<ul class="dashed" data-apple-notes-indent-amount="0"><li><span style="font-family: '.PingFangUITextSC-Regular'">文章标题:</span>RoPECraft: Training-Free Motion Transfer with Trajectory-Guided RoPE Optimization on Diffusion Transformers</li><li><span style="font-family: '.PingFangSC-Regular'">文章地址:</span><a href="https://arxiv.org/abs/2505.13344">https://arxiv.org/abs/2505.13344</a> </li><li>arxiv</li></ul> <img src="https://res.cloudinary.com/montaigne-io/image/upload/v1751893419/F6D83E34-1A97-4F38-941E-EA8590C55A93.png" style="background-color:initial;max-width:min(100%,1808px);max-height:min(878px);;background-image:url(https://res.cloudinary.com/montaigne-io/image/upload/v1751893419/F6D83E34-1A97-4F38-941E-EA8590C55A93.png);height:auto;width:100%;object-fit:cover;background-size:cover;display:block;" width="1808" height="878"> 文章利用DiTFlow的发现,利用位置编码的优化进行动作转移。其本质就是去优化一个位置编码,使其学习到动作转移的模式,从而在DiT中完成动作转移。 此外,作者还提出了一种新的评估动作转移质量的方法