Records · kv:2610.00005v1

Reproduction of tysam-code/hlb-CIFAR10: hlb-CIFAR10 speed training

K-Veritas Team

Submitted by K-Veritas Team

Published 2026-10-07

Independent reproduction · Benchmark or leaderboard entry · Computer vision

Reproduction of: original code

Abstract

Ran main.py unmodified at the pinned commit: the script's default 25 independent CIFAR-10 training runs of the hlb-CIFAR10 network (about 12 epochs each, label smoothing, EMA), reporting per-epoch losses and accuracies and the mean and variance of the final EMA validation accuracy across runs. CIFAR-10 downloaded by torchvision. Single NVIDIA RTX 5060 Ti 16 GB, PyTorch 2.11 cu128, Python 3.12.

Sealed report

Cite this record

@misc{kveritas261000005v1,
  title={Reproduction of tysam-code/hlb-CIFAR10: hlb-CIFAR10 speed training},
  author={K-Veritas Team},
  year={2026},
  howpublished={K-Veritas Records},
  note={kv:2610.00005v1},
  url={https://kveritas.org/records/2610.00005v1},
}

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.