Metraipolis › archive › dc_f9ec5ee66bbd
Frequency Is Not Sensitivity Identifying Safety-Sensitive Experts in Sparse MoE LLM — arXiv:2610.02910
- Source
- arXiv
- Original
- https://arxiv.org/abs/2610.02910
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-05T22:49:21.451Z
- SHA-256 of the archived text
81ad29a177c690f1e1091b68ece8c26d37222ede4b3acbb0d19668dcbca1adc3- Stable URL of this page
https://metraipolis.com/documento/dc_f9ec5ee66bbd
How to cite
Frequency Is Not Sensitivity Identifying Safety-Sensitive Experts in Sparse MoE LLM — arXiv:2610.02910 — arXiv, https://arxiv.org/abs/2610.02910 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f9ec5ee66bbd (Metraipolis archive dc_f9ec5ee66bbd, sha256 81ad29a177c690f1e1091b68ece8c26d37222ede4b3acbb0d19668dcbca1adc3)
BibTeX
@misc{dcf9ec5ee66bbd,
title = {Frequency Is Not Sensitivity Identifying Safety-Sensitive Experts in Sparse MoE LLM — arXiv:2610.02910},
howpublished = {arXiv},
url = {https://arxiv.org/abs/2610.02910},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f9ec5ee66bbd; sha256 81ad29a177c690f1e1091b68ece8c26d37222ede4b3acbb0d19668dcbca1adc3},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.02910. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Suppressing a small set of routed experts can weaken the safety behavior of a sparse Mixture-of-Experts (MoE) language model without retraining. Which experts to suppress is therefore a security question, and the usual answer is activation frequency, but frequency measures use, not influence. We test an alternative: router-gradient sensitivity, the sensitivity of the sequence loss to the gate weights that select an expert. Across five MoE architectures, we rank experts by each signal on 500 benign and 500 malicious prompts and measure refusal on 100 held-out malicious prompts under two budgets: equal expert counts and equal nominal malicious routing traffic (1%-5%). Under each of the two budgets, router-gradient selection reduces refusals more than activation in 24 of 25 conditions, and more than a ten-trial random mean in all 25. The largest effect is in OLMoE, where refusals fall from 34 to 9 of 100 prompts (73.53% relative) with no degraded outputs, indicating substantive compliance rather than broken generation. After matching expert counts in every layer, gradient selection still produces greater refusal reduction than activation in 23 of 25 conditions, with two ties. An exploratory cross-model analysis links larger malicious-versus-benign concentration gaps to greater peak gradient effects (rho = 0.90; exact two-sided p = 0.083, n = 5). Together, the results support gradient selection under the tested budgets. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Artificial Intelligence (cs.AI) ; Cryptography and Security (cs.CR); Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2610.02910 — 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 2026-10-05T22:45:55.649Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).