🚩 Report: Copyright infringement
Directly cloned my leaderboard (Open_SLM_Leaderboard), including the design, metrics, and scores, which is not allowed by the license of the leaderboard, nor is it permitted by the license of the upstream models that are listed (all were submitted/added to the open_slm_leaderboard either by or with express permission from the model owners). With no credit apart from a dark text on black background (in the bottom right of the page) with the only change being changing "By axiomic labs" -> "By AtomixLabs" and adding a single filter
Hi @Datdanboi25 ,
Using our Discussions tab for performative takedown threads instead of submitting an official ticket to Hugging Face Trust and Safety is pure theater. Let's address your claims directly, because every single legal argument in your post is completely fictional:
You chose the Apache 2.0 license for your Space. Under Section 2 and Section 4 of Apache 2.0, anyone has an explicit, worldwide, irrevocable legal right to reproduce, modify, prepare derivative works of, and distribute the code. Claiming an open-source fork is "not allowed by the license" proves you don't understand the license you published under.
Claiming you need "express permission from upstream model owners" to display public benchmark scores is legally embarrassing. Under established Supreme Court precedent (Feist Publications, Inc. v. Rural Telephone Service Co.), evaluation metrics, parameter counts, and benchmark percentages are raw mathematical facts. Facts cannot be copyrighted. You don't own a monopoly on public evaluation data, and model authors cannot copyright a benchmark score.
We originally submitted every single one of these improvements (architecture transparency filters, URL deep-linking, schema verification, and dark mode UI fixes) directly to your Space in PR #97 (https://huggingface.co/spaces/AxiomicLabs/Open_SLM_Leaderboard/discussions/97). You closed it within hours because you "didn't see a need" for users to filter by native deployability versus custom code requiring
trust_remote_code=True.
Forking an open-source repository when the upstream maintainer blocks deployability transparency to protect their own custom models is the foundational principle of open source. Attribution to Axiomic Labs is explicitly preserved in the footer and repository under Section 4 of the Apache 2.0 license.
If anyone genuinely believes that derivative open-source software or public benchmark tables violate copyright law, stop using discussion threads for performative claims and file a formal DMCA takedown with Hugging Face Trust and Safety. We welcome the review.
Haha well enjoy the permanent ban and removal from the slm leaderboard and a loss of followers. Good luck with the 4 likes, terrible models and the whole community against you❤️
Removing models from a public leaderboard over an open-source fork proves exactly why an independent, transparent leaderboard needed to exist in the first place.
Best of luck with your Space.
@Datdanboi25 He just added a nice feature to your leaderboard that didn't get into the main, that's what open source does. But I guess your new license prohibits it now. Im guessing you didn't know you set it to apache 2
@Banaxi-Tech Appreciate the objective take! That's literally the core definition of open source.
And the Git commit log settles the entire debate: commit 5a8c6af shows the metadata being manually changed from license: apache-2.0 to license: other 12 hours ago, well after our fork was already live.
Under Section 2 of the Apache 2.0 license, the grant on the codebase is perpetual and irrevocable. Relicensing a repository only applies to future commits, it cannot retroactively revoke rights on an existing Apache 2.0 fork.
The fork is legally compliant, fully attributed, and here to stay. Thanks for keeping the focus on open-source principles.
@Datdanboi25 He just added a nice feature to your leaderboard that didn't get into the main, that's what open source does. But I guess your new license prohibits it now. Im guessing you didn't know you set it to apache 2
But, multivex didn't asked and in NOWHERE i see something saying "Derivated from AxiomicLabs Open SLM Leaderboard", he just claims the leadeboard is yours
I still think you should take it down or edit it alot that makes it different though, while its legally compliant, a other leader board would be more original and just better.
@Datdanboi25 He just added a nice feature to your leaderboard that didn't get into the main, that's what open source does. But I guess your new license prohibits it now. Im guessing you didn't know you set it to apache 2
But, multivex didn't asked and in NOWHERE i see something saying "Derivated from AxiomicLabs Open SLM Leaderboard", he just claims the leadeboard is yours
Two quick points:
You don't need to "ask" under Apache 2.0. Section 2 grants open, irrevocable permission to everyone worldwide from the second the code is published.
The attribution is literally in the footer of the page right now: "Forked and modified from the Open SLM Leaderboard by Axiomic Labs (Apache 2.0 License)."
Naming a fork (like Neovim from Vim, or MariaDB from MySQL) while preserving upstream attribution in the footer is standard open-source practice.
@Datdanboi25 He just added a nice feature to your leaderboard that didn't get into the main, that's what open source does. But I guess your new license prohibits it now. Im guessing you didn't know you set it to apache 2
But, multivex didn't asked and in NOWHERE i see something saying "Derivated from AxiomicLabs Open SLM Leaderboard", he just claims the leadeboard is yours
Two quick points:
You don't need to "ask" under Apache 2.0. Section 2 grants open, irrevocable permission to everyone worldwide from the second the code is published.
The attribution is literally in the footer of the page right now: "Forked and modified from the Open SLM Leaderboard by Axiomic Labs (Apache 2.0 License)."
Naming a fork (like Neovim from Vim, or MariaDB from MySQL) while preserving upstream attribution in the footer is standard open-source practice.
Dan probably didn't know he set it under apache 2.
@Banaxi-Tech Most likely. It's a common mistake with default licenses, but that's open source for you.
@Banaxi-Tech Most likely. It's a common mistake with default licenses, but that's open source for you.
Yea thats why i said its legally compliant but its still a good practice to make it original if he says that.
@Banaxi-Tech Most likely. It's a common mistake with default licenses, but that's open source for you.
Yea thats why i said its legally compliant but its still a good practice to make it original if he says that.
Totally fair. That's the plan moving forward as we add more unique tooling and features. Thanks for keeping it level headed!
Removing models from a public leaderboard over an open-source fork proves exactly why an independent, transparent leaderboard needed to exist in the first place.
Best of luck with your Space.
I publish my stuff under apache 2.0 because I place trust in the SLM community, to show decency and use my projects in a way that is transformative and interesting. Never did I expect a member to behave so poorly as to create a direct copy of all my work, because of a rejected PR. Thus, we choose to not provide publicity to your uncompetitive models nor your organization. If you are so unable to create a unique leaderboard, your own metrics/benchmarks, or even your own gui (as seen in great examples such as @Banaxi-Tech 's bananamindbench leaderboard or the CLMI) by all means blatantly copy my code, however as a consequence, you have alienated yourself from the SLM community (as shown by almost all major members contributing to these threads) and our leaderboard (as addressed earlier).
Dan, Axiomic Labs, and a good chuck of the SLM community.
P.S please remove all of our models from the leaderboard and make the accreditation clear, or don't, we are on top regardless!
Removing models from a public leaderboard over an open-source fork proves exactly why an independent, transparent leaderboard needed to exist in the first place.
Best of luck with your Space.
I publish my stuff under apache 2.0 because I place trust in the SLM community, to show decency and use my projects in a way that is transformative and interesting. Never did I expect a member to behave so poorly as to create a direct copy of all my work, because of a rejected PR. Thus, we choose to not provide publicity to your uncompetitive models nor your organization. If you are so unable to create a unique leaderboard, your own metrics/benchmarks, or even your own gui (as seen in great examples such as @Banaxi-Tech 's bananamindbench leaderboard or the CLMI) by all means blatantly copy my code, however as a consequence, you have alienated yourself from the SLM community (as shown by almost all major members contributing to these threads) and our leaderboard (as addressed earlier).
Dan, Axiomic Labs, and a good chuck of the SLM community.
P.S please remove all of our models from the leaderboard and make the accreditation clear, or don't, we are on top regardless!
Signing your post as "Dan and a good chunk of the SLM community" is hilarious self-importance. Nobody elected you the spokesperson of open source, and Hugging Face is not your private aristocracy.
Let's look at the actual facts:
"We are on top regardless": Look at the leaderboard. You aren't on top. Hugging Face's SmolLM2-135M is sitting at #1 (27.13 Int Index), SmolLM-135M is #2 (25.74), and Axiomic is sitting at #3. And the second a developer clicks the Native filter to find models that run in vLLM and GGUF without custom code, your models disappear from the rankings entirely.
Openly admitting that you delist models to "deny publicity to uncompetitive models" confirms what the community already saw: your Space is a personal PR showcase, not an impartial benchmark. An independent leaderboard that cannot be manipulated by personal grudges is precisely what the ecosystem needed.
Open source is a meritocracy, not a fiefdom. Apache 2.0 exists specifically to prevent maintainers from holding public software hostage behind personal egos. When an upstream maintainer blocks deployability filters to protect their custom models, forking is the textbook open-source solution.
Keep managing your private club and threatening delistings. The community will eventually use the leaderboard that reports benchmark numbers transparently and shows which models are actually deployable in production.
Signing your post as "Dan and a good chunk of the SLM community" is hilarious self-importance. Nobody elected you the spokesperson of open source, and Hugging Face is not your private aristocracy.
I sign with Dan too, at least. If you think community isn't there with him.
Lol holy semantics, reads like chatgpt, we are on top for any company worth less then a billion dollars, I think a good chunk of the SLM community is validated given the 7 report threads haha. We delist your models because we dont believe you nor your org deserves publicity, which is both entirely within our right as the creator of this leaderboard, and quite frankly doesnt impact rankings nor user choices at all (like I said your models are uncompetitive). But like I said, if you truly lack the creativity and capability to create anything of your own, enjoy the code, but you've lost all the people that make it valuable (ie the labs).
Signing your post as "Dan and a good chunk of the SLM community" is hilarious self-importance. Nobody elected you the spokesperson of open source, and Hugging Face is not your private aristocracy.
I sign with Dan too, at least. If you think community isn't there with him.
Ty axion!
Lol holy semantics, reads like chatgpt, we are on top for any company worth less then a billion dollars, I think a good chunk of the SLM community is validated given the 7 report threads haha. We delist your models because we dont believe you nor your org deserves publicity, which is both entirely within our right as the creator of this leaderboard, and quite frankly doesnt impact rankings nor user choices at all (like I said your models are uncompetitive). But like I said, if you truly lack the creativity and capability to create anything of your own, enjoy the code, but you've lost all the people that make it valuable (ie the labs).
Moving the goalpost from "we are on top regardless" to "we are on top if you exclude companies worth a billion dollars" is the most desperate qualification on Hugging Face. Inventing a corporate valuation clause to disqualify the models that beat you says everything.
And citing 9 report threads as proof of community backing is equally comical when more than half of those threads were opened by you and your partner Axion spamming duplicates. Six accounts opening 9 threads across a partner circle is not "the community", it is astroturfing.
Thank you for putting it on record that you delist models because you "don't believe an org deserves publicity." You have officially confirmed that your Space is a private PR vehicle where rankings are dictated by personal grudges rather than objective benchmarks.
Keep your billion-dollar exceptions, your delisting threats, and your private club. The community will eventually use the leaderboard that reports benchmark numbers transparently and shows which models are actually deployable in production.