{"slug": "llm-evaluators-are-biased-across-languages", "title": "LLM Evaluators are Biased across Languages", "summary": "Researchers found that LLM evaluators, including trained reward models and prompted judges, assign significantly different scores to semantically identical content across 23 languages, with lower-resource languages receiving more generous scores. This bias persists in frontier models and is invisible to standard pairwise accuracy metrics, meaning harmful content in lower-resource languages is more likely to pass safety filters. The study attributes the bias to structural, language-level misalignment rather than content difficulty alone.", "body_md": "# Computer Science > Computation and Language\n\n[Submitted on 16 Jul 2026]\n\n# Title:LLM Evaluators are Biased across Languages\n\n[View PDF](/pdf/2607.14480)\n\n[HTML (experimental)](https://arxiv.org/html/2607.14480v1)\n\nAbstract:LLM evaluators (trained reward models and prompted LLM-as-a-Judge) are routinely validated via pairwise accuracy. In a multilingual setting, this operates under the premise that high pairwise accuracy implies reliable, language-neutral scoring. We show that this assumption does not hold. We conduct experiments with semantically identical instruction-response pairs across 23 languages, and find that multilingual evaluators assign significantly different scores to different evaluation languages. The bias is statistically significant and consistent across eight open-weight evaluators of different architectures and training paradigms, persists in frontier judges, and is strongly correlated with language resource level: lower-resource languages are scored more generously. Meanwhile, these biases are invisible to pairwise accuracy: evaluators achieve above 90% pairwise accuracy, yet have up to 43% difference in acceptance rate across languages under a global decision threshold, meaning, for instance, that harmful content in lower-resource languages is more likely to pass safety filters. Per-language thresholds would require language identification, which can be defeated by code-switched prompts. We then investigate why lower-resource languages receive higher rather than lower scores, and we find that model uncertainty is linked with the effect: models tend to give higher scores when less confident, both under negative log-likelihood and under token-free uncertainty measures; however, language identity remains a significant predictor after controlling for uncertainty, and the bias cannot be explained away by content difficulty alone, but is a structural, language-level misalignment.\n\n### References & Citations\n\nLoading...\n\n# Bibliographic and Citation Tools\n\nBibliographic Explorer\n\n*(*[What is the Explorer?](https://info.arxiv.org/labs/showcase.html#arxiv-bibliographic-explorer))\nConnected Papers\n\n*(*[What is Connected Papers?](https://www.connectedpapers.com/about))\nLitmaps\n\n*(*[What is Litmaps?](https://www.litmaps.co/))\nscite Smart Citations\n\n*(*[What are Smart Citations?](https://www.scite.ai/))# Code, Data and Media Associated with this Article\n\nalphaXiv\n\n*(*[What is alphaXiv?](https://alphaxiv.org/))\nCatalyzeX Code Finder for Papers\n\n*(*[What is CatalyzeX?](https://www.catalyzex.com))\nDagsHub\n\n*(*[What is DagsHub?](https://dagshub.com/))\nGotit.pub\n\n*(*[What is GotitPub?](http://gotit.pub/faq))\nHugging Face\n\n*(*[What is Huggingface?](https://huggingface.co/huggingface))\nScienceCast\n\n*(*[What is ScienceCast?](https://sciencecast.org/welcome))# Demos\n\n# Recommenders and Search Tools\n\nInfluence Flower\n\n*(*[What are Influence Flowers?](https://influencemap.cmlab.dev/))\nCORE Recommender\n\n*(*[What is CORE?](https://core.ac.uk/services/recommender))# arXivLabs: experimental projects with community collaborators\n\narXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.\n\nBoth individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.\n\nHave an idea for a project that will add value for arXiv's community? [ Learn more about arXivLabs](https://info.arxiv.org/labs/index.html).", "url": "https://wpnews.pro/news/llm-evaluators-are-biased-across-languages", "canonical_source": "https://arxiv.org/abs/2607.14480", "published_at": "2026-07-17 07:07:06+00:00", "updated_at": "2026-07-17 07:21:04.531101+00:00", "lang": "en", "topics": ["large-language-models", "ai-safety", "ai-ethics", "natural-language-processing"], "entities": [], "alternates": {"html": "https://wpnews.pro/news/llm-evaluators-are-biased-across-languages", "markdown": "https://wpnews.pro/news/llm-evaluators-are-biased-across-languages.md", "text": "https://wpnews.pro/news/llm-evaluators-are-biased-across-languages.txt", "jsonld": "https://wpnews.pro/news/llm-evaluators-are-biased-across-languages.jsonld"}}