Metraipolis › archive › dc_a05dcd6161fb
Mitigating LLM Over-Refusal via Dynamic Semantic Routing Calibration — arXiv:2609.25049
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.25049
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T13:45:58.884Z
- SHA-256 of the archived text
79ace4d8e606fe55b7816fd31d44130db9f105764c7007768e33942a5c5d6182- Stable URL of this page
https://metraipolis.com/documento/dc_a05dcd6161fb
How to cite
Mitigating LLM Over-Refusal via Dynamic Semantic Routing Calibration — arXiv:2609.25049 — arXiv (en), https://arxiv.org/abs/2609.25049 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_a05dcd6161fb (Metraipolis archive dc_a05dcd6161fb, sha256 79ace4d8e606fe55b7816fd31d44130db9f105764c7007768e33942a5c5d6182)
BibTeX
@misc{dca05dcd6161fb,
title = {Mitigating LLM Over-Refusal via Dynamic Semantic Routing Calibration — arXiv:2609.25049},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.25049},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_a05dcd6161fb; sha256 79ace4d8e606fe55b7816fd31d44130db9f105764c7007768e33942a5c5d6182},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.25049. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Large language models (LLMs) aligned for safety often suffer from over-refusal, incorrectly rejecting benign yet safety-related instructions. Prior studies primarily attribute this to static representation overlap, largely overlooking the underlying dynamic mechanisms. In this paper, we present the mechanistic analysis of over-refusal through the lens of internal routing conflicts within transformer attention. We discover that a sparse subset of Hypersensitive Safety Heads misfires on Hard-Safe prompts, exhibiting abnormal attention entanglement that forcefully binds harmless target entities to refusal semantics. This triggers a severe, high-entropy routing conflict that deprives target entities of necessary attention. To counteract this, we propose Semantic Routing Calibration (SRC), a lightweight, training-free inference framework. SRC precisely localizes and dynamically suppresses these hypersensitive safety heads at the inference stage. Coupled with a dual-branch logits fusion that acts as a safety regularizer during subsequent decoding, SRC seamlessly restores trustworthy reasoning. Extensive experiments demonstrate that SRC alleviates over-refusal, with intrinsic safety performance preserved as much as feasible. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-23 in cs.AI). Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.25049 — 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 2026-10-08T13:18:38.189Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-23?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).