SEDATE：大语言模型是否跨语言共享攻击性回路？
SEDATE: Do Large Language Models Share an Aggression Circuit Across Languages?
AI 导读
研究者提出 SEDATE（Selective Detection And Targeted Editing），一种免训练的检测—编辑—验证方法，定位大语言模型中负责攻击性内容的神经元并按需编辑，再验证模型回复。该方法在 7 种语言、4 个模型家族上评估，并在 4 个分布外语言和 4 个额外数据集上验证，安全得分显著优于最强基线，同时几乎不增加开销且不损失通用能力。
正文 · 原文
Abstract:As the applications of AI continue to expand across various multimodal and multilingual domains, ensuring safe and responsible deployment of these systems has become an important concern. A wide-range of interventions have been proposed to make AI responses safer and more reliable. They often incur substantial computational or annotation costs, or cannot be extended to multilingual settings. In this work, we propose Selective Detection And Targeted Editing (SEDATE), a training-free approach based on detect-edit-verify strategy. SEDATE first detects the responsible neurons, edits them as and when necessary and finally verifies the model responses; all without any additional model training. We evaluate SEDATE across 7 languages and 4 model families. We further evaluate its effectiveness across 4 out-of-distribution languages and 4 additional datasets. We show that SEDATE achieves substantially superior safety scores across all languages and models over the best-performing baseline while retaining the general capabilities and incurring negligible overhead. Our results are robust, with the trends remaining similar for out-of-distribution languages and datasets.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.33542 [cs.CL] |
| (or arXiv:2610.33542v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.33542 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Somnath Banerjee [view email]
[v1]
Sat, 24 Oct 2026 08:38:52 UTC (1,275 KB)
来源：arXiv cs.AI · arxiv.org