# HiSC: Hierarchical Spatial Clustering Token Compression for Efficient 3D Scene Understanding

arXiv: https://arxiv.org/abs/2608.04610 · HTML: https://mlsanity.com/paper/2608.04610/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Spatial and embodied reasoning, one of 17 directions in LLM reasoning.
- Builds on 29 of its 52 references that are in the map.
- First posted on arXiv 5 Aug 2026.

## Placed in

- LLM reasoning / [Spatial and embodied reasoning](https://mlsanity.com/lr/spatial-and-embodied-reasoning/)
- Multimodal LLMs / vision-language models / [Embodied and 3D vision-language models](https://mlsanity.com/mllm/embodied-and-3d-vision-language-models/)
- Efficient LLM inference and serving / [Visual token pruning in multimodal models](https://mlsanity.com/inference/visual-token-pruning-in-multimodal-models/)

## Builds on (most cited first, 29 of 29)

- [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)
- [Token Merging: Your ViT But Faster](https://mlsanity.com/paper/2210.09461/) (2022)
- [DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsification](https://mlsanity.com/paper/2106.02034/) (2021)
- [An Image is Worth 1/2 Tokens After Layer 2: Plug-and-Play Inference Acceleration for Large Vision-Language Models](https://mlsanity.com/paper/2403.06764/) (2024)
- [Point Transformer V3: Simpler, Faster, Stronger](https://mlsanity.com/paper/2312.10035/) (2023)
- [Efficient Content-Based Sparse Attention with Routing Transformers](https://mlsanity.com/paper/2003.05997/) (2020)
- [ScanRefer: 3D Object Localization in RGB-D Scans using Natural Language](https://mlsanity.com/paper/1912.08830/) (2019)
- [ScanQA: 3D Question Answering for Spatial Scene Understanding](https://mlsanity.com/paper/2112.10482/) (2021)
- [PointLLM: Empowering Large Language Models to Understand Point Clouds](https://mlsanity.com/paper/2308.16911/) (2023)
- [SQA3D: Situated Question Answering in 3D Scenes](https://mlsanity.com/paper/2210.07474/) (2022)
- [Scan2Cap: Context-aware Dense Captioning in RGB-D Scans](https://mlsanity.com/paper/2012.02206/) (2020)
- [LLaVA-3D: A Simple yet Effective Pathway to Empowering LMMs with 3D-awareness](https://mlsanity.com/paper/2409.18125/) (2024)
- [Evo-ViT: Slow-Fast Token Evolution for Dynamic Vision Transformer](https://mlsanity.com/paper/2108.01390/) (2021)
- [Chat-Scene: Bridging 3D Scene and Large Language Models with Object Identifiers](https://mlsanity.com/paper/2312.08168/) (2023)
- [Beyond Text-Visual Attention: Exploiting Visual Cues for Effective Token Pruning in VLMs](https://mlsanity.com/paper/2412.01818/) (2024)
- [Multi3DRefer: Grounding Text Description to Multiple 3D Objects](https://mlsanity.com/paper/2309.05251/) (2023)
- [Unaligned RGB Guided Hyperspectral Image Super-Resolution with Spatial-Spectral Concordance](https://mlsanity.com/paper/2505.02109/) (2025)
- [VisionTrim: Unified Vision Token Compression for Training-Free MLLM Acceleration](https://mlsanity.com/paper/2601.22674/) (2026)
- [Beyond First Impressions: Integrating Joint Multi-modal Cues for Comprehensive 3D Representation](https://mlsanity.com/paper/2308.02982/) (2023)
- [LoopGaussian: Creating 3D Cinemagraph with Multi-view Images via Eulerian Motion Field](https://mlsanity.com/paper/2404.08966/) (2024)
- [Flow-Anything: Learning Real-World Optical Flow Estimation from Large-Scale Single-view Images](https://mlsanity.com/paper/2506.07740/) (2025)
- [Short-LVLM: Compressing and Accelerating Large Vision-Language Models by Pruning Redundant Layers](https://mlsanity.com/paper/2507.23362/) (2025)
- [SeqVLM: Proposal-Guided Multi-View Sequences Reasoning via VLM for Zero-Shot 3D Visual Grounding](https://mlsanity.com/paper/2508.20758/) (2025)
- [Real Noise Decoupling for Hyperspectral Image Denoising](https://mlsanity.com/paper/2511.17196/) (2025)
- [Supervise-assisted Multi-modality Fusion Diffusion Model for PET Restoration](https://mlsanity.com/paper/2602.11545/) (2026)
- [Enhancing Unregistered Hyperspectral Image Super-Resolution via Unmixing-based Abundance Fusion Learning](https://mlsanity.com/paper/2603.07918/) (2026)
- [LVLM_CSP: Accelerating Large Vision Language Models via Clustering, Scattering, and Pruning for Reasoning Segmentation](https://mlsanity.com/paper/2504.10854/) (2025)
- [ActionReasoning: Robot Action Reasoning in 3D Space with LLM for Robotic Brick Stacking](https://mlsanity.com/paper/2602.21161/) (2026)

## Built on by (most cited first, 1 of 1)

- [GeoGR: Spatial-Semantic Token Compression and Group Routing for Efficient 3D VLMs](https://mlsanity.com/paper/2610.13035/) (2026)

## Abstract (arXiv)

3D vision-language models (3D VLMs) enable spatial reasoning over multi-view scenes but suffer from substantial token redundancy due to duplicated observations and large uninformative regions, leading to high computational cost. Although visual token compression has shown promise in accelerating 2D VLMs, it fails to capture the structured nature of 3D scenes and leads to incomplete spatial coverage and loss of fine-grained details. In this paper, we propose HiSC, a training-free framework for hierarchical spatial clustering token compression in 3D VLMs. HiSC lifts token compression from token-level selection to cluster-level processing by organizing tokens into spatially grounded clusters using joint geometric and semantic cues. Specifically, we first introduce a spatial graph-based merging (SGraM) strategy that models cross-view redundancy as spatial connectivity and consolidates physically consistent regions, effectively merging extremely similar redundant tokens prior to LLM inference. We then propose a spatial clustering-based pruning (SCluP) paradigm within LLM inference, which performs hierarchical compression across clusters and within clusters, preserving object instance completeness while retaining fine-grained details for important regions. Extensive experiments on diverse 3D reasoning benchmarks show validate the effectiveness of HiSC, particularly under high visual token pruning ratios. Besides, HiSC achieves over 90% token reduction with minimal performance degradation. Code is accessible at https://github.com/elecreak/HiSC.