StyleForge: Indoor Furniture Styling by Counterfactual Reasoning in a Hypergraph Field
StyleForge picks furniture as a room-level compatibility problem, not a slot-by-slot retrieval task.
The paper targets fixed layouts where furniture categories, positions, orientations, and scales cannot change. Its system uses a frozen multimodal LLM to extract style priors, then models furniture slots with a dynamic hypergraph that weighs higher-order dependencies across the room. Candidate pieces are tested as local substitutions against the evolving scene context, with training and inference aimed at reducing cross-slot style conflicts. The authors report state-of-the-art retrieval and stronger scene-level coherence on 3D-FRONT. HF Daily Papers' note
The paper targets fixed layouts where furniture categories, positions, orientations, and scales cannot change. Its system uses a frozen multimodal LLM to extract style priors, then models furniture slots with a dynamic hypergraph that weighs higher-order dependencies across the room. Candidate pieces are tested as local substitutions against the evolving scene context, with training and inference aimed at reducing cross-slot style conflicts. The authors report state-of-the-art retrieval and stronger scene-level coherence on 3D-FRONT. HF Daily Papers' note
score 4