Unlocking Latent Discourse Translation in LLMs Through Quality-Aware Decoding
This addresses a specific bottleneck in machine translation for users needing accurate document-level translations, but it is incremental as it builds on existing decoding methods.
The paper tackled the problem of LLMs struggling with discourse phenomena like pronoun resolution in machine translation, and demonstrated that quality-aware decoding (QAD) effectively extracts latent discourse knowledge, improving translation quality and aligning with human preferences.
Large language models (LLMs) have emerged as strong contenders in machine translation.Yet, they still struggle to adequately handle discourse phenomena, such as pronoun resolution and lexical cohesion at the document level. In this study, we thoroughly investigate the discourse phenomena performance of LLMs in context-aware translation. We demonstrate that discourse knowledge is encoded within LLMs and propose the use of quality-aware decoding (QAD) to effectively extract this knowledge, showcasing its superiority over other decoding approaches through comprehensive analysis. Furthermore, we illustrate that QAD enhances the semantic richness of translations and aligns them more closely with human preferences.