Cotton
Translation workflow
Turning generated translations into reviewable results.
pinxlab AI R&D: a game-dialogue translation workflow.
Turning generated translations into reviewable results.
pinxlab AI R&D: a game-dialogue translation workflow.
- Verdict rule
When the judge model reports a hard-gate failure such as a glossary violation, the translation cannot pass however high its mean score on other dimensions. Generation and judgment get the same glossary, and code aggregates the verdict.
Whether a gate failed is what the judge model reports; this does not guarantee that every banned term is found.