Spaces:

open-llm-leaderboard

/

open_llm_leaderboard

like 10.9k

Running on CPU Upgrade

Sorry for reporting this space itself but as a new user I can't make comments (but I can make reports). This is a followup to:
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard/discussions/287

The guy they're talking about (AIDC-ai-business) is a con artist.

The user @xxyyy123 pastes a screenshot of the user's reply in Marcoroni-70B-GPTQ but ignores the edit history involved. AIDC-ai-business literally edited and renamed the issue title from an ethical concern/report to "just a bug", and then felt that wasn't gentle enough on his ego, so renamed it again and called it a "config issue". Click on Edit history to see this:
https://i.imgur.com/G92Pytp.png

I'm not saying it's 100% proof, but that it's par-the-course typical behavior for scammers. In conjunction with other flags/factors, I'm saying it's 90% likely. The red flags (which are consistent with other scam artist attempts):

• doesn't provide detailed training information;
• vague on how they trained it;
• new user with no genuine history;
• tracing them back to Github repos shows the majority of their repositories aren't sources but clones/copies and the ones that are sources are filled with just readme's or basic HTML/CSS.

This stuff is just bizarre, like some form of narcissism / egotism where they don't care about sharing code, they care about being seen as owning popular repositories..... to hell with long-term thinking, it's that rush they get for 3 days before they're caught.

For posterity, pasting a link to my post before I submitted it:
https://i.imgur.com/Oa7d6Bw.png

Oh look what I just found. The user xxyyy123 who came out in support of this guy happened to accidentally update the README file with very very intimate knowledge of the model - and his PR was immediately merged by AIDC-ai-business. Not saying they are necessarily related, but he at minimum has a vested interest here:
https://i.imgur.com/bBy2wu8.png