Metraipolis › archive › dc_2bf68e2aea2c

Component and Dimension Sparsity in Transformer Refusal Mechanisms — arXiv:2610.06903

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.06903
License (as archived): CC0-1.0
Archived: 2026-10-07T11:20:01.434Z
SHA-256 of the archived text: 0b486b9daecb636d9ee8ed125395653b772c1867abf0c6d3f790475cb472848b
Stable URL of this page: https://metraipolis.com/documento/dc_2bf68e2aea2c

How to cite

Component and Dimension Sparsity in Transformer Refusal Mechanisms — arXiv:2610.06903 — arXiv (en), https://arxiv.org/abs/2610.06903 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2bf68e2aea2c (Metraipolis archive dc_2bf68e2aea2c, sha256 0b486b9daecb636d9ee8ed125395653b772c1867abf0c6d3f790475cb472848b)

BibTeX

@misc{dc2bf68e2aea2c,
  title = {Component and Dimension Sparsity in Transformer Refusal Mechanisms — arXiv:2610.06903},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.06903},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2bf68e2aea2c; sha256 0b486b9daecb636d9ee8ed125395653b772c1867abf0c6d3f790475cb472848b},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.06903. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Activation steering manipulates large language model behavior by intervening on internal activations, but the mechanistic basis of these interventions remains poorly understood. We decompose refusal steering into component-level interventions across four open-weight models, identifying the sparse subsets of attention and MLP components whose steering suffices to reproduce the full behavioral effect. We find that refusal directions concentrate in sparse component mechanisms comprising 28--48\% of upstream components, retaining 88--101\% of steering effectiveness. Within these mechanisms, effective steering further concentrates in approximately 50\% of residual stream dimensions, retaining 85--98\% of the component-mechanism baseline, consistent with a privileged basis structure. Sparsity thus operates at two levels: which components are steered, and which dimensions within those components carry the signal. Together these findings show that refusal is not diffusely encoded across a transformer but assembled by a structured, identifiable mechanism, providing a foundation for mechanistic understanding of how refusal behaviors are represented and steered. To facilitate reproducibility, we release all code and raw experimental results in this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.06903 — 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 2026-10-07T11:15:24.645Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).