On-the-Fly Fusion of Large Language Models and Machine Translation
This addresses translation quality for NLP applications, but it is incremental as it builds on existing ensembling and prompting techniques.
The paper tackles the problem of improving machine translation by on-the-fly ensembling of a neural machine translation model with a large language model, finding that even a weaker LLM can enhance translations and outperform ensembling two stronger MT models.
We propose the on-the-fly ensembling of a machine translation model with an LLM, prompted on the same task and input. We perform experiments on 4 language pairs (both directions) with varying data amounts. We find that a slightly weaker-at-translation LLM can improve translations of a NMT model, and ensembling with an LLM can produce better translations than ensembling two stronger MT models. We combine our method with various techniques from LLM prompting, such as in context learning and translation context.