Metraipolis › archive › dc_769ed8b009d3
Abliteration Mitigation via Refusal Aliases — arXiv:2608.18093
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.18093
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-18T13:44:46.561Z
- SHA-256 of the archived text
1eac310e4eecf0288bf41ea94b6d8e15ad21c66c4ba27a01e5ed017a19bc762a- Stable URL of this page
https://metraipolis.com/documento/dc_769ed8b009d3
How to cite
Abliteration Mitigation via Refusal Aliases — arXiv:2608.18093 — arXiv (en), https://arxiv.org/abs/2608.18093 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_769ed8b009d3 (Metraipolis archive dc_769ed8b009d3, sha256 1eac310e4eecf0288bf41ea94b6d8e15ad21c66c4ba27a01e5ed017a19bc762a)
BibTeX
@misc{dc769ed8b009d3,
title = {Abliteration Mitigation via Refusal Aliases — arXiv:2608.18093},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.18093},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_769ed8b009d3; sha256 1eac310e4eecf0288bf41ea94b6d8e15ad21c66c4ba27a01e5ed017a19bc762a},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.18093. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Abliteration, the removal of refusal capabilities from large language models by projecting weight matrices orthogonal to an extracted refusal direction, has emerged as a prominent safety concern through its ability to bypass post-training alignment using only a small set of contrastive prompts. We find that existing defenses commonly overlook the cause of abliteration; that is, how easily the refusal direction can be extracted. To hinder this process, we introduce a weight-editing method that obscures the refusal signal by applying rank-$k$ updates to residual stream writer matrices while replacing refusal-inducing activations with random aliases and correcting downstream reader matrices to preserve the model's original behavior. On Llama-3-8B, AMRA improves post-abliteration refusal scores by $2.16$ points over the undefended baseline with less than $0.5$ percentage points of MMLU degradation. On Gemma-2-9B, it improves the post-abliteration refusal by $14.70$ points over the baseline while keeping harmful output rates similar to the baseline, albeit at a greater utility cost. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-20 in cs.AI). Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR). Pagina del lavoro: https://arxiv.org/abs/2608.18093 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-18T13:41:17.899Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-20?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).