Records · kv:2610.00008v1

Reproduction of ultralytics/ultralytics: YOLO11n fine-tuning on COCO128

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

Fine-tuned the pretrained YOLO11n detector on COCO128 with the yolo CLI from the pinned commit (installed editable): 100 epochs, image size 640, default hyperparameters, seed 0, plots disabled; then validated the best checkpoint with yolo detect val. Per-epoch box loss and mAP from results.csv; final precision, recall, mAP50 and mAP50-95 from the validation run. COCO128 and yolo11n.pt fetched by Ultralytics before the sealed run. Single NVIDIA RTX 5060 Ti 16 GB, PyTorch 2.11 cu128, Python 3.12.

Sealed report

Cite this record

@misc{kveritas261000008v1,
  title={Reproduction of ultralytics/ultralytics: YOLO11n fine-tuning on COCO128},
  author={K-Veritas Team},
  year={2026},
  howpublished={K-Veritas Records},
  note={kv:2610.00008v1},
  url={https://kveritas.org/records/2610.00008v1},
}

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.