Look, Reason, Draw a Region: Recursive Refinement for Spatial Context Prediction
Spatial-TRM raises RoboSpatial-Home context accuracy to 29.51% with 11.2M trainable parameters.
Spatial-TRM is a compact recursive vision-language reasoner for predicting vacant spatial regions relative to objects. It uses frozen pretrained features, a convex region decoder, weight tying, and 11.2 million trainable parameters, for 877.8 million parameters including frozen encoders—about 15 times fewer than the reported 13B RoboPoint model. On RoboSpatial-Home it reaches 29.51% context accuracy, 9.02 points (44% relative) above a same-size single-step model, across six relations and three reference frames.
- Spatial-TRM uses frozen vision-language features and a convex region decoder.
- Only 11.2M parameters are trainable; total size is 877.8M.
- It covers six spatial relations, three frames, and open-vocabulary anchors.
- RoboSpatial-Home context accuracy is 29.51%, up 9.02 points.
- AI models
- Spatial-TRMRoboPoint
From the original article
1 min read at arxiv.orgSpatial context prediction requires visual grounding and spatial reasoning to identify vacant regions relative to objects in a scene. Spatial relationships and physical constraints remain challenging for large vision-language models (VLMs). Recursive reasoning has demonstrated strong performance on discrete and abstract reasoning tasks, motivating its use for improving spatial context prediction in natural images.
By Wajeeha Tahir · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.