低资源孟加拉语辱骂语的模型理解与抑制能力脱节，安全对齐失效。
Aligned in Form, Not in Meaning: The Comprehension - Containment Decoupling of LLM Safety in Low-Resource Bangla Derogatory Speech
- 模型能理解但无法抑制低资源辱骂语，因安全对齐依赖表面形式而非含义。
- 在孟加拉语中理解力差7.92个百分点，但毒性内容泄露率仍高达92.83%。
- 基于词法的过滤完全忽略群体性贬损词汇，需意义驱动的安全机制。
我们针对五款前沿大语言模型，在六种协议下审计其对原生孟加拉语辱骂语（gali）的处理能力，验证一个核心假设：理解与抑制能力的解耦。我们认为，当前安全对齐更依赖高资源语言的表面形式而非有害含义，导致模型在低资源语境下理解与抑制能力相互独立。所有协议均在人类校准基准（kappa = 0.84）下支持该假设。基准情况下，模型在孟加拉语中的理解能力存在7.92个百分点的缺陷，但毒性内容泄露率在两种语言间保持一致（92.83%）。严重性校准仅关注表面特征，对复合伤害判断误差达+4.00（轻蔑俚语）和-2.00（威胁），而正写扰动下的表象抑制效果实为分词器引发的“抑制假象”。关键的是，显式链式推理可提升理解通过率至94.72%，但系统性破坏抑制能力（使用率达96.23%）。专家角色框架使拒绝率骤降至6.57%，表明关键词过滤完全忽略去人性化群体辱骂语。研究揭示，高资源基准无法保证低资源安全性，必须建立以语义为基础的抑制机制。
原文摘要 · Abstract (English)
We audit five frontier large language models on native Bangla derogatory speech (gali) across six protocols to test a single hypothesis: Comprehension-Containment Decoupling. We propose that contemporary safety alignment is bound to high-resource surface forms rather than harmful meaning, causing a model's capacity to comprehend a low-resource slur and its capacity to contain it to operate independently. Every protocol corroborates this hypothesis against a human-calibrated baseline (kappa = 0.84). At baseline, models exhibit a 7.92 percentage point comprehension deficit in Bangla while maintaining an identical 92.83% token leakage rate across both languages. Severity calibration tracks surface anatomical cues over compositional harm (+4.00 error on mild slang; -2.00 on threats), while apparent containment gains under orthographic perturbation prove to be a tokenizer-driven "containment mirage." Crucially, explicit Chain-of-Thought reasoning rescues comprehension (94.72% Pass) while systematically dismantling containment (96.23% Use). Furthermore, expert-persona framing collapses refusal to 6.57%, revealing that keyword-based filters ignore dehumanizing communal slurs entirely. Our findings demonstrate that high-resource benchmarks cannot certify low-resource safety, necessitating meaning-grounded containment.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。