wxrks
Studies from wxrks on people, AI and translation, with the method and the data in every paper.
Research we publish so you can check our work.
Working paper · WP-2026-02
The economics of AI post-editing, run two ways for two years. In the base case the expert is paid about twice the post-editing rate per word, and by year two the program costs about half and ships a quarter of the MQM penalty points.
Updated 2026-11-10Read
Working paper · WP-2026-01
How much humans change machine suggestions, measured across 27 million translated segments. Context-sensitive output needs 47% less editing than neural machine translation.
Updated 2026-10-14Read