A newly revised arXiv preprint analyzes how quantization affects large language models deployed for machine translation in server settings. The work weighs reduced memory use and faster inference against the quality loss that compression can introduce. It aims to help practitioners pick quantization settings that balance latency and translation accuracy.
A new arXiv paper examines how to decide which machine translations of classical texts need expert review when no human reference translation exists. Using Pali-to-English as a test case, it combines source-novelty measures, GEMBA-style quality scoring, and a limited review budget. The goal is a practical way to route scarce expert attention to the most error-prone outputs.
A new arXiv paper analyzes how value is defined and negotiated across the increasingly automated language and translation sector. It draws on interviews with 29 industry stakeholders gathered through the LT-LiDER project. The author discusses what these stakeholder perspectives could mean for how translation is taught.
North Small Translate is a new open-weight translation model that also follows instructions, described as being trained on the same base as Cohere's Command A Plus mixture-of-experts system with 25 billion active parameters. The authors position it as a cost-effective option for machine translation workloads that need instruction-following behaviour.
A new arXiv preprint describes the CRITICS project, which combines large language model machine translation with educational technology to improve access to scientific content and literacy. The work focuses on translation systems tuned for scientific material so that learners can engage critically with research across language barriers. The announcement is an abstract-only listing of a new submission in the cs.CL category.
Researchers have assembled IndicQE-APE, a single benchmark that brings together scattered Indic-language resources for quality estimation and automatic post-editing. The dataset draws on WMT shared task data from 2020 through 2024, allowing models to be trained and evaluated across multiple tasks and language pairs under consistent conditions. The goal is to remove the fragmentation that previously made cross-task and cross-language comparison difficult.
Researchers present a new annotated corpus for classifying sentence functions in Bangla, a resource previously lacking for the language. The work benchmarks several models on the task and adds interpretability analysis of their predictions. Such sentence-type identification supports dialogue systems, speech synthesis, and machine translation.
Researchers describe the process used to create pseudo-references for the WMT26 General Machine Translation task, where ten language pairs lack any human translations or post-edited outputs. The work also covers six additional language pairs that do have references, aiming to give systems a consistent basis for automatic scoring. The paper details how these synthetic references were constructed and evaluated.
A new arXiv paper analyzes how meta-evaluation of machine translation must balance alignment with adequacy versus fluency, noting that the preferred balance shifts depending on which translation systems are included in the evaluation set. Because those system sets are typically small and filtered, the authors propose parameterizing this balance explicitly. The work appears in the cs.AI and cs.LG cross-listings.
Researchers describe SalamandraTA, their entry for the WMT 2026 shared task on terminology-aware translation, where outputs must exactly match glossary-prescribed terms. The paper argues that fine-tuning on all glossary-annotated translation pairs is inefficient and proposes prioritizing hard examples during training. The work is available as an arXiv preprint.
A computational linguistics paper investigates how readers actually experience AI-translated literary works, looking beyond surface accuracy to immersion and literary impact. The findings suggest that while machine translations convey the content acceptably, readers still favor human versions, indicating that common automatic evaluation metrics overlook qualities that matter most in literary reading.