The paper presents BUMP, a method for synthesizing human-object interactions (HOIs) that involve multiple objects. The abstract identifies a core difficulty: the motions in such scenes arise from different underlying causes, so a single generative model may struggle to capture them all.
Human motion, the authors note, reflects semantic intent, while intentionally manipulated objects must follow a different set of constraints. BUMP is described as body-driven, suggesting the human body acts as the central reference point for coordinating these distinct physical and intentional dynamics.
The work is positioned as a unified approach to multi-object physics for HOI synthesis. Because the abstract is truncated, full evaluation details and comparisons are not available in the source, but the proposed direction is clear: treat intent and physics together rather than separately.