cn,abstract);world models(summary_cn,abstract);embodied world model(summary_cn,abstract);world model(summary_cn,abstract)
AI总结 本文提出Primitive Embodied World Models,通过限制视频生成时间范围,实现语言与视觉表示的细粒度对齐,降低学习复杂度,提升数据效率,并减少推理延迟,支持复杂任务的组合泛化。
Comments The authors identified issues in the 4D generation pipeline and evaluation that affect result validity. To ensure scientific accuracy, we will revise the methodology and experiments thoroughly before resubmitting. This version should not be cited or relied upon