Embodied3DBench: Benchmarking Low-Level Embodied Spatial Intelligence of Vision Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Jiyao, Zhang, Mingxu, Peng, Yitong, Liu, Haoxuan, Wang, Chenshuo, Long, Yuxing, Huang, Haoyang, Li, Dongjiang, Duan, Nan, Shen, Hui, Dong, Hao
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918528802619392
author Zhang, Jiyao
Zhang, Mingxu
Peng, Yitong
Liu, Haoxuan
Wang, Chenshuo
Long, Yuxing
Huang, Haoyang
Li, Dongjiang
Duan, Nan
Shen, Hui
Dong, Hao
author_facet Zhang, Jiyao
Zhang, Mingxu
Peng, Yitong
Liu, Haoxuan
Wang, Chenshuo
Long, Yuxing
Huang, Haoyang
Li, Dongjiang
Duan, Nan
Shen, Hui
Dong, Hao
contents Are current Vision Language Models (VLMs) ready to comprehend and reason about complex embodied interactions in 3D environments? We introduce Embodied3DBench, a robot-centric benchmark targeting low-level spatial intelligence in embodied 3D environments. To systematically evaluate these foundational perceptual capabilities, the benchmark includes 6 task categories divided into two core groups: Spatial Structural Understanding (Grounding, Spatial Relation Prediction, and Multi-view Correspondence) and Interaction-Oriented Perception (Affordance Prediction, Grasp Point Prediction, and Trajectory Prediction). The benchmark spans 12 subcategories and contains over 21k high-quality question-answer pairs. We evaluate 13 state-of-the-art models, and the results show that while current models exhibit relatively strong high-level spatial reasoning, such as understanding object-to-object positional relations, they remain fragile in interaction-oriented perception, highlighting a significant lack of robust 3D-aware interaction priors. To actively bridge this capability gap revealed by our benchmark, we further synthesize a large-scale training dataset comprising 1.3M QA pairs. Notably, fine-tuning on this dataset yields significant improvements in low-level spatial intelligence. Ultimately, Embodied3DBench fills a critical gap by providing both a systematic evaluation framework and a scalable data solution, setting a clear target for the development of interaction-aware multimodal systems.
format Preprint
id arxiv_https___arxiv_org_abs_2605_29074
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Embodied3DBench: Benchmarking Low-Level Embodied Spatial Intelligence of Vision Language Models
Zhang, Jiyao
Zhang, Mingxu
Peng, Yitong
Liu, Haoxuan
Wang, Chenshuo
Long, Yuxing
Huang, Haoyang
Li, Dongjiang
Duan, Nan
Shen, Hui
Dong, Hao
Computer Vision and Pattern Recognition
Robotics
Are current Vision Language Models (VLMs) ready to comprehend and reason about complex embodied interactions in 3D environments? We introduce Embodied3DBench, a robot-centric benchmark targeting low-level spatial intelligence in embodied 3D environments. To systematically evaluate these foundational perceptual capabilities, the benchmark includes 6 task categories divided into two core groups: Spatial Structural Understanding (Grounding, Spatial Relation Prediction, and Multi-view Correspondence) and Interaction-Oriented Perception (Affordance Prediction, Grasp Point Prediction, and Trajectory Prediction). The benchmark spans 12 subcategories and contains over 21k high-quality question-answer pairs. We evaluate 13 state-of-the-art models, and the results show that while current models exhibit relatively strong high-level spatial reasoning, such as understanding object-to-object positional relations, they remain fragile in interaction-oriented perception, highlighting a significant lack of robust 3D-aware interaction priors. To actively bridge this capability gap revealed by our benchmark, we further synthesize a large-scale training dataset comprising 1.3M QA pairs. Notably, fine-tuning on this dataset yields significant improvements in low-level spatial intelligence. Ultimately, Embodied3DBench fills a critical gap by providing both a systematic evaluation framework and a scalable data solution, setting a clear target for the development of interaction-aware multimodal systems.
title Embodied3DBench: Benchmarking Low-Level Embodied Spatial Intelligence of Vision Language Models
topic Computer Vision and Pattern Recognition
Robotics
url https://arxiv.org/abs/2605.29074