How to Balance Cost and Quality in AI Translation Evaluation
Researchers show how to reduce computational overhead in LLM-based translation evaluation while maintaining accuracy, offering...
AI translation
From enterprise translation AI to automated localization.
Android’s Instant Translation Gets Enhanced in ‘Circle to Search’
Instant translation will soon be available without navigating away from the “Circle to Search” function...
Which Parts of a Prompt Should Be Translated to Improve Large Language Models?
Researchers at Bar-Ilan University show that translating only specific parts of a prompt improves large...
Meet Tradutor, the First Open-Source AI Translation Model for European Portuguese
Researchers introduce Tradutor, the first open-source AI translation model for European Portuguese, which is outperforming...
Google, Unbabel Expand Key AI Translation Benchmark to 55 Languages with WMT24++
Google, Unbabel unveil WMT24++, expanding the WMT24 benchmark to 55 languages. Automatic metrics suggest LLMs...
Google Expands Low-Resource AI Translation with SMOL Dataset
Google releases SMOL, a professionally translated dataset designed to improve machine translation for 115 low-resource...
Makers of Open Source Model Hibiki Promise ‘High-Fidelity’ Speech Translation
French AI research lab Kyutai presents Hibiki, a model for real-time, “high-fidelity” speech translation, designed...
Cascaded Speech Translation Systems Outperform End-to-End Models, Research Finds
SpeechT mentorship brings researchers and practitioners together to examine speech translation, finding cascaded architectures deliver...
Xiaomi’s Training Recipe for Better Multilingual AI Translation
Xiaomi researchers propose a new data strategy that they say significantly boosts the translation performance...
Meta’s BOUQuET Benchmark Brings Linguistic Diversity to AI Translation Evaluation
Meta’s BOUQuET dataset moves AI translation evaluation beyond English, offering a multilingual benchmark to enhance...
Researchers Present DOLFIN, a New Test Set for AI Translation for Financial Content
Lingua Custodia and Grenoble Alpes University introduce DOLFIN, a new document-level test set designed to...
How Apple Wants to Fix Hallucinations in AI Translation
Apple researchers present an “easily scalable” and “highly effective” method to mitigate translation hallucinations in...
China’s AI Researchers Triple Down on AI Speech Translation
Speech-to-speech translation is a priority in China. Within less than 24 hours China-based AI researchers...
Enter OpenEuroLLM, an ‘Open, Compliant, Diverse’ Series of Foundation Models
Having secured close to USD 40m in funding, the EU’s OpenEuroLLM project kicks off under...