Cross-Lingual LLM-Judge Transfer via Evaluation Decomposition
This addresses the problem of costly and scarce human-annotated judgments for non-English languages in AI evaluation, offering a practical solution for multilingual deployment.
The paper tackles the challenge of extending automated evaluation of large language models beyond English by introducing a decomposition-based framework with a Universal Criteria Set, achieving consistent improvements on faithfulness tasks across languages and model backbones without needing target-language annotations.
As large language models are increasingly deployed across diverse real-world applications, extending automated evaluation beyond English has become a critical challenge. Existing evaluation approaches are predominantly English-focused, and adapting them to other languages is hindered by the scarcity and cost of human-annotated judgments in most languages. We introduce a decomposition-based evaluation framework built around a Universal Criteria Set (UCS). UCS consists of a shared, language-agnostic set of evaluation dimensions, producing an interpretable intermediate representation that supports cross-lingual transfer with minimal supervision. Experiments on multiple faithfulness tasks across languages and model backbones demonstrate consistent improvements over strong baselines without requiring target-language annotations.