Cross-Lingual LLM-Judge Transfer via Evaluation Decomposition

TL;DR

This paper presents a universal, language-agnostic evaluation framework for cross-lingual assessment of large language models, enabling effective transfer without extensive target-language annotations.

Contribution

It introduces a decomposition-based evaluation method utilizing a Universal Criteria Set (UCS) for cross-lingual transfer with minimal supervision.

Findings

Consistent improvements over baselines across multiple languages.

Effective transfer without target-language annotations.

Supports diverse faithfulness evaluation tasks.

Abstract

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.