Event
Relational Geometry Forcing method proposed to improve world action models
1 reports · 1 sources
Yuan Xu and other researchers propose the Relational Geometry Forcing (RGF) method, which injects the geometric priors of frozen 3D foundation models into world action models (WAM) through pairwise feature-relationship alignment rather than pointwise feature alignment, in order to avoid the degradation of action learning caused by dense geometric objectives over-constraining compact action features. The method is used only during training and requires no explicit geometry branch or additional inference module. The paper states that simulation and real-world experiments show that RGF yields consistent gains across a variety of manipulation scenarios and WAM paradigms.
Report timeline
- Relational Geometry Forcing: Improving Robot Manipulation Learning in World Action Models (WAMs) via Relational Geometry Constraints
The paper proposes Relational Geometry Forcing (RGF), which injects geometric priors from a frozen 3D foundation model into World Action Models (WAMs) by aligning pairwise feature relations instead of pointwise feature alignment, avoiding the degradation of action learning caused by dense geometric targets over-constraining compact action features.
arXiv cs.RO ·