Cross-Lingual Stability of LLM Judges Under Controlled Generation: Evidence from Finno-Ugric Languages

TL;DR

This study examines the stability of large language model evaluations across Finno-Ugric languages, revealing that surface metrics are stable while pragmatic judgments vary, highlighting the need for language-specific calibration.

Contribution

It introduces a controlled evaluation framework to diagnose cross-lingual stability issues in LLM judges, emphasizing the unreliability of zero-shot transfer for discourse-level assessments.

Findings

Surface metrics are stable across languages.

Pragmatic judgments show rank inversions across languages.

Evaluation instability reflects judge scoring behavior, not true model differences.

Abstract

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.