Records · kv:2610.00020v1

Reproduction of yandex-research/tabular-dl-tabr: TabR on California housing, 15 seeds

K-Veritas Team

Submitted by K-Veritas Team

Published 2026-10-08

Independent reproduction · Artifact evaluation · Machine learning

PaperReproduction of: original code

Abstract

Re-ran TabR on the California housing dataset at the pinned commit with the paper's 15 tuned evaluation configs (exp/tabr/california/0-evaluation, seeds 0 to 14), and compared each test RMSE with the report stored in the repository for the same seed. Environment: PyTorch 2.11 with CUDA 12.8 in place of the pinned 1.13 (the GPU needs it), conda-forge faiss-gpu 1.14, TORCH_FORCE_NO_WEIGHTS_ONLY_LOAD=1 for the checkpoint format. Datasets and stored results for other experiments were moved out of the working copy; Python's hostname lookup returns localhost so it does not appear in log file names. One NVIDIA RTX 5060 Ti.

Sealed report

Cite this record

@misc{kveritas261000020v1,
  title={Reproduction of yandex-research/tabular-dl-tabr: TabR on California housing, 15 seeds},
  author={K-Veritas Team},
  year={2026},
  howpublished={K-Veritas Records},
  note={kv:2610.00020v1},
  url={https://kveritas.org/records/2610.00020v1},
}

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.