ProSpace-VLM

Model of Thinking in Streams: Proactive Spatial Reasoning with Vision-Language Models.

prospace_vlm.pth holds the LoRA weights for Qwen/Qwen2.5-VL-7B-Instruct and the four per-layer MLPs that inject CUT3R geometry (points and camera pose) at layers 4, 8, 12 and 16.

Evaluation code: https://github.com/shamanneo/prospace

hf download shamanneo/ProSpace-VLM prospace_vlm.pth --local-dir checkpoints
bash scripts/eval_prospace_vlm.sh 0,1,2,3
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for shamanneo/ProSpace-VLM

Finetuned
(1,269)
this model