arXiv cs.CL· Anastasiia Ivanova·· 13 小时前AI 评分31
拒绝方向在低资源语言中是否普遍适用？研究扩展至白俄罗斯语、巴什基尔语和塔吉克语
Is Refusal Direction Universal across Low-Resourced Languages?
AI 导读
研究将 PolyRefuse 扩展至白俄罗斯语、巴什基尔语和塔吉克语三种低资源语言，并在 Qwen2.5-14B-Instruct 上复现 Wang et al. (2025) 的流程，发现拒绝方向的普遍性是分级的而非二元的：来自安全对齐和部分对齐语言的拒绝方向可广泛迁移，而来自严重错位语言的方向在排除天花板效应后无法可靠调控任何目标语言的拒绝行为。
正文 · 原文
Abstract:Refusal directions have been shown to transfer cross-lingually with near-perfect effectiveness across safety-aligned languages (Wang et al., 2025), reportedly regardless of resource level. However, the only truly low-resourced language evaluated is Yoruba, which is strongly misaligned and shows limited transfer, leaving open how refusal directions behave for other low-resourced languages. We extend PolyRefuse to three additional low-resourced languages (Belarusian, Bashkir, and Tajik) and replicate the pipeline of (Wang et al., 2025) on Qwen2.5-14B-Instruct. We find that universality is graded rather than binary: directions from safety-aligned and partially-aligned languages transfer broadly, while those from strongly misaligned languages fail to reliably modulate refusal behavior on any target tested, including other misaligned languages, once ceiling effects are accounted for. Geometric alignment and behavioral transfer are further shown to be partially dissociated. These findings complicate the universality claim and suggest that users of the least-resourced, most strongly misaligned languages may face weaker safety guarantees.
| Comments: | BlackboxNLP@EMNLP 2026: Reproducibility Track |
| Subjects: | Computation and Language (cs.CL); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.28897 [cs.CL] |
| (or arXiv:2610.28897v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.28897 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Anastasiia Ivanova [view email]
[v1]
Tue, 20 Oct 2026 20:19:40 UTC (1,603 KB)
来源：arXiv cs.CL · arxiv.org