Direct Preference Optimization for Neural Machine Translation with Minimum Bayes Risk Decoding

arXiv:2311.08380v2 Announce Type: replace
Abstract: Minimum Bayes Risk (MBR) decoding can significantly improve translation performance of Multilingual Large Language Models (MLLMs). However, MBR decoding is computationally expensive. We show how the recently developed Reinforcement Learning technique, Direct Preference Optimization (DPO), can fine-tune MLLMs to get the gains of MBR without any additional computation in inference. Our method uses only a small monolingual fine-tuning set and yields significantly improved performance on multiple NMT test sets compared to MLLMs without DPO.

abstract arxiv bayes cs.cl decoding direct preference optimization however language language models large language large language models machine machine translation mllms multilingual neural machine translation optimization performance reinforcement reinforcement learning risk show translation type