# LEO-VL: Efficient Scene Representation for Scalable 3D Vision-Language Learning

arXiv: https://arxiv.org/abs/2506.09935 · HTML: https://mlsanity.com/paper/2506.09935/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Placed in Embodied and 3D vision-language models, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 111 of its 130 references that are in the map.
- First posted on arXiv 11 Jun 2025.

## Placed in

- Multimodal LLMs / vision-language models / [Embodied and 3D vision-language models](https://mlsanity.com/mllm/embodied-and-3d-vision-language-models/)

## Builds on (most cited first, 30 of 111)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation](https://mlsanity.com/paper/1612.00593/) (2016)
- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space](https://mlsanity.com/paper/1706.02413/) (2017)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes](https://mlsanity.com/paper/1702.04405/) (2017)
- [ShapeNet: An Information-Rich 3D Model Repository](https://mlsanity.com/paper/1512.03012/) (2015)
- [LLaVA-OneVision: Easy Visual Task Transfer](https://mlsanity.com/paper/2408.03326/) (2024)
- [InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks](https://mlsanity.com/paper/2312.14238/) (2023)
- [Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains](https://mlsanity.com/paper/2006.10739/) (2020)
- [Point Transformer](https://mlsanity.com/paper/2011.00931/) (2020)
- [Matterport3D: Learning from RGB-D Data in Indoor Environments](https://mlsanity.com/paper/1709.06158/) (2017)
- [Kimi k1.5: Scaling Reinforcement Learning with LLMs](https://mlsanity.com/paper/2501.12599/) (2025)
- [4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks](https://mlsanity.com/paper/1904.08755/) (2019)
- [PointCNN: Convolution On X-Transformed Points](https://mlsanity.com/paper/1801.07791/) (2018)
- [ScanNet++: A High-Fidelity Dataset of 3D Indoor Scenes](https://mlsanity.com/paper/2308.11417/) (2023)
- [3D Semantic Segmentation with Submanifold Sparse Convolutional Networks](https://mlsanity.com/paper/1711.10275/) (2017)
- [Self-Rewarding Language Models](https://mlsanity.com/paper/2401.10020/) (2024)
- [SFT Memorizes, RL Generalizes: A Comparative Study of Foundation Model Post-training](https://mlsanity.com/paper/2501.17161/) (2025)
- [PointConv: Deep Convolutional Networks on 3D Point Clouds](https://mlsanity.com/paper/1811.07246/) (2018)
- [Point-BERT: Pre-training 3D Point Cloud Transformers with Masked Point Modeling](https://mlsanity.com/paper/2111.14819/) (2021)

## Built on by (most cited first, 5 of 5)

- [DreamArt: Generating Interactable Articulated Objects from a Single Image](https://mlsanity.com/paper/2507.05763/) (2025)
- [SceneCOT: Eliciting Grounded Chain-of-Thought Reasoning in 3D Scenes](https://mlsanity.com/paper/2510.16714/) (2025)
- [3D-RFT: Reinforcement Fine-Tuning for Video-based 3D Scene Understanding](https://mlsanity.com/paper/2603.04976/) (2026)
- [PAR3D: A Unified 3D-MLLM with Part-Aware Representation for Scene Understanding](https://mlsanity.com/paper/2606.06485/) (2026)
- [What Makes a Good Visual Representation for Spatial Understanding in MLLMs?](https://mlsanity.com/paper/2610.17844/) (2026)

## Abstract (arXiv)

Developing vision-language models (VLMs) capable of understanding 3D scenes has been a longstanding research goal. Despite recent progress, 3D VLMs still struggle with spatial reasoning and robustness. We identify three key obstacles hindering their progress: (1) scene representation is constrained by a capacity-efficiency trade-off, which impedes scalable learning; (2) training data lacks a comprehensive scheme, with limited diversity across tasks and scene domains; and (3) models exhibit robustness deficiencies and lack effective post-training. To address these challenges, we first propose condensed feature grid (CFG), an efficient scene representation that significantly reduces token overhead while preserving strong perceptual capacity. Building on CFG, we introduce LEO-VL, a 3D VLM trained on over 700k 3D vision-language (3D-VL) data spanning four real-world indoor domains and five tasks such as captioning and dialogue. To further improve robustness, we propose SceneDPO, a novel post-training objective that incorporates contrastive signals across both answers and scenes. LEO-VL achieves state-of-the-art performance on various 3D-VL benchmarks, such as SQA3D, Beacon3D, and Scan2Cap. Extensive analyses highlight the efficiency of CFG and provide key insights such as the importance of task and scene diversity, the priority of data quality for effective scaling, and the advantages of SceneDPO.