WMT24++ (en→xx)

WMT24++ (en→xx) is a machine-translation benchmark: English source texts must be translated into around 55 target languages, including many non-commercial ones. The score reflects translation quality measured against references.

WMT24++ (en→xx) measures performance in languages other than english. In detail, it checks:

• translation quality into a wide set of languages
• faithfulness to the source text
• quality for less common and lower-resource languages
• consistency across language pairs

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better.

What is the WMT24++ (en→xx) benchmark?

WMT24++ (en→xx) is a machine-translation benchmark: English source texts must be translated into around 55 target languages, including many non-commercial ones. The score reflects translation quality measured against references.

What does WMT24++ (en→xx) test?

WMT24++ (en→xx) mainly measures performance in languages other than English. Concretely, it checks: translation quality into a wide set of languages; faithfulness to the source text; quality for less common and lower-resource languages.

How is the WMT24++ (en→xx) score calculated?

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better. The declared maximum score of this benchmark is 100.0.

Where do I get WMT24++ (en→xx) data and scores?

The official benchmark page/reference is linked on this page. Every score on Open LLM Radar is also available as open data: download the scores JSON (http://open-llm-radar-api:8000/api/v1/benchmarks/wmt24-plus-plus-en-xx/scores) or pull it directly from the Open LLM Radar API with any HTTP client.