Aixiu An

Reinforcement learning can significantly boost the translation quality of small language models in the complex legal domain, outperforming traditional fine-tuning methods.

Including reasoning in NMT inference boosts translation quality but comes with a significant increase in computational cost.