属性控制器在预训练多语言翻译模型上的可迁移性如何？
How Transferable are Attribute Controllers on Pretrained Multilingual Translation Models?
AI总结：
本文以 NLLB-200 为基础，利用基于梯度的推理时属性控制器，将翻译属性控制零样本迁移到无标注语言，并证明其在远距离语言和领域鲁棒性上可补充微调。
摘要：
定制机器翻译模型以符合所需属性（例如，正式程度或语法性别）是一个已被充分研究的课题。然而，当前大多数方法依赖带有属性标注的（半）监督数据。这种数据稀缺性阻碍了将此类定制能力普及到更广泛的语言，尤其是资源较低的语言。这一差距与预训练大规模多语言翻译模型的最新进展不同步。为此，我们以 NLLB-200 模型为基础，将属性控制能力迁移到没有属性标注数据的语言。受可控生成技术的启发，我们采用基于梯度的推理时控制器来引导预训练模型。该控制器能很好地迁移到零样本条件，因为它作用于预训练的多语言表示，并且是属性特定的，而非语言特定的。通过与基于微调的控制进行全面比较，我们证明，尽管微调在监督设置中具有明显优势，但在转向零样本条件时，尤其是面对新的和远距离目标语言时，它与推理时控制之间的差距会缩小。后者还表现出更强的领域鲁棒性。我们进一步表明，我们的推理时控制可补充微调。在真实的低资源语言孟加拉语上进行的人工评估证实了我们的发现。我们的代码位于 https://github.com/dannigt/attribute-controller-transfer
英文摘要：
Customizing machine translation models to comply with desired attributes (e.g., formality or grammatical gender) is a well-studied topic. However, most current approaches rely on (semi-)supervised data with attribute annotations. This data scarcity bottlenecks democratizing such customization possibilities to a wider range of languages, particularly lower-resource ones. This gap is out of sync with recent progress in pretrained massively multilingual translation models. In response, we transfer the attribute controlling capabilities to languages without attribute-annotated data with an NLLB-200 model as a foundation. Inspired by techniques from controllable generation, we employ a gradient-based inference-time controller to steer the pretrained model. The controller transfers well to zero-shot conditions, as it operates on pretrained multilingual representations and is attribute -- rather than language-specific. With a comprehensive comparison to finetuning-based control, we demonstrate that, despite finetuning's clear dominance in supervised settings, the gap to inference-time control closes when moving to zero-shot conditions, especially with new and distant target languages. The latter also shows stronger domain robustness. We further show that our inference-time control complements finetuning. A human evaluation on a real low-resource language, Bengali, confirms our findings. Our code is https://github.com/dannigt/attribute-controller-transfer
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
