XLGoBench: Detecting cross-lingual skill gaps with algorithmic tasks

In three linesXLGoBench is a benchmark of synthetic algorithmic tasks to detect cross-lingual gaps in LLM abilities. The benchmark is commensurate across languages, scalable (variable complexity), quantifiable (objective correctness), and transparent (auditable templates). Experiments reveal persistent cross-lingual gaps in multiple state-of-the-art models.

Summary generated by Claude — human-verified