Metraipolis › archive › dc_560640d80564

From Detection to Refusal: Safer LLMs via Circuit-Guided Weight Scaling — arXiv:2609.00051

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.00051
License (as archived)
CC0-1.0
Archived
2026-10-17T06:32:36.106Z
SHA-256 of the archived text
7278f69c7be02c5ffb150e837cc1edcd83ab7f502c2f1f27811393915010d270
Stable URL of this page
https://metraipolis.com/documento/dc_560640d80564

How to cite

From Detection to Refusal: Safer LLMs via Circuit-Guided Weight Scaling — arXiv:2609.00051 — arXiv (en), https://arxiv.org/abs/2609.00051 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_560640d80564 (Metraipolis archive dc_560640d80564, sha256 7278f69c7be02c5ffb150e837cc1edcd83ab7f502c2f1f27811393915010d270)
BibTeX
@misc{dc560640d80564,
  title = {From Detection to Refusal: Safer LLMs via Circuit-Guided Weight Scaling — arXiv:2609.00051},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.00051},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_560640d80564; sha256 7278f69c7be02c5ffb150e837cc1edcd83ab7f502c2f1f27811393915010d270},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.00051. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Despite extensive alignment efforts, Large Language Models (LLMs) remain vulnerable to generating unsafe content under adversarial prompting, yet the internal mechanisms by which safety behaviors are implemented remain poorly understood. We study LLM safety from a mechanistic interpretability perspective and characterize a multi-stage *safety circuit* that organizes refusal behavior, consisting of (i) $\textbf{Harmful Detection Heads}$ that respond to harmful inputs, (ii) $\textbf{Safety Neurons}$ that mediate and stabilize safety signals in the residual stream, and (iii) $\textbf{Refusal Heads}$ that translate these signals into safe response generation. Using targeted attention-head and neuron-level interventions, we provide causal evidence consistent with this circuit organization, showing that suppressing upstream Harmful Detection Heads disrupts downstream refusal behavior and that safety neurons mediate this interaction. We validate that this decomposition recurs across multiple LLM architectures and adversarial attack settings, and use simple, architecture-preserving weight scaling as a mechanistic probe to test its functional relevance. Across six LLMs, circuit-guided scaling improves safety rates under attacks by 26.5%, while incurring only a 1.7% accuracy drop across four standard benchmarks. Overall, our results support a circuit-level interpretation of LLM safety and suggest that mechanistic abstractions can reveal stable and transferable patterns underlying aligned behavior.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-02 in cs.AI).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI); Computers and Society (cs.CY).
Pagina del lavoro: https://arxiv.org/abs/2609.00051 — 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 2026-10-17T06:15:12.717Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).