ViewFusion

Official resources for ViewFusion: Structured Spatial Thinking Chains for Multi-View Reasoning.

Paper

Overview

ViewFusion is a framework for multi-view reasoning that introduces structured spatial thinking chains to improve cross-view understanding and spatial reasoning.

This repository provides resources related to the paper, such as model weights, datasets, or other project materials.

Citation

If you find this work useful, please cite:

@misc{tao2026viewfusionstructuredspatialthinking,
  title={ViewFusion: Structured Spatial Thinking Chains for Multi-View Reasoning},
  author={Xingjian Tao and Yiwei Wang and Yujun Cai and Yifan Song and Jing Tang},
  year={2026},
  eprint={2603.06024},
  archivePrefix={arXiv},
  primaryClass={cs.CL},
  url={https://arxiv.org/abs/2603.06024},
}
Downloads last month
7
Safetensors
Model size
570k params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for xjtao/ViewFusion

Finetuned
(209)
this model

Paper for xjtao/ViewFusion