Evaluation sourcesOfficial evaluation
RWS M-GATE
RWS · 30 languages including Chinese, testing grammar and translation
In NewsObserving
Evidence budgetNot scored
Upstream data as ofTo be confirmed
Last syncedNot collected yet
What it measures, and how
Grammar MCC and translation are stored separately; tokenisation efficiency does not enter the capability score.
How this evidence is used
Under observation: a new model is visible in the browser, but the machine interface returns empty content; there is no verifiable publication time for results or stable export.
Limits and data attribution
Under observation: a new model is visible in the browser, but the machine interface returns empty content; there is no verifiable publication time for results or stable export.
Data licence: Leaderboard data usage boundaries to be confirmed
Scores published by RWS. Raw scores and the News consensus score use different scales and cannot be added directly.