# Unified Scene Representation and Reconstruction for 3D Large Language Models

arXiv: https://arxiv.org/abs/2404.13044 · HTML: https://mlsanity.com/paper/2404.13044/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Builds on 40 of its 50 references that are in the map.
- First posted on arXiv 19 Apr 2024.

## Builds on (most cited first, 30 of 40)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [Segment Anything](https://mlsanity.com/paper/2304.02643/) (2023)
- [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://mlsanity.com/paper/2301.12597/) (2023)
- [PaLM: Scaling Language Modeling with Pathways](https://mlsanity.com/paper/2204.02311/) (2022)
- [PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space](https://mlsanity.com/paper/1706.02413/) (2017)
- [OPT: Open Pre-trained Transformer Language Models](https://mlsanity.com/paper/2205.01068/) (2022)
- [ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes](https://mlsanity.com/paper/1702.04405/) (2017)
- [PaLM-E: An Embodied Multimodal Language Model](https://mlsanity.com/paper/2303.03378/) (2023)
- [Depth Map Prediction from a Single Image using a Multi-Scale Deep Network](https://mlsanity.com/paper/1406.2283/) (2014)
- [Direct Voxel Grid Optimization: Super-fast Convergence for Radiance Fields Reconstruction](https://mlsanity.com/paper/2111.11215/) (2021)
- [PIFu: Pixel-Aligned Implicit Function for High-Resolution Clothed Human Digitization](https://mlsanity.com/paper/1905.05172/) (2019)
- [Deep Hough Voting for 3D Object Detection in Point Clouds](https://mlsanity.com/paper/1904.09664/) (2019)
- [Habitat 2.0: Training Home Assistants to Rearrange their Habitat](https://mlsanity.com/paper/2106.14405/) (2021)
- [ScanRefer: 3D Object Localization in RGB-D Scans using Natural Language](https://mlsanity.com/paper/1912.08830/) (2019)
- [Searching Efficient 3D Architectures with Sparse Point-Voxel Convolution](https://mlsanity.com/paper/2007.16100/) (2020)
- [An Embodied Generalist Agent in 3D World](https://mlsanity.com/paper/2311.12871/) (2023)
- [ScanQA: 3D Question Answering for Spatial Scene Understanding](https://mlsanity.com/paper/2112.10482/) (2021)
- [Deep Modular Co-Attention Networks for Visual Question Answering](https://mlsanity.com/paper/1906.10770/) (2019)
- [Neural-Symbolic VQA: Disentangling Reasoning from Vision and Language Understanding](https://mlsanity.com/paper/1810.02338/) (2018)
- [NeuralRecon: Real-Time Coherent 3D Reconstruction from Monocular Video](https://mlsanity.com/paper/2104.00681/) (2021)
- [Scan2Cap: Context-aware Dense Captioning in RGB-D Scans](https://mlsanity.com/paper/2012.02206/) (2020)
- [Atlas: End-to-End 3D Scene Reconstruction from Posed Images](https://mlsanity.com/paper/2003.10432/) (2020)
- [3D-VisTA: Pre-trained Transformer for 3D Vision and Text Alignment](https://mlsanity.com/paper/2308.04352/) (2023)
- [Simple but Effective: CLIP Embeddings for Embodied AI](https://mlsanity.com/paper/2111.09888/) (2021)
- [SAT: 2D Semantics Assisted Training for 3D Visual Grounding](https://mlsanity.com/paper/2105.11450/) (2021)
- [DPSNet: End-to-end Deep Plane Sweep Stereo](https://mlsanity.com/paper/1905.00538/) (2019)
- [EDA: Explicit Text-Decoupling and Dense Alignment for 3D Visual Grounding](https://mlsanity.com/paper/2209.14941/) (2022)
- [Semantic Abstraction: Open-World 3D Scene Understanding from 2D Vision-Language Models](https://mlsanity.com/paper/2207.11514/) (2022)

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

- [Unveiling the Mist over 3D Vision-Language Understanding: Object-centric Evaluation with Chain-of-Analysis](https://mlsanity.com/paper/2503.22420/) (2025)
- [SceneCOT: Eliciting Grounded Chain-of-Thought Reasoning in 3D Scenes](https://mlsanity.com/paper/2510.16714/) (2025)
- [LEO-VL: Efficient Scene Representation for Scalable 3D Vision-Language Learning](https://mlsanity.com/paper/2506.09935/) (2025)
- [Vid-LLM: A Compact Video-based 3D Multimodal LLM with Reconstruction-Reasoning Synergy](https://mlsanity.com/paper/2509.24385/) (2025)
- [3D-RFT: Reinforcement Fine-Tuning for Video-based 3D Scene Understanding](https://mlsanity.com/paper/2603.04976/) (2026)

## Abstract (arXiv)

Enabling Large Language Models (LLMs) to interact with 3D environments is challenging. Existing approaches extract point clouds either from ground truth (GT) geometry or 3D scenes reconstructed by auxiliary models. Text-image aligned 2D features from CLIP are then lifted to point clouds, which serve as inputs for LLMs. However, this solution lacks the establishment of 3D point-to-point connections, leading to a deficiency of spatial structure information. Concurrently, the absence of integration and unification between the geometric and semantic representations of the scene culminates in a diminished level of 3D scene understanding. In this paper, we demonstrate the importance of having a unified scene representation and reconstruction framework, which is essential for LLMs in 3D scenes. Specifically, we introduce Uni3DR2 extracts 3D geometric and semantic aware representation features via the frozen pre-trained 2D foundation models (e.g., CLIP and SAM) and a multi-scale aggregate 3D decoder. Our learned 3D representations not only contribute to the reconstruction process but also provide valuable knowledge for LLMs. Experimental results validate that our Uni3DR2 yields convincing gains over the baseline on the 3D reconstruction dataset ScanNet (increasing F-Score by +1.8%). When applied to LLMs, our Uni3DR2-LLM exhibits superior performance over the baseline on the 3D vision-language understanding dataset ScanQA (increasing BLEU-1 by +4.0% and +4.2% on the val set and test set, respectively). Furthermore, it outperforms the state-of-the-art method that uses additional GT point clouds on both ScanQA and 3DMV-VQA.