# 3DTurboQuant: Training-Free Near-Optimal Quantization for 3D Reconstruction Models

arXiv: https://arxiv.org/abs/2604.05366 · HTML: https://mlsanity.com/paper/2604.05366/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 34 of its 44 references that are in the map.
- First posted on arXiv 7 Apr 2026.

## Builds on (most cited first, 30 of 34)

- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [3D Gaussian Splatting for Real-Time Radiance Field Rendering](https://mlsanity.com/paper/2308.04079/) (2023)
- [Instant Neural Graphics Primitives with a Multiresolution Hash Encoding](https://mlsanity.com/paper/2201.05989/) (2022)
- [DUSt3R: Geometric 3D Vision Made Easy](https://mlsanity.com/paper/2312.14132/) (2023)
- [TensoRF: Tensorial Radiance Fields](https://mlsanity.com/paper/2203.09517/) (2022)
- [K-Planes: Explicit Radiance Fields in Space, Time, and Appearance](https://mlsanity.com/paper/2301.10241/) (2023)
- [KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cache](https://mlsanity.com/paper/2402.02750/) (2024)
- [Nerfstudio: A Modular Framework for Neural Radiance Field Development](https://mlsanity.com/paper/2302.04264/) (2023)
- [KVQuant: Towards 10 Million Context Length LLM Inference with KV Cache Quantization](https://mlsanity.com/paper/2401.18079/) (2024)
- [LightGaussian: Unbounded 3D Gaussian Compression with 15x Reduction and 200+ FPS](https://mlsanity.com/paper/2311.17245/) (2023)
- [Compact 3D Gaussian Representation for Radiance Field](https://mlsanity.com/paper/2311.13681/) (2023)
- [Compressed 3D Gaussian Splatting for Accelerated Novel View Synthesis](https://mlsanity.com/paper/2401.02436/) (2023)
- [HAC: Hash-grid Assisted Context for 3D Gaussian Splatting Compression](https://mlsanity.com/paper/2403.14530/) (2024)
- [EAGLES: Efficient Accelerated 3D Gaussians with Lightweight EncodingS](https://mlsanity.com/paper/2312.04564/) (2023)
- [Compact 3D Scene Representation via Self-Organizing Gaussian Grids](https://mlsanity.com/paper/2312.13299/) (2023)
- [CompGS: Smaller and Faster Gaussian Splatting with Vector Quantization](https://mlsanity.com/paper/2311.18159/) (2023)
- [TurboQuant: Online Vector Quantization with Near-optimal Distortion Rate](https://mlsanity.com/paper/2504.19874/) (2025)
- [Variable Bitrate Neural Fields](https://mlsanity.com/paper/2206.07707/) (2022)
- [ContextGS: Compact 3D Gaussian Splatting with Anchor Level Context Model](https://mlsanity.com/paper/2405.20721/) (2024)
- [3DGS.zip: A survey on 3D Gaussian Splatting Compression Methods](https://mlsanity.com/paper/2407.09510/) (2024)
- [Compressing Volumetric Radiance Fields to 1 MB](https://mlsanity.com/paper/2211.16386/) (2022)
- [HAC++: Towards 100X Compression of 3D Gaussian Splatting](https://mlsanity.com/paper/2501.12255/) (2025)
- [LP-3DGS: Learning to Prune 3D Gaussian Splatting](https://mlsanity.com/paper/2405.18784/) (2024)
- [QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead](https://mlsanity.com/paper/2406.03482/) (2024)
- [Fast Feedforward 3D Gaussian Splatting Compression](https://mlsanity.com/paper/2410.08017/) (2024)
- [SHACIRA: Scalable HAsh-grid Compression for Implicit Neural Representations](https://mlsanity.com/paper/2309.15848/) (2023)
- [How Far Can We Compress Instant-NGP-Based NeRF?](https://mlsanity.com/paper/2406.04101/) (2024)
- [PolarQuant: Quantizing KV Caches with Polar Transformation](https://mlsanity.com/paper/2502.02617/) (2025)
- [Compression of 3D Gaussian Splatting with Optimized Feature Planes and Standard Video Codecs](https://mlsanity.com/paper/2501.03399/) (2025)
- [FlexGaussian: Flexible and Cost-Effective Training-Free Compression for 3D Gaussian Splatting](https://mlsanity.com/paper/2507.06671/) (2025)

## Abstract (arXiv)

Every existing method for compressing 3D Gaussian Splatting, NeRF, or transformer-based 3D reconstructors requires learning a data-dependent codebook through per-scene fine-tuning. We show this is unnecessary. The parameter vectors that dominate storage in these models, 45-dimensional spherical harmonics in 3DGS and 1024-dimensional key-value vectors in DUSt3R, fall in a dimension range where a single random rotation transforms any input into coordinates with a known Beta distribution. This makes precomputed, data-independent Lloyd-Max quantization near-optimal, within a factor of 2.7 of the information-theoretic lower bound. We develop 3D, deriving (1) a dimension-dependent criterion that predicts which parameters can be quantized and at what bit-width before running any experiment, (2) norm-separation bounds connecting quantization MSE to rendering PSNR per scene, (3) an entry-grouping strategy extending rotation-based quantization to 2-dimensional hash grid features, and (4) a composable pruning-quantization pipeline with a closed-form compression ratio. On NeRF Synthetic, 3DTurboQuant compresses 3DGS by 3.5x with 0.02dB PSNR loss and DUSt3R KV caches by 7.9x with 39.7dB pointmap fidelity. No training, no codebook learning, no calibration data. Compression takes seconds. The code will be released (https://github.com/JaeLee18/3DTurboQuant)