Paper on open-vocabulary 3D scene graphs accepted at CVPR 2024