2026-05-29原文

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

Embodied3DBench targets low-level spatial intelligence in embodied 3D environments, with 6 task categories and over 21k QA pairs. Evaluations of 13 models show strong high-level reasoning but weak interaction-oriented perception. A synthesized dataset of 1.3M QA pairs significantly improves performance after fine-tuning.

Article intelligence

InvestorsAdvanced

Key points

Benchmark focuses on low-level embodied spatial intelligence for VLMs
Includes spatial structural understanding and interaction-oriented perception
13 models evaluated; interaction perception remains fragile
Fine-tuning on 1.3M QA pairs yields substantial improvements

Why it matters

This matters because benchmark focuses on low-level embodied spatial intelligence for VLMs.

Technical impact

May affect model selection, inference cost, product capability, and evaluation benchmarks.

[2605.29074] Embodied3DBench: Benchmarking Low-Level Embodied Spatial Intelligence of Vision Language Models

[Submitted on 27 May 2026]

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

View a PDF of the paper titled Embodied3DBench: Benchmarking Low-Level Embodied Spatial Intelligence of Vision Language Models, by Jiyao Zhang and 10 other authors

View PDF

Abstract: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.

Subjects:

Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)

Cite as: arXiv:2605.29074 [cs.CV]

(or arXiv:2605.29074v1 [cs.CV] for this version)

https://doi.org/10.48550/arXiv.2605.29074

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Jiyao Zhang [view email] [v1] Wed, 27 May 2026 20:28:56 UTC (5,868 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Embodied3DBench: Benchmarking Low-Level Embodied Spatial Intelligence of Vision Language Models, by Jiyao Zhang and 10 other authors

View PDF

TeX Source

view license

Current browse context:

cs.CV

new | recent | 2026-05

Change to browse by:

cs cs.RO

References & Citations

NASA ADS

Google Scholar

Semantic Scholar

Data provided by:

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)