plan to release training-corpus and checkpoints

[youngwanLEE]

Hi,
first of all, I would like to extend my warmest congratulations on your paper, "SpatialScore," being selected as a CVPR 2026 Highlight! This is a truly impressive achievement and speaks volumes about the impact of your work on spatial reasoning in Vision-Language Models.

I have found the results regarding the "SpatialCorpus" and the fine-tuned models (Qwen3-VL-4B, Qwen3-VL-8B) particularly compelling. However, I noticed that while the benchmark dataset is available, the specific training corpus and model checkpoints mentioned in the paper do not seem to be present in the current GitHub repository or on Hugging Face.

Regarding the HuggingFace repository (https://huggingface.co/datasets/haoningwu/SpatialScore), I would like to clarify if it currently only hosts the benchmark set. I have looked through the files but was unable to find training-related files such as train.json or the SpatialCorpus itself.

Could you please let me know if there are plans to release the SpatialCorpus and the trained checkpoints? Having access to these resources would be immensely helpful for my ongoing research and for further developments in this field.

Thank you for your time and for contributing such valuable work to the community. I look forward to hearing from you.

[haoningwu3639]

Hi Youngwan,

As I mentioned in my email reply, we are planning to release the updated CVPR version of the SpatialScore benchmark, the SpatialCorpus dataset, the model checkpoints, and the new codebase in about 2 to 3 weeks. Please stay tuned, and thank you again for your interest in our work!