Fang Liu^*   Yuhao Liu^*   Ke Xu^†   Gerhard Petrus Hancke   Rynson W.H. Lau^†

City University of Hong Kong

^*Equal Contribution   ^†Corresponding Authors

Unlike previous methods that either achieve cross-scene generalization by being bounded to a predefined vocabulary or handle free-form language by overfitting to individual scenes, GenSplat is robust to free-form language queries and generalizable across 3DGS scene representations. Our key insight is to formulate a structured learning process to progressively align linguistic concepts with 3D Gaussians. It contains two novel technical contributions: Progressive Language Grounding Curriculum that structurally guides the model through learning category-level semantics to instance-level concepts and free-form language, preventing overfitting by building a generalizable language feature space. MLLM-guided Reasoning Module that leverages Multi-modal Large Language Models' semantic and spatial priors to enhance 3D localization and reasoning. Extensive cross-task evaluations — including 3D referring segmentation, 3D visual question answering, and 3D open-vocabulary understanding — demonstrate state-of-the-art performances and strong generalization capability.

• Release inference code
• Release training code
• Release pre-trained model weights
• Release evaluation scripts

Code will be released soon. Stay tuned!

If you find this work useful, please consider citing: