这篇论文搞了个G-IdiomAlign基准，专门测AI能不能理解不同语言的习语。结果模型爱直译，加了注释能好点但还差得远，值得一看。
G-IdiomAlign是一个新的基准测试，以Wiktionary的英语注释为锚点，专门评估跨语言习语对齐能力。它包含两个协议：带类型干扰项的多项选择习语等价测试，以及通过有无注释对比生成来隔离语义锚点效果。实验发现，多种LLM在低资源语言上普遍倾向于直译，这是主要失败模式。添加注释后，在基于嵌入的语义度量下生成质量有所提升，但绝对性能仍然有限，说明开放输出空间仍有较大改进余地。对Qwen3-8B的分析表明，差异更多集中在注意力头而非层上，且优质生成与更强的注释锚定相关。
G-IdiomAlign: A Gloss-Pivoted Benchmark for Cross-Lingual Idiom Alignment
Idioms are difficult to transfer across languages due to their non-compositionality and weak surface-form grounding, making literal mappings unreliable. We present G-IdiomAlign, a gloss-pivoted benchmark where each idiom is anchored by an English gloss from Wiktionary. We further construct a high-confidence reference alignment set for reproducible evaluation. G-IdiomAlign supports two protocols: (1) a controlled Multiple-Choice Idiom Equivalence with typed distractors for error attribution; and (2) a Gloss-Contrastive Generation contrasting No-gloss and With-gloss inputs to isolate the effect of an explicit semantic pivot. Across diverse LLMs, a bias to literal translation is a dominant failure mode, especially when the target is a low-resource language. Glosses consistently improve Gloss-Contrastive Generation under an embedding-based semantic proxy, but performance remains modest, indicating substantial headroom in the open output space. Subsequent analysis on Qwen3-8B further suggests that cross-condition differences are concentrated more in attention heads than in layers, while better With-gloss generations coincide with stronger gloss anchoring.