Guardado en:
| Autores principales: | , , , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | https://arxiv.org/abs/2604.20570 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866915949202898944 |
|---|---|
| author | Zhu, Muzhi Jiang, Shunyao Zheng, Huanyi Luo, Zekai Zhong, Hao Li, Anzhou Wang, Kaijun Rong, Jintao Liu, Yang Chen, Hao Lin, Tao Shen, Chunhua |
| author_facet | Zhu, Muzhi Jiang, Shunyao Zheng, Huanyi Luo, Zekai Zhong, Hao Li, Anzhou Wang, Kaijun Rong, Jintao Liu, Yang Chen, Hao Lin, Tao Shen, Chunhua |
| contents | Spatial intelligence is essential for multimodal large language models, yet current benchmarks largely assess it only from an understanding perspective. We ask whether modern generative or unified multimodal models also possess generative spatial intelligence (GSI), the ability to respect and manipulate 3D spatial constraints during image generation, and whether such capability can be measured or improved. We introduce GSI-Bench, the first benchmark designed to quantify GSI through spatially grounded image editing. It consists of two complementary components: GSI-Real, a high-quality real-world dataset built via a 3D-prior-guided generation and filtering pipeline, and GSI-Syn, a large-scale synthetic benchmark with controllable spatial operations and fully automated labeling. Together with a unified evaluation protocol, GSI-Bench enables scalable, model-agnostic assessment of spatial compliance and editing fidelity. Experiments show that fine-tuning unified multimodal models on GSI-Syn yields substantial gains on both synthetic and real tasks and, strikingly, also improves downstream spatial understanding. This provides the first clear evidence that generative training can tangibly strengthen spatial reasoning, establishing a new pathway for advancing spatial intelligence in multimodal models. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_20570 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Exploring Spatial Intelligence from a Generative Perspective Zhu, Muzhi Jiang, Shunyao Zheng, Huanyi Luo, Zekai Zhong, Hao Li, Anzhou Wang, Kaijun Rong, Jintao Liu, Yang Chen, Hao Lin, Tao Shen, Chunhua Computer Vision and Pattern Recognition Spatial intelligence is essential for multimodal large language models, yet current benchmarks largely assess it only from an understanding perspective. We ask whether modern generative or unified multimodal models also possess generative spatial intelligence (GSI), the ability to respect and manipulate 3D spatial constraints during image generation, and whether such capability can be measured or improved. We introduce GSI-Bench, the first benchmark designed to quantify GSI through spatially grounded image editing. It consists of two complementary components: GSI-Real, a high-quality real-world dataset built via a 3D-prior-guided generation and filtering pipeline, and GSI-Syn, a large-scale synthetic benchmark with controllable spatial operations and fully automated labeling. Together with a unified evaluation protocol, GSI-Bench enables scalable, model-agnostic assessment of spatial compliance and editing fidelity. Experiments show that fine-tuning unified multimodal models on GSI-Syn yields substantial gains on both synthetic and real tasks and, strikingly, also improves downstream spatial understanding. This provides the first clear evidence that generative training can tangibly strengthen spatial reasoning, establishing a new pathway for advancing spatial intelligence in multimodal models. |
| title | Exploring Spatial Intelligence from a Generative Perspective |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2604.20570 |