The paper investigates gender bias in machine translation evaluation metrics using an occupation-balanced subset of GAMBIT+ across seven English‑source language pairs, including a new German extension. It finds that masculine translations tend to receive higher scores and that biases align with stereotypical gender representations, though the strength varies by evaluator and language. The study highlights that assessing bias requires multiple dimensions beyond a single aggregate measure.
arXiv:2610.33749v1 Announce Type: new Abstract: As machine translation systems approach human-level quality, traditional evaluation methodologies struggle to detect subtle translation errors. We crit...
The paper presents a multilingual audit of stereotype bias in large language models (LLMs) using 12,000 locally written Reddit posts in English, Spanish, French, and German. It evaluates gender, racial, and religious bias across six models (three open-weight and three proprietary APIs) with a two-blank design to reduce grammatical cues, finding that religion shows the highest imbalance, Spanish leads in gender bias, and German in race and religion. The study highlights that aggregate scores can mask category- and context-specific biases, underscoring the need for detailed inspection before deploying LLMs in consequential settings.
arXiv:2608. 08606v1 Announce Type: cross Abstract: Machine translation (MT) systems often fail to correctly translate gender, especially when converting from a gender-neutral language like English to a gendered target language such as Romanian.
arXiv:2609.06634v1 Announce Type: cross Abstract: LLMs' performance on machine translation (MT) tasks is often dependent on the data availability in the specific domains and language pairs that they...
arXiv:2607. 20241v1 Announce Type: cross Abstract: Culturally loaded translation poses unique challenges for machine translation (MT), as meanings are deeply embedded in socio-cultural contexts beyond surface linguistic forms.
The paper introduces a unified framework that simultaneously measures intrinsic (encoded) and extrinsic (expressed) gender bias in large language models using identical neutral prompts. It finds a consistent link between latent gender information and output bias, but shows that alignment via supervised fine‑tuning reduces expressed bias while leaving internal gender associations largely intact and reactivatable by adversarial prompts. The study also demonstrates that debiasing gains on structured benchmarks may not transfer to realistic tasks such as story generation.
The paper audits 51 recent multilingual evaluation benchmarks covering 242 datasets and 219 languages, focusing on coverage, representativeness, and rigor. It finds that while coverage is broad, many languages appear in only one benchmark and entire regions are underrepresented; low‑resource languages face a task equity gap. Representativeness is compromised by heavy reliance on English translations, and rigor is weakened by benchmark contamination and parallel‑corpus overlap. The authors propose concrete recommendations for building evaluation frameworks that are natively constructed, culturally grounded, contamination‑aware, and community‑aligned.
arXiv:2603.12050v3 Announce Type: replace Abstract: Translated texts exhibit systematic differences from comparable texts originally written in the target language. Explaining this phenomenon, common...
Large Language Models (LLMs) are being tested as a cost‑effective alternative to human judgment for evaluating machine translation (MT). This study compares LLM agreement with human annotators on two MT quality schemes—Multidimensional Quality Metrics (MQM) and Error Span Annotation (ESA)—across 70 language pairs and public WMT23/25 data. Results show that while LLMs sometimes match or exceed human agreement, performance varies widely by scheme, language, and domain, and remains unreliable in many settings. The work also identifies specific challenges for both humans (e.g., fine‑grained MQM, low‑resource languages) and LLMs (e.g., minor errors, wrong language variants, span annotation).
The paper audits 51 multilingual evaluation benchmarks covering 242 datasets and 219 languages, focusing on coverage, representativeness, and rigor. It finds that while coverage is broad, many languages appear in only one benchmark and entire regions are underrepresented, with low-resource languages evaluated on far fewer tasks than high-resource ones. The study also highlights that most datasets rely on English translations, creating artifacts and limiting cultural relevance, and that benchmark contamination through translation leakage and corpus overlap undermines rigor. Recommendations call for natively constructed, culturally grounded, contamination-aware evaluation frameworks that truly serve the communities they aim to assess.
arXiv:2603. 23485v2 Announce Type: replace-cross Abstract: Standard evaluation practices assume that large language model (LLM) outputs are stable when prompts are embedded in contextually equivalent discourses.
MBBQ (Multilingual Bias Benchmark for Question-answering) is a curated extension of the English BBQ dataset that includes Dutch, Spanish, and Turkish, designed to measure stereotypes across these languages while controlling for cultural differences and task accuracy. The dataset also provides a parallel control set to assess question-answering performance independently of bias. Experiments with several open-source and proprietary LLMs show that non‑English languages can exhibit greater bias than English, and cross‑lingual bias differences persist except in the most accurate models.