Point Cloud as a Foreign Language for Multi-modal Large Language Model
This work addresses a bottleneck in 3D understanding for AI applications, offering a novel method that enhances computational efficiency and generalization, though it is incremental in advancing multi-modal capabilities.
The paper tackles the problem of semantic misalignment and computational inefficiency in 3D multi-modal large language models by introducing SAGE, an end-to-end model that directly processes raw point clouds without pre-trained encoders, achieving superior performance on benchmarks with improved efficiency and robustness.
Multi-modal large language models (MLLMs) have shown remarkable progress in integrating visual and linguistic understanding. Recent efforts have extended these capabilities to 3D understanding through encoder-based architectures that rely on pre-trained 3D encoders to extract geometric features. However, such approaches suffer from semantic misalignment between geometric and linguistic spaces, resolution sensitivity, and substantial computational overhead. In this work, we present SAGE, the first end-to-end 3D MLLM that directly processes raw point clouds without relying on a pre-trained 3D encoder. Our approach introduces a lightweight 3D tokenizer that combines geometric sampling and neighbourhood aggregation with vector quantization to convert point clouds into discrete tokens--treating 3D data as a foreign language that naturally extends the LLM's vocabulary. Furthermore, to enhance the model's reasoning capability on complex 3D tasks, we propose a preference optimization training strategy with a semantic alignment-based reward, specifically designed for open-ended 3D question answering where responses are descriptive. Extensive experiments across diverse 3D understanding benchmarks demonstrate that our end-to-end approach outperforms existing encoder-based methods while offering significant advantages in computational efficiency, generalization across LLM backbones, and robustness to input resolution variations. Code is available at: github.com/snehaputul/SAGE3D.