Records · kv:2610.00016v1

Reproduction of Lightning-AI/litgpt: Pythia-410M LoRA fine-tuning and ARC-Easy evaluation

K-Veritas Team

Submitted by K-Veritas Team

Published 2026-10-07

Independent reproduction · Other · Natural language processing

Reproduction of: original code

Abstract

Ran the README fine-tuning example at the pinned commit: LoRA fine-tuning of EleutherAI Pythia-410M on the Alpaca2k dataset for 2 epochs with litgpt's default LoRA settings, followed by a zero-shot ARC-Easy evaluation of the merged model with litgpt evaluate (lm-evaluation-harness). One NVIDIA RTX 5060 Ti.

Sealed report

Cite this record

@misc{kveritas261000016v1,
  title={Reproduction of Lightning-AI/litgpt: Pythia-410M LoRA fine-tuning and ARC-Easy evaluation},
  author={K-Veritas Team},
  year={2026},
  howpublished={K-Veritas Records},
  note={kv:2610.00016v1},
  url={https://kveritas.org/records/2610.00016v1},
}

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.