Add Both Links to Paper | Generated by AI

Home 2026.06

Question: Should I add the GitHub repo link (https://github.com/lzwjava/sec-edgar-gpt) and the Hugging Face model link (https://huggingface.co/lzwjava/sec-edgar-gpt-124m-hf) into the paper (.tex), given the content won’t change much in the near future — or should I skip the paper and just add them to a Twitter post instead?

Answer:

Yes, put both links in the paper. This is the right call, and the reasoning is concrete:

A paper is the canonical, archival artifact. A tweet is ephemeral. Six months from now, someone who finds your .tex/PDF (via arXiv, Google Scholar, a citation, or your notes site) needs the code and weights reachable from the document itself. Twitter posts decay — they’re unsearchable after a week, die if the account changes, and nobody cites a tweet. The paper is what gets indexed, forwarded, and referenced. Reproducibility links belong where the claims live.

Convention strongly supports it. Modern ML papers routinely include \footnote{Code: ...} on the first page and a dedicated reproducibility/artifacts section. Reviewers and readers expect “Code available at” and “Weights available at.” A GPT-2 124M reproduction paper without a repo + checkpoint link is strictly weaker.

The “content won’t change” point actually argues for adding it, not against. Stable links are exactly what you want in an archival document. The risk with papers is linking to something that rot-changes; since your repo and the HF model are stable, the paper link stays valid. If you’re worried about future churn, pin a release tag and a specific HF revision:

That way the paper points at an immutable snapshot even if main evolves.

Concrete placement in the .tex:

First-page footnote on the title (most visible):

\title{SEC-EDGAR-GPT: ...%
\thanks{Code: \url{https://github.com/lzwjava/sec-edgar-gpt} \quad
Model: \url{https://huggingface.co/lzwjava/sec-edgar-gpt-124m-hf}}}

And a short reproducibility section near the end:

\section{Code and Model Availability}
The training code, configuration, and dataset pipeline are released at
\url{https://github.com/lzwjava/sec-edgar-gpt}. Pretrained weights
(124M parameters, HF format) are available at
\url{https://huggingface.co/lzwjava/sec-edgar-gpt-124m-hf}.

Make sure you have \usepackage{hyperref} (and url is pulled in by it) so the links are clickable in the PDF.

Do both — it’s not either/or. Paper = canonical record (must have the links). Twitter = distribution/marketing (drives traffic to the paper and repo). The tweet links to the paper; the paper links to the artifacts. That’s the correct directionality. Cross-link them: the GitHub README should also link the paper PDF and the HF model, so each entry point reaches the other two.

So: add both links to the .tex now, pin them to a tag/revision if you want immutability, then announce on Twitter with a link back.

Back Donate