研究人员开发了一个新的框架，通过结合自蒸馏和强化学习来训练视频扩散模型以解决通用任务。该方法允许模型从无标签数据中学习解决任务的能力，绕过了昂贵的、经过策划的任务-视频监督的需求。该方法使用视觉语言模型生成任务和解决方案，然后指导视频扩散模型学习执行，并通过来自视觉语言模型的反馈进行强化学习进一步增强。 AI
影响 使视频扩散模型能够在没有明确的任务-视频数据的情况下执行复杂任务，有可能加速机器人和规划应用。
排序理由 该集群包含一篇详细介绍训练AI模型新方法的论文。
在 Hugging Face Daily Papers 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
A scalable framework combines self-distillation and reinforcement learning to transfer task-solving abilities from vision-language models to video diffusion models without requiring labeled task-video data.
arXiv:2606.12072v1 Announce Type: new Abstract: Pretrained video generators are promising visual world models that exhibit emergent task-solving abilities; however, their reliance on detailed textual descriptions limits their direct use for planning and decision-making. Existing …
Pretrained video generators are promising visual world models that exhibit emergent task-solving abilities; however, their reliance on detailed textual descriptions limits their direct use for planning and decision-making. Existing approaches either outsource this reasoning to la…