发表机构
Korea University; Hankuk University of Foreign Studies; Noah’s Farm; AIGEN Sciences（高丽大学; 韩国外国语大学; 诺亚农场; AIGEN科学公司）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究提出跨语言特征迁移机制框架，通过分离高资源语言的任务推理特征注入低资源语言，以提升低资源语言的推理能力，将部分跨语言推理差距归因于机制引出失败。
AI 中文摘要
大型语言模型在不同语言上的性能存在显著差异，即便解决语义等价任务时亦是如此。现有分析常将该现象视为预训练数据、分词或基准覆盖差异导致的观测性差距。本文研究一个互补假设：高资源语言（HRLs）能更可靠地引出对特定任务（即数学）推理有用的潜在计算，而低资源语言（LRLs）即便表达相同任务，也可能未激活这些计算。为验证该假设，本文提出一种机制性干预框架，用于识别并跨语言迁移特定任务相关的稀疏潜在特征。通过在残差流激活上使用稀疏自编码器，本文分离出在成功的高资源语言特定任务推理中富集的特征，同时过滤掉源语言和通用生成特征。随后，从这些特征构建引导方向并在低资源语言推理过程中注入。所得干预措施用于测试所选特征是否在观测到的推理差距中发挥功能：抑制这些特征应损害源语言推理，而激活它们应能在随机和非任务对照之外部分恢复目标语言推理。本文框架将部分跨语言推理差距重新定义为机制引出失败而非能力缺失，并提供了一条可因果验证的、无需翻译、微调或改变用户面向语言的特征介导迁移路径。
英文摘要
Large language models exhibit substantial performance variation across languages, even when solving semantically equivalent tasks. Existing analyses often treat this phenomenon as an observational disparity caused by differences in pretraining data, tokenization, or benchmark coverage. We study a complementary hypothesis: high-resource languages (HRLs) may more reliably elicit latent computations useful for task-specific (i.e. mathematical) reasoning, while lower-resource languages (LRLs) may under-activate those computations despite expressing the same task. To test this hypothesis, we introduce a mechanistic intervention framework for identifying and transferring task-relevant sparse latent features across languages. Using sparse autoencoders over residual-stream activations, we isolate features enriched in successful HRL task-specific reasoning while filtering out source-language and generic-generation features. We then construct steering directions from these features and inject them during LRL inference. The resulting interventions test whether the selected features are functionally involved in the observed reasoning gap: suppressing them should impair source-language reasoning, while activating them should partially recover target-language reasoning beyond random and non-task controls. Our framework reframes some cross-lingual reasoning gaps as failures of mechanism elicitation rather than capability absence, and offers a causally testable route to feature-mediated transfer without translation, fine-tuning, or changing the user-facing language.
CommentsAccepted to EMNLP 2026 Findings
