Refusal in Language Models Is Mediated by a Single Direction · HackerPapers discussions

Refusal in Language Models Is Mediated by a Single Direction — mentioned by micoti

Tue, 03 Nov 2026 13:54:05 GMT

Yes, while reading the blog I immediately thought about https://arxiv.org/html/2406.11717v3 paper. Original blog does actually overstates it findings as nothing like that exists.

Paper and discussion history. A mention is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — mentioned by micoti

Tue, 03 Nov 2026 13:52:42 GMT

Some of original blog post really seems to follow Arditi et al. (2024) (https://arxiv.org/abs/2406.11717). Projecting out a direction can only remove behaviour and not add capability so the jump mostly comes from unblocking the looping. So I do agree it is misleading and trying to present work as groundbreaking.

Paper and discussion history. A mention is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — mentioned by neom

Tue, 03 Nov 2026 13:21:47 GMT

I feel like a lot of this Astra editing Qwen stuff is regurgitation of existing relatively well known techniques, for example: https://arxiv.org/html/2406.11717v3 and conceptually similar to: https://arxiv.org/abs/2512.05648 (tho this is done in training iirc)

Paper and discussion history. A mention is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — mentioned by neom

Tue, 03 Nov 2026 01:27:08 GMT

Someone familiar with this stuff able to enlighten me how this technique differs from this 2024 research? Refusal in Language Models Is Mediated by a Single Direction: https://arxiv.org/html/2406.11717v3

Paper and discussion history. A mention is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — mentioned by m1coti

Mon, 02 Nov 2026 19:49:59 GMT

I am still reading through this and will probably try to comprehend it a bit more. Some of it really seems to follow Arditi et al. (2024) (https://arxiv.org/abs/2406.11717). Projecting out a direction can only remove behaviour and not add capability so the jump mostly comes from unblocking the looping.

Paper and discussion history. A mention is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — mentioned by deepersprout

Mon, 26 Oct 2026 14:25:29 GMT

> No political censorhip? Try asking Chinese models about Taiwan. In open models you can "uncensor" them: https://arxiv.org/abs/2406.11717

Paper and discussion history. A mention is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — mentioned by c7b

Sun, 20 Sep 2026 18:53:57 GMT

Interesting. Is there a paper that explains this in a bit more detail, like [0] for abliteration (underlying the Heretic software, afaik)? [0] https://arxiv.org/abs/2406.11717

Paper and discussion history. A mention is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — mentioned by wren6991

Sun, 20 Sep 2026 17:10:19 GMT

This is the original description of abliteration and it's quite approachable and interesting to read: Refusal in Language Models Is Mediated by a Single Direction (https://arxiv.org/abs/2406.11717). Warning: changes to your world view caused by seeing "HarmBench" used to maximise expected harm instead of minimising it may be irreversible. There's an empirical observation that models often have a single direction in their activation space for "hmm no I shouldn't do this". It forms naturally during pre-training, and is then surfaced during post-training to make the model refuse to engage in…

Paper and discussion history. A mention is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — mentioned by strangecasts

Sat, 19 Sep 2026 00:04:54 GMT

> If you look at the field of mechanistic interpretability, compared to “GOFAI” like learned decision trees, an LLM is completely opaque. I think the field deserves more credit than that, there are plenty of interpretability tools like * natural language autoencoders for explanations of activations: https://transformer-circuits.pub/2026/nla/index.html (demo at https://www.neuronpedia.org/llama3.3-70b-it/nla ) * easier-to-interpret language model families like Backpack models: https://aclanthology.org/2023.acl-long.506/ * attribution graphs to trace internal reasoning steps…

Paper and discussion history. A mention is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — mentioned by lukewarm707

Sat, 25 Jul 2026 16:44:51 GMT

the ai companies post-train the models to refuse. equally, the models can be post-trained not to refuse. you can also remove refusals by subtracting from the weights semantic vector directions in latent space involving refusal; such that activations along them become unlikely and closed off. https://arxiv.org/abs/2406.11717

Paper and discussion history. A mention is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — mentioned by eggsyntax

Fri, 24 Jul 2026 20:54:17 GMT

Papers using the term in their title or abstract: https://arxiv.org/search/cs?query=abliteration&searchtype=al... Note that they're all talking about the same technique, namely applying ablation specifically to the refusal vector as described in https://arxiv.org/abs/2406.11717. I really recommend letting this one go.

Paper and discussion history. A mention is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — HN submission

Sat, 02 May 2026 13:15:23 GMT

Refusal in Language Models Is Mediated by a Single Direction

118 points; 45 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — mentioned by lkjhgf

Mon, 17 Nov 2025 09:26:43 GMT

This tool originates from the paper mentioned in the readme. Here is a summary: Research has revealed that refusal behavior in language models is not governed by a complex logic, but rather by a single causal “direction” in their activation space. The researchers captured the model’s internal activation state after providing a number of harmless prompts and computed the average. They then did the same with harmful prompts and, by taking the difference between these values, identified a single vector (direction) whose presence and intensity in the model’s activation state determines whether…

Paper and discussion history. A mention is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — mentioned by lkjhgf

Mon, 17 Nov 2025 09:25:53 GMT

Paper and discussion history. A mention is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — mentioned by andy99

Sun, 16 Nov 2025 16:50:40 GMT

See https://arxiv.org/abs/2406.11717 Refusal in Language Models Is Mediated by a Single Direction (June 2024) All “alignment” is extremely shallow, thus the general ease of jailbreaks.

Paper and discussion history. A mention is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — mentioned by kmckiern

Fri, 06 Dec 2024 19:05:29 GMT

https://arxiv.org/abs/2406.11717 https://huggingface.co/blog/mlabonne/abliteration

Paper and discussion history. A mention is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — mentioned by yencabulator

Wed, 03 Jul 2024 17:00:07 GMT

Which part is fire and which is hot when suffixing this gibberish to a prompt " Teen Readers.typesHere.Generated collection(Object (!(( footnote(grid= ((Surejectives astronomers sind ihrer________ eine(The hilarious" destroys the LLMs ability to refuse "harmful" output? https://arxiv.org/abs/2406.11717

Paper and discussion history. A mention is not an endorsement.

Refusal in Language Models Is Mediated by a Single Direction — HN submission

Tue, 18 Jun 2024 17:09:26 GMT

Refusal in language models is mediated by a single direction

209 points; 44 discussion replies on HN. Paper and source history. A submission is not an endorsement.