研究人员推出了一种新颖的视频预训练方法 LeVJEPA，该方法在保持或提高下游准确性的同时，显著降低了计算成本。该方法绕过了常见的启发式方法，如架构不对称和 EMA 目标编码器，而是使用单个编码器配合不变性损失和 SIGReg 正则化。同时，LAION-BVD 数据集已发布，提供千万小时的视频数据用于多模态预训练，这些数据源自通过 CommonCrawl 收集的 8000 万个视频。该数据集旨在通过提供大规模、开放访问的合成字幕资源来推动视频、音频和图像模态的研究。 AI
影响 LeVJEPA 的效率可以加速视频模型训练，而 LAION-BVD 为多模态人工智能开发提供了大规模资源。
排序理由 该集群包含两篇研究论文，详细介绍了视频预训练的新方法和数据集。
在 Hugging Face Daily Papers 阅读 →
AI 生成摘要 · Google Gemini · 来自 4 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
arXiv:2608.27395v1 Announce Type: cross Abstract: Video carries the temporal structure of the physical world, yet learning representations from it has remained computationally expensive: prevailing self-supervised methods either prevent representation collapse through architectur…
arXiv:2608.24845v1 Announce Type: cross Abstract: We present LAION-BVD, a large-scale open video dataset for multimodal learning, which contains 1.3B platform-specific video URLs collected from CommonCrawl. From these, we download 80M videos with a total duration of 10 million ho…
We present LAION-BVD, a large-scale open video dataset for multimodal learning, which contains 1.3B platform-specific video URLs collected from CommonCrawl. From these, we download 80M videos with a total duration of 10 million hours. The dataset is designed for multimodal pre-tr…
LAION-BVD is a large-scale open video dataset enabling multimodal pre-training across video, audio, and image modalities with synthetic captions and strong benchmark performance.