핵심 개념
ACT-MNMT introduces a novel mechanism for Multilingual Neural Machine Translation, addressing off-target issues and improving translation performance.
통계
LLMs have achieved promising performance in multilingual machine translation tasks.
ACT-MNMT achieves substantially improved performance across multiple translation directions.
인용구
"Large language models have demonstrated remarkable capabilities in various scenarios."
"ACT-MNMT introduces trigger tokens to represent different task semantics and improve translation quality."