Asal Mehrabi, Mohammad Aliannejadi, Clemencia Siro · 2026 · arXiv (Cornell University)
DOI: 10.48550/arxiv.2610.31172 · cited by 0 works (OpenAlex) · open-access PDF
Оценка LLM-as-a-Judge все чаще применяется к многоязычным задачам, но такие стратегии, как Chain-of-Thought и Role Conditioning, рекомендуются исключительно на основе английских ориентиров. Остается неизвестным, эффективность их передачи по моделям и языкам судьи: исследования пробной чувствительности в основном ориентированы на английский язык, в то время как существующие многоязычные исследования судьи фиксируют пробную модель. В этой работе мы устраняем этот разрыв, измеряяя это взаимодействие посредством бенчмарка из 9 пробных стратегий, 4 открытых судей, 27 языков и 4 задач по 13 наборам данных, давая более 1М очковых баллов по результатам золота-референции, подтвержденных по отношению к человеческим су
Machine translation (Meta M2M100); the English original follows.
LLM-as-a-Judge evaluation is increasingly applied to multilingual tasks, yet prompting strategies such as Chain-of-Thought and Role Conditioning are recommended based solely on English benchmarks. Whether their effectiveness transfers across judge models and languages remains unknown: prompt-sensitivity studies are largely English-centric, while existing multilingual judge studies fix the prompt template. In this work, we address this gap by measuring this interaction through a benchmark of 9 prompting strategies, 4 open-weight judges, 27 languages, and 4 tasks across 13 datasets, yielding over 1M pointwise scores on gold-reference outputs, validated against human judgments. We find that prompting strategies are not transferable: the choice of the judge model dominates the choice of prompting strategy, strategy effects reverse direction across judges, and stability degrades in low-resource languages, where these effects are concealed by aggregate reporting. These findings show that prompting effects in multilingual evaluation are conditioned on the judge, task, and language. Consequently, prompting strategies should be validated for the intended evaluation setting, and cross-study comparisons should account for both judge and prompt configuration.
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
Topics: Robustness (evolution), Pointwise, Computer science, Artificial intelligence, Cognitive psychology, Machine learning
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.31172, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange