From Thousands to Billions: 3D Visual Language Grounding via Render-Supervised Distillation from 2D VLMs

TL;DR

LIFT-GS introduces a differentiable rendering-based distillation method that leverages 2D foundation models to improve 3D vision-language grounding without requiring 3D annotations, achieving state-of-the-art results.

Contribution

It proposes a novel render-supervised distillation approach that bridges 3D and 2D models, enabling effective training of 3D vision-language models with limited 3D data.

Findings

Achieves 25.7% mAP on open-vocabulary instance segmentation.

Demonstrates 10-30% improvements on referential grounding tasks.

Pretraining doubles effective dataset size, showing strong scaling.

Abstract

3D vision-language grounding faces a fundamental data bottleneck: while 2D models train on billions of images, 3D models have access to only thousands of labeled scenes--a six-order-of-magnitude gap that severely limits performance. We introduce $\textbf{LIFT-GS}$, a practical distillation technique that overcomes this limitation by using differentiable rendering to bridge 3D and 2D supervision. LIFT-GS predicts 3D Gaussian representations from point clouds and uses them to render predicted language-conditioned 3D masks into 2D views, enabling supervision from 2D foundation models (SAM, CLIP, LLaMA) without requiring any 3D annotations. This render-supervised formulation enables end-to-end training of complete encoder-decoder architectures and is inherently model-agnostic. LIFT-GS achieves state-of-the-art results with $25.7\%$ mAP on open-vocabulary instance segmentation (vs. $20.2\%$ prior SOTA) and consistent $10-30\%$ improvements on referential grounding tasks. Remarkably, pretraining effectively multiplies fine-tuning datasets by 2X, demonstrating strong scaling properties that suggest 3D VLG currently operates in a severely data-scarce regime. Project page: https://liftgs.github.io