---
title: ConsiSpace
canonical_url: "https://www.modelscope.cn/models/BAAI/ConsiSpace"
md_url: "https://www.modelscope.cn/models/BAAI/ConsiSpace.md"
repository: BAAI/ConsiSpace
last_updated: 2026-09-05
license: cc-by-4.0
model_type:
  - qwen2_5_vl
architectures:
  - Qwen2_5_VLForConditionalGenerationWithVGGT
library_name:
  - pytorch
frameworks:
  - pytorch
downloads: 47
stars: 0
tags:
  - vision-language-model
---

# ConsiSpace

> ConsiSpace - BAAI 在 ModelScope 开源的模型。This model is the result of joint research conducted by Hao Tang's team at the School of Computer Science, Peking University, and the Beijing Academy of Artificial Intelligence (BAAI).

BAAI/ConsiSpace 是 ModelScope 魔搭社区上的机器学习模型，采用 cc-by-4.0 许可。

- **Repository**: BAAI/ConsiSpace
- **License**: cc-by-4.0
- **Tags**: vision-language-model
- **Downloads**: 47
- **Stars**: 0
- **Last updated**: 2026-09-05

Source: https://www.modelscope.cn/models/BAAI/ConsiSpace

---

# This model is the result of joint research conducted by Hao Tang's team at the School of Computer Science, Peking University, and the Beijing Academy of Artificial Intelligence (BAAI).

## ConsiSpace

Official model weights for **ConsiSpace**.

ConsiSpace is a geometry-consistent multimodal framework for
spatial reasoning in long-form visual observations.

📄 Paper: https://arxiv.org/abs/2607.17599

```bibtex
@article{huang2026consispace,
      title={ConsiSpace: Learning Geometric Consistency Matters for Video Spatial Reasoning}, 
      author={Ting Huang and Zhenyu Zhang and Wenyuan Huang and Jian Yang and Hao Tang},
      year={2026},
      eprint={2607.17599},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2607.17599}, 
}

```