RoboRefer: Towards Spatial Referring with Reasoning in Vision-Language Models for Robotics

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhou, Enshen, An, Jingkun, Chi, Cheng, Han, Yi, Rong, Shanyu, Zhang, Chi, Wang, Pengwei, Wang, Zhongyuan, Huang, Tiejun, Sheng, Lu, Zhang, Shanghang
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917181778821120
author Zhou, Enshen
An, Jingkun
Chi, Cheng
Han, Yi
Rong, Shanyu
Zhang, Chi
Wang, Pengwei
Wang, Zhongyuan
Huang, Tiejun
Sheng, Lu
Zhang, Shanghang
author_facet Zhou, Enshen
An, Jingkun
Chi, Cheng
Han, Yi
Rong, Shanyu
Zhang, Chi
Wang, Pengwei
Wang, Zhongyuan
Huang, Tiejun
Sheng, Lu
Zhang, Shanghang
contents Spatial referring is a fundamental capability of embodied robots to interact with the 3D physical world. However, even with the powerful pretrained vision language models (VLMs), recent approaches are still not qualified to accurately understand the complex 3D scenes and dynamically reason about the instruction-indicated locations for interaction. To this end, we propose RoboRefer, a 3D-aware VLM that can first achieve precise spatial understanding by integrating a disentangled but dedicated depth encoder via supervised fine-tuning (SFT). Moreover, RoboRefer advances generalized multi-step spatial reasoning via reinforcement fine-tuning (RFT), with metric-sensitive process reward functions tailored for spatial referring tasks. To support SFT and RFT training, we introduce RefSpatial, a large-scale dataset of 20M QA pairs (2x prior), covering 31 spatial relations (vs. 15 prior) and supporting complex reasoning processes (up to 5 steps). In addition, we introduce RefSpatial-Bench, a challenging benchmark filling the gap in evaluating spatial referring with multi-step reasoning. Experiments show that SFT-trained RoboRefer achieves state-of-the-art spatial understanding, with an average success rate of 89.6%. RFT-trained RoboRefer further outperforms all other baselines by a large margin, even surpassing Gemini-2.5-Pro by 17.4% in average accuracy on RefSpatial-Bench. Notably, RoboRefer can be integrated with various control policies to execute long-horizon, dynamic tasks across diverse robots (e,g., UR5, G1 humanoid) in cluttered real-world scenes.
format Preprint
id arxiv_https___arxiv_org_abs_2506_04308
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle RoboRefer: Towards Spatial Referring with Reasoning in Vision-Language Models for Robotics
Zhou, Enshen
An, Jingkun
Chi, Cheng
Han, Yi
Rong, Shanyu
Zhang, Chi
Wang, Pengwei
Wang, Zhongyuan
Huang, Tiejun
Sheng, Lu
Zhang, Shanghang
Robotics
Artificial Intelligence
Computer Vision and Pattern Recognition
Spatial referring is a fundamental capability of embodied robots to interact with the 3D physical world. However, even with the powerful pretrained vision language models (VLMs), recent approaches are still not qualified to accurately understand the complex 3D scenes and dynamically reason about the instruction-indicated locations for interaction. To this end, we propose RoboRefer, a 3D-aware VLM that can first achieve precise spatial understanding by integrating a disentangled but dedicated depth encoder via supervised fine-tuning (SFT). Moreover, RoboRefer advances generalized multi-step spatial reasoning via reinforcement fine-tuning (RFT), with metric-sensitive process reward functions tailored for spatial referring tasks. To support SFT and RFT training, we introduce RefSpatial, a large-scale dataset of 20M QA pairs (2x prior), covering 31 spatial relations (vs. 15 prior) and supporting complex reasoning processes (up to 5 steps). In addition, we introduce RefSpatial-Bench, a challenging benchmark filling the gap in evaluating spatial referring with multi-step reasoning. Experiments show that SFT-trained RoboRefer achieves state-of-the-art spatial understanding, with an average success rate of 89.6%. RFT-trained RoboRefer further outperforms all other baselines by a large margin, even surpassing Gemini-2.5-Pro by 17.4% in average accuracy on RefSpatial-Bench. Notably, RoboRefer can be integrated with various control policies to execute long-horizon, dynamic tasks across diverse robots (e,g., UR5, G1 humanoid) in cluttered real-world scenes.
title RoboRefer: Towards Spatial Referring with Reasoning in Vision-Language Models for Robotics
topic Robotics
Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2506.04308