MultihopSpatial: Multi-hop Compositional Spatial Reasoning Benchmark for Vision-Language Model
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912974287929344 |
|---|---|
| author | Lee, Youngwan Jang, Soojin Cho, Yoorhim Lee, Seunghwan Lee, Yong-Ju Hwang, Sung Ju |
| author_facet | Lee, Youngwan Jang, Soojin Cho, Yoorhim Lee, Seunghwan Lee, Yong-Ju Hwang, Sung Ju |
| contents | Spatial reasoning is foundational for Vision-Language Models (VLMs), particularly when deployed as Vision-Language-Action (VLA) agents in physical environments. However, existing benchmarks predominantly focus on elementary, single-hop relations, neglecting the multi-hop compositional reasoning and precise visual grounding essential for real-world scenarios. To address this, we introduce MultihopSpatial, offering three key contributions: (1) A comprehensive benchmark designed for multi-hop and compositional spatial reasoning, featuring 1- to 3-hop complex queries across diverse spatial perspectives. (2) Acc@50IoU, a complementary metric that simultaneously evaluates reasoning and visual grounding by requiring both answer selection and precise bounding box prediction - capabilities vital for robust VLA deployment. (3) MultihopSpatial-Train, a dedicated large-scale training corpus to foster spatial intelligence. Extensive evaluation of 37 state-of-the-art VLMs yields eight key insights, revealing that compositional spatial reasoning remains a formidable challenge. Finally, we demonstrate that reinforcement learning post-training on our corpus enhances both intrinsic VLM spatial reasoning and downstream embodied manipulation performance. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_18892 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | MultihopSpatial: Multi-hop Compositional Spatial Reasoning Benchmark for Vision-Language Model Lee, Youngwan Jang, Soojin Cho, Yoorhim Lee, Seunghwan Lee, Yong-Ju Hwang, Sung Ju Computer Vision and Pattern Recognition Artificial Intelligence Spatial reasoning is foundational for Vision-Language Models (VLMs), particularly when deployed as Vision-Language-Action (VLA) agents in physical environments. However, existing benchmarks predominantly focus on elementary, single-hop relations, neglecting the multi-hop compositional reasoning and precise visual grounding essential for real-world scenarios. To address this, we introduce MultihopSpatial, offering three key contributions: (1) A comprehensive benchmark designed for multi-hop and compositional spatial reasoning, featuring 1- to 3-hop complex queries across diverse spatial perspectives. (2) Acc@50IoU, a complementary metric that simultaneously evaluates reasoning and visual grounding by requiring both answer selection and precise bounding box prediction - capabilities vital for robust VLA deployment. (3) MultihopSpatial-Train, a dedicated large-scale training corpus to foster spatial intelligence. Extensive evaluation of 37 state-of-the-art VLMs yields eight key insights, revealing that compositional spatial reasoning remains a formidable challenge. Finally, we demonstrate that reinforcement learning post-training on our corpus enhances both intrinsic VLM spatial reasoning and downstream embodied manipulation performance. |
| title | MultihopSpatial: Multi-hop Compositional Spatial Reasoning Benchmark for Vision-Language Model |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2603.18892 |