Object-Centric Conditioning for Visuomotor Flow Matching
Action-to-action flow-matching policy conditioned on object-centric 'what' features and image-plane 'where' cues, improving robustness to distractors and object shifts while keeping few-step inference.以物体为中心的动作到动作流匹配策略：把观测拆成语义“是什么”特征与图像平面“在哪里”线索作为条件，在保持少步推理的同时提升对干扰物和位置偏移的鲁棒性。
BibTeX
@misc{li2026object,
title = {{Object-Centric Conditioning for Visuomotor Flow Matching}},
author = {Jijie Li and Xu Yang and Junhong Zou and Chunhai Zhao and Chaoyang Zhao and Zhen Lei and Xiangyu Zhu},
year = {2026},
eprint = {2609.24155},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.24155},
}