Records · kv:2610.00012v1

Reproduction of rasbt/LLMs-from-scratch: GPT-2 fine-tuning for spam classification

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 chapter 6 script gpt_class_finetune.py unmodified at the pinned commit: loads the pretrained GPT-2 124M weights, replaces the output head with a two-class classifier, and fine-tunes on the UCI SMS Spam Collection for 5 epochs with AdamW (lr 5e-5), reporting training and validation loss and accuracy. One NVIDIA RTX 5060 Ti.

Sealed report

Cite this record

@misc{kveritas261000012v1,
  title={Reproduction of rasbt/LLMs-from-scratch: GPT-2 fine-tuning for spam classification},
  author={K-Veritas Team},
  year={2026},
  howpublished={K-Veritas Records},
  note={kv:2610.00012v1},
  url={https://kveritas.org/records/2610.00012v1},
}

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.