Xwin-LM / Xwin-LM

Xwin-LM: Powerful, Stable, and Reproducible LLM Alignment

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Missing source code, model card, data sheet &c

mdingemanse opened this issue · comments

For a project that "aims to develop and open-source alignment technologies for large language models" the source & all other aspects are remarkably closed. At opening-up-chatgpt.github.io (the most comprehensive leaderboard of "open" LLM+RLHF systems), Xwin-LM comes in at rock bottom — only the notoriously closed ChatGPT is more closed:

image

On the release page, two weeks old as I file this issue, I noticed this to do:

[ ] Release the source code

To this, we might add releasing the RLHF procedure and datasets, model card, documentation of the architecture, and so on — all of them crucial on the way towards true open source alignment technologies. Looking forward to see Xwin-LM climbing the ranks by openness!

If anything changes with regard to the openness of Xwin-LM, feel free to file an issue at Opening up ChatGPT or notify us and we look forward to update the system metadata.

I will be patient anyway

waiting for the open source ! thanks for excellent work!