Cross-Lingual Stability of LLM Judges Under Controlled Generation: Evidence from Finno-Ugric Languages
This addresses evaluation instability for morphologically rich languages, highlighting the unreliability of zero-shot judge transfer for discourse-level assessment, which is incremental as it builds on existing cross-lingual evaluation methods.
The study investigated cross-lingual evaluation reliability of LLM judges by controlling generation conditions across Estonian, Finnish, and Hungarian, finding that pragmatic judgments like coherence and instruction-following showed rank inversions and near-zero correlations, indicating instability, while surface-level metrics remained stable.
Cross-lingual evaluation of large language models (LLMs) typically conflates two sources of variance: genuine model performance differences and measurement instability. We investigate evaluation reliability by holding generation conditions constant while varying target language. Using synthetic customer-support dialogues generated with identical parameters across Estonian, Finnish, and Hungarian, we test whether automatic metrics and LLM-as-a-judge scoring produce stable model rankings across these morphologically rich, related Finno-Ugric languages. With a small set of Estonian native speaker annotations as a reference point, we find systematic ranking instabilities: surface-level metrics (lexical diversity, surface and semantic similarity) maintain cross-language stability, but pragmatic judgments (coherence, instruction-following) exhibit rank inversions and near-zero correlations. Because generation is controlled, these inconsistencies reflect how judge scoring behaves differently across languages rather than true model differences. This controlled design provides a diagnostic probe: evaluation methods that fail to maintain stability under identical generation conditions signal transfer failure before deployment. Our findings suggest that zero-shot judge transfer is unreliable for discourse-level assessment in morphologically rich languages, motivating language-specific calibration against targeted human baselines. We release our controlled generation protocol, synthetic data, and evaluation framework to enable replication across language families at https://github.com/isaac-chung/cross-lingual-stability-judges.