# G-IdiomAlign: A Gloss-Pivoted Benchmark for Cross-Lingual Idiom Alignment

arXiv: https://arxiv.org/abs/2606.18989 · HTML: https://mlsanity.com/paper/2606.18989/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 12 of its 33 references that are in the map.
- First posted on arXiv 17 Jun 2026.

## Builds on (most cited first, 12 of 12)

- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [EPIE Dataset: A Corpus For Possible Idiomatic Expressions](https://mlsanity.com/paper/2006.09479/) (2020)
- [Translate Meanings, Not Just Words: IdiomKB's Role in Optimizing Idiomatic Translation with Language Models](https://mlsanity.com/paper/2308.13961/) (2023)
- [Improving LLM Abilities in Idiomatic Translation](https://mlsanity.com/paper/2407.03518/) (2024)
- [LIDIOMS: A Multilingual Linked Idioms Data Set](https://mlsanity.com/paper/1802.08148/) (2018)
- [Crossing the Threshold: Idiomatic Machine Translation through Retrieval Augmentation and Loss Weighting](https://mlsanity.com/paper/2310.07081/) (2023)
- [A Hard Nut to Crack: Idiom Detection with Conversational Large Language Models](https://mlsanity.com/paper/2405.10579/) (2024)
- [Rolling the DICE on Idiomaticity: How LLMs Fail to Grasp Context](https://mlsanity.com/paper/2410.16069/) (2024)
- [Comparative Study of Multilingual Idioms and Similes in Large Language Models](https://mlsanity.com/paper/2410.16461/) (2024)
- [Chengyu-Bench: Benchmarking Large Language Models for Chinese Idiom Understanding and Use](https://mlsanity.com/paper/2506.18105/) (2025)
- [Probing Semantic Alignment, Lexical Invariance, and Syntactic Influence in LLM Metaphor Processing](https://mlsanity.com/paper/2510.04120/) (2025)
- [Evaluating LLMs on Chinese Idiom Translation](https://mlsanity.com/paper/2508.10421/) (2025)

## Abstract (arXiv)

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.