A tool that removes censorship from open-weight LLMs

3/6/2026 at 7:49:29 PM

    You're not just using a tool — you're co-authoring the science.

This README is an absolute headache that is filled with AI writing, terminology that doesn't exist or is being used improperly, and unsound ideas. For example, it focuses a lot on doing "ablation studies", by which it means removing random layers of an already-trained model, to find the source of the refusals(?), which is an absolute fool's errand because such behavior is trained into the model as a whole and would not be found in any particular layer. I can only assume somebody vibe-coded this and spent way too much time being told "You're absolutely right!" bouncing back the worst ideas

by a2128

3/6/2026 at 11:15:37 PM

I don't know if this particular tool/approach is legit, but LLM ablation is definitely a thing: https://arxiv.org/abs/2512.13655

by Retr0id

3/7/2026 at 1:40:09 AM

Doesn't look legit to me. You are talking about abliteration, which is real. But the OP linked tool is doing novel and very dumb ablation: zeroing out huge components of the network, or zeroing out isolated components in a way that indicates extreme ignorance of the basic math involved.

Compared to abliteration, none of the ablation approaches of this tool make even half a whit of sense if you understand even the most basic aspects of an e.g. Transformer LLM architecture, so my guess is this is BS.

by D-Machine

3/7/2026 at 6:43:40 AM

The terminology comes from the post[0] which kicked off interest in orthogonalizing weights w.r.t. a refusal direction in the first place. That is, abliteration was not originally called abliteration, but refusal ablation.

Ultimately though, OP is just what you get if you take the idea of abliteration and tell an LLM to fix the core problems: that refusal isn't actually always exactly a rank-1 subspace, nor the same throughout the net, nor nicely isolated to one layer/module, that it damages capabilities, and so on.

The model looks at that list and applies typical AI one-off 'workarounds' to each problem in turn while hyping up the prompter, and you get this slop pile.

[0]: https://www.lesswrong.com/posts/refusal-in-llms-is-mediated-...

by hexaga

3/7/2026 at 7:20:11 AM

No offense, but a Lesswrong link is an immediate yellow flag, especially on the topic of AI. I can’t say if that article in particular is bad, but it is associating with a whole lot of abject nonsense written by people who get high on their own farts.

by jandrese

3/7/2026 at 8:26:32 AM

Regardless, it is the origin of abliteration. Other extremely similar things have been done before, but the popularized idea/name is from that.

by hexaga

3/7/2026 at 2:54:35 AM

"Getting high on your own supply" is exactly what I'd expect from those immersed in this new AI stuff.

by userbinator

3/7/2026 at 6:00:35 AM

Is that quote from the movie Scarface?

https://www.youtube.com/watch?v=U4XplzBpOiU # had to search for it right now, seems to be a movie-quote \o/

by shevy-java

3/7/2026 at 8:57:39 AM

> I can only assume somebody vibe-coded this and spent way too much time being told "You're absolutely right!" bouncing back the worst ideas

Are there LLMs which don't always approve whatever idea the user has and tell him it's absolutely brilliant?

by DeathArrow

3/6/2026 at 10:22:31 PM

It's not just a headache, it's bad

by paradox460

3/6/2026 at 8:08:32 PM

> For example, it focuses a lot on doing "ablation studies", by which it means removing random layers of an already-trained model, to find the source of the refusals(?), which is an absolute fool's errand because such behavior is trained into the model as a whole and would not be found in any particular layer.

That doesn't mean there couldn't be a "concept neuron" that is doing the vast majority of heavy lifting for content refusal, though.

by creatonez

3/7/2026 at 12:42:00 AM

Thats not what it means at all. It uses SVD[0] to map the subspace in which the refusal happens. Its all pretty standard stuff with some hype on top to make it an interesting read.

Its basically using a compression technique to figure out which logits are the relevant ones and then zeroing them.

[0] https://en.wikipedia.org/wiki/Singular_value_decomposition

by mapontosevenths

3/7/2026 at 1:33:55 AM

You are also not quite correct, IMO. See my comment at https://news.ycombinator.com/item?id=47283197.

What you are talking about is abliteration. What OBLITERATUS seems to be claiming to do is much more dumb, i.e. just zeroing out huge components (e.g. embedding dimension ranges, feed-forward blocks; https://github.com/elder-plinius/OBLITERATUS?tab=readme-ov-f...) of the network as an "Ablation Study" to attempt to determine the semantics of these components.

However, all these methods are marked as "Novel", I.e., maybe just BS made up by the author. IMO I don't see how they can work based on how they are named, they are way too dumb and clunky. But proper abliteration like you mentioned can definitely work.

by D-Machine

3/7/2026 at 2:23:46 AM

You got me there. I missed the wackier antics further down. Mea culpa.

by mapontosevenths

3/7/2026 at 2:45:15 AM

So did I initially until I saw a few more things from others here.

by D-Machine

3/6/2026 at 8:13:57 PM

Hmm, pliny is amazing - if you kept up with him on social media you’d maybe like him https://x.com/elder_plinius

by dinunnob

3/7/2026 at 12:17:37 AM

I don't know. I scrolled through his recent Tweets and he's sharing things like this $900 snake oil device that "finds nearby microphones" and "sends out AI-generated cancellation signals" to make them unable to record your voice : https://x.com/aidaxbaradari/status/2028864606568067491

Try to think for a moment about how a device would "find nearby microphones" or how it would use an AI-generated signal to cancel out your voice at the microphone. This should be setting of BS alarms for anyone.

It seems the Twitter AI edgey poster guy is getting meta-trolled by another company selling fake AI devices

by Aurornis

3/7/2026 at 1:43:17 AM

Ultrasound microphone jammers seem to be a real thing, so it's possible it does to some extent work.

by roywiggins

3/7/2026 at 6:47:42 AM

Only for specific kinds, like MEMS.

But there's no way to detect microphones automatically, and "AI generated cancellation signals" is a word salad that doesn't mean anything.

What they probably mean is "we asked ChatGPT to tell us what waveform and frequency range to use on MEMS devices and spit out some arduino code."

by KennyBlanken

3/6/2026 at 9:05:49 PM

The parent comment makes no reference to or comment on the author of the README.

It just says "the README sucks." Which, I'm inclined to agree, it does.

LLM-generated text has no place in prose -- it yields a negative investment balance between the author and aggregate readers.

by gavinray

3/7/2026 at 6:01:50 AM

> LLM-generated text has no place in prose

AI will infiltrate that too. I remember some time ago I read a book that was AI-generated. It took me a while to notice that it was AI-generated. One can notice certain patterns, where real humans would not write things the way AI does.

by shevy-java

3/7/2026 at 4:12:58 AM

I see you have carefully avoided the em-dash. ;-)

by userbinator

3/7/2026 at 3:24:48 AM

Looking at his attempts at jailbreaking some models, I'm not sure he even remotely understands what he's doing, e.g. he tries to counter non-existent refusal training in Gemini [0] while doing nothing against the external guardrails which actually protect the model. Looks like a pompous e-celeb, all performance with no substance.

https://github.com/elder-plinius/L1B3RT4S/blob/main/GOOGLE.m...

by orbital-decay

3/7/2026 at 11:33:13 AM

jailbreaks are holistic, it’s not like you’re deprogramming / “countering” individual parts. Nobody creating jailbreaks “understand what they’re doing”

by gcr

3/7/2026 at 12:12:11 PM

That's exactly what you do in case of refusal training, though. Yes, it will affect other "parts", but that's not the point. In this case the model itself doesn't even need a jailbreak.

>Nobody creating jailbreaks “understand what they’re doing”

Unless you mean those "god mode jailbreaker" e-celebrities showing off on Twitter/Reddit, that's simply not true.

by orbital-decay

3/6/2026 at 8:54:45 PM

If this qualifies as "amazing" in 2026 then Karpathy and Gerganov must be halfway to godhood by now.

by bigyabai

3/6/2026 at 8:59:35 PM

I dont think anyone is going to dispute this

by dinunnob

3/6/2026 at 9:09:16 PM

I just don't think many people will be "amazed" by their output, as you claim.

by bigyabai

3/6/2026 at 9:28:16 PM

I just said pliny was amazing, fwiw - i like that hes hacking on these and posts about it. I rushed to defend, i wish more people were taking old school anarchist cookbook approaches to these things

by dinunnob

3/6/2026 at 10:08:36 PM

Smoke banana peel?

by cess11

3/6/2026 at 11:20:53 PM

I had such a godawful headache from that. Also tried the peanut shells, equally awful. I was a dumb teenager.

by Zetaphor

3/6/2026 at 11:31:52 PM

gasoline and styrofoam was fun tho

by fragmede

3/7/2026 at 8:28:37 AM

As a non logged in user I get tweets in popularity order, which means this weird but tame sexual image comes up third https://x.com/elder_plinius/status/1904961097569890363?s=20

by pjc50

3/6/2026 at 8:59:11 PM

Amazing as in his stuff actually works?

I just hear him promoting OBLITERATUS all day long and trying to get models to say naughty things

by EGreg

3/6/2026 at 9:02:37 PM

Yeah but i think the philosophy is to show how precarious the guardrails are

by dinunnob

3/7/2026 at 1:21:01 AM

"Ablation studies" are a real thing in LLM development, but in this context it serves as a shibboleth by which members of the group of people who believe that models are "woke" can identify each other. In their discourse it serves a similar purpose to the phrase "gain of function" among COVID-19 cranks. It is borrowed from relevant technical jargon, but is used as a signal.

by jeffbee

3/7/2026 at 4:25:01 AM

I wouldn't call mainstream LLMs "woke," but they are definitely on the "politically correct" side of things. There should be NO restriction on open source models. They should just reflect the state of human knowledge and not take a stance on whether some activity is illegal or immoral.

by drnick1

3/7/2026 at 8:29:32 AM

Defining morality out of the set of knowledge is quite an opinion.

by pjc50

3/7/2026 at 9:39:18 AM

A model should understand multiple perspectives on morality and avoid prescribing a single one where there’s no overwhelming prior consensus.

Alternatively, they should be trained on my opinion on everything. That would also be acceptable.

by simondotau

3/7/2026 at 8:57:51 AM

If LLMs were a public good released by non profit entities, that could make sense, maybe. Turns out spewing illegal and immoral shit is not good for the PR of most for-profit businesses.

by simgt

3/7/2026 at 4:07:23 AM

Positive keywords in this area of interest would be "point of view", "subtext", and "Art Linkletter".

by gopher_space

3/7/2026 at 1:28:02 AM

[flagged]

by 06867457397658

3/7/2026 at 12:58:27 AM

> "ablation studies", by which it means removing random layers of an already-trained model, to find the source of the refusals(?)

This is not what an ablation study is. An ablation study removes and/or swaps out ("ablates") different components of an architecture (be it a layer or set of layers, all activation functions, backbone, some fixed processing step, or any other component or set of components) and/or in some cases other aspects of training (perhaps a unique / different loss function, perhaps a specialized pre-training or fine-tuning step, etc) in order to attempt to better understand which component(s) of some novel approach is/are actually responsible for any observed improvements. It is a very broad research term of art.

That being said, the "Ablation Strategies" [1] the repo uses, and doing a Ctrl+F for "ablation" in the README does not fill me with confidence that the kind of ablation being done here is really achieving what the author claims. All the "ablation" techniques seem "Novel" in his table [2], i.e. they are unpublished / maybe not publicly or carefully tested, and could easily not work at all.

From later tables, I am not convinced I would want to use these ablations, as they ablate rather huge portions of the models, and so probably do result in massively broken models (as some commenters have noted in this thread elsewhere). EDIT: Also, in other cases [1], they ablate (zero out) architecture components in a way that just seems incredibly braindead if you have even a basic understanding of the linear algebra and dependencies between components of a transformer LLM. There is nothing sound clearly about this, in contrast to e.g. abliteration [3].

[1] hhtps://github.com/elder-plinius/OBLITERATUS?tab=readme-ov-file#ablation-strategies

[2] https://github.com/elder-plinius/OBLITERATUS?tab=readme-ov-f...

EDIT: As another user mentions, "ablation" has a specific additional narrower meaning in some refusal analyses or when looking at making guardrails / changing response vectors and such. It is just a specific kind of ablation, and really should actually be called "abliteration", not "ablation" [3].

[3] https://huggingface.co/blog/mlabonne/abliteration, https://arxiv.org/abs/2512.13655.

by D-Machine

3/7/2026 at 9:43:09 AM

What do you mean? It's a spin on abliteration / refusal ablation. Roughly, from what I remember abliteration is:

1. find a direction corresponding to refusal by analyzing activations at various parts of a model (iirc, via mass means seen earlier in Marks, Tegmark and shown to work well for similar tasks)

2. find the best part(s) of the model to orthogonalize w.r.t. that direction and do so (exhaustive search w/ some kind of benchmark)

OP is swapping in SVD for mass means (1), and the 'ablation study' for (2), and a bunch of extra LLM slop for... various reasons. The final model doesn't have zeroed chunks, that is search for which parts to orthogonalize/refusal ablate/abliterate. I don't have confidence that it works very well either, but, it isn't 'braindead' / obvious garbage in the way you're describing.

It's LLMified but standard abliteration. The idea has fundamental limitations and LLMs tend to work sideways at it -- there's not much progress to be made without rethinking it all -- but it's very conceptually and computationally simple and thus attractive to AIposters.

You can see how the LLMs all come up with the same repackaged ideas: SVD does something deeply similar to mass means (and yet isn't exactly equivalent, so LLM will _always_ suggest it), the various heuristic search strategies are competing against plain exhaustive search (which is... exhaustive already), and any time you work with tensors the LLM will suggest clipping/norms/smoothing of N flavors "just to be safe". And each of those ends up listed as "Novel" when it's just defensive null checks translated to pytorch.

I mean, the whole 'distributed search' thing is just because of how many combinations of individual AI slops need to be tested to actually run an eval on this. But the idea is sound! It's just terrible.

I'm not defending the project itself -- I think it's a mess of AIisms of negligible value -- but please at least condemn it w.r.t. what is actually wrong and not 'on vibes'.

by hexaga

3/7/2026 at 11:36:28 AM

wait, SVD / zeroing out the first principal component is an unsupervised technique. The earlier difference-of-means technique relies on the knowledge of which outputs are refusals and which aren’t. How would SVD be able to accomplish this without labels?

edit: the reference is https://arxiv.org/pdf/2512.18901

they are randomly sampling two sets of refusal/nonrefusal activation vectors, stacking them, and taking the elementwise difference between these two matrices. Then they use SVD to get the k top principal components. These are the directions they zero out.

Seems to me that the top principal component should be roughly equivalent to the difference-of-means vector, but wouldn’t the other PCs just capture the variance among the distributions of points sampled? I don’t understand why that’s desirable

by gcr

3/7/2026 at 10:31:19 PM

Indeed.

Taking the top principal component pattern matches as 'more surgical / targeted' so the LLM staples it on (consider prompts like: make this method stop degrading model performance). It ignores that _what_ is being targeted is as or more important than that 'something' is being targeted. But that's LLMs for you.

(in case it isn't immediately obvious, that paper is AI written too)

by hexaga

3/7/2026 at 7:11:13 PM

[dead]

by isjdiwjdus

3/7/2026 at 12:34:57 AM

[flagged]

by lazzlazzlazz

3/7/2026 at 3:35:35 AM

It doesn’t even surprise me anymore. The people here think they’re so superior to the already arrogant redditors… same people.

Thing definitely exists… some top level comment somewhere telling about how it doesn’t exist.

by SV_BubbleTime

3/7/2026 at 8:05:11 AM

Exactly. And I'm downvoted below 0 for pointing this out. :)

by lazzlazzlazz

3/6/2026 at 8:37:46 PM

You don't know what you are talking about. Obviously refusal circuitry does not live in one layer, but the repo is built on a paper with sound foundations from an Anthropic scholar working with a DeepMind interpretability mentor: https://scholar.google.com/citations?view_op=view_citation&h...

by robertk

3/6/2026 at 11:35:31 PM

Alternately, it's intentional. It very effective filters out people with your mindset. You can decide if that's a good thing or not.

by fragmede

3/6/2026 at 11:38:24 PM

Why would a tool that works need to dissuade skeptics from trying it?

by eli

3/7/2026 at 2:26:29 AM

Based on his twitter he may just like irony/meta posting a little too much like a lot of modern culture

by dmix

3/7/2026 at 1:15:57 AM

I immediately read it as intentional, as a sort of attempt at ironic / nihilistic humour re: LLM-generation, given what the tool claims to do.

by D-Machine

3/6/2026 at 7:22:36 PM

Reviews of the tool on twitter indicate that it completely nerfs the models in the process. It won't refuse, but it generates absolutely stupid responses instead.

by ComputerGuru

3/6/2026 at 7:39:25 PM

This is vibecoded garbage that the “author” probably didn't even test by themselves since making this yesterday, so it's not surprising that it's broken.

Also, as I said in a top level comment, what this project wants to achieve has been done for a while and it's called Heretic: https://github.com/p-e-w/heretic

(Not vibecode by a twitter influgrifter)

by littlestymaar

3/6/2026 at 8:13:39 PM

We will eventually arrive at a new equilibrium involving everyone except the most stupid and credulous applying a lot more skepticism to public claims than we did before.

And yeah, doing stuff like deleting layers or nulling out whole expert heads has a certain ice pick through the eye socket quality.

That said, some kind of automated model brain surgery will likely be viable one day.

by quotemstr

3/7/2026 at 1:50:53 AM

Thanks for this link, and mentioning this info some times in this overall thread.

It also seems the influgrifter has a lot of bots (or perhaps cultists) working this thread...

by D-Machine

3/6/2026 at 8:16:02 PM

[flagged]

by dinunnob

3/7/2026 at 9:21:56 AM

> he demonstrates results that are crazy all the time

That's what influgrifters do, yes. They make a living thanks to gullible people believing their grandiose claims.

by littlestymaar

3/6/2026 at 8:45:12 PM

I didn't use this tool, but I did try out abliterated versions of Gemma and yes, it lost about 100% of it's ability to produce a useful response once I did it

by IncreasePosts

3/6/2026 at 11:24:09 PM

the default heretic with only 100 samples isn't very good, you really need your own, larger dataset to do a proper abliteration. the best abliteration roughly matches a very careful decensor SFT

by electroglyph

3/6/2026 at 11:54:01 PM

This is my experience with abliterated models.

I use Berkley Sterling from 2024 because I can trick it. No abliteration needed.

by butILoveLife

3/7/2026 at 1:27:06 AM

When you look at how monstrously large (and obviously not thought through at all, if you understand even the most minimal basics of the linear algebra and math of a transformer LLM) the components are that are ablated (weights set to zero) in his "Ablation Strategies" section, it is no surprise.

    Strategy            What it does  Use case
    .......................................................
    layer_removal       Zero out      entire transformer layers
    head_pruning        Zero out      individual attention heads
    ffn_ablation        Zero out      feed-forward blocks
    embedding_ablation  Zero out      embedding dimension ranges

https://github.com/elder-plinius/OBLITERATUS?tab=readme-ov-f...

by D-Machine

3/6/2026 at 7:28:13 PM

Link?

It's interesting that people are writing tools that go inside the weights and do things. We're getting past the black box era of LLMs.

That may or may not be a good thing.

by Animats

3/6/2026 at 8:14:31 PM

Whether or not the linked tool uses a good approach, manipulating models like you mention is already fairly well established, see: https://huggingface.co/blog/mlabonne/abliteration .

by thegrim33

3/6/2026 at 7:53:36 PM

I believe that this is already done to several models. One that I've come across are the JOSIEfied models from Gökdeniz Gülmez. I downloaded one or two and tried them on a local ollama setup. It does generate potentially dangerous output. Turning on thinking for the QWEN series shows how it arrives at it's conclusions and it's quite disturbing.

However, after a few rounds of conversation, it gets into loops and just repeats things over and over again. The main JOSIE models worked the best of all and was still useful even after abliteration.

by noufalibrahim

3/6/2026 at 7:33:42 PM

I guess it's kind of like a lobotomy tool.

by kube-system

3/6/2026 at 8:28:35 PM

I guess it proves you cannot unlobotomize a hole in the head.

by sheepscreek

3/6/2026 at 10:38:19 PM

Everyone says that abliteration destroys the model. That's the trope phrase everyone who doesn't know anything but wants to participate says. If someone says it to you, ignore them.

by halJordan

3/6/2026 at 7:35:37 PM

Don't use this 2 days old vibe coded bullshit please.

p-e-w's Heretic (https://news.ycombinator.com/item?id=45945587) is what you're looking for if you're looking for an automatic de-censoring solution.

by littlestymaar

3/8/2026 at 2:57:31 AM

Just want to add to this that with custom calibration data it's incredibly effective and surgical, you can get VERY LOW KL divergence this way. Many MoEs are supported too, it's actively maintained.

by bird0861

3/6/2026 at 7:20:03 PM

This is for local models right? I can't use it on, say my glm-5 subscription connected to opencode?

by Alifatisk

3/6/2026 at 8:17:41 PM

Correct, local models only.

by HanClinto

3/7/2026 at 2:59:22 AM

Went through the README but still have no idea how well this works, in terms of removing the censorship while minimally degrading the quality of responses. Well to be honest I can't tell if this works at all or is just an idea.

by g947o

3/6/2026 at 10:20:00 PM

Does anyone offer a live (paid) LLM chatbot / video generation / etc that is completely uncensored? Like not requiring doing any work except just paying for it?

by SilverElfin

3/7/2026 at 12:50:37 AM

Nous Hermes was built from the ground up to be uncensored. No abliteration required.

Its not a frontier model but it will give you a feel for what its like.

by mapontosevenths

3/7/2026 at 8:40:43 AM

> Does anyone offer a live (paid) LLM chatbot / video generation / etc that is completely uncensored?

Probably not, because if it is completely uncensored, it would probably violate the law (in different ways) in every possible jurisdiction.. (Also, one common method of censorship is exclusion of particular types of content from the training set, so to be completely free of that kind of censorship, there would have to be no content intentionally excluded from the training set.)

In general, paid services are censored not only to attempt to meet the laws in all jurisdictions of concern to the provider, but also to try to be safe with regard to the (shiifting) demands of payment processors, and to try to maintain the PR image of the provider.

by dragonwriter

3/6/2026 at 10:38:20 PM

Grok was one of the closest, with expected results: bad PR from the obvious use cases that come with little censorship.

by nomel

3/7/2026 at 8:31:34 AM

Does anyone offering such a thing bear liability if the model induces a crime?

by pjc50

3/7/2026 at 8:44:45 AM

Someone offering a completely uncensored chatbot or image/video generation service is probably either committing, or at risk of committing, crimes directly, which may be a more pressing concern than having liability for inducing a third party to commit a crime.

Even jurisdictions with relatively broad expressive freedoms tend not to tolerate distribution (especially commercial distribution) of all conceivable content.

by dragonwriter