EVALUATIONS
Browse published results by task domain. Read conclusions alongside the evaluated objects, configurations and evidence scope.
2 reports
Compare 300 final deliveries from Qianwen Office, WPS Lingxi, WorkBuddy and Doubao Work across office files, spreadsheets, PPT and document questions.
Five suites with 15 tasks each. Assistance, continuation and retries were permitted; this measures final delivery quality, not unassisted first-attempt success.
Explore configuration-level dimension scores and sample sizes, with six representative task deliveries and verification summaries.
Tasks and sample sizes differ between configurations. These scores do not form a controlled cross-product ranking or measure plugin effects.
Task-level delivery checks and usage costs. Full articles are in Chinese.
Three task deliveries illustrate field checks, range completeness and the difference between stored values and percentage formatting.
Compare edit locations, text scope, source alignment and presentation quality using deliveries from the same tasks.
Recorded charges by task type, differences between PPT generation and editing, and limits to billing coverage.