Multilingual GSM-Symbolic: What determines capability transfer across languages?

• 类型：arxiv
• 标识：2610.03367
• 链接：https://arxiv.org/abs/2610.03367
• 主分类：evaluation
• 形态：method
• TLDR：We understand little about how capabilities acquired in one language carry over to another, or what governs this transfer: evaluations rely on incomparable, saturation-prone datasets and rarely examine its determinants jointly. Identifying what predicts transfer would let us avoid exhaustive evaluation across all language pairs and let developers target the factors that limit performance in low-resource languages. To evaluate cross-lingual capability transfer, we introduce Multilingual GSM-Symbolic, an extensible multilingual mathematical dataset covering 30,000 item-matched question-answer pa
• 待LLM分类：否
• 标题中文：多语言 GSM-Symbolic：跨语言能力迁移由什么决定？
• TLDR中文：我们对一种语言中习得的能力如何迁移到另一种语言知之甚少，也不清楚支配这种迁移的因素：评估依赖不可比且易饱和的数据集，且很少联合考察其决定因素。识别出迁移的预测因素将使我们免于对所有语言对进行穷举评估，并帮助开发者针对性改善限制低资源语言性能的因素。为评估跨语言能力迁移，我们提出 Multilingual GSM-Symbolic，一个可扩展的多语言数学数据集，覆盖 30,000 道题项匹配的问答对
• 来源文件：
• /inbox/tom/_candidates/2026-10-05-agent-rag-longcontext-candidates.json