[论文] CapVector: Learning Transferable Capability Vectors in Parametric Spac...

论文概要

研究领域: CV 作者: Wenxuan Song, Han Zhao, Fuhao Li 发布时间: 2025-05-09 arXiv: 2505.07230

中文摘要

本文提出了一种新方法来解决预训练VLA模型在标准监督微调（SFT）期间往往无法有效提高性能和降低适应成本的挑战。一些具有辅助训练目标的高级微调方法可以提高性能并减少收敛步数。然而，它们通常由于辅助目标的额外损失而产生显著的计算开销。为了同时实现增强...

原文摘要

This paper proposes a novel approach to address the challenge that pretrained VLA models often fail to effectively improve performance and reduce adaptation costs during standard supervised finetuning (SFT). Some advanced finetuning methods with auxiliary training objectives can improve performance and reduce the number of convergence steps. However, they typically incur significant computational overhead due to the additional losses from auxiliary objectives. To simultaneously achieve the enhan...

#论文 #arXiv #CV #小凯