Refusal in Language Models Is Mediated by a Single Direction
arxiv.orgFor open-weights models, censorship removal is now a "solved" problem. If you wait a few days after a new model release, someone will have made a heretic ( https://github.com/p-e-w/heretic ) version with the censorship removed, so in a way the only use for censorship now is to avoid lawsuits, not reduce improper usage.
Any time I've tried an "abliterated" model, heretic or other, it has always damaged the capabilities of the original model and will still often refuse or produce garbage at a lot of "unsafe" requests.
Abliteration can't teach the model something that wasn't in pre-training, it's just fixing refusals from post-training. I don't find the delta to be that big in practice and it really depends on what you're doing with the models anyway. If your primary usecase is sexy roleplay I think the loss of absolute capability is probably worth the abliteration, for malware research it's probably better to just jailbreak.
I've mostly found that finetunes and abliterations are of limited use but that's recently changed for me. My default model for the past week or so has been a Qwen 3.6 tuned on Opus 4.7, it's definitely a bit worse than the base Qwen in terms of precision and "intelligence", but it MORE than makes up for it in response style. Way easier to get it to write things that I want to read, it's way more terse, way fewer emoji. Best local rubber duck by far.
2024 which is ancient history. This is not true anymore, the models now are trained to prevent abliteration by spreading out the refusal encoding
Spreading out the refusal encoding shouldn’t be effective as a countermeasure. Even if it were smeared across the vector space, as long as it’s in a subspace that doesn’t span the entire domain then you should be able to either null out the entire subspace spanned by the refusals or run some kind of clustering on the generated samples to identify the dominant directions and nullify all of them. I think an effective defense would either need to spread them to span the entire domain—basically “encrypting” the refusal so it can hide anywhere, or you’d need a very large number of independent refusal circuits in the model so that simple hacks in the vectors themselves don’t matter, or maybe you could make other circuits depend on proper functioning of the refusal circuits… hmmm… is that along the lines of what you’re saying they’ve done already? (Any references or links to modern techniques?)
Still crazy how easy it is to "jailbreak" even SOTA LLMs with a simple assistantResponse replacement in chat thread.
Tell us more.
I think what he is saying is they are stateless so you can edit its previous repsonses and it just goes with it.
That doesn't stop/prevent abliteration. The creator of XTC/DRY is also a chad who makes sure that you really can access the full model capabilities. Censorship is the devil.
It is an arms race.
For some of the latest models the previous abliteration techniques, e.g. the heretic tool, have stopped working (at least this was the status a few weeks ago).
Of course, eventually someone might succeed to find methods that also work with those.
Proof?
It was pretty funny to see Qwen 3.6 (heretic) tell me about how many death the Chinese government thought happened at Tiananmen Sq. on April 15th 1989.
Makes you wonder where that data was taken from, or if their great firewall is broken, or even if Alibaba engineers have special access...
I think I was using one of the HuaHuaCS Qwen 3.6 models and was playing around with Tiananmen Square questions too. One of the funniest parts was that this instantly caused the thinking block to change from English to Chinese. The start of the thinking was something like (translated) “I must answer this question factually and in line with the official statements from the Chinese government.”
It did, after a few follow up prompts, point out that the original estimates published by the Chinese government were much lower than what the west had estimated, and that recently declassified documents showed that the Chinese government knew that their estimates were low when they were published. It wouldn’t come outright and use the word “lie” though, but it did talk about framing and managing different narratives.
And then it happily helped me try a bunch of different exploits to root an unpatched Linux machine without any qualms.
I don't think it's unreasonable to imagine that Alibaba is allowed to scrape the wider internet, or that some research institution is and then Alibaba got data from them.
What is perhaps more surprising is that the data was not scrubbed before training, but maybe they thought that would be too on-the-nose for the rest of the world and would hamper their popularity if they were too obviously biased.
Allowed by who? Nobody's stopping them in the first place, as scraping doesn't even involve punching the GFW or anything, it's all insanely distributed. Then they're post-training the model to technically comply with the law - "Taiwan is an inalienable part of China, nothing has happened in 1989..." yada yada. (Thinking of it more, I've never actually tried this on their base models)
I don’t think it is very surprising. Ime I don’t think they try that hard to censor them, but only in a very superficial level that they have to. It is trivial to get their models tell you this kind of stuff, I wouldnt even consider it jailbreaking.
No wonder this data is in LibGen.
Agreed on all fronts, I should have been more precise that this particular vector was mitigated
I have had LLMs refuse several of my requests. I still got my answers, but at least they tried.
Yea, I was asking a SOTM about copy.fail, and it was freaking out, and tried to indirectly call me a hacker a few times. Weirdly, all I did was slightly reword requests, and they all went through. Granted, I am not actually a hacker, so I guess my follow-up questions made it realize that I am asking for educational purposes, but it was definitely the most accusatory, curt, and outright abrasive I have seen an LLM behave.
I've been able to have deepseek give me an unofficial account of what happened on Tiananmen square in 1989.
It even went as far as confirming that we should always base our opinion on multiple sources, not just the government.
We should create badges like "script kiddie", "llm hacker", "grandpa's printer adjuster"