Records · kv:2610.00022v1

Reproduction of nerfstudio-project/gsplat: 3D Gaussian Splatting on Mip-NeRF 360 garden

K-Veritas Team

Submitted by K-Veritas Team

Published 2026-10-08

Independent reproduction · Artifact evaluation · Computer vision

PaperReproduction of: original code

Abstract

Trained 3D Gaussian Splatting with gsplat's simple_trainer.py default strategy at the pinned commit on the Mip-NeRF 360 garden scene at quarter resolution, 30,000 steps, with the built-in held-out evaluation (PSNR, SSIM, LPIPS) at 7,000 and 30,000 steps. gsplat and its CUDA dependencies were compiled for sm_120 with CUDA 12.8 and PyTorch 2.9.1 as pinned. One NVIDIA RTX 5060 Ti.

Sealed report

Cite this record

@misc{kveritas261000022v1,
  title={Reproduction of nerfstudio-project/gsplat: 3D Gaussian Splatting on Mip-NeRF 360 garden},
  author={K-Veritas Team},
  year={2026},
  howpublished={K-Veritas Records},
  note={kv:2610.00022v1},
  url={https://kveritas.org/records/2610.00022v1},
}

References

1. [1] Mamadou K. Keita and Christopher Homan. Computer Science Conferences Should Require Nonrepudiable Experimental Results. NeurIPS 2026 Position Paper Track. arXiv:2605.08586.

Sealed with K-Veritas [1]. Listed, not endorsed. A record shows these results came from the sealed code, unchanged. Whether the work is sound is for the reader to judge.