mlsanityData to 7 Oct 2026

XStreamVGGT: Extremely Memory-Efficient Streaming Vision Geometry Grounded Transformer with KV Cache Compression

arXiv 2601.01204 ↗First posted 3 Jan 2026cs.CVcited by 6

On the map: Efficient LLM inference and serving → Token pruning and merging for ViTs

Abstract (arXiv)

Learning-based 3D visual geometry models have benefited substantially from large-scale transformers. Among these, StreamVGGT leverages frame-wise causal attention for strong streaming reconstruction, but suffers from unbounded KV cache growth, leading to escalating memory consumption and inference latency as input frames accumulate. We propose XStreamVGGT, a tuning-free approach that systematically compresses the KV cache through joint pruning and quantization, enabling extremely memory-efficient streaming inference. Specifically, redundant KVs originating from multi-view inputs are pruned through efficient token importance identification, enabling a fixed memory budget. Leveraging the unique distribution of KV tensors, we incorporate KV quantization to further reduce memory consumption. Extensive evaluations show that XStreamVGGT achieves mostly negligible performance degradation while substantially reducing memory usage by 4.42× and accelerating inference by 5.48×, enabling scalable and practical streaming 3D applications. The code is available at https://github.com/ywh187/XStreamVGGT/.

Key facts. Cited by 6 papers in the map (data through 7 Oct 2026). Placed in Token pruning and merging for ViTs, one of 12 directions in Efficient LLM inference and serving. Builds on 27 of its 42 references that are in the map. First posted on arXiv 3 Jan 2026.

Builds on

27 of 42 references.

References · 27

  1. 2017ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenescited by 3,088
  2. 2023FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioningcited by 2,551
  3. 2023Efficient Streaming Language Models with Attention Sinkscited by 2,115
  4. 2023DUSt3R: Geometric 3D Vision Made Easycited by 1,526
  5. 2025VGGT: Visual Geometry Grounded Transformercited by 1,475
  6. 2023H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Modelscited by 805
  7. 2024KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cachecited by 604
  8. 2024KVQuant: Towards 10 Million Context Length LLM Inference with KV Cache Quantizationcited by 571
  9. 2025Continuous 3D Perception Model with Persistent Statecited by 527
  10. 2021A White Paper on Neural Network Quantizationcited by 453
  11. 2024MonST3R: A Simple Approach for Estimating Geometry in the Presence of Motioncited by 431
  12. 2019Cascade Cost Volume for High-Resolution Multi-View Stereo and Stereo Matchingcited by 414
15 more
  1. 2021Neural RGB-D Surface Reconstructioncited by 312
  2. 2019ReFusion: 3D Reconstruction in Dynamic Environments for RGB-D Cameras Exploiting Residualscited by 180
  3. 2025Streaming 4D Visual Geometry Transformercited by 147
  4. 2024Fit and Prune: Fast and Training-free Visual Token Pruning for Multi-modal Large Language Modelscited by 125
  5. 2024A Survey on Large Language Model Acceleration based on KV Cache Managementcited by 125
  6. 2025TTT3R: 3D Reconstruction as Test-Time Trainingcited by 122
  7. 2025Point3R: Streaming 3D Reconstruction with Explicit Spatial Pointer Memorycited by 82
  8. 2023Milestones in Autonomous Driving and Intelligent Vehicles: Survey of Surveyscited by 75
  9. 2025LongCat-Flash Technical Reportcited by 54
  10. 2025RotateKV: Accurate and Robust 2-Bit KV Cache Quantization for LLMs via Outlier-Aware Adaptive Rotationscited by 42
  11. 2023Robotic Vision for Human-Robot Interaction and Collaboration: A Survey and Systematic Reviewcited by 25
  12. 2025Introducing LongCat-Flash-Thinking: A Technical Reportcited by 21
  13. 2025KVSink: Understanding and Enhancing the Preservation of Attention Sinks in KV Cache Quantization for LLMscited by 18
  14. 2024Accelerating Multimodal Large Language Models via Dynamic Visual-Token Exit and the Empirical Findingscited by 15
  15. 2025AKVQ-VL: Attention-Aware KV Cache Adaptive 2-Bit Quantization for Vision-Language Modelscited by 9

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 6.

6 papers.

  1. 2026OScaR: The Occam's Razor for Extreme KV Cache Quantization in LLMs and Beyondcited by 5
  2. 2026FrameVGGT: Coherence-Preserving Memory for Bounded Streaming Geometrycited by 2
  3. 2026GeoFF3D: Coordinate-Anchored Feed-Forward Reconstruction for Large-Scale UAV Mappingcited by 1
  4. 20263DTurboQuant: Training-Free Near-Optimal Quantization for 3D Reconstruction Modelscited by 0
  5. 2026Stream3D-VLM: Online 3D Spatial Understanding with Incremental Geometry Priorscited by 0
  6. 2026RegVGGT: Sustainable Visual Geometry Grounding for Streaming via Regulated Memorycited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026XStreamVGGT: Extremely Memory-Efficient Streaming Vision Geometry Grounded Transformer with KV Cache Compression
  2. 2026OVGGT: O(1) Constant-Cost Streaming Visual Geometry Transformer
  3. 2026InfiniteVGGT: Visual Geometry Grounded Transformer for Endless Streams
  4. 2025Evict3R: Training-Free Token Eviction for Memory-Bounded Streaming Visual Geometry Transformers
  5. 2026LongStream: Long-Sequence Streaming Autoregressive Visual Geometry
  6. 2025LiteVGGT: Boosting Vanilla VGGT via Geometry-aware Cached Token Merging
  7. 2026Attention Itself Could Retrieve.RetrieveVGGT: Training-Free Long Context Streaming 3D Reconstruction via Query-Key Similarity Retrieval
  8. 2025FastVGGT: Training-Free Acceleration of Visual Geometry Transformer